II / DISTILLATION / Energy Management 1005
Energy Management
T. P. Ognisty, The M. W. Kellogg Company,
Houston, TX, USA
Copyright ^ 2000 Academic Press
Introduction
Distillation dominates separation processes in the
petrochemical industry and consumes nearly 95%
of the total energy used in commercial separations.
The energy used in distillation is approximately
3% of the total energy consumed in US industry.
Distillation columns provide reliable, cost-effective
separation for large throughputs and high purity
product requirements. Distillation is an energy-
intensive process and economics drive the need to
optimize energy consumption in all aspects of its
design.
In the pioneer years of distillation development,
energy was comparatively cheap compared to capital
costs. ReSux ratios of 1.2}1.5 times the minimum
were commonly used to obtain an economical
balance between operating and Rxed costs. Columns
were designed with large design margins to allow
for feed changes and the lack of accuracy in physical
properties, tray hydraulics and tray efRciency.
By providing excess exchanger and tray capacity,
product purity could be maintained by adjusting
the tower operation to compensate for design
uncertainty.
With the advent of the oil crises of the 1970s and
1980s, the energy costs became the major factor in
column costs and created an urgency to Rnd ways to
reduce the energy requirements of distillation. Several
techniques for analysing columns and plants were
extensively developed and published in this era.
Among the notable advances is ‘exergy’ analysis,
developed primarily in Europe, and ‘pinch’ techno-
logy. This trend has subsided somewhat as more
efRcient methods of design and experience have
eventually replaced the older technologies. The more
exciting opportunities are in revamping older plants
and designing new products. The introduction of high
speed computers has also played a signiRcant role in
producing more accurate and more extensively eva-
luated designs. Physical and transport properties
continue to be obtained more accurately. As the regu-
latory requirements for pollutant emissions and
wastewater reduction became more stringent, many
new studies has been directed towards reducing emis-
sions from Rred heaters and steam boilers used to
supply heating requirements in columns. Wastewater
reductions of steam condensate blowdowns and cool-
ing water make-up have become subject to tighter
restrictions. Distillation column design will become
more extensive and complex as the stringency of
requirements continues to increase.
The primary objective of distillation column design
is to produce the desired products with the minimum
amount of energy expenditure and capital cost.
The energy consumption in distillation columns is
affected by the physical properties of the system,
the utilities available for heating and cooling,
the internals for contacting the liquid and vapour
and the arrangement of the column separation
sequence.
Measurement of Energy Performance
In order to determine the energy performance of
a distillation column, two thermodynamic principles
are applied to give a measurement method for deter-
mining the value of design improvements.
Availability or Exergy Analysis
The concept of availability or available useful work is
a convenient way to determine the thermodynamic
minimum amount of energy required for doing work
in a steady Sow process. The availability is referred to
as exergy in Europe. The availability function is de-
Rned as:
A,H!TaS
where H is the enthalpy, S is the entropy and Ta is the
reference temperature (298 K). A higher value of
availability indicates that more work can be extracted
as compared to a lower value. For example, high
pressure steam is more valuable than steam at atmos-
pheric pressure, although the latent heat values are
about equal.
The change in availability represents the amount
of shaft work that can be extracted from a system
as it Sows from an initial to Rnal state. It also
deRnes the minimum work required to achieve
a change in a Sowing system. This available shaft
work, Ws, is deRned in terms of the availability
change, or:
!Ws,
A"
H!Ta
S
These properties are readily available from data refer-
ences or process simulators and can be applied to all
types of plant equipment in a process.
1006 II / DISTILLATION / Energy Management
Heat Engines
A Carnot heat engine is useful in determining the
theoretical maximum heat efRciency of an ideal,
reversible distillation. The heat engine analogy con-
sists of the simple column arrangement depicted in
Figure 1. Heat is added at the bottom of the tower at
a high level of temperature and then removed at the
top of the tower at a lower temperature. This degra-
dation in heat provides the work to separate the feed
into the products. If we consider a distillation column
as a heat engine, then the theoretical minimum work
required for the heat engine is:
Wheat"Qreboiler(1!(Ta/Treboiler))
!Qcondenser(1!(Ta/Tcondenser))
The availability change of the streams in and out of
the column is equal to this heat work minus a lost
work term to account for thermodynamic inefRc-
iency, or:
Aproducts!Afeeds"Wheat!Wlost
To account for real processes, a lost work term is
included to account for irreversible changes within
the column. This extra work is required to overcome
pressure, temperature and composition driving for-
ces. There are a number of commercial algorithms for
determining each driving force contribution to lost
Figure 1 Distillation heat engine analogy.
work. These lost work contributions are generally
classiRed as:
1. Momentum loss due to pressure drop.
2. Heat transfer loss from temperature difference.
3. Mixing losses due to composition differences.
As the lost work term increases, the reboiler and
condenser loads must increase in order to supply the
same amount of useful work. As the driving forces
approach zero, the lost work approaches zero and less
energy is wasted.
For a particular column design, the availability of
the feed and products and the theoretical minimum
work are Rxed. The lost work energy is solved by
difference. The actual heat duty required for the dis-
tillation column will be the sum of the theoretical
minimum plus the lost work, or (Wheat#Wlost). By
comparing values of this sum, the designer has a
yardstick in which to rate different column
designs. For existing columns, a simple calculation
of the total lost work will quickly focus as
approach towards the most potentially beneRcial
areas. For example, if the lost work is less than 5%
of the total heat duty, then this is probably within
the design margin of the equipment and not worth
pursuing.
Distillation Design Guidelines
A thermodynamic analysis provides information
about the theoretical limits and does not address the
issues of expense and implementation. Exergy studies
should include capital costs with a payback period for
a proper economic assessment.
The following list is a collection of design guide-
lines and practical constraints to be considered in
searching for better energy schemes:
1. In order to distil vapour and liquids, the pressure
must be below the critical pressure.
2. Accurate physical properties are required; proper
models for nonideal solutions and azeotropes
should be checked.
3. Thermally unstable components may limit the
reboiler temperatures.
4. Air and/or water-cooling are preferred to the
more expensive refrigeration.
5. Operational Sexibility is usually required for feed
changes and plant turndown.
6. Distillation columns do not work well at less
than the minimum reSux ratio and at pinch
points.
7. Heat integration may not be practical; transport-
ing Suid over large distances may not be econ-

omical; excessive temperatures result in Rlm
boiling; interaction between columns should be
controlled.
8. Shipping weight and total height restrictions tend
to limit the maximum number of trays or packing
in one tower shell.
9. Lowering the operation pressure is a simple way
to reduce energy consumption. Increasing pres-
sure is the simpler way to increase capacity.
10. Lower energy consumption usually requires
higher capital costs.
McCabe+Thiele Diagrams
One of the best visual aids for distillation is the
ubiquitous McCabe}Thiele diagram. Despite its
many limitations, the diagrams show the basic rela-
tionship between the equilibrium curve and the com-
position proRles throughout the column. A typical
McCabe}Thiele diagram is illustrated in Figure 2.
Heating or cooling changes the internal vapour and
liquid Sows and this effect can be seen as a shift
in the slope of the operating lines in the diagram. For
energy management, we are interested in seeking
ways to minimize the difference (the composition
driving forces) between the operating and equilibrium
curves.
The absolute minimum energy expenditure will
occur if external heating or cooling is applied to each
stage to adjust the operating line so that it coincides
with the equilibrium curve. This condition is imprac-
tical, since an inRnite number of stages would be
Figure 2 Typical McCabe}Thiele diagram.
II / DISTILLATION / Energy Management 1007
required. The goal is to design a column with an
approach as close as possible to equilibrium, but with
an acceptable number of stages. A few successful
designs exist that provide internal continuous
cryogenic cooling in a small tray section. These types
of columns that have continuous, internal heat ex-
changers are called dephlegmators.
There are numerous practical applications in the
industry where adding just a few select external heat-
ing or cooling locations will provide a cost-effective
reduction in energy.
Intermediate Heating and Cooling
Pumparounds
One of the earliest methods for saving signiRcant
amounts of energy was developed for crude units.
Pumparounds are used extensively in atmospheric
and vacuum towers to remove heat at selected stages.
Liquid is withdrawn from an intermediate product
stage, externally cooled and then pumped back to the
column at a higher elevation. This arrangement is
shown in Figure 3. The hot liquid from the tower is
used to supply heat to several sources. The remaining
cooled liquid is returned to the column and used to
cool the column vapour. Pumparounds have the ef-
fect of shifting the operating lines at select intervals
closer to the equilibrium curve to improve energy
consumption. The sub-cooled liquid provided in the
pumparound reduces the cold utility requirements
and size of the overhead system. The cooling tower
and condensers are less expensive and thermal and
wastewater pollution is reduced. The hot pum-
paround liquid has the additional role of recovering
heat supplied in the feed furnaces by transferring it to
feed pre-heat, steam generation and heating for
downstream units.
Intercondensers and Interreboilers
Another useful technique for reducing the utility
loads is to add intermediate exchangers. Figure 4
shows the effect of adding an inter-reboiler on
a McCabe}Thiele diagram. The beneRt of this modiR-
cation is that the reboiler heat duty is reduced by the
amount of duty used in the inter-reboiler. Since the
temperature level will be less than the reboiler tem-
perature, it may be possible to use a process stream
for the heating. If this heat integration is feasible, then
signiRcant savings of the reboiler utility, such as high
pressure steam, can be achieved. The internal column
Sows below the interreboiler stage will be reduced,
but the stage requirement will increase since the oper-
ating line will be closer to the equilibrium curve. For
this reason, the addition of intermediate exchangers is
1008 II / DISTILLATION / Energy Management
Figure 3 Pumparound section.
usually economical only when a few extra stages will
be required. An analogous result will be obtained
with an inter-condenser. The condenser utility load
will be reduced, the internal Sows in the stages above
the condenser will be reduced, but more stages will be
required. The beneRts of intermediate cooling or con-
densing are particularly economical when an expen-
sive refrigerant is used as the condenser cooling
medium.
There are semi-rigorous calculation methods for
determining how much heat can be added or removed
Figure 4 McCabe}Thiele diagram with interreboiler.
Figure 5 Column grand composite curve.
at a particular temperature level. One technique,
which is an extension of ‘pinch’ technology, is to
generate a column grand composite curve (CGCC).
These curves plot the heat transferred between
the vapour and liquid versus stage temperature.
A typical example is shown in Figure 5. The best
opportunities for using intermediate heat exchangers
exist in the regions where large amounts of heat
are exchanged at temperatures signiRcantly different
to that of the reboiler or condenser. When deciding
upon how much heat will be added or removed,
one should consider the extra stages that will be
required and allow for a suitable design margin for
Sexibility. If this type of region exits near the feed
locations, then feed conditioning may be an option.
Adding a feed exchanger will generally be less expen-
sive than adding an intermediate exchanger on the
column.
Feed Conditioning
The feed is introduced into the column where the
temperature and composition roughly match the col-
umn proRle. The feed is at higher pressure in order to
Sow into the column. The location of the feed stream
can be optimized with a few judicious choices with
a process simulator. The feed quality and location can
also be adjusted within limits to lower the reboiler or
the condenser duty, but not both. The best opportuni-
ties for utilizing feed exchangers are during revamps
when there is excess vapour- and liquid-handling ca-
pacity in one section of a column. For example, if the
trays above the feed can handle more Suid trafRc
without modiRcation, but the bottom section
trays are capacity-limited, then installing a feed pre-
heater can be used to unload the bottom trafRc. The
 II / DISTILLATION / Energy Management 1009

energy penalty will be increased condenser duty, but

internal modiRcations or tower replacement may be
avoided.

Prefractionation
In a multi-component column separation, the pres-
ence of lighter nonkey components will limit the
purity of the light key. At the top of the tower, the
light key will actually be the heavier component and
its composition will reach a maximum value and then
decrease as it is fractionated from the lightest compo-
nents. By removing these lighter non-key components
from the feed in a pre-absorber, the column can now
produce a higher purity product. Removing the light
components from the feed reduces the Sow to the
column, so that pre-absorbers can be used to unload
the rectifying section of an existing column. In ana-
logous fashion, the addition of a pre-stripper will
unload heavy components from the stripping section
of a column. Figure 6 shows an illustration of a pre-
stripper column arrangement. The additional feed to
the last column will usually provide a modest energy
saving for that column.
Heat Pumping
A signiRcant saving in condenser duty can be
achieved by compressing the overhead vapour to
a temperature that can be used to reboil the bottoms
liquid. A typical heat-pumped column is shown in
Figure 7. The compressor is a signiRcant addition to
the total capital costs. This scheme is feasible if the
distillation involves a close boiling mixture where the
top and bottom temperatures are not signiRcantly
different. The most common industrial application is
propane/propylene splitters, which require large
Figure 7 Heat-pumped column.
amounts of heating and cooling energy. A large re-
duction in cooling water and wastewater products
provides a large incentive to heat pump these types of
columns.
Extractive and Azeotropic Distillation
Instead of modifying the heat exchange in columns,
a reduction in energy is possible by altering the phys-
ical properties of the mixtures. For mixtures that are
difRcult to distil, a solvent may be added to increase
the relative volatility of the mixture. If a suitable
solvent can be found with a relatively low volatility,
then the separation process is known as extractive
distillation. A typical extraction distillation column
arrangement is illustrated in Figure 8. This
process is economical to use when a small amount of
inert solvent can permit an easier (less expensive)
separation of the original products. An additional
tower is required to remove the solvent from the
heavier product.

Figure 6 Prestripper column sequence. Figure 8 Extractive distillation configuration.

1010 II / DISTILLATION / Energy Management

If the addition of a light component can alter an acting control of the column reduces the energy waste
azeotrope, then the process is termed azeotropic dis- of recycling off-spec product.
tillation. The entrainer often forms a new low boiling
azeotrope that is relatively easy to separate from the
heavier product. The entrainer is separated from the
Column Sequencing
lighter product with an additional column and/or The sequencing and arrangement of columns to ob-
condensed to two liquid phases that are separated in tain multiple products has a major impact on energy
a settling drum. Azeotropes are common in the and capital costs. In order to separate X amount of
specialty chemicals industry and most conRgurations products, a minimum of X!1 simple, conventional
are proprietary. columns are required. If the products are removed as
Extractive or azeotropic distillation should be con- distillates, the arrangement is called a direct se-
sidered a last resort option. Finding a suitable solvent quence. The direct sequence is favoured when the
or entrainer that is inexpensive and relatively inert is distillations are at high pressure and require refriger-
the key to a practical design. ation, such as oleRn production. An indirect sequence
is used if the products are withdrawn as bottoms.
Indirect sequencing favours lower heat consumption.
Column Internals Selection In many petrochemical plant conRgurations, both se-
The choice of internals can have impact on energy quences are used or mixed as dictated by the many
efRciency, control response and downstream equip- product requirements. An illustration of the direct
ment. The function of trays or packing is to promote and indirect sequence arrangement is given in
efRcient heat and mass transfer by intimate contact- Figures 9 and 10.
ing of the vapour and liquid. For trays, a portion of A number of heuristics have evolved from experi-
the vapour energy is expended in the form of pressure ence of sequencing columns. A few of the more com-
drop to produce liquid droplets and to overcome the mon guidelines are listed below:
liquid head on the tray. In contrast, the contact area
1. Leave the most difRcult separations to the end.
in packing is mainly created by the spreading of liquid
The binary separation of high purity products re-
over the metal surfaces and by rivulet Sow. For these
quires larger towers and higher energy consump-
reasons, the pressure drop in packing is signiRcantly
tion when non-key components are present.
less than in trays.
2. Direct sequences are favoured when heating costs
This becomes an important characteristic in high
are less than cooling costs.
vacuum distillation, where the pressure and temper-
3. Less energy is expended when the top and bottom
ature changes dramatically from top to bottom as
product Sows are about equal.
pressure drop increases. A tower pressure change
from 50 mm Hg to 25 mm Hg will double the vol- The column sequencing depends upon numerous
umetric Sow of vapour and will have a major impact factors that make each case different. ReRning, ethy-
on the tower size. The normal design pressure drop lene, aromatic and specialty chemical plants all have
for one tray is 3}5 mm Hg, which will limit the num- unique properties and idiosyncrasies that make
ber of trays that can be used. Low pressures and generalizations difRcult. The column arrangements
temperatures in the top of the tower have an impact
on the cost and performance of the condenser and
vacuum system. Using lower pressure drop internals
reduces the exergy losses associated with momentum
and temperature difference. For these reasons, packed
ns are favoured for vacuum distillations.
There also is an advantage to lowering pressure
drop in a column when the overhead vapour is fed to
a compressor. Changing from trays to packing during
a capacity increase revamp is a convenient way to
avoid compressor changes. If the pressure drop
through the tower is reduced, the increase in
compressor suction pressure may just compensate for
the new capacity increase.
By virtue of lower pressure drop and liquid hold-
up, packed columns have relatively faster response to
changes in Sow and composition changes. The faster Figure 9 Direct column sequence.

Figure 10 Indirect column sequence.
are commonly integrated with the total plant
optimization to minimize overall capital costs, energy
consumption and pollution reduction. Techniques
such as pinch technology and exergy analysis are
used to evaluate the global effect of equipment
changes to arrive at the best plant design, which may
not coincide with the best column design.
Distributed Distillation
Distributed distillation is a separation technique that
minimizes the lost work due to mixing and recycling
of Suids within a column. The energy consumption
can be reduced by making a separation between the
most volatile and least volatile components with the
rest of the components distributed between the top
and bottom. An easy separation between components
Figure 11 Distributed column sequence.
II / DISTILLATION / Energy Management 1011
with the highest relative volatility will yield the min-
imum boil-up and reSux for any column. This con-
cept is illustrated in Figure 11. A distributed column
sequence will theoretically consume less energy, but
will require more columns. The energy savings have
been reported to be upwards of 30%.
To lower the capital costs, the column can be
thermally coupled as described in the Petlyuk conRg-
uration shown in Figure 12. Thermal coupling is used
to describe the situation when liquid from a column is
used to supply reSux and vapour is used to supply
boil-up to another column. Several variations of
this arrangement are used to eliminate heat ex-
changers in column designs. Combining the two
columns together and providing a partition to form
a divided wall column reduces the capital cost. Three
products can be produced from one column, one
reboiler and one condenser, as compared to a conven-
tional sequence that requires two columns and four
exchangers. The commercial use of divided wall col-
umns tends to be limited to separations with signiR-
cant amounts of intermediate components.
The Future of Energy and Distillation
Distillation is a mature separation technology and
will remain dominant in the near future. Despite the
high energy requirements, distillation is a cost-
effective method for separating large quantities of
material into high purity products. Other separation
methods are speciRc in application and lack the
1012 II / DISTILLATION / Energy Management
Figure 12 Petlyuk configuration.
versatility of distillation. Current research efforts
have been evolutionary and the majority of the indus-
try is reluctant to test any technology that has not
been commercially proven. Therefore, even promis-
ing new technology changes are slow and cautious.
Many research efforts have started to combine
the best features of different separation methods
to save energy and capital costs. The future of
distillation will always depend upon energy costs
and change will occur when these costs become unac-
ceptable.
See also: II/Distillation: High and Low Pressure Distilla-
tion; Historical Development; Modelling and Simulation;
Multicomponent Distillation; Packed Columns: Design and
Performance; Pilot Plant Batch Distillation; Theory of Dis-
tillation; Tray Columns: Design; Vapour-Liquid Equilib-
rium: Theory.
Further Reading
Dhole VR and Linhoff B (1993) Distillation column targets.
Computers in Chemical Engineering 17 (56): 549}560.
Dodge BF (1994) Chemical Engineering Thermodynamics.
New York: McGraw-Hill.
Fonyo Z (1974) Thermodynamic analysis of rectiRcation}I.
Reversible model of rectiRcation. International Chem-
ical Engineer 14(1): 18}27.
Kaibel G Blass E and Kohler J (1990) Thermo-
dynamics}guidelines for the development of distillation
column arrangements. Gas Separation and PuriTcation
4: 109}114.
Kaibel G (1992) Energy integration in thermal process-
engineering. International Chemical Engineer 32 (4):
631}641.
King CJ (1971) Separation Processes. New York: McGraw-
Hill.
Liebert T (1993) Distillation feed preheat } is it energy
efRcient? Hydrocarbon Processing 3: 37}42.
Linhoff B, Dunford H and Smith R (1983) Heat integration
of distillation columns into overall process. Chemical
Engineering Science 38: 1175}1188.
Petterson WC and Wells TA (1977) Energy-saving schemes
in distillation. Chemical Engineer (September): 78}86.
Sussman MV (1980) Availability (Exergy) Analysis. Lexin-
gton, MA: Mulliken House.
Terranova BE and Westerberg AW (1989) Temper-
ature}heat diagrams for complex columns. I. Inter-
cooled/interheated distillation columns. Ind. Eng.
Chem. Res. 28: 1374}1379.
Treybal RE (1968) Mass-Transfer Operations, 2nd edn.
New York: McGraw-Hill.
US Department of Energy (1994) Assessment of Potential
Energy Savings in Fluid Separation Technologies. Docu-
ment DOE/ID/124763-1, December. Washington, DC:
US Department of Energy.
Vogler TC (1988) Thermodynamic availability analysis for
maximizing a systems efRciency. C.E.P. 25}42.

Extractive Distillation
F.-M. Lee, GTC Technology Corporation,
Houston, TX, USA
Copyright ^ 2000 Academic Press
Introduction
In the Rrst half of the twentieth century, extractive
distillation (ED) became an important industrial
process when World War II demanded high purity
toluene for explosive production and butadiene for
synthetic rubber production. Over the years, substan-
tial developments in ED have been carried out in
terms of novel solvent discovery for a particular sep-
aration, as well as the development of more sophisti-
cated ED tower internal designs. In the petroleum and
petrochemical industries, ED has been found effective
in separating mixtures of aromatics/nonaromatics,
dioleRns/oleRns, oleRns/parafRns and naphthenes/
parafRns.
This article will brieSy review the basic concept of
ED, and summarize the development of ED technolo-
gies for the applications in the following areas:
1. Aromatic puriRcation from reRning and petro-
chemical streams.
2. CycloparafRn (cyclohexane or cyclopentane) re-
covery from naphtha or natural gas liquid.
3. Light oleRns from light hydrocarbon mixtures.
The basis of ED is the increase of relative volatility
between the close-boiling components caused by in-
troducing a selective solvent, which has stronger af-
Rnity with one type of the components in the mixture.
If there is a single liquid phase (no phase separation),
the solvent selectivity can be measured from the ex-
perimentally observed relative volatility (
) between
the key components in the presence of solvent as:

"(Y1/X1)/(Y2/X2)
where X1 and X2 are the mole fractions of compo-
nents 1 and 2, respectively, in the liquid phase, and
Y1 and Y2 are those in the vapour. All compositions
are measured on a solvent-free basis.
In some cases, liquid-phase separation may occur
in the ED tower, especially in the upper portion of the
tower where less polar components are concentrated.
Under this condition, the solvent phase can reject
a second liquid phase, which can be deRned as the
rafRnate liquid phase. The liquid in the solvent-rich
phase is deRned as the extract liquid phase. The
solvent selectivity is determined by the relative vola-
II / DISTILLATION / Extractive Distillation 1013
tilities of the key components in the two interrelated
liquid phases and the common vapour phase, accord-
ing to the following formulae:

r"(1r p1r)/(2r p2r)

e"(1e p1e)/(2e p2e)
where
r and
e are the relative volatilities of compo-
nents 1 and 2, respectively, in the rafRnate phase and
extract phase; 1r and 2r are the activity coefRcients in
the rafRnate phase, and 1e and 2e are the activity
coefRcients in the extract phase; and p1r, p2r, p1e, and
p2e are the vapour pressures of the pure components
(which can be estimated from an Antoine equation).
A schematic diagram of a typical ED process is
presented in Figure 1. During a normal run, a polar,
high-boiling (low volatility) solvent is introduced to
near the top of the ED tower. As the nonvolatile
solvent Sows down the column, it preferentially asso-
ciates the more polar components in the ascending
vapour mixture, thus increasing the relative volatility
between the polar and less polar components. The
process feed stream is introduced to the middle por-
tion of the ED tower. The more polar components are
concentrated in the rich solvent, exiting the bottom of
the ED tower, while the less polar components are
concentrated in the overhead rafRnate stream. The
tower reSux stream is provided to knock down the
entrained solvent from the overhead rafRnate stream.
The solution, rich in polar compounds from the
bottom of the ED tower, is fed to the solvent stripper,
where the polar components are stripped from the
solvent by heat alone or by heat and a stripping gas,
such as steam. The lean solvent is then recycled to the
ED tower from the bottom of the stripper.
Aromatic Puri\cation from Re\ning
and Petrochemical Streams
Advantages and Principle of ED Technology
Although liquid}liquid extraction (LLE) technologies
have dominated the industrial processes for purifying
benzene, toluene, xylene (BTX) aromatics from reRn-
ing and petrochemical streams, ED technologies have
gained ground quickly since the 1980s for more re-
cent grassroots plant installations. In comparison to
LLE, ED has the following advantages:
1. Lower capital costs. ED requires two major pro-
cess units (ED tower and solvent stripper), while
1014 II / DISTILLATION / Extractive Distillation
Figure 1 Configuration of an ED process.
the popular LLE, using sulfolane as the solvent,
requires four major process units, including
LLE tower, extractive stripper, solvent
recovery column, and rafRnate wash tower (see
Figure 2).
2. Higher operational Uexibility. LLE uses only sol-
vent selectivity (polarity) for separation, while ED
uses both solvent selectivity and boiling point for
separation, so it has one extra dimension for op-
erational Sexibility.
Figure 2 Configuration of liquid}liquid extraction using sulfolane for aromatic recovery.
3. Less physical property restrictions. Interfacial ten-
sion and density difference between the liquid
phases are important concerns for LLE, but not for
ED.
The principle of ED for aromatic puriRcation was
studied as early as 1944. One example was the recov-
ery of toluene from parafRns using phenol as the
selective solvent. The effect of phenol on a paraf-
Rn}toluene mixture is plotted in liquid}vapour

Figure 3 Effect of phenol on the vapour}liquid equilibrium of
paraffin and toulene. Numbers on curves refer to mol% solvent in
liquid.
diagrams as shown in Figure 3, in which the parafRn
is considered as a hypothetical octane having the
same boiling point as toluene. In the absence of phe-
nol, there exists an azeotrope of parafRn and toluene.
However, at 50 mol% phenol, the azeotrope is de-
stroyed and the mixture is easily separated; at
100 mol% phenol, the separation between parafRn
and toluene becomes very easy. Figure 4 illustrates
the effect of phenol on the change in relative volatility
between parafRn and toluene. Phenol causes an in-
crease of activity coefRcient for both parafRn and
toluene, but the activity coefRcient of the parafRn
increases to a greater extent than that of toluene.
Therefore, the relative volatility of parafRn over tol-
uene can be increased from 1.0 (no separation) to 3.7
(easy separation) at near zero hydrocarbon concen-
tration in phenol (inRnite dilution).
Figure 4 Effect of phenol on the activity coefficient of paraffin
and toluene.
II / DISTILLATION / Extractive Distillation 1015
The vapour}liquid equilibria of the para-
fRn}toluene}phenol system were applied to test
a commercial ED tower for toluene puriRcation. As
shown in Figure 5, the McCabe}Thiele diagram,
drawn on a phenol-free basis, was used to carry out
the theoretical calculations from tray to tray in the
ED tower.
The calculated results were then compared with the
actual results generated from a commercial ED tower
with 2.1 m diameter and 65 trays. The hydrocarbon
feed tray and the solvent feed tray are located at trays
19 and 39 (counted from the bottom of the tower),
respectively. The tower was operated at a solvent-to-
feed ratio (S/F) of 2.5, a reSux-to-overhead ratio
(R/D) of 2.75, and reboiler temperature at 1703C
under 1.3 atm bottom pressure. On the basis of the
charge, overhead and bottoms analyses, tray-to-tray
calculations were made.
Figure 6 shows the calculated concentration pro-
Rles for each component plotted against theoretical
tray number. It also shows the plot of the tray ana-
lyses against actual tray number. The overall efRcien-
cies calculated over small sections of the tower are
given in Table 1. The average of the overall tray
efRciencies throughout the tower is about 50%.
Based on the above principle, much more rigorous
algorithms for tray-to-tray calculation of ED towers
for multicomponent systems have been developed in
Figure 5 McCabe}Thiele diagram for paraffin and toulene sep-
aration in the presence of phenol. Part (B) is an enlargement of
part of the diagram in (A).
1016 II / DISTILLATION / Extractive Distillation

Figure 6 Calculated versus actual concentration profile of the componenets in an ED tower. Key: 䊏, methylcyclohexane;
䉱, toluene; ; phenol; ***, calculated values.

recent years, with the help of advanced vapour}liquid
equilibrium theories and high-speed computers.
Handling Two Liquid Phases in ED Towers
One of the challenges of ED technology for aromatics
puriRcation is the handling of the possible formation
of two liquid phases in the upper portion of the ED
tower where nonaromatics are concentrated. The oc-
currence of a second liquid phase is due to the fact
that the nonaromatics, such as parafRns, naphthenes
and oleRns, have signiRcantly lower solubility in the
polar solvent than aromatics.
One way to solve the problem of two liquid phases
in the ED tower is to select a polar solvent that has
Table 1 Tray efficiency of ED tower for toulene purification
Section of Theoretical Actual Overall
tower trays trays trays efficiency
(%)
enough solvency to dissolve both aromatics and
nonaromatics in the mixture under process condition.
In general, however, solvents with a high selectivity
for compounds to be separated will have a reduced
solvency (capacity), and vice versa. The selectivity
versus solvency of the common commercial solvents
for aromatic extraction is shown in Figure 7. There-
fore, in order to eliminate two liquid phases, one may
have to compromise the solvent selectivity, some-
times to a great extent.
A better way is to cope with two liquid phases in
the ED tower, without sacriRcing the solvent selectiv-
ity, for the following reasons:
1. Two liquid phases normally reduce the solvent
selectivity in the three-phase equilibrium (va-
pour}liquid}liquid) condition in the ED tower.
However, this can be compensated by intrinsic
selectivity of a highly selective solvent. For
example, the performance of sulfolane was

Below phenol feed tray

1}3 1.8 3 60.0
4}7 2.7 4 67.5
8}11 2.1 4 52.5
12}15 1.5 4 37.5
15}18 1.8 3 60.0
23}27 2.1 4 52.5
27}30 2.5 4 62.5
31}34 2.65 4 66.0
35}39 2.35 5 47.0

Above phenol feed tray

43}65 10.8 23 47.0
49}65 8.6 17 50.7
57}65 4.5 9 50.0
61}65 2.8 5 56.0 Figure 7 Selectivity versus solvency (solubility) of the common
commercial solvents. DEG, diethylene glycol; TEG, triethylene
All trays numbered from bottom of tower. glycol; DMS, dimethyl sulfoxide; NMP, N-methyl pyrrolidone.

compared with those of N-formyl morpholine
(NFM) and N-methyl pyrrolidone (NMP). The
ability of these solvents to enhance the relative
volatility of n-heptane over benzene (an aromatic
and nonaromatic separation) in a one-stage equi-
librium cell was determined. Table 2 shows that,
although two liquid phases were observed using
sulfolane as the solvent, sulfolane still gave a bet-
ter performance than the other solvents where
a single liquid phase existed in the mixture.
2. Two liquid phases have no ill effects on the efR-
ciency of small tray or packed towers with dia-
meter from 0.08 m to 0.46 m. However, in larger
towers, the heavy liquid phase tends to accumulate
on the tray if the liquid phases are not well mixed.
This problem can be eliminated by tray designs
promoting gas agitation, forcing the two liquid
phases to behave as a homogeneous liquid. For
larger packed columns, liquid}liquid redis-
tributors are specially designed to allow separate
distribution of the two liquid phases.
Computer simulations have been developed which
are capable of accurately predicting the development
of two liquid phases in the ED tower. In one
approach, the simulation algorithm starts from
linearized pressure, temperature and concentration
proRles and feed conditions given by the program
operator. New estimates of composition are solved,
using the material balance and equilibrium relation-
ship for each tray. Then the equilibrium constants are
re-estimated and a new temperature gradient is estab-
lished to calculate a tray-by-tray energy balance. Ac-
cumulated errors are calculated for the energy, mater-
ial and equilibrium balances. Appropriate column
operation restraints are factored in at this point.
A correction factor is found for the temperature, rate
proRles, and liquid composition proRle by inverting
the accumulated error matrix. These correction fac-
tors are used to form new iterative estimates of com-
position to start the process again until the correction
factors are small enough to call the components
converged.
Table 2 ED solvent screening for aromatics recovery
Solvent Relative volatility Number of
(n-heptane/benzene) liquid phases
Sulfolane 3.9 2 cosolvents may be blended into the base solvent to
DMSO 3.6 1 enhance the performance in speciRc applications.
NFM 3.1 1 Table 3 summarizes the key performance para-
NMP 2.6 1
Feed: 20% n-heptane and 80% benzene; pressure 1 atm;
DMSO, dimethyl sulfoxide; NFM, N-formyl morpholine; NMP,
N-methyl pyrrolidone.
II / DISTILLATION / Extractive Distillation 1017
Multicomponent vapour}liquid and liquid}liquid
equilibria solutions are required for the algorithm.
Two activity coefRcient models, NRTL (nonrandom
two liquids) and UNIQUAC (universal quasichemi-
cal), are readily extendable to multicomponent sys-
tems and capable of such solutions. Experimental
activity coefRcients, , at inRnite dilution are used for
calculating binary parameters for the NRTL equa-
tion. These parameters are then tested using experi-
mental liquid}liquid ternary data, experimental va-
pour}liquid equilibrium data, and data from pilot
plant or commercial plant. The NRTL equation is
used in the algorithm to calculate activity coefRcients
and is given in the following equations:
ln 1"x22[21(G21/(x1#x2G21))2
#12(G12/(x2#x1G12)2]
ln 2"x21[12G12/(x2#x1G12))2
#21G21/(x1#x2G21)2]
where
ln G12"!1212
ln G21"!2121
12"(12#S12T)/RT
21"(21#S21T)/RT
and where Gij, ij, Sij and ij are empirical constants,
i is activity coefRcient, R is the gas universal con-
stant, T is absolute temperature, and xi is liquid phase
mole fraction of component i.
The simulation uses a Newton}Raphson-based
Sash algorithm that checks for two liquid phases by
checking Gibbs free energies for components the pro-
gram operator lists as possible second liquid phase
formers. If two liquid phases are indeed present, regu-
lar solution theory provides a method of combining
the liquid-phase activity coefRcients.
State-of-the-art ED Technologies
The modern state-of-the-art ED technologies for BTX
aromatic puriRcation are based on several solvent
systems: sulfolane, NFM and NMP. Proprietary
meters of LLE and ED for aromatics recovery. ED
process can provide up to 25% savings in capital
investment as compared with the commercially avail-
able LLE processes. This saving is attributable to the
1018 II / DISTILLATION / Extractive Distillation

Table 3 Comparison of aromatic recovery technologies

System UDEX TM ArosolvanTM SulfolaneTM Morphylane威 GT-AromexTM

Solvent Tetra-EG NMP Sulfolane N-formyl morpholine Proprietary

ISBL capital cost (million US$) 9.5 11.0 10.7 11.0 8.1

Utilities
Power (kWh per t of feed) * 6.6 2.4 4.6 4.8
Steam (kcal per kg of feed) 211 225 177 250 194
Cooling H2O (gal per lb of feed) 19 * * 16 21

Aromatics recovery
Benzene 99.9% 99.9% 99.9% 99.9% 95.0%
Toluene 99.5% 99.5% 99.0% 99.5% 99.5%
Xylene 95% 95% 95% 97% 100%

Solvent-to-feed ratio (v/v) 4:1 0.4 : 1 2:1 3:1 3:1

Data are for 1994 construction, extraction section only; all processes are pro rata for 1600 metric tons day\1 reformate feed; sources
include SRI International, Handbook of Solvent Extraction, Petroleum Refining Technologies & Economics, and licensor literature.

smaller number of operating units, as mentioned
above. The ED process recovers more xylenes but less
benzene than LLE processes.
The Morphylane威 process offered by Krupp Kop-
pers uses NFM as the selective solvent. A schematic
diagram of the Morphylane威 process is given in Fig-
ure 8. The diagram is very similar to the general ED
process scheme as shown in Figure 1, except the
nonaromatic vapour exiting the top of the ED tower
contains a small amount of NFM solvent (0.9 wt%),
which must be recovered. Two methods are used for
this solvent recovery, both of which require addi-
tional equipment and expense: (1) a separate solvent
recovery column; and (2) additional trays or packings
Rtted to the top of the ED tower (above the solvent
tray), using nonaromatics as the reSux to Sush NFM
back into the ED tower. To use the second method,
the feedstocks to the ED tower must contain only
very small amounts of critical components, such as
methylcyclohexane and dimethylcyclopentane in py-
rolysis gasoline feedstock, or C7 oleRns in reformate
feedstock.
The Morphylane威 process is available in commer-
cial applications for recovering high purity benzene
from C6 fraction, or pure benzene and toluene from
the C6}C7 fraction of reformate or pyrolysis gasoline.
For example, the process has been commercially tes-
ted with a feedstock from a mixture of C6 reformate
fraction and a C6 fraction of a pyrolysis gasoline. The
plant had a top-Rtted solvent recovery system. The
results are summarized in Table 4. Approximately
98% benzene recovery with 99.9% benzene purity
was achieved with this process. This process, how-
ever, has not been applied to the recovery of higher

Figure 8 Schematic diagram of the Morphylane威 process for aromatic purification.

 II / DISTILLATION / Extractive Distillation 1019

Table 4 ED performance for benzene recovery from the C6 and C# 9 and higher aromatics. Unlike the Mor-
fraction phylane威 process, the overhead nonaromatics stream
from the ED tower in the GT-BTXSM process contains
Parameter Units Value
essentially no solvent, and does not require a separate
Throughput t h\1 23.0 solvent recovery tower.
(approx. 116%) A hybrid concept of the GT-BTXSM process can be
Benzene production t h\1 12.89 used to increase substantially the capacity of the
Benzene purity wt % 99.98
liquid}liquid extraction unit and improve the quality
Benzene yield wt % 98.11
Solvent consumption g t\1 aromatics 6.0 of the benzene product, through retroRtting of the
Solvent in benzene ppm Not ascertainable existing unit. The retroRtting can be carried out using
product this hybrid concept without requiring extensive modi-
Steam consumption kg t\1 feed 564 Rcations, investment or lengthly shutdown time.
(16;105 Pa)
Figure 9 shows a new process using a hybrid of the
(including benzene
column) sulfolane liquid}liquid extraction process with the
Energy consumption for kg t\1 feed 349 GT-BTXSM process that bypasses part of the feed
extractive distillation only Gcal t\1 feed 0.161 around the original extraction section. In the hybrid
16;105 Pa steam scheme, the ED tower is better suited to purifying the
benzene-rich feed than the liquid}liquid extraction
unit, and it is not subject to the maximum aromatics
boiling aromatics, such as mixed xylenes or C# 9 and limit in the hydrocarbon charge. The ED tower
higher aromatics, probably because of the relatively nonaromatic stream (rich in cyclohexane) may be
low boiling point of NFM. recycled to the reformer unit for producing more
The GT-BTXSM process offered by GTC Techno- benzene, while the rafRnate stream from the
logy Corporation is available for recovering not only liquid}liquid extractor (rich in parafRns) could be
benzene and benzene/toluene, but also a full range of routed to gasoline blending or used as a feedstock for
aromatics (benzene, toluene and xylenes) with high naphtha cracking to produce ethylene and pyrolysis
purity and recovery. The process uses a proprietary gasoline. The major changes are modiRcations of the
sulfolane-based solvent blend. Due to the high boiling solvent system to be compatible with both extraction
point of the solvent, the process is very effective in operations and to make the appropriate tie-ins to the
recovering higher boiling aromatics, such as xylenes ED tower.

Figure 9 Hybrid scheme for aromatic recovery process expansion.

1020 II / DISTILLATION / Extractive Distillation

Cyclohexane Recovery from Naphtha Table 6 Computer simulation for solvent comparison
or Natural Gas Liquid Solvent 2,4-DMP CyC6 purity Separation
recovery (%) (wt%) factor a
In recent years, ED technology has also been applied
to the separation of parafRns and cyclopara- MIST 85.5 97.5 586
fRns, a much more difRcult separation than aromatics TEG 53.7 92.4 115
and nonaromatics. One of the major developments Sulfolane 51.7 92.1 106
NMP 45.9 91.2 84
was cyclohexane recovery from naphtha or natural
EGb 0.0 85.0 0
gas liquid (NGL) streams. Cyclohexane, an important
raw material for the nylon industry, exists naturally a
Separation factor
in naphtha and NGL streams. However, recovery of mole fraction 2,4-DMP raffinate/mole fraction CyC6 raffinate
" .
high purity cyclohexane from naphtha or NGL mole fraction 2,4-DMP extract/mole fraction CyC6 extract
b
through conventional distillation is virtually imposs- Simulation failed to converge.
Premises: 99% CyC6 recovery, overhead product allowed to vary;
ible, owing to the close-boiling C7 isomers
S/F weight ratio"16; 25 equilibrium stages (solvent fed on stage
in the streams. Since the polarity difference between 24, hydrocarbon fed on stage 12); reflux fixed at 0.48 (hydrocar-
cyclohexane and C7 isomers is substantially smaller bon feed).
than that for aromatic and nonaromatic compounds,
no extractive solvent has been found that can effect
the separation. However, through the use of a cosol-
vent (to enhance the solvency of the mixed solvent), VLE cell for the mixed and single solvents shown in
an ED process has been commercialized to recover Table 5. These simulations were based on experi-
high purity cyclohexane directly from an NGL frac- mental physical property data, such as the inRnite
tion containing 85% cyclohexane. dilution activity coefRcients of binary solvent}
Many solvent blends show synergistic improve- hydrocarbon mixtures. Again, NRTL thermodyn-
ment over what would be expected by pure compon- amic correlations were used to predict the occurrence
ent mixing. To test the concept, experiments were of two liquid phases and a Newton}Raphson
conducted in a one-stage vapour}liquid equilibrium convergence method was used to carry out the simu-
(VLE) cell to compare the selectivity of Rve solvents. lations.
To a hydrocarbon mixture of 85 wt% cyclohexane Simulations of a ED process separating an
(CyC6) and 15 wt% 2,3 dimethylpentane (2,3-DMP), 85/15 wt% CyC6/2,4-DMP mixture were made to
a selective solvent or a mixed solvent was added, at compare the MIST solvent with four common extrac-
a solvent-to-feed ratio (S/F) of 7.0. The relative vola- tion solvents, ethylene (EG), triethylene glycol (TEG),
tility of 2,3-DMP over CyC6 was measured in the sulfolane and N-methyl pyrrolidone (NMP). The
equilibrium cell with various solvents. Table 5 pres- simulations were for a 25 theoretical stage ED tower
ents a comparison of relative volatilities obtained for at a S/F ratio of 16. The CyC6 recovery in the extract
Rve solvents tested, including a proprietary mixed stream was speciRed at 99.0% and the overhead
solvent (MIST) from Phillips Petroleum Company. rafRnate product was allowed to vary. Table 6 shows
MIST solvent, discovered by investigating the combi- that the MIST solvent has a separation factor 5 times
nations of many other solvents, has a signiRcantly greater than TEG, which has the highest separation
higher relative volatility than the other single factor of the single solvents.
solvents. The MIST solvent was Rrst tested in a 150 mm
Computer simulations were carried out to conRrm diameter ED pilot plant using as the feedstock a re-
the results on solvent screening from the one-stage Rnery stream that had an average composition as
shown in Table 7. Based on the successful pilot plant
study, a commercial plant purifying 100 metric
Table 5 Equilibrium cell study for CyC6 and 2,3-DMP separation tonnes per day cyclohexane was designed, construc-
ted and commissioned in 1991.
Solvent No. of liquid phases Relative volatility
(2,3-DMP/CyC6)

(No solvent) 1 0.84 Light Ole\n and Paraf\n

EG 2 1.02
TEG 2 1.06
Separations
Sulfolane 2 1.07 The synthetic rubber process, brought to a successful
NMP 1 1.07
MIST 1 1.22
culmination during World War II, required large
quantities of butadiene; consequently normal
 II / DISTILLATION / Extractive Distillation 1021

Table 7 Average ED pilot plant feedstock withdrawn from the bottom of this section together
with hydrocarbon bottoms. The solvent and bottoms
Components wt% are separated in a smaller 20-tray stripper tower,
Cyclohexane 89.1 the solvent-free bottoms being removed as overhead
2,2-Dimethylpentane 1.3 and stripped solvent circulated back to the ED
2,4-Dimethylpentane 4.0 tower. Hydrocarbon feed is charged at some
3,3-Dimethylpentane 0.1 point below the solvent feed, near the bottom of the
2,3-Dimethylpentane 0.9
Rrst section or top of the second section of the ED
2-Methylhexane 1.6
3-Methylhexane 1.1 tower.
2,2,3-Trimethylbutane 0.8 For n-butane and 2-butene separation, the purity of
Dimethylcyclopentanes 1.0 2-butene was 94.6 vol% with only 39.4 vol% recov-
n-Heptane 0.1 ery. For mixed butanes and mixed butenes separ-
ation, the purity of mixed butenes was 88.7 vol%
with 96.7 vol% recovery, and for butadiene and
butenes separation, the purity of butadiene was
butenes, the feedstock to butadiene units, were also in 96.9 vol% with 89.7 vol% recovery. Obviously,
great demand. ED process technology was developed these results did not meet the industrial require-
to recover high purity n-butenes suitable for pro- ments for producing high purity product with high
ducing butadiene to feed the synthetic rubber process. recovery.
In this case, the selective solvent, developed by Further studies were carried out to screen solvents
Shell Development Company in Houston, Texas, for oleRn and parafRn separations. For example,
USA, was a mixture of 85% acetone and 15% a comprehensive solvent screening study was conduc-
water. ted for n-butane and butene-2 separations. Eighty
Later, furfural was used as an ED solvent for separ- solvents were evaluated, including ester-type solvents
ating isobutane from butene-1, n-butane from bu- containing hydroxyl groups, aldehyde groups, amine
tene-2, and butene-1 from butadiene. As shown in groups, nitrile groups, nitro groups, ketone groups,
Figure 10, furfural was tested in an ED tower consist- nitrogen; ether-type solvents; and miscellaneous sol-
ing of two 50-tray sections in series for separating vents. It was found that aniline and furfural were the
butene-2 from n-butane. Solvent was charged to most selective solvents. The VLE data for n-butane
a tray, which was several trays from the top of the and 2-butenes in furfural and aniline solvents are
Rrst section (A), and Sowed to the bottom of this given in Figures 11 and 12. Although N-formyl mor-
section. It was pumped together with dissolved hy- pholine was also tested among the nitrogen-contain-
drocarbons to the top of the second section (B), and ing solvents for n-butane and 2-butenes separation,

Figure 10 Schematic ED process diagram for separating 2-butene and n-butane.

1022 II / DISTILLATION / Extractive Distillation
Figure 11 Vapour}liquid equilibrium of n-butane and 2-butene
in Furfural. Solvent dosage: 䢇, 3.7; 䉺, 4.5. Pressure,
3862 mmHg.
for unknown reasons the solvent was not considered
for commercialization until recently.
Krupp Koppers has offered the BUTUNEX pro-
cess, an ED process using N-formyl morpholine as the
selective solvent, for recovering 1-butene and 2-bu-
tene from C4 hydrocarbon streams. On the basis of
such a feedstock with the composition of 25.6%
isobutane, 32.7% n-butane, 26.6% 1-butene and
15.1% 2-butene, the following yields can be
Figure 12 Vapour}liquid equilibrium of n-butane and 2-butene
in aniline. Solvent dosage, 5.0; pressure, 3862 mmHg.
achieved:
1-Butene 95.6%
2-Butene 99.1%
Butanes 98.9%
The purity of 1-butene and 2-butene products can be
99.6% and 95.9%, respectively.
Conclusions
Since the 1940s, ED technology has gone through
extensive development for solving many difRcult sep-
aration problems in the petroleum and petrochemical
industries. The development in cosolvent selection
tailored for a speciRc separation and the advancement
in tower internal design have made ED a competitive
process. In many cases, ED processes can be more
efRcient and economical than conventional LLE
in terms of capital investment, energy consumption
and ease of operation. It is anticipated that the ED
technology will be selected more frequently in the
future for the petroleum and petrochemical
industries.
See also: II/Distillation: Theory of Distillation. III/React-
ive Distillation.
Further Reading
Benedict M and Rubin LC (1945) Extractive and azeotropic
distillation: I. Theoretical aspect. Transactions of the
American Institution of Chemical Engineers 41:
353d370.
Brown RE and Lee FM (1991) Way to purity cyclohexane.
Hydrocarbon Processing May: 83d86.
Doherty MF and Knapp JP (1993) Distillation, azotropic
and extractive. In: Kirk}Othmer Encyclopedia of Chem-
ical Technology, 4th edn, vol. 8, pp. 358d398. New
York: John Wiley and Sons.
Drickamer HG and Hummel HH (1945) Application of
experimental vapor}liquid equilibria to an analysis
of the operation of a commercial unit for the
puriRcation of toluene from petroleum. Transactions of
the American Institute of Chemical Engineers 41:
607d629.
Gentry JC and Kumar CS (1998) Improve BTX Process-
ing Economics. Hydrocarbon Processing (March):
69}74.
Lee FM and Gentry JC (1997) Don’t overlook extractive
distillation. Chemical Engineering Progress 93 (10):
56}64.

Freeze-Drying
G.-W. Oetjen, Chemical Engineering,
Lu( beck, Germany
Copyright ^ 2000 Academic Press
Introduction
Freeze-drying is a process, in which a product is Rrst
frozen and then dried by sublimation of the ice. The
total process involves four steps: freezing; sublima-
tion of the ice, called main drying (MD); desorption
of the water bound to the solid, called secondary
drying (SD); and packing in containers to exclude
absorption of water and/or oxygen from the atmos-
phere. By freeze-drying a product unstable in water is
transformed into a dry, stable product. The process
has to be developed to satisfy four demands on the
Rnished product: its volume remains that of the
frozen substance; the structure and the biological
activity of the dried solid correspond as far as
possible to those of the original substance; the
dried product remains stable during storage, if
possible at temperatures up to #403C and for
up to 2 years; and with the addition of water the
original product is quickly reconstituted. This article
summarizes the problems and solutions to achieve
these aims.
Theoretically, sublimation of ice can be done at
atmospheric pressure; however, the vapour pressure
of ice between !103C and !403C is approximately
2.6}0.13 mbar and 1 kg of ice has a volume of ap-
proximately 470 m3 at !103C and 8400 m3 at
!403C. To transport these volumes at atmospheric
pressure the gas volume has to be approximately
400}8000 times larger than that of the vapour.
Therefore all freeze-drying plants today are vacuum
plants, in which the air is reduced to some 10% of the
vapour pressure, to allow the free Sow of vapour at
velocities up to 100 m s\1.
The Rrst organ tissues were freeze-dried in 1890
and in 1932 a vacuum freeze-drying plant was built,
but only after 1940 did it become an industrial pro-
cess with the freeze-drying of blood plasma and
penicillin.
Freezing
Structure of Water and Ice
In a water molecule the two H atoms form an almost
tetrahedral angle with a strong dipole moment.
A shell of about four water molecules exists at a dis-
tant of 28 nm, followed by a second at approximately
II / DISTILLATION / Freeze-Drying 1023
45 nm, and beyond 80 nm no shell can be identiRed.
In addition to this short range order, a network of
hydrogen bonds exits with a very short lifetime of
10\12 to 10\13 s.
In sub-cooled, very pure water the nucleation of ice
crystals (homogeneous nucleation) starts around
!413C. Normally water contains about 106 particles
which act as nuclei for crystallization (heterogeneous
crystallization). They become more effective as their
structure approaches that of ice. Ice crystals exist in
nine forms, of which the cubic and hexagonal at under
atmospheric pressure. The growth rate of crystals
depends on the diffusion of molecules to the nuclei,
on Rnding a suitable place, and on the transportation
of the freed energy to the heat sink. With extreme
cooling rates (of the order of 1053C min\1), crystalli-
zation can be avoided and water solidiRes into an
amorphous, glass-like phase. Figure 1 shows that the
amorphous phase of ice is only stable below
!1603C; in the range !1603C to!1303C amorph-
ous and cubic phases can exist, and between !1303C
and !653C all the forms can be present, depending
on the speed of warming. It is technically impossible
to cool pure water quickly enough, to produce glassy
ice; even small droplets (1 mm) injected into liquid
nitrogen may freeze at a rate less than 103 3C min\1.
The freezing behaviour of water changes completely
if other substances are present in the water, e.g.
cryoprotective agents (CPAs). The most widely used
CPAs are:
E protein: human serum albumin, gelatin
E amino acids: glycine, arginine, alanine
E alcohols: mannitol, polyethylene glycol (PEG)
E carbohydrates
} monosaccharides: glucose, fructose
} disaccharides: lactose, maltose, sucrose,
trehalose
} polysaccharides: dextran, cyclodextrins
E others
} metals
} surfactants
} polymers
} buffer salts
They all protect in one way or another, alone or in
combination, the original quality of the product to be
freeze-dried.
As an example of a CPA the inSuence of glycerol
concentration is shown on the right-hand side of
Figure 1. The temperature of homogenous nucleation
1024 II / DISTILLATION / Freeze-Drying

Figure 1 Phase diagram, water}glycerine. On the left-hand side the dependence of the phase transformation time from the ice
temperature is shown: at !1403C amorphous ice transforms into cubic ice in approximately 10 min. (From Umrath, W. Kurzbeitrag fuK r
die Tagung Raster-Elektronenmikroskopie in Medizin und Biologie, unpublished, BruK hl.)

(Thn) can be reduced by about 203C, the devitriRca-
tion temperature (Tg) can rise by almost 353C and
sub-cooled liquids can exist down to very low tem-
peratures. Polyvinyl pyrrolidone reduces Thn only
a few degrees, while Tg can be changed by more than
703C. The freezing of a product is mostly done so
quickly that no equilibrium state is reached during
the process. The structure of the frozen product de-
pends therefore not only on its components but also
very much on the freezing rate and the temperature at
the end of the freezing process. Generally speaking,
during slow freezing the nuclei have time to grow and
the solution in between the ice crystals becomes
increasingly concentrated. During quick freezing
only small crystals can grow and the remaining
solution can become so viscous that the water
molecules cannot diffuse to the crystals and they
become part of the solidiRed liquid (glass) between
the ice crystals.
The freeze concentration may, among other effects,
change the pH, the water structure around proteins
and extract water from cells. The unfrozen water
will crystallize during warming when the mobility of
the molecules is sufRciently increased. The crystalliza-
tion can be very abrupt, warm the surroundings,
melt it at least partially and destroy the structure.
The freezing of a product is as critical as the dry-
ing } in some respects even more so. The structure
achieved during freezing and solidiRcation deter-
mines the main and secondary drying process, the
reconstitution of the dried product and its storage
capability. Therefore it is mandatory to analyse the

formation of the structure and the factors which inSu-
ence it.
Methods of Structure Analysis
A number of methods have been described to supply
information during freezing. Electrical resistance dur-
ing cooling and warming (ER) is measured in a test
vial at different freezing and warming rates. For
a more accurate interpretation of the function log
(ER)"f(T) the Rrst derivative of the plot is cal-
culated as shown in Figure 2. The advantages of the
method are: sample size is of the order of a product in
Figure 2 (A) Electrical resistance (ER) of a pharmaceutical
product as a function of temperature during cooling at 13C min\1
and warming at 33C min\1. Heat transfer medium and product are
approximately uniformly heated. (B) Measurement of the electri-
cal resistance as in A, but with the wall of the vial insulated by
a plastic tape up to the filling height of the product. Heat is
therefore mostly removed through the bottom of the vial. (A and
B from Willemer, H., KoK ln, unpublished measurements.)
II / DISTILLATION / Freeze-Drying 1025
a vial; it simulates heat transfer from the shelf to the
vial/product; and the equipment is relatively inexpen-
sive and easy to operate. The disadvantages are that
interpretation needs some experience, the measured
data reSect the mobility of ions and the amount of
energy used or freed during an event cannot be cal-
culated. In Figure 2A, the heat transfer medium and
product are at a similar temperature; in Figure 2B,
the wall of the test vial is isolated from the heat
transfer medium to simulate the freezing of a product
in a vial on the shelf. In Figure 2A, the effect of sub-
cooling during freezing can be seen at about !103C,
but the derivative shows it more clearly between
!33C and !103C. During warming at event
1 (!123C) the structure softens, allowing unfrozen
water to crystallize, represented by the increase in
resistance. In Figure 2B the crystallization energy
cannot be quickly removed: freezing occurs in two
steps. During warming, events 1 and 2 are not found,
all freezable water is crystallized during cooling.
Differential scanning calorimetry (DSC) compares
heat Sows, one to and from the sample and the other
to and from a substance with no transitions in the
measuring range. Roos and Karel showed by DSC
(Figure 3) the inSuence of unfrozen water on Tg of
fructose (1) and glucose solutions (2). After rapid
freezing (303C min\1) to !1003C Tg of fructose and
glucose is at !883C and !843C respectively; at
!483C and !443C respectively the unfrozen water
crystallizes, followed by the melting of ice. If the
products are thermally treated or annealed (after
freezing the product is warmed to !483C for 15 min
and then cooled again to !1003C), Tg, called Tg if
Y
all freezable water is frozen, is raised to !583C and
!573C and no crystallization event is measurable.
Time and temperature of annealing must be carefully
determined to achieve a certain mobility of the mol-
ecules without collapse of the structure (see Fig-
ure 9B). The advantages of DSC are the quantitative
measurement of the changes in the heat capacity of
the sample and the energy freed or used in an event.
The disadvantage is the small sample (milligrams),
which can behave differently from a product in vials
(grams) and the cost of the equipment.
In a cryomicroscope the sample can be optically
observed during cooling and warming at different
rates. Some models also permit freeze-drying of the
sample. Willemer has shown (Figure 4) the structure
of a Factor VIII solution during warming after quick
freezing. This product must be freeze-dried at a tem-
perature of the sublimation front of the ice (Tice)
below !443C and if annealing is necessary it may be
possible at !433C to !423C for several minutes but
a longer time at !453C is recommended. The ad-
vantage is the visual conRrmation of data gained by
1026 II / DISTILLATION / Freeze-Drying

at a Tico(!203C, otherwise the structure would

collapse. The unique advantage of NMR is the ability
to discriminate between free, crystallized and bound
water.
More details concerning these methods and addi-
tional procedures have been reported by Oetjen.

Freezing Rates
The freezing time can be estimated by an equation
developed by Steinbach:

tf"
J/
T g(2/2g#d/Ksu)

where tf"freezing time;
J"enthalpy difference be-

tween the initial freezing point and the Rnal temper-
ature;
T"difference of temperature between the
freezing point and the cooling medium; d"thickness
of the product parallel to the direction of prevailing
heat transfer; g"density of the frozen product;
g"thermal conductivity of the frozen product; and
Ksu"surface heat transfer coefRcient between cool-
Figure 3 Results of annealing (thermal treatment) on the ing medium and the freezing zone.
formation of ice in a 60% fructose solution (1) and in a 60% In this equation Ksu has to be measured for each
glucose solution (2). Curve A: after cooling at 303C min\1 down to
type of vial or tray used. The Satness of the bottoms
!1003C, the DSC plots have been recorded during warming at
53C min\1. Tg approximately !853C and !883C, respectively,
for fructose and glucose. At approximately !483C and !443C
respectively, ice crystallization starts clearly, followed by the be-
ginning of the melting of ice. (During freezing only a part of the
water has been crystallized.) Curve B: after cooling down to
!1003C, the product has been warmed at 103C min\1 to!483C,
kept for 15 min at this temperature (thermal treatment), cooled
down again at 103C min\1 to !1003C, and the DSC plot (B)
measured during rewarming. During thermal treatment all
freezable water is crystallized, and Tg is increased to !583C and
Y
!573C, respectively. During warming, no crystallization can be
detected. (Reproduced with permission from Roos and Karel,
1991.)

other methods, e.g. ER or DSC, and the possibility to

analyse the image quantitatively by computer. The
disadvantages are high cost and the relatively small
region of the sample that can be observed.
Nuclear magnetic resonance (NMR) provides in-
formation, among other things about free or bound
water (e.g. to protein molecules), the inSuence of
unfrozen water on the collapse temperature and the
crystallization of amorphous dry products. Hanafusa
has shown by NMR (Figure 5) how the amount of
unfrozen (bound) water in a 0.57% ovalbumin solu-
tion is reduced by the addition of a 0.01 M solution of
sucrose or glycerol. Similar information can be gained
for a coffee extract with 25% solids: during freezing
Figure 4 Photographs taken with a cryomicroscope of Factor
and rewarming at !703C 0.01 g water per gram
VIII solutions at four temperatures. At !403C the structure is still
solid are unfrozen, at !403C 0.1 g per gram and at visible, but is more coarse compared with the appearance at
!203C approximately 30%; thereafter the amount !443C. At !353C the structure is collapsed. (Reproduced with
increases rapidly. This extract has to be freeze-dried permission from Willemer, 1996.)
 II / DISTILLATION / Freeze-Drying 1027

developed by Steinbach is used:

tMD"(gw LS
m d)/Ttot(1/Ktot)#(d/2.g)
#(d/2 LS b/u)

where g"density of the frozen product (kg m\3);

w"amount of water (kg kg\1); LS"sublimation
energy (2.805 kJ kg\1); Ttot"temperature difference
(Tsh!Tice); Ktot"total heat transmission coefRcient
from the shelf to the sublimation front of the ice;
g"thermal conductivity of the frozen product;
d"thickness of the layer (m);
m"content of
frozen water; and b/"permeability (kg m\1
h\1 mbar\1) for water vapour through the dried
product.
Ttot is known, if Tice is measured and Ktot has to be
measured once for each type of container; one can

Figure 5 Unfreezable water (UFW) in a 0.57% ovalbumin solu-

tion as a function of the freezing temperature with different CPAs.
(Reproduced with permission from Hanafusa, 1992.)

can change Ksu by a factor of two } if the vials are in

trays without machined bottom surfaces by a factor
four and more. With Ksu measured, tf can be esti-
mated with an error of 10}15%.

Main or Sublimation Drying

The main drying process (MD) has been photo-
graphed in a cryomicroscope by Kochs et al. as
shown in Figure 6. The ice crystals grow extremely
uniformly using a special freezing method. The
ice sublimes and the remaining solids show their
original structure after freezing. During sublimation
the temperature of the ice at the sublimation front
(Tice) has to be kept well below the collapse temper-
ature Tc.
As can be seen in Figure 6 Tice cannot be measured
by a sensor because the ice front travels. If the valve
between chamber and condenser (8 in Figure 11) is
closed for a short time ((3 s) the water vapour
pressure in the chamber rises until the saturation
pressure (g) of the ice front is reached. The rising
pressure is measured 100 times per second and the
change in the slope (after 2.14 s), if saturation is
reached, is determined as 0.286 mbar, corresponding
to !32.73C as shown by Haseley and Oetjen in
Figure 6 Course of main drying observed using a cryomicro-
Figure 7. This procedure is called barometric temper-
scope, in which freeze-drying is carried out. The hydroxyethyl
ature measurement (BTM). It permits checking starch solution is optimally frozen. The dark lines show the form of
Tice during MD (e.g. every 10 min). To estimate the the sublimated ice crystals. (Reproduced with permission from
main drying time (tMD) the following equation Kochs et al., 1991.)
1028 II / DISTILLATION / Freeze-Drying
Figure 7 Pressure rise as a function of time. 1, Pressure rise
in the chamber after the valve is closed; 2, first derivative of 1.
The maximum of 2 is reached at 2.14 s; the related equilibrium
vapour pressure ps"0.286 mbar, corresponding to Tice"
!32.73C. (Reproduced with permission from Haseley and
Oetjen, 1998.)
expect values between 60 and 120 kJ m\2 h\13C\1.
Ktot is only slightly dependent on the operation pres-
sure up to 0.1 mbar; then it increases up to 1 mbar
by a factor of two. g is, in most cases, the Rgure for
pure ice.
m has to be determined for each product
by methods described above. b/"1.3;10\2
(kg m\1 h\1 mbar\1) is an average which is often
found in practice, it can vary by a factor of two, but
Figure 8 Density of water vapour flow (g cm\2 h\1) as function
of pch with jet flow and different l / d as parameter. (Reproduced
with permission from Oetjen, 1999.)
the term with b/ in most freeze-drying processes
only has an inSuence of a few per cent on tMD. The
standard deviation of Tice, measurements during MD
in the range of !15 to !453C should be (0.53C,
if measured automatically. tMD is in most cases
governed by the value of Ttot and the term (1/Ktot).
The term (d/2 ) g) is, for d values below 10 mm, of the
order of 10% or less of l/Ktot, growing to approxim-
ately 50% at d"35 mm. Tice is the result of a ther-
modynamic equilibrium between heat transfer to the
sublimation front and energy consumption for subli-
mation. Both depend on several factors, but the heat
supply and vapour transport to the condenser are
most important during MD. Therefore the operation
pressure is a very effective tool to control Tice, if the
shelf temperature is kept constant and the condenser
temperature is always below a maximum, which de-
pends on the water vapour pressure in the chamber
and the design of the plant. By changing the operation
pressure, e.g. from 0.1 mbar to 0.8 mbar, Tice can be
controlled between !303C and !203C. For another
product, a different product thickness or a different
plant, the pressure range and its controlled range are
different, but the dependence is reproducible. Since
Tice depends also on the structure of the frozen prod-
uct it can be used to prove that the structure of
products in different runs is homogeneous and sufR-
ciently identical. If the structure contains unfrozen
water, Tice data will from time to time jump by 13C or
more (when the water evaporates) and the data will
be different for a product frozen at different freezing
rates.
At the end of MD the ice is mostly sublimed and the
measured Tice decreases below the standard deviation
of Tice during MD. This effect can be used to change
automatically from main to secondary drying (SD),
e.g. if the measured Tice becomes 2}33C smaller than
the average during MD. Other criteria are often sug-
gested, such as an increase of product temperature,
a decrease in operating pressure or a decrease in
partial pressure of water vapour, but it is more difR-
cult to use these other methods quantitatively.
Besides heat transfer, the water vapour transport
from the chamber to the condenser is often critical in
a freeze-drying process. The length (l) and diameter
(d) of the connection between chamber and conden-
ser in a freeze-drying plant as shown in Figure 11
determine the vapour Sow, assuming that the other
Sow resistances in the chamber are relatively small by
comparison. The vapour Sow can be estimated using
the GuK nther}Jaeckel}Oetjen equation. Figure 8 shows
that: the vapour Sow density decreases in a nonlinear
manner with the chamber pressure; the relation of
l/d is of increasing importance with decreasing
pressure; for example, at 4;10\2 mbar the vapour

Sow density at l/d"5 is only 30% that at l/d"1.
Right-angle bends contribute to the length not only
by their physical dimensions but, depending on the
design, by a factor of four or more of the measured
length. For operation pressures below about
10\1 mbar, l/d'2.5 should be avoided.
Secondary or Desorption Drying
Desorption Rates
During secondary drying (SD), water that is removed
is more less bound to solid molecules. The amount of
water removed is small (e.g. 10% of solids), com-
pared to 10 times the weight of solids during MD.
The behaviour of water molecules close to a protein
surface has been described by Bellissent-Funel and
Teixeira. The water molecules are in a monolayer
around the protein with a reduced mobility compared
with bulk water.
The desorption of bound water can be measured
during SD by measuring the pressure rise in the cham-
ber after closing the valve to the condenser for
60}120 s. The length of time is not critical since the
temperature does not change quickly in this phase.
The pressure rise (dp s\1) can be converted into the
desorption rate (DR) using:
DR"2.89;102 (Vch/mso) (dp/dt)
where DR"desorption of water vapour in per cent
of solids per hour; Vch"chamber volume (L);
dp"pressure rise (mbar); dt"time of dp (s); and
mso"mass of solids (g).
The course of DR describes not only the progress of
the secondary drying quantitatively, but also reSects
the structure of the frozen product as shown by
Haseley and Oetjen in Figure 9. In Figure 9A DR
data are shown for a 10% mannitol solution frozen in
vials on the shelves of the freeze-drying plant at a rate
of 0.5}0.83C min\1. In Figure 9B the same solution
in the same vials is frozen in liquid nitrogen and in
Figure 9C the solution is frozen in liquid nitrogen but
annealed before freeze-drying. From these Rgures the
following conclusions can be drawn. (1) Slow
freezing of 10% mannitol solutions results in struc-
tures in which the water is bound in several forms.
The DR plots as a function of time are not single-
valued. (2) Freezing at rates of more than
303C min\1 produces structures with a more uniform
desorption behaviour. DR plots show the inSuence of
the operating conditions during MD: (1) run 6 in
Figure 9B is collapsed, and the water has dissolved
part of the solids, forming a sticky cake. The water
II / DISTILLATION / Freeze-Drying 1029
vapour of 378 vials resulted in an unstable Tice, which
is 5}73C higher than in all other runs. (2) The system-
atic inSuence of Tsh is shown in Figure 9B and C. (3)
The inSuence of pc is shown in Figure 9C, and the
inSuence of annealing or low Tice is demonstrated by
comparing Figure 9B and C: without annealing, DR
plots bend between 3% and 5% per hour; with
annealing this effect practically disappears. The ex-
ceptions prove the sensitivity of the measurements:
run 1 (Figure 9B) is freeze-dried at Tice"!36.93C,
others at approximately !34.93C; run 5 (Figure 9B)
is annealed at !403C but for 8 h; runs 2 and 3 (Fig-
ure 9C) are annealed at a temperature 13C too high
and 1.53C too low for 18 min. Annealing reduces the
amount of unfrozen water.
Residual Moisture Content
The integration of DR over time results in the amount
of water which can still be desorbed; this is called
desorbable water, dW or residual moisture content.
In Figure 10 Haseley and Oetjen show the calculated
plots from the DR data of Figure 9B and C. From the
lower part of Figure 10, it follows that the heat con-
ductivity of the annealed product during SD is almost
100% higher and more uniform than that of the
unannealed product. The upper part shows that
the correctly treated products and the one dried at
a low Tice during MD will dry more quickly than the
others.
Storage
The storage capability of a dried product depends
generally on its chemical and structural qualities. The
complexity of the problem is highlighted, for
example, by the storage stability of therapeutic pro-
teins. The degradation of a protein, the irreversible
change in primary structure, conformation or state of
aggregation in a glassy surrounding depends on the
thermodynamic behaviour of the glass as well as on
the qualities of the protein produced during freezing
and freeze-drying, as shown by Pikal. The storage
temperature of such products has to be well below
Tg of the dried formulation; nevertheless unfolding or
aggregation of unfolded molecules can occur because
of poor interaction between the stable glass structure
and movements in protein conRgurations. From this
example some simpliRed guidelines can be proposed.
There are no general rules to estimate the maximum
storage time at a maximum tolerable temperature
} both depend even on small changes in the formula-
tion of a drug or the variations between two types of
fruits or the processing methods of extracts (e.g. cof-
fee). In many cases the maximum storage time is
1030 II / DISTILLATION / Freeze-Drying

Figure 9 (A) Desorption rate (DR) as a function of time of a 10% mannitol solution frozen on the shelves of the freeze-drying plant at
a rate of 0.5}0.83C min\1. In all runs: 300 vials at an operation pressure (pc)"0.3 mbar during MD. Runs 1 and 2 at a shelf temperature
(Tsh)"203C; runs 3 and 4, Tsh"53C. Plot 5 egg albumin and plot 6 saccharose for comparison. (B) DR as in A, but the solution is
frozen in liquid nitrogen at a rate between 353C and 663C min\1. During MD all runs at pc"0.15 mbar, except run 1"0.08 mbar and
Tsh in run 1"!53C; in run 2"03C; in runs 3}6"03C for the first 11 h, thereafter until end of MD 103C. In runs 1}5, 126 vials; in run 6,
378 vials. Run 5 intentionally changed from MD to SD 7 h earlier than in runs 3 and 4. (C) DR as in B, but the frozen mannitol was
annealed at slightly different temperatures and times:

Run Annealing temperature (3C) Annealing time (min)

1 !24 18
2 !23.5 18
3 !26 18
4 !24.5 18
5 !24 20

All runs at pc"0.15 mbar, Tsh in the first 11 h"03C, thereafter 103C, during SD"303C except run 1 pc"0.08 mbar and Tsh"!53C
in the first 11 h. (Reproduced with permission from Haseley and Oetjen, 1999.)

inversely related to the maximum temperature and change during storage and for glassy products the
depends strongly on the residual moisture content maximum storage temperature has to be well below
($1% or less can be decisive). For crystallized prod- Tg (see above). The main difference between the stres-
ucts (e.g. antibiotics) the crystal structure must not ses during drying and storage is the length of the
 II / DISTILLATION / Freeze-Drying 1031

Figure 10 Residual moisture content shown as desorbable water (dW) during secondary drying. Plot 1"plot 1 in Figure 9B; 2"2
(9B); 3"3 (9B); 4"4 (9B); 9"5 (9B); 5"2 (9C); 6"3 (9C); 7"4 (9C); 8"1 (9C); 10"5 (9C); except run 1, pc"0.08 mbar and
Tsh"!53C in the first 11 h. (Reproduced with permission from Haseley and Oetjen, 1999.)

effective time: a few hours as opposed to many
months up to years.
Relaxation time in molecular conRgurations may
be large compared with the drying cycle, but this may
be totally different for the long-term storage time.
Besides temperature-induced changes, the residual
moisture content (RM) can increase the mobility of
molecules and promote chemical reactions. The RM
at the end of drying can be as speciRed; nevertheless
moisture can diffuse from the stoppers closing the
vials into the product, raising the RM by several per
cent during storage. Stoppers and the gas in the con-
tainer of the product have to be dried carefully. On
the other hand, ‘the drier the better’ is unjustiRed for
many products. The Maillard reaction increases with
decreasing water activity (aw"p/ps, where p"va-
pour pressure of the product and ps"saturation va-
pour pressure) as well as the oxidation of fats. InSu-
enza virus in a freeze-dried formulation shows the
largest decrease in infectivity at 0.4% and 3.2% RM,
while at 1.7% it is about 30 times less. Tissue plas-
minogen activator and human growth factor in cer-
tain formulations are optimally stabilized if they are
surrounded by a monolayer of water molecules
(which may not be distributed evenly).
Freeze-Drying Equipment
Figure 11 shows a freeze-drying plant, designed for
maximum current demands. The condenser is cooled
by liquid nitrogen controllable between !703C and
!1003C (Tco); the brine for the shelves is temper-
ature-controlled (Tsh) between #603C and !803C
(cooling by liquid nitrogen); the four-stage pump set
can reach about 5;10\5 mbar (pe); the vials can be
closed shelf by shelf by a hydraulically operated
1032 II / DISTILLATION / Freeze-Drying

Figure 11 Freeze-drying plant condenser and shelves cooled with liquid nitrogen. Clean In Place system in chamber and condenser.
1, liquid nitogen inlet to condenser and heat exchanger; 2 nitrogen outlet from the condenser and heat exchanger; 3, heat exchanger for
the brine in the shelves; 4, brine to and from the shelves; 5, pressure plate for closing vials; 6, piston rod with bellows; 7, hydraulic piston
for 5 and 6; 8, hydraulically operated valve; 9, hydraulic system; 10 and 13, water and steam inlets; 11, pumping system; 12, water
outlet. (Courtesy of Steris GmbH, HuK rth, Germany.)

pressure plate; the connection between chamber and
condenser is as short as technically possible; the valve
between chamber and condenser is operated by a fast
hydraulic piston; chamber and condenser can be
cleaned by a pressure spray and cleaning system
(clean-in-place); chamber, condenser and all compo-
nents within them can be sterilized by the pressureless
Vaporized Hydrogen Peroxide (VHP)威 process; load-
ing and unloading of the plant is fully automatic; the
documentation and control of the total process from
loading to freezing, to drying, to closing of the valves,
to venting and unloading can be automatic with no
thermocouples in the product; this includes the
change from MD to SD, the calculation of the moist-
ure content during SD and the termination of the
secondary drying at a speciRed moisture content.
If no extreme temperatures are required, refriger-
ant compressors can be used for the condenser down
to Tco+!803C and for the brine down to
Tsh+!603C, and a three-stage pump set is sufR-
cient. If steam sterilization is mandatory, the equip-
ment has to be built for pressures up to 2.5 bar and
temperatures up to 1253C.
At the other end of the line of freeze-drying equip-
ment laboratory installations are found, of which
a typical example is shown in Figure 12. Usually the
product is frozen in vials or trays in a separate freezer
or in the condenser of the plant and the shelves are
only heated. The chamber is often a bell jar,
Tco+!453C, pe+0.05 mbar. It is not advisable to
use this type of plant as a pilot plant for process
development, because the product temperature is not
sufRciently uniform and cannot be controlled accu-
rately, especially in the low temperature area.
Regulatory Issues
In the Validation Documentation Inspection Guide,
US Department of Health and Human Services, Food
and Drug Administration, 1993, process validation is
deRned as follows:
E Establishing documented evidence, which provides
a high degree of assurance that a speciRc process
will consistently produce a product meeting its
predetermined speciRcations and quality at-
tributes.
E The Guide to Inspections of Lyophilization of Par-
enterals, published by the US Food and Drug Ad-
ministration, July 1993, contains among others the
chapters ‘Lyophilization Cycle and Controls’,
‘Cycle Validation’ and ‘Lyophilizer Steriliza-
tion/Design’.
 II / DISTILLATION / Freeze-Drying 1033

Figure 12 Schema of a laboratory freeze-drying plant: 1, two-stage vacuum pump; 2, exhaust filter; 3, valve; 4, refrigeration
compressor; 5, liquefaction of refrigerant; 6, valve; 7, filter; 8, injection valve; 9, drain valve; 10, ice condenser; 11, pressure switch;
12, ventilator; 13, drying chamber with heated shelves and closing system for stopper of vials. (Lyovac威 GT 2, Courtesy of Steris
GmbH, HuK rth, Germany.)

In the European Union, the directive 91/356 EEC Conclusion

provides the principles and guidelines of Good Manu-
facturing Practice (GMP). In a series of annexes,
supplementary guidelines are covered, but up until
1996 only ‘Annex 1: Manufacture of Sterile Medical
Products’ has been revised. In spite of all these guide-
lines and annexes, Monger summarized the situation
for the user of freeze-drying processes and installa-
tions as follows: ‘It might be expected that some
substantial guidance would be provided. Regrettably,
this is not so’.
Powell-Evans provided a range of advice on how to
‘streamline validation’, which he calls ‘one of the
most time-consuming and costly exercises faced by
pharmaceutical manufacturers’. The qualiRcation
and validation of freeze-drying installations and pro-
cesses for the production of pharmaceuticals cannot
be summed up in this section. For cosmetic and
food products regulatory issues, depending on the
country of manufacturing and use, have also to be
followed.
Freeze-drying is the most complex and costly conser-
vation process of all drying methods. However, it is
the only way for many pharmaceutical products to
maintain their original qualities for an acceptable
time at readily available temperatures or even at
room temperature and above. For food and cosmetic
products it provides an opportunity to supply the
customers with stable high quality products which
can be easily used. For many products, e.g. some
antibiotics and some food ingredients, simpler
methods of preservation have been developed but in
pharmaceuticals there is an increase in the number of
products which have to be frozen and freeze-dried at
low temperatures using tightly controlled processes.
The tendency to automate the whole procedure is
promoted by three goals: (1) to have little or no per-
sonnel in the sterile areas; (2) to restrict the volume of
sterile areas as much as possible, for example by en-
closing the whole production line from vial Rlling to
1034 II / DISTILLATION / Freeze-Drying
Figure 13 (See Colour Plate 37). Isolator, Class 100, for filling, transportation and loading of vials into the freeze-drying plant.
Decontamination of the isolator and the equiment therein is accomplished by vaporized hydrogen peroxide (VHP). The VHP 100威
generator can be seen in the centre in front of the isolator. (Courtesy of Steris GmbH, HuK rth, Germany.)
unloading from the chamber in isolators as shown in
Figure 13; (3) to exclude human error as much as pos-
sible and to have each step documented by computer.
To automate an existing process can be more difR-
cult than to develop a new automated process. This is
based on several factors. The formulation of the drug
has to reSect the automation, e.g. Rlling and loading
can require hours, during which the solution has to be
stable, possibly at room temperature. Freezing of the
product on the shelves and drying in the chamber
have to be executed without temperature sensors in
the product; other methods of temperature control
have to be used, tested and installed. Criteria have to
be deRned for the automatic change from main to
secondary drying. Automatic termination of the sec-
ondary drying has to be effected when certain
measurable events are accomplished. Besides these
main points several others have to be evaluated. More
accurate and independent sensor systems will inSu-
ence freezing and drying procedures. The required
data processing and the actuators to fulRl the com-
mands are available today.
See Colour Plate 37.
Further Reading
Bellissent-Funel MC and Teixera J (1999) Structural and
dynamic properties of bulk and conRned water. In: Rey
L and May JC (eds) Freeze-Drying/Lyophilization of
Pharmaceutical and Biological Products, pp. 53}77.
New York: Marcel Dekker.
Hanafusa N (1992) The behavior of hydration water
of protein with the protectant in the view of HNMR.
In: May JC and Brown F (eds) Developments in
Biological Standardization, vol. 74, pp. 241}253. Basel:
Karger.
Kochs M, KoK rber Ch, Nunner B and Heschel I (1991). The
inSuence of the freezing process on vapor transport
during sublimation in vacuum-freeze-drying. Interna-
tional Journal of Heat and Mass Transfer 34:
2395}2408.
Monger P (1997) Freeze dryer validation. In: Cameron
P (ed.) Good Pharmaceutical Freeze-Drying Practice,
p. 157. Buffalo Grove, IL: Interpharm Press.
Oetjen GW (1999) Freeze Drying, ch. 1.1.5. Weinheim:
Wiley}VCH.
Oetjen GW (1999) Freeze Drying, ch. 1.2.4. Weinheim:
Wiley}VCH.
Pikal MJ (1999) Mechanisms of protein stabilization
during freeze-drying and storage: the relative import-
ance of thermodynamic stabilization and glassy state
relaxation dynamics. In: Rey L and May JC (eds)
Freeze-Drying/Lyophilization of Pharmaceutical and
Biological Products, pp. 161}198. New York: Marcel
Dekker.
Steinbach G (1971) Equations for the Heat and Mass
Transfer in Freeze-Drying of Porous and Non-Porous
Layers and Bodies, pp. 674-683. Washington, DC: In-
ternational Institute of Refrigeration.
 II / DISTILLATION / High and Low Pressure Distillation
High and Low Pressure Distillation
A. R. Jose and J. Lopez-Toledo,
Instituto Tecnologico de Celaya,
Celaya, Gto. Mexico
Copyright ^ 2000 Academic Press
Introduction
Engineers and scientists frequently have the problem
of separating a mixture into its components or need
to purify a speciRc product. To solve the problem they
usually apply a heuristic procedure, such as the fol-
lowing McMaster Rve-stage method:
1. Identify and deRne the problem.
2. Propose and develop several alternatives to solve
the problem.
3. Based on available resources, choose one or two of
the best alternatives.
4. Work on the chosen alternatives in greater detail
and compare them before selecting the most ap-
propriate solution.
5. Evaluate and decide if the problem has been
solved.
Let us assume that, for step 1, the problem is
identiRed as the recovery of one or more organic
compounds from a mixture, or the puriRcation of
a speciRc chemical.
In order to apply the second step, the differences in
physical properties are taken into account. Sometimes
the physical properties of the components are very
different, so that, if two phases already exist in the
mixture, mechanical separations such as Rltration or
decanting may then be used. But more often the
components of the mixture form a single phase and
other differences in physical properties need to be
found.
When a homogeneous mixture is formed with
components of different vapour pressure or
different boiling points, then distillation may be
one of the several alternatives proposed in step 2. As
will be seen later, the relative volatility (
) of a
mixture is:
Ki yi/xi p0i i j ate two species to a desired degree is strongly depen-

i,j" " + [1]
Kj yj/xj p0j j i dent on the value of
.
where Ki and Kj are the equilibrium constant for
components i and j respectively. These K values pro-
vide for each component a linear relationship for the
1035
mole fraction in the gas phase and the mole fraction
of the same component in the liquid phase. Eqn [1]
shows that the equilibrium constant and relative vola-
tility depend on the ratios of vapour pressure, liquid-
phase activity coefRcients () and vapour-phase
fugacities ( ).
The relative volatility
is a measure of the ease of
the separation.
E If
"1.0 the separation is not possible by distilla-
tion.
E If
51.2, distillation will probably be a good
alternative, and should be chosen as the next
step.
In step 4, the distillation system should be de-
signed. The design and analysis of these columns
operating at high and low pressure form the core of
this manuscript and are presented in the next sec-
tions.
Finally, in step 5, the scientist or engineer debates
whether the problem have been solved satisfactorily.
If this is the case, the problem is Rnished; if not, it is
necessary to go back and start again.
Distillation
Distillation is based on diffusion of one or more
components through a mixture operating at a temper-
ature, pressure and composition that assures the pres-
ence of liquid and vapour phases. In distillation, the
mass transfer is due to a concentration difference
moving from a place of high concentration to one of
low concentration; it is not bulk movement as a result
of a pressure gradient, like pumping liquid through
a pipe.
Relative Volatility
The key separation factor in distillation is the relative
volatility, deRned by eqn [1]. As the value of relative
volatility increases, the easier it is for components to
be separated by distillation.
The number or theoretical stages required to separ-
The variation of this parameter with pressure is
shown in Figure 1 for the system ethane}propane. As
seen,
is greater at low pressure than at high
pressure. Therefore, at low pressure (e.g. 1 atm), for
1036 II / DISTILLATION / High and Low Pressure Distillation

The design or sizing of distillation equipment re-

Figure 1 Variation of relative volatility with pressure.
quires the calculation of diameter and height of the
column. Diameter depends on volumetric Sow rates
of liquid and vapour inside the column, and these are
functions of the total amount of the mixture feed to
the column and the reSux or boil-up ratios. The
desired purity of the components at the top and bot-
tom of the tower dictates the height of the column.
First, the theoretical or ideal plates are calculated,
then efRciency is estimated to convert from ideal to
real stages. By specifying the distance between plates
(30}60 cm), and providing space for about four
plates at the top of the column and six for the
bottom for disengagement of the phases, the total
height of the shell is determined. Figure 4 shows
a block diagram for a typical design of a distillation
column.
Distillation at high and low pressures involves
special characteristics, summarized in Table 2.

With Internal Devices of Distillation Columns

There are three types of internal devices which
a speciRed separation, the number of theoretical provide the intimate contact between phases in a
stages is less than at high pressure (e.g. 10 atm). In distillation column. These are tray, random packing
this case, why not use low pressure for this separ- and structured packing.
ation? As will be seen later, the temperature at
the top and bottom of the column plays an important
role in this decision. At low pressure (0.05 atm) Table 1 Different classifications of distillation equipment or
top temperature !1303C, while at high operation
pressure (30 atm) top temperature is 103C. Therefore,
Amount of compound to be separated Solute recovery
distillation at high pressure is used since it is much
Fractionation
easier and more economical to reach 103C than Mode of operation Steady-state
!1303C. Unsteady-state
Batch
Semibatch
Classi\cation, Equipment and Design of Distillation Start and shut-down
Columns Mixing between phases Stagewise contact
Continuous contact
Distillation is the separation process most used in
Internal device used None
the chemical and petrochemical industry; as shown Plates
in Table 1, its operation is classiRed into several Bubble cups
forms. Sieve
Most of the time, distillation is carried out in verti- Valve
cal columns or towers (like the packed column shown Packing
in Figure 2) where the liquid descends while the va- Random
Structured
pour ascends to the top of the column. The vapour
System characteristics Flash
left at the top of the tower is condensed and at least Fractionation
a fraction is returned back to the top of the tower as Azeotropic
liquid reSux. Part of the liquid leaving the bottom of Extractive
the column is vaporized in a reboiler and returned to Operating pressure Low
the column as boil-up. High vacuum
How the distillation equipment operates and how Medium vacuum
Low vacuum
it is calculated has been modiRed over the years.
Medium
Figure 3 shows some developments related to distilla- High
tion.
 II / DISTILLATION / High and Low Pressure Distillation
Figure 2 Distillation column equipped with structured packing.
Each has advantages and disadvantages. Trays
have been used for many years. Random packings
have also been used over three generations of design.
Structured packings have replaced trays, especially in
applications at low and atmospheric pressures. At
high pressure, trays perform better than packings.
For any applcation one must determine whether
tray or packing is the most appropriate. The follow-
ing factors are an indication of when trays or pack-
ings are favoured.
1037
Factors favouring trays
E High liquid rate (this occurs when high column
pressures are involved);
E large diameter (packing prone to maldistribution);
E complex columns with multiple feed/take-offs;
E feed composition variation;
E scale-up less risky;
E columns equipped with tray weigh less than those
equipped with some packings.
Factors favouring packings
E vacuum conditions;
E low pressure drop required;
E in smaller diameter columns (where trays are more
difRcult to install, diameters 0.6}0.9 m or less);
E corrosive system (more construction materials
available);
E foaming;
E low liquid hold-up.
High Pressure Distillation
As seen from Table 3, distillation at high pressure
cover a wide range of applications that have some of
the following characteristics:
1. The compounds have low molecular weight (like
C2, C3, C4 hydrocarbons).
2. Cooling water can be used in the condenser.
3. A change in pressure could change the azeotropic
point by more than 10% (in mole fraction).
4. Energy is integrated between condensers and re-
boilers of different columns.
Sometimes the increase in pressure is limited by the
heat sensitivity of the bottom product (it could poly-
merize or degrade) or by its critical temperature or
pressure. As is known, at the critical conditions only
one phase exists. If the mixture reaches this point two
phases cannot exist anymore and therefore separation
is not possible. It is necessary, therefore, to be careful
when one is thinking of the operating column pres-
sure.
The upper pressure could be limited by economical
conditions. If water is used as cooling medium at
the top, the pressure required may be too high
and therefore it would be necessary to use a refri-
gerant. Some authors recommend a upper pressure
value of 1.48 MPa using a total condensation with
water, 2.52 MPa using partial condensation with
water, and if the pressure required is higher than
this value it is recommended to Rx the top pressure
at 2.86 MPa using partial condensation with a refrig-
erant.
1038 II / DISTILLATION / High and Low Pressure Distillation

Figure 3 Paradigm shifts related to distillation. (With permission from Chemical Engineering Process (1972) 68 (8): 16).
 II / DISTILLATION / High and Low Pressure Distillation 1039

Figure 4 Procedure for designing a distillation column.

The upper pressure also could be limited by cost. If
pressure is increased above 1 MPa, all equipment
costs begin to go up. For instance, a stainless steel
column with sieve trays could double in cost when the
pressure is raised from 0.1 to 3 MPa. A similar rela-
tionship exists for other equipment and materials of
construction.
Some speciRc applications of high pressure distilla-
tion are given in Table 3. Most applications are in the
petrochemical industry, but another very important
1040 II / DISTILLATION / High and Low Pressure Distillation

Table 2 Characteristics of high and low distillation operation

Feature Low pressure High pressure

When is it used?
Liquid density
Vapour density
Liquid diffusivity
Vapour diffusivity
Liquid viscosity
Vapour viscosity
Relative volatility
Products sensitive to temperature (thermal Gas feeds (C2/C3/C4 separation, air separation
degradation) and ammonia recovery)
Avoid using high pressure steam Allow water to be used in the condenser
Break azeotropes and allow separation of the system; also when a significant change occurs at the
azeotropic composition (about 5%) with change in pressure
When trains of distillations columns with energy integration are desired allow energy of overhead vapour
partially to vaporize the bottom of the other column)
High molecular weight of components Low molecular weight of components
No significant change Decreases slowly when pressure increases
Low; decreases if pressure decreases Increases quickly if pressure increases
Low; increases slowly if pressure increases Increases quickly if pressure increases
High; decreases if pressure increases Low; decreases quickly if pressure increases
High; decreases quickly if pressure increases Low; decreases quickly if pressure increases
High; no significant change at low or high High; no significant change at low or high
pressure pressure
Increases if pressure decreases Decreases if pressure increases


L G (0.1 '0.3
Flow parameter Fp"
G L

application is cryogenic air separation. This is based signing a new one. High pressure distillation begins
on a low and high pressure distillation column. The about 1 MPa, where liquid density, liquid viscosity
reboiler of the low pressure upper column (0.13 MPa) and surface tension are unusually low, while vapour
cools the condenser of the high pressure lower col- density is high. Trays are recommended as internal
umn (0.60 MPa). As can be seen, the high pressure devices, especially above 2 MPa; below this value,
column does not have a high pressure value but nei- trays and packings must be evaluated. Many prob-
ther is it at atmospheric pressure. lems in high pressure distillation could be avoided
The Further Reading section cites a good report on with an appropriate design. Available design
high pressure distillation by Brierley which gives methods for low pressure can be applied to high
some tips on improving an existing column and de- pressure but with some corrections, speciRcally in
downcomer design because this is where liquid Sood-
ing usually starts in high pressure distillation, while in
Table 3 Specific applications of high pressure distillation low pressure, the bottleneck is the vapour Sow
through the active area. Downcomer design should
Application Pressure Range
consider both downcomer back-up and downcomer
Ethylene plant velocity.
Demethanizer 32 MPa Three Sow regimes may exist in industrial columns:
Deethanizer 27 MPa spray, froth and emulsion (see Further Reading). In
C-2 splitter 16 MPa high pressure distillation, tray operation is usually in
Depropanizer 19 MPa the emulsion regime, where liquid loads are high and
Debutanizer 5 MPa vapour velocities comparatively low. However, in
Dimethyl ether (DME) production
small diameter column (less than 1.5 m) at low liquid
(via the dehydration of methanol)
DME column 1.60 MPa loads or at the low end of the high pressure range
Water column 0.75 MPa (about 1 MPa), froth and spray regimes can be found.
Production of heptenes from propylene In the spray regime, Sooding is caused by excessive
and butenes: entrainment of liquid from the active area to the tray
C3#C" 3 /mixture separation 0.60 MPa above. It increases the tray pressure drop and the
Miscellaneous hydrocarbons entrained liquid recirculates to the tray below. The
Propylene/pronane separation 2.1 MPa
larger liquid load in the downcomer and the increased
Cryogenic air separation
Lower column 0.6 MPa tray pressure drop together cause the downcomer to
overRll, so that the tray Soods. In the emulsion
 II / DISTILLATION / High and Low Pressure Distillation
regime, there is little entrainment of liquid by the
vapour; instead, the high liquid load causes the down-
comer to overRll and the tray to Sood. In the froth
regime, which is between the spray and the emulsion
ones, Sooding may be by either mechanism, depend-
ing on tray spacing and the particular combination of
vapour and liquid loads.
A new model for the design or analysis of sieve
tray columns has been developed at the University
of Texas at Austin, in its Separation Research
Program. This model applies to both low and high
pressure columns. The model was adjusted with
a wide experimental database from the open litera-
ture and also from the Separation Research Program
facilities.
The equipment used to perform distillation at high
pressure is basically the same as that used for close to
atmospheric pressure, but the following special con-
siderations should be taken into account.
1. The physical properties move in the direction in-
dicated in Table 2 (liquid density, liquid viscosity
and surface tension are low and vapour density is
high).
2. The thickness of the shell and column peripheral
equipment must be greater.
3. The advantage in capacity and/or efRciency of
structured packing over random packing and
plates decreases as the pressure increases.
4. At very high pressure the efRciency of distillation
decreases, due to back-mixing.
5. At very high pressure the use of plates (standard
and high capacity) is more reliable than packing.
The design equations for packed columns are basi-
cally the same, with one possible correction to the
height of a transfer unit (HTU) value, to take into
account the deviation from plug Sow:
HTUtotal"HTUplug flow#HTUback-mixing [2]
Remember that at high pressure the density of
gases increases by several orders of magnitude; the
opposite is true for the diffusivity of the gas phase,
and the surface tension decreases to very low values.
Low Pressure Distillation
Modern society is becoming ever more demanding in
the quality of the products it uses and for health and
environmental reasons a better removal of some com-
ponents is required. This better puriRcation of many
products sometimes requires operating at very low
pressure. Fortunately, the development of industrial
1041
equipment to obtain better and lower vacuum has
been maintained. The cost of carrying out these
operations is of course more expensive than at
around atmospheric pressure. The increase in cost
is inversely proportional to the absolute operating
pressure.
Distillation at low pressure is used for special cases
with one or more of the following characteristics:
1. heat sensitive products;
2. liquid feeds or liquid residue with high viscosity;
3. liquids with fouling and/or foaming tendencies;
4. low operating pressure (medium and high vac-
uum);
5. low residence time.
The applications may be classiRed into four groups:
E distillation or evaporation of sensitive organic
chemicals;
E concentration of foods, chemicals, polymers and
biological compounds;
E recovery of organic solvents;
E desolventing, devolatilization and Rnishing of
polymer solutions.
Table 4 shows the levels of vacuum used and also
lists representative equipment. Some advantages of
vacuum and molecular distillation are:
E low residence time;
E high selectivity due to the higher values for relative
volatility;
E cheaper heating requirements.
Some of these special kinds of distillation are dis-
cussed below.
Agitated Thin-\lm or Wiped-\lm Evaporators (WFE)
and Short Path Distillation Equipment
For medium vacuum distillation, thin-Rlm evapor-
ators are used with or without agitation, but evapor-
ators with scraping blades provide better perfor-
mance and Sexibility. Where there are heat-sensitive
substances, thermal decomposition may occur during
evaporation. Decomposition increases exponentially
with temperature and linearly with duration of ther-
mal load. A gentle distillation method therefore re-
duces the evaporation temperature and the residence
times at high temperature.
Since the evaporation temperature depends on
pressure, evaporation is performed under vacuum at
considerably lower temperatures. If, in addition to
applying vacuum, the thickness of the material on the
1042 II / DISTILLATION / High and Low Pressure Distillation

Table 4 Classification of low pressure distillation

Approximate pressure Typical equipment

Low vacuum 760 to 1 Torr (1000 to 1.315789 mbar) Vacuum column internals
Plates
Random packing
Structured packings
Medium vacuum 1 to 10\3 Torr (1.315789 to 0.013157 mbar) Thin-film evaporators
Agitated thin-film evaporators
High vacuum 10\3 to 10\7 Torr (0.013157 to 0.0000013 mbar) Short path distillation
Molecular distillation
Rotary stills
Falling-film stills
Wiped-film stills
Centrifugal stills

Adapted with permission from Eckles AJ (1997) Difficult to process? Vacuum it! Chemical Engineering 94}100.

evaporator wall is reduced, lower evaporation operate at the lowest pressure and provide the lowest
temperature and shorter residence times can be pressure drop.
achieved. The double-walled evaporator jacket is heated con-
Most WFE are vertical cylinders where the feed tinuously with a heating medium. A vacuum system
material is distributed to the inner surface; as the (often a combination of several individual pumps)
liquid Sows downward, axially arranged blades or reduces the pressure in the distillation chamber.
roller wipers distribute the liquid as a thin Rlm which
is constantly mixed. In Figures 5 and 6, two types of
WFE are shown. Figure 5 illustrates a WFE with
rotating blades, while Figure 6 shows a unit with
roller wipers and condenser. The last feature is a dis-
tinct characteristic of short path evaporators and mo-
lecular distillation stills. These types of equipment

Figure 5 Agitated thin- or wiped-film evaporator (WFE) with Figure 6 Wiped-film evaporator or short-path distillation equip-
rigid blade rotor. ment, with roller wiper system.
 II / DISTILLATION / High and Low Pressure Distillation
Depending on the temperature and the pressure in the
distillation chamber, vapours leave through the va-
pour discharge and travel to an external condenser.
Involatile substances are discharged at the lower end
of the evaporator.
The cylindrical evaporation chamber is externally
heated with hot pressurized water, steam or heat
transfer oils.
Essential parts of a WFE are the rotating blades or
roller wiper system, which are axially arranged in the
evaporator. The blades (or roller wiper system) inSu-
ence the following aspects:
E Film thickness: the components to be evaporated
are more easily separated from a thin Rlm.
E Uniformity of distribution: uniform distribution of
the feed material on the evaporator surface pro-
motes excellent heat transfer and avoids overheat-
ing.
E Mixing of the Rlm: optimal mixing within the
liquid Rlm increases the complete separation of the
components, thus enhancing evaporation.
E Residence time: if gentle distillation is required,
heat-sensitive materials can be heated in the evap-
orator for only a short period of time. With a well-
constructed blade or wiper system, the residence
time can be considerably reduced.
Thin-Rlm evaporators can be totally vacuum-
sealed, thus avoiding any oxidation of product caused
by penetration of air.
Discharge of the distillate vapours takes place
above the feed nozzle. This is why undistilled liquid
droplets, which may occur during Sash evaporation,
are trapped in the head of the evaporator and then
Sow back to the evaporator surface.
McKenna (see Further Reading) provides equations
to design or analyse operation of a WFE. Some of the
important equations to calculate the Rnal concentra-
tion CF of the residual liquid are:
C0!CH
CF" #CH [3]
(1#
)ntotal
w0.5de2 tan

n"21.5Dif 0.5
QN0.5b


2NbQ0.5 de 0.5
; 1# 2.5 2 0.5 !lb sin [4]
 de w tan Nb
Hede HeNb
ntot" " [5]
deHs de tan
1043
Advantages of the process of WFE and short path
distillation Thin-Rlm evaporators offer a number of
advantages. If higher operating pressure and temper-
ature are required for economic reasons, thermally
sensitive materials can still be processed because of
the short residence times utilized in WFE equipment.
The thin liquid Rlm and turbulent mixing of the
Rlm result in very quick attainment of equilibrium.
This is especially important if a complete separation
of a low boiling, volatile component out of the resi-
due is required. This is why thin-Rlm evaporators
are successfully used as reboilers for rectiRcation
columns.
Thin-Rlm evaporators are excellent degassers. If,
for example, small quantities of a volatile component
have to be removed down to only a few p.p.m., the
evaporation capacity is not very important. The main
goal in this case is to transfer the portions to be
separated as completely as possible to the surface of
the Rlm. The transport is achieved by the roller wiper
or blade system.
Molecular Distillation
Molecular distillation may be considered as a special
version of evaporative distillation in which the liquid
is evaporated without boiling, but in such circum-
stances that the evaporating molecules reach the con-
denser surface without obstruction. Three conditions
for molecular distillation are:
1. Pressure must be lower than 0.001 mmHg. This
low pressure is required to ensure that the mole-
cules do not collide with each other.
2. The distance between evaporation and condensa-
tion surfaces is of the same order of magnitude
as the mean free path of the molecules and the
free motion of the molecule is not mechanically
hindered.
3. The temperature of the condenser surface should
be between 50 and 1003C lower than that of the
evaporation surface to prevent re-evaporation of
molecules.
The mean free path  is given by:
1
" [6]
(2N
where  is the diameter of the molecule in centimetres
and N is the number of molecules in 1 cm3. By reduc-
ing the pressure to very low values, N decreases,
giving values for  of 1}3 cm, in the range of
the distance between evaporator and condenser
surface.
1044 II / DISTILLATION / High and Low Pressure Distillation

In molecular distillation the maximum or theoret- equipment or molecular distillation are shown in
ical rate of evaporation was proposed by Langmuir in Table 5.
1916:

Technologies to Improve Distillation


M
We"0.0583Pmm [7] Processes
T
There are several ways of improving the separation of
where M is the molecular mass and T is the absolute mixtures into the desired products. Some of these
temperature. combine distillation with other processes (like ad-
The relative volatility is given by: sorption, stripping, pervaporation, reverse osmosis,
membranes, etc.). Others are improvements to the
p01 1 (M internal devices (such as high efRciency trays or high

1,2" 0  [8]
p2 2 (M efRciency packings). There are also the so-called en-
 hanced distillation methods (like extractive distilla-
tion, homogeneous and heterogeneous azeotropic dis-
From these two equations it may be seen that tillation, reactive distillation, heat integration, high
molecular weight of the compounds involved is an gravity distillation, spinning cone distillation, mech-
important consideration. anical vapour recompression, and pressure swing dis-
Equipment for molecular distillation and medium tillation). Further information about these technolo-
vacuum is expensive, but it is economically justiRed gies can be found in the texts cited in the Further
for the separation of high value products such as Reading section.
vitamins, fats, essential oils and hormone concentra-
tion.
Typical unit operations and applications for me- Conclusion
dium and high vacuum using WFE, short path
In the 21st century, distillation at high, medium and
low pressure, will continue to be a much used method
for the separation of components from homogeneous
Table 5 Specific applications of low pressure distillation
mixtures.
Unit operations where wiped-film evaporators (WFE) are used
Evaporation Drying See also: I/Distillation. II/Distillation: Historical Develop-
Concentration Degassing
ment; Theory of Distillation. Membrane Separations:
Distillation Reactions
Stripping Deodorizing
Filtration.
Heating

Distillation Concentration Further Reading

Application of WFE for distillation and concentration at pressure Billet R (1979) Distillation Engineering. New York:
of 1}1000 mbar Chemical.
Acrylonitrile Extract solutions Brierley RJP (1994) High-pressure distillation is different!
Essential oils Insulin Chemical Engineering Progress 90 (7): 68}77.
Amines caprolactam Peroxides Dean JA (1995) Analytical Chemistry Handbook. New
Quinoline derivatives Phospholipids York: McGraw-Hill.
Dioctyl phthalate Pyrethrum Hewitt GF, Shires GL and Polezhaev YV (eds) (1997) Ency-
clopedia of Heat and Mass Transfer. Boca Raton, FL:
WFE for distillation and concentration of viscous products
Fatty alcohols Resins CRC Press.
Waxes Honey HolloH J, Kurucz E and BoroH di (1971) The Applications
Glue of Molecular Distillation. Budapest: Akademiai
Polymers KiadoH .
Lubricant oils Humphrey JL and Keller GE II (1997) Separation Process
Silicone fluids Technology. New York: McGraw-Hill.
Kister HZ (1990) Distillation Operation. New York:
WFE for distillation and concentration of products liable to encrust McGraw-Hill.
the heated surface Kister HZ (1992) Distillation Design. New York:
Used oils Contaminated effluents
McGraw-Hill.
Solvents containing impurities Residue products from rectifi-
cation and evaporation plants
Lockett MJ (1986) Distillation Tray Fundamentals. Cam-
bridge: Cambridge University Press.

McKenna TF (1995) Design model of a wiped Rlm
evaporator. Application to the devolatilization of poly-
mer melts. Chemical Engineering Science 50 (3):
453}467.
Historical Development
M. S. Ray, Curtin University of Technology,
Perth, Western Australia
Copyright ^ 2000 Academic Press
Introduction
Distillation is one of the oldest and most widely
studied unit operations in chemical engineering. It is
familiar as a separation technique to chemical, pro-
cess and petroleum engineers and to chemists. The
common techniques, design methods and numerous
applications have been extensively documented in
monographs and in the journal literature (and confer-
ence proceedings) over many decades. Specialist texts
and more general handbooks should be familiar to
those working in related Relds, therefore these stan-
dard sources are not documented here. The import-
ance of distillation and its future directions have been
discussed by Kunesh et al. (1995) and by Porter
(1995) (see Further Reading). This article presents
a state-of-the-art overview of distillation by concen-
trating on recent advances and possible future devel-
opments.
Sources of Information and Data
For a subject as old as distillation, and with such
a wide range of applications, there is an extensive
collection of published information. With the advent
of electronic databases and online web-based re-
sources it has become much easier to perform litera-
ture searches on particular topics and to keep abreast
of the current literature and recent developments. For
this reason an extensive reference list to journal arti-
cles is not provided in the Further Reading at the end
of this article. Selected papers are included that can be
used to locate related references. Author or subject
searches of the journal literature (from 1956 to the
present, with six-monthly updates) can easily be
performed by using the CHERUB2+ Chemical Engin-
eering Database (complied by M. S. Ray) which is
included on the Engineering & Applied Science CD-
Rom Database. The ability to easily search the chem-
ical engineering literature is a recent development,
and is an important advantage for distillation re-
II / DISTILLATION / Historical Development 1045
Seader JD and Henley EJ (1998) Separation Process
Principles. New York: Wiley.
Treybal RE (1980) Mass-Transfer Operations, 3rd edn.
New York: McGraw-Hill.
searchers. Another useful reference source is a series
of annual update bibliographic papers on ‘Equilib-
rium-staged Separations’ (e.g. Separation Science and
Technology 32(18): 3067}3083, 1997). Patent
searches can also be performed on the web, e.g.
www.ibm.com/patents, and www.uspto.gov (the
website of the US Patent & Trademark OfRce). Sev-
eral handbooks and monographs (and CD-ROMs)
containing property data useful for distillation sys-
tems have also been compiled, e.g. C. L. Yaws (trans-
port properties data and thermodynamic diagrams);
J. Gmehling and co-workers (including VLE data,
heats of mixing, azeotropic data); and the American
Institute of Chemical Engineers’ Design Institute for
Physical Property Data (DIPPR) publications.
Prediction of Vapour^Liquid Equilibria
(VLE) Data
SigniRcant advances in the interpretation and predic-
tion of vapour}liquid equilibria (VLE) data have been
made since the 1970s. These advances developed
from the publication of a range of equations of state
(EOS) based upon the application of traditional ther-
modynamic principles and relationships. The EOSs
provide interpretation or evaluation of available ex-
perimental VLE data. The Wilson model (1964) is
probably the most popular for dealing with liquid-
phase activity coefRcients because it has only two
adjustable parameters, and it works well for both
binary and multicomponent systems. The prediction
of nonidealities in binary mixtures using the
UNIQUAC model (1975) is rather more complex.
Subsequent and related studies led to the develop-
ment and use of the Group Contribution Methods
such as ASOG and UNIFAC for the prediction of
VLE data. The latter is widely used when actual
system data are not available, provided that an ap-
proximate nonideality correction is acceptable. New
methods are being developed and probably the one
showing most promise and of general applicability is
known as: A Generalized Approach to Phase Equilib-
ria (AGAPE, 1995).
There are many EOS models described in the litera-
ture but only a few have wide use for engineering
1046 II / DISTILLATION / Historical Development
applications. The Redlich}Kwong equation (1949)
and its modiRcation by Soave (1972), and the
Peng}Robinson equations (1976) have broad general
application for nonpolar components and are the
most widely used, especially in the reRning and gas
processing industries. For highly nonideal compo-
nents, use of the EOS requires an appropriate mixing
rule. The alternative approach is direct application of
liquid activity model parameters in the EOS. A collec-
tion of papers relating to the development of the
Peng}Robinson EOS over 20 years was published in
1998 (Industrial and Engineering Chemistry Re-
search 37(5): 1579}1706).
The availability of commercial software packages,
e.g. Sowsheeting design packages such as HYSYS2+
and PROTISS2+ (see below), have made the predic-
tion and evaluation of equilibrium data quicker and
easier. These packages were generally developed for
use by the oil and gas industry and reRning com-
panies, and they include extensive VLE prediction
equations integrated within the design methods.
Some of these prediction methods have been de-
veloped speciRcally for use with common petroleum
systems (e.g. Chao-Seader; Grayson-Streed).
Applications of the Design Methods
Distillation design methods are well established and
described in detail in the traditional reference texts
(see Kister, 1992). The original methods were for-
mulated in the 1920s and 1930s such as the
McCabe}Thiele (1925) and Ponchon-Savarit (1921,
1922) methods for binary mixtures, and the rigorous
multicomponent analogues of the Lewis}Matheson
(1932) and Thiele}Geddes (1933) procedures. Use of
the latter trial-and-error methods emphasized the
need for the incorporation of suitable numerical tech-
niques to ensure that the solution (of number of
stages) would actually converge, and also to reduce
the time spent performing the calculations. Designers
later came to rely on short-cut design methods, e.g.
Fenske (1932), Underwood (1945, 1946, 1948),
Smith}Brinkley (1960), etc., to provide ‘ballpark’ re-
sults before embarking upon the detailed rigorous
calculations. Easier access to mainframe computers in
the 1960s, and desktop machines in the late 1970s
meant that the time required to perform the numer-
ical calculations was reduced, and developments
then centred on design methods which simpliRed the
problem formulation, e.g. matrix manipulation
techniques.
The arrival in the 1980s of general Sowsheeting
design packages such as HYSIM2+ and PRO/II2+
(replaced by HYSYS2+ and PROTISS2+ in the 1990s)
shifted the design emphasis away from the develop-
ment of speciRc design methods to the use of commer-
cially available packages which could provide quick
and easy short-cut designs. These developments
meant that the designer was liberated from tedious
calculation but still required a sound knowledge of
distillation principles and the ability to analyse the
simulation results in order to avoid serious errors.
Many papers have been published in the mid-1990s
concerning the limitations of the simulation methods
(e.g. Chemical Engineering Progress 91(6): 63}75,
1995; Chemical Engineering Education 31(1): 46}51,
1997), and the problems that can occur if too great an
emphasis is placed upon their use with too little
feedback from experienced designers (e.g. Chemical
Engineering Progress 94(6): 63}77, 1998).
Other design packages are available, the most re-
cent developments being the Computational Fluid
Dynamics (CFD) modelling software. Such packages
may be useful for modelling effects within distillation
equipment rather than straightforward applications
of the equilibrium stage calculations. CFD is dis-
cussed later in this article in relation to possible future
developments.
Advances in Column Design
The basic equipment used to achieve a separation has
not changed signiRcantly within the last half century.
A tray (or packed) column is still used to provide
contact area between the liquid and vapour phases in
order to achieve mass transfer and hence the desired
separation. Tray columns are generally preferred
(packed columns are used for particular types of sep-
arations, or speciRc situations) and the equipment
consists of tall vertical towers with a large percentage
of internal free space. Distillation is also character-
ized by its thermal design requirements and is an
energy-intensive process. Advances in column design
have tended to focus on improvements in energy
usage and/or improved separation efRciencies. Two
examples from the literature are the Higee distillation
unit and the integral dual column. The Higee high-
performance distillation or extraction unit described
by Ramshaw (The Chemical Engineer (IChemE)
June: 17}21, 1987) attempted to utilize centrifugal
Relds to improve the separation efRciency and to
reduce the column size. The dual distillation columns
(see US Patent no. 4 681 661 (1987) and The Chem-
ical Engineer (IChemE) December: 21, 1987) consists
of two concentric columns (the stripping and rectify-
ing sections) arranged one within the other, the rea-
son being an attempt to better utilize heat effects and
also to reduce the column height. Neither of these
developments (or numerous others described in the
literature) has been widely adopted by industry,

possibly because they do not offer signiRcant cost
advantages (or because of the conservative nature of
the chemical processing industry).
Column Developments
These have centred mainly on a better understanding
of the Suid dynamics within the column and speciR-
cally across the trays (the Rrst book on this topic was
by Lockett in 1986), and also on improved prediction
methods for determining plate and overall column
efRciencies. A better knowledge of the interrelation of
these two aspects is beginning to emerge from several
research groups, e.g. Biddulph and co-workers, and
their studies of the relationship between Marangoni
surface tension effects and plate efRciencies (Ameri-
can Institute of Chemical Engineers Journal 37(8):
1261}1264, 1991). A development since the 1970s is
the preference for sieve trays, rather than bubble or
valve trays which were prevalent up to that time. The
AIChE Bubble Tray Design Manual (1958) is still
used (with modiRcations) and quoted, even for the
calculation of sieve tray efRciencies for which it was
not intended and for which it provides rather poor
results. However, this older method has provided
a starting point for recent studies which examine the
signiRcance of Suid-related variables (such as surface
tension) and how this knowledge can be used to
design more efRcient tray columns.
Many papers have been published in the more
industry-orientated journals (e.g. Chemical Engineer-
ing (NY), Chemical Engineering Progress and Hydro-
carbon Processing) concerning column operation and
performance problems. These papers also discuss par-
ticular aspects of internal column design and use of
the newer structured packings, now available as alter-
natives to tray systems, in relation to column perfor-
mance and operation. The use of high efRciency pack-
ings requires a better understanding of the hy-
drodynamic conditions and the mass transfer pro-
cesses that occur in the packing. Fouling and plugging
within distillation equipment can become serious
problems and they are areas in need of better under-
standing. High surface area packings are popular
because they promote efRciency, however they are
also more prone to fouling problems.
If solids are present in a stream then design solu-
tions generally act to keep the solids moving. There-
fore any liquid maldistribution within the column or
tray channelling (due to initial vapour maldistribu-
tion) must be corrected to avoid plugging problems.
Plugging and fouling have varying effects depending
upon the actual in-service conditions, and hence it is
difRcult to devise generic strategies or solutions. Gen-
eral advice tends to focus upon the need for good
II / DISTILLATION / Historical Development 1047
distributor design, good wetting (and wettability) of
the mass transfer surface, and good distribution of the
Sow streams within the equipment.
Papers describing column operational problems,
such as product draw limitations especially for col-
umn revamps (Chemical Engineering Progress 94(6),
63}77, 1998), are particularly useful for designers
and process operators.
Energy-intensive Nature of Column Design
The energy-intensive nature of column design and
operation is unlikely to change signiRcantly, mainly
because the requirements for the reboiler to provide
vapour Sow and the condenser to provide liquid
product are essential aspects of the separation. An
additional constraint is that the heat removed by the
condenser is typically low grade and of little use
elsewhere in the plant. Traditional solutions have
been to use vacuum operation in the column to lower
the boiling point of the mixture (especially for heat-
sensitive materials), and to utilize low-grade heat
from other plant operations to provide the reboiler
duty. Attention has been directed towards lowering
the energy requirements of a column, e.g. by reducing
the reSux ratio. Energy-saving revamps have replaced
trays with packings in order to create more theoret-
ical stages within a column, and hence reduce the
reSux rate and boil-up rate. The integration of col-
umns within a sequence and the use of overhead
vapours in another column’s reboiler (e.g. by use of
a Rankine cycle) have also been considered. Energy
saving and integration techniques have become stan-
dard design procedures since the 1970s (including the
use of pinch technology and heat exchanger network
design). However, it is difRcult to achieve (or to
envisage) large scale reductions in the thermal re-
quirements of distillation due to the inherent basis of
the separation itself, i.e. latent heat of vaporization is
required for the essential partial vaporization at each
stage. The most likely developments in this area are
the use of hybrid systems, e.g. membranes#distilla-
tion, which effect a part of the separation in a less
energy-intensive operation (see the section below on
developments and applications).
Particular Techniques and Situations
This section considers only two speciRc aspects of
distillation operations, namely process control and
difRcult separations (e.g. azeotropes).
Chemical Process Control
This was mainly in the hands of electrical/instrument
engineers until the mid-1960s, and most of the
1048 II / DISTILLATION / Historical Development
published literature reSected their particular expert-
ise. Since that time process control has developed as
a signiRcant area of chemical engineering expertise
and publication. Many academics have adopted and
published in this Reld (see Luyben, 1992) and there
are also several handbooks and practical texts dealing
with plant operations (e.g. Shinskey, 1984). Distilla-
tion was one of the primary chemical engineering unit
operations to be researched and developed in depth
for process control applications. This was because of
the scope offered by distillation as a control problem
and the range of options and alternatives available.
The focus of distillation control is usually the product
speciRcation, but the critical operations of reboiler
and condenser performance, the uncertainty of feed-
point location, the reSux rate, and the need for cor-
rect internal functioning (e.g. Suid Sow on and be-
tween the trays) means that several complex relation-
ships and problems need to be considered together.
The number of publications dealing spe-
ciRcally with distillation control attests to the
complex nature of the problem and the number of
approaches and applications that are possible
within any single situation. An additional considera-
tion for the designer is the possible number of column
arrangements, utilizing several columns, and their
possible alternative speciRcations (hence the advant-
ages offered by simulation packages), and therefore
the integration of individual column control
within the overall control of a set of distillation
operations.
Developments in distillation control have focused
on several areas, generally searching for optimal solu-
tions to the following:
1. Individual column control (including reSux
streams, and reboiler and condenser operation
and performance).
2. Control of a column required to perform a speciRc
separation (e.g. heat-sensitive mixtures) or a par-
ticular type of application (e.g. separation of azeo-
tropes or close-boiling mixtures; reactive distilla-
tion).
3. Control of sequences and arrangements of several
columns.
Distillation control has developed into a multi-
faceted problem incorporating mass transfer and sep-
arations, process modelling and optimization tech-
niques, instrumentation and control functions, and
the application of simulation packages in order to
evaluate a range of possible problem solutions. Con-
trol theory has now developed to include a range of
advanced techniques such as adaptive control, model-
based control, and the use of neural networks which
supplement the basic approaches offered by combina-
tions of proportional, integral and derivative (P-I-D)
control functions.
Dif\cult Separations
DifRcult separations using distillation techniques de-
pend primarily upon the nature and properties of the
components in the mixture, rather than the physical
arrangement of the equipment such as bubble caps
versus sieve trays, tray versus packed columns, etc.
Reactive distillation is an exception to this generaliz-
ation (see next section). Distillation contrasts with
other separation techniques such as membranes and
adsorbents where the characteristics of the actual
separation media have a signiRcant effect upon the
ability to separate the components. This is why gen-
eric distillation methods have been developed and
used effectively whereas there is no single theory or
method available for the design of either membrane
or adsorption systems. However, distillation was de-
veloped in order to separate mixtures of components
exhibiting signiRcant differences in relative volatil-
ities and boiling points and problems arise where this
is not the case. In particular, azeotropes exhibit no
difference in volatility (at certain conditions) and
hence no change in the composition of the mixture
obtained. Traditional approaches have been either to
avoid the conditions where an azeotrope forms, or
(more likely) the addition of a solvent or entrainer
which ‘breaks’ the azeotrope but which also requires
an additional column(s) to remove and recycle the
solvent. The common azeotropes are well known and
documented, e.g. ethanol/water, acetone/chloroform,
etc. Recent developments have centred on the ability
to predict the occurrence of an azeotropic mixture (or
a mixture with very little difference in component
volatilities), and also on the application of the tradi-
tional design methods to the design of a range of
column conRgurations. These conRgurations aim to
produce an optimum separation in terms of product
speciRcation, minimum use and recycle of solvent,
effective and feasible control schemes, minimum
capital and operating costs. Computer-based prop-
erty packages have been developed speciRcally to
predict azeotrope formation, and the vapour}liquid
equilibria data can then be used in an appropriate
column simulation package to evaluate the alterna-
tive equipment arrangements. Researchers have de-
veloped prediction and property packages by the ap-
plication of basic thermodynamic principles, and also
utilizing speciRc equations of state (see section on
VLE data above). The property packages have also
included data on available solvents. The aim is gener-
ally to combine azeotrope prediction with solvent
selection, and full speciRcation of the combined mix-
ture properties for use in column design simulations.

Developments and Applications
Reactive or Catalytic Distillation
Reactive or catalytic distillation has emerged as a sig-
niRcant development in recent years, the original pat-
ents were obtained by L.A. Smith in the early 1980s.
Several research groups have considered a range of
speciRc systems and applications and a substantial
body of literature has been published. The main sys-
tem that has been investigated and reported is the
production of methyl tert-butyl ether (MTBE) mainly
due to the high efRciency of the process, and also the
related systems of ethyl tert-butyl ether (ETBE) and
tert-amyl methyl ether (TAME). The technique in-
cludes reactive (homogeneous catalyst) or catalytic
effects (heterogeneous catalyst acts as the packing)
within the traditional distillation equipment. The de-
sign requires a combination of the well-known mass
and energy balance equations with the individual
reaction mechanisms, and consideration of heats of
reactions. The design of such systems requires model-
ling of the mass transfer, reaction and separation,
thermal effects, and the inherent control systems. An
additional requirement is the knowledge of any pos-
sible side reactions and by-product formation, and
their effect on the Rnal separation of the mixture
produced.
If a catalyst is required then the physical installa-
tion of this material, its removal and replacement,
and its effect upon the Suid dynamics within the
column become important additional operational and
design considerations. The distillation column now
becomes a countercurrent two-phase Sow, Rxed-bed
reactor. Packed beds typically have a void fraction of
0.7 (even up to 0.95), whereas small catalyst particles
result in a voidage of 0.3}0.4 which makes counter-
current operation impractical. Therefore catalyst sup-
port structures must be designed and incorporated to
provide a voidage of at least 0.5, and in addition
allow for expansion and contraction of the bed, and
provide a uniform catalyst spatial distribution. An
alternative is to manufacture the catalyst in the shape
of a distillation packing. For certain systems there are
distinct advantages from employing reactive distilla-
tion, but their dynamics are complex and careful
consideration of the operation and control are re-
quired in order to avoid potential difRculties.
Hybrid Separation Systems
Hybrid separation systems combining two or more
unit operations have become popular in the 1990s,
and there are numerous examples in the journal liter-
ature. Many of the examples have combined distilla-
tion with a membrane-type technique, thus aiming to
II / DISTILLATION / Historical Development 1049
utilize the well-known data and methods (and separa-
tion efRciency) of distillation with the low tempera-
ture advantages offered by membrane systems. Os-
motic distillation (or isothermal membrane distilla-
tion) has been developed speciRcally for applications
requiring the retention of Savours and fragrances
(Chemical Engineering Progress 94(7): 49}61, 1998).
This hybrid process can concentrate solutes to very
high levels at low temperatures and pressures, and it
is particularly suitable for sensitive solute materials. It
can also effect the selective removal of a single vol-
atile solute from an aqueous solution. The process
involves the transfer of volatile components between
two inherently miscible liquid streams, separated by
a semipermeable membrane, the driving force being
provided by differences in component activity be-
tween the streams. The strip solution becomes diluted
during separation/transfer and must be reconcen-
trated by distillation (or evaporation). Many other
examples of membrane}distillation systems can be
found in the literature. Alternative hybrid systems
include crystallization}distillation, solvent extrac-
tion}distillation, pervaporation}distillation, and
Suidized reaction}distillation. These systems have
been described in some detail including patents, al-
though generally the work is directed towards a par-
ticular application or problem.
The Future: Developments and
Applications
Future developments would seem to be most closely
linked to the areas already outlined in this chapter
rather than sudden and unexpected applications. Dis-
tillation is a mature separation technology and devel-
opments are likely to be in incremental advances in
our knowledge and understanding of the process
itself, and in the underlying principles that determine
its ultimate effectiveness for separating components.
The most likely areas for signiRcant advances and
developments are:
1. Further combinations of mass transfer effects
within distillation equipment (tray or packed col-
umns), developing the current trends of reaction
with distillation (reactive and/or catalytic distilla-
tion), and hybrid systems of membranes (or other
techniques) with distillation.
2. Improvements in separation effectiveness (and
costs) including attempts to reduce the size of the
equipment, increased efRciencies, and reductions
in energy requirements.
3. More reliable prediction and modelling techniques
directed towards VLE predictions, efRciency mod-
els and predictions, and improvements in the CAD
1050 II / DISTILLATION / Instrumentation and Control Systems
modelling packages in order to identify practical
limitations of the simulations at an early stage.
4. Development of separation systems incorporating
distillation in order to address speciRc environ-
mental problems and applications.
SigniRcant applications are expected in the use of
computational Uuid dynamics (CFD) packages for
prediction of effects occurring within distillation
equipment. This is a different area of research from
the use of the Sowsheeting packages and the calcu-
lation of equilibrium stages. The CFD approach (gen-
erally using commercial packages such as PHOE-
NIX2+ and FLUENT2+) has been used to predict
single-phase Sow patterns (of a vapour phase) from
numerical solutions of the Navier}Stokes equation,
turbulence equations, and the continuity equation. If
the equations of momentum and mass transfer are
inserted into the CFD methodology then it may be
possible to predict the Sow patterns and their effects
upon tray performance. However, the major chal-
lenge is the consideration and modelling of the three-
dimensional froth height and its shape.
Reviews of the state-of-the-art in distillation and the
need for and possible directions of future research have
been discussed by Fair (1988), Kunesh et al. (1995)
and Porter (1995). Assessments of advances and devel-
opments in distillation equipment regularly appear in
the journal literature, e.g. Chemical Engineering (NY),
December 1992; Hydrocarbon Processing, February
1989; The Chemical Engineer (IChemE), September
1987. Fouling and plugging in equipment and a better
understanding of the internal Sow mechanisms and
regimes are areas receiving and requiring further atten-
tion, as discussed earlier. Most new ideas tend event-
ually to become either an academic curiosity, or niche
applications, and approximately every 10 years a new
technique gains attention and prominence, e.g. react-
ive distillation, membrane}distillation.
Instrumentation and Control Systems
B. Roffel, University of Twente, Faculty of Chemical
Engineering, Enschede, The Netherlands
Copyright ^ 2000 Academic Press
Introduction
Distillation columns have been widely used in the
past to separate mixtures of liquids into individual
components. And even though new separation tech-
niques are being developed, distillation remains the
See Colour Plate 38.
See also: II/Distillation: Energy Management; Instrumen-
tation and Control Systems; Modelling and Simulation;
Theory of Distillation.
Further Reading
CHERUB2+ d CHemical Engineering Reference User
Bibliography on the Engineering & Applied Science
CD-ROM from INFORMIT, Melbourne, Victoria,
Australia (published semi-annually by subscription; de-
tails available from this author).
CAD Design Packages: HYSIM2+ (1987) and HYSYS2+
(1996), Hyprotech Ltd, Alberta, Canada; PRO/II2+
(1984) and PROTISS2+ (1996). California, USA: Simu-
lation Sciences Inc.
Fair JR (1988) Distillation: whither, not whether. Chemical
Engineering Research and Design 66: 363}370.
Kister HZ (1990) Distillation Operation. New York:
McGraw-Hill.
Kister HZ (1992) Distillation Design. New York:
McGraw-Hill.
Kunesh JG, Kister HZ, Lockett MJ and Fair JR (1995)
Distillation: Still towering over other options. Chemical
Engineering Progress 91(10): 43}54.
Luyben WL (1992) Practical Distillation Control. New
York: Van Nostrand Reinhold.
Porter KE (1995) Why research is needed in distillation.
Chemical Engineering Research and Design 73(4):
357}362.
Shinskey FG (1984) Distillation Control, 2nd edn. New
York: McGraw-Hill.
Sneesby MG, TadeH MO, Datta R and Smith TN (1998)
Detrimental inSuence of excessive fractionation on
reactive distillation. American Institute of Chemical
Engineers Journal 44(2): 388}393.
TadeH MO, Sneesby MG, Datta R and Smith TN (1997)
ETBE synthesis via reactive distillation. Part 1: Steady-
state simulation and design aspects; Part 2: Dynamic
simulation and control aspects. Industrial and Engine-
ering Chemistry Research 36(5): 1855}1869 and
1870}1881.
most important separation method applied in the
process industries today.
The layout of a simple distillation column is shown
in Figure 1. A single feed enters the column at the side
and two products are produced: the light or most
volatile components are withdrawn from the top and
heavy components are removed from the bottom.
Heat (in the case shown, steam) for evaporation of
the liquid is supplied to the reboiler, and heat is
removed (in this case, through cooling water) at the
top in the condenser. The nomenclature used in this
 II / DISTILLATION / Instrumentation and Control Systems
Figure 1 Distillation column layout.
chapter is shown in Figure 1, where V" vapour
Sow, F"feed Sow, R"reSux Sow, D"distillate
Sow, B"bottom Sow, x"composition, M"mass
hold-up and W and S are cooling water and steam
Sow respectively. In order to maintain constant sep-
aration in the distillation column, it should be well
instrumented and controlled. The treatment of instru-
mentation and control techniques will focus primarily
on packed columns or columns with trays.
A number of excellent books and review articles
have been written about distillation control. Shin-
skey’s book (1984) is a very practical one and pro-
vides a good introduction to the subject of distillation
control, although a detailed explanation of different
approaches to control alternatives is always clear to
the novice. The book by Buckley et al. (1985) pro-
vides a comprehensive treatment of pressure, level
and protective controls; control of composition is
restricted to a short treatment of composition dynam-
ics in binary columns. The book edited by Luyben
(1992) is probably the best starting point: It has been
written by several experts in the Reld of distillation
dynamics, instrumentation and control. It provides
a comprehensive treatment of distillation models, dis-
tillation simulation, identiRcation of distillation pro-
cesses and selection and comparison of control struc-
tures. In addition, several chapters are devoted to
1051
particular case studies. The Process Control Instru-
ment Engineers’ Handbook (Liptak, 1995) also pro-
vides three interesting sections on distillation control:
one section discusses basic controls, another section
discusses advanced controls and in a subsequent sec-
tion the subject of relative gain calculations is re-
viewed. The handbook also provides a wealth of
information on instrumentation.
A book which gives a good introduction to batch
distillation control is the one by Fisher (1990) and
Luyben (1992) gives a good overview. Three review
articles should be mentioned as a starting point for
further reading: the Rrst one is by Tolliver and Wag-
goner (1980: 195 references), another is written by
McAvoy and Yang (1986: 270 references), and the
one by Skogestad (1992) also provides easy-to-under-
stand material: it has 206 references.
Because vapour and liquid with a certain energy
content are present in the column, basic instrumenta-
tion includes measurements of the vapour hold-up
(column pressure), liquid hold-ups (column bottom
level and reSux drum level) and generally a number of
temperatures along the column. In addition, the in-
and outgoing Sows are usually measured, as shown in
Figure 1.
Since the feed to the distillation column is often
Rxed by a preceding process, no control valve is
1052 II / DISTILLATION / Instrumentation and Control Systems
shown in this Sow, although sometimes the feed pre-
heater control valve is used to control the amount of
feed that vaporizes. This means that Rve manipulable
Sows remain and there are essentially Rve degrees of
freedom for control. However, the vapour and liquid
hold-ups have to be controlled, which means that
5!3"2 degrees of freedom remain, which are gen-
erally used for composition control.
Distillation columns pose some interesting control
problems. First of all, the process is often highly
nonlinear. Secondly, there are Rve variables to be
controlled (pressure, P, level, hB, level, hR, composi-
tion, xD, composition, xB) and Rve variables which
can be manipulated (Sows R, D, B, S and W). How
the coupling between these controlled and manipu-
lated variables should be established has been an
interesting Reld of study for many years. In our ap-
proach a matrix (Figure 2) is constructed where all
variables are listed; for each combination of control-
led output and manipulated input the control quality
is determined. The control quality can be established
on the basic of speed of control, power of control and
the requirement of minimal interaction between con-
trol loops. The speed of control is related to the
period of oscillation of the control loop at the limit of
stability. The shorter this period, the higher the speed
of control. The power of control relates to the range
over which control is effective. For an acceptable
combination, the speed of control should be large, as
Figure 2 Control matrix, showing controlled and manipulated
variables.
should the power of control } in other words, the
controlled output should respond quickly to changes
in the manipulated process input. However, there are
not always Rve variables to be controlled. It could be
that there is no strict requirement for the column
pressure, in which case it is often optimal to minimize
the pressure and open the cooling water valve of the
condensor completely. It could also be that there is
a strict requirement for the top product composition
but no requirement for the bottom composition,
in which case there would be an extra degree of
freedom.
The issue of column operation, column instru-
mentation and selection of the right pairing between
controlled process outputs and manipulated inputs
will be considered in more detail in the following
sections.
Objectives for the Separation Process
The main objective of the distillation process is usu-
ally the recovery of a valuable component from the
feed. In that case there is also often a quality require-
ment for this valuable component. If the purity of the
valuable component is low, then the product has little
value. If the concentration meets the speciRcation,
then the product value is high.
There may be no distinct quality requirement, in
which case the economic value of the product could
be a continuous function of the product properties.
Let us assume that both top and bottom product
represent an economic value. Then an economic ob-
jective for the operation of the process could be de-
Rned as:
J"cDD#cBB!cFF!cSS [1]
where cD is the top product value ($ kg\1), cB the
bottom product value ($ kg\1), cF the cost of the feed
($ kg\1) and cS the cost of the steam ($ kg\1). The use of
cooling water usually involves little cost and Rxed costs
do not play a role in the optimization of the operation.
If F is Rxed and the overall material balance
F"B#D is substituted in eqn [1], then the variable
part Jd of eqn [1] can be written as:
cS
Jd"D! S [2]
cD!cS
Figure 3 shows both terms as a function of the
steam Sow to the column reboiler. When S increases,
the yield of the valuable product approaches
asymptotically a value which is equal to the amount
of top product which is present in the feed:
Dlim"F XF,lk/XD,lk, where the subscript lk refers to
the light key component.
 II / DISTILLATION / Instrumentation and Control Systems
Figure 3 Optimal steam usage in distillation process.
The maximum value of Jd is found when the tan-
gent to the curve for D parallels the straight line for
cSS/(cD!cS), as shown in Figure 3. The optimum
point may be located within the operation area. But
at different values of the feed, however, some con-
straint may make the optimum point unreachable, in
which case the optimum lies on the constraint.
Control of Vapour Inventory
There are a number of ways to control the vapour
inventory or column pressure. Even though one of the
Rve Sows indicated in Figure 2 could be used for
pressure control, only the use of W and S gives sufR-
cient power of control. If S were used for pressure
control, a step increase in S would result (via higher
pressure and top temperature) in an increase in the
vapour Sow in the top. Consequently, the concentra-
tion of the less volatile components in the top would
increase, resulting a higher top temperature and conse-
quently a higher vapour Sow in the top. This positive
feedback can sometimes be so strong that the pressure
Rnally attains a lower value (inverse response). There-
fore, the most common method of pressure control is
through manipulation of the coolant Sow, W.
If a water-cooled condenser is used, the water Sow
rate is manipulated to control pressure: if an air-
cooled condenser is used, the fan speed is generally
changed. The attractiveness of this pressure control
option is that the condensed liquid is at its bubble
point and is not subcooled, as may be the case with
other pressure control options.
If there are incondensable gases in the system, pres-
sure could also be controlled through manipulation
of a bleed valve, through which the gases are bled
from the column. The bleed valve is often installed on
top of the reSux drum.
It should be emphasized that the pressure has
a large degree of self-regulation. If the pressure in
a column increases, the temperature will also increase
1053
and because of this the heat transfer in the condenser
will increase; consequently the pressure will decrease
again to a certain extent.
The general recommendation for pressure control is:
1. if the condenser performs partial condensation,
control the column pressure by manipulating the
vapour Sow leaving the column;
2. if there is no net vapour Sow from the column, the
next preferred option is to control pressure by
manipulating condenser duty, e.g. by changing the
coolant Sow.
Control of Liquid Inventory
Liquid inventory of the distillation process can be
controlled by controlling the liquid levels: the
base level in the column, hB, and the level in the
reSux drum, hR. Often the levels serve the purpose
of smoothing disturbances, hence for control of the
reSux drum level the combination (hR, D) or (hR, R)
is suitable, since W is already used for pressure con-
trol and B and S provide insufRcient power of control.
For bottom-level control, D is unsuitable since the
power of control is nil. Hence both (hB, S) and (hB, B)
are suitable combinations, since they both have
a large power of control.
Under normal circumstances the most logical com-
binations are to use the distillate Sow for controlling
the reSux drum level and the bottom Sow for control-
ling the column base level.
In columns with a small distillate Sow, D, and
a large reSux Sow, R, this scheme does not work so
well. This problem can be partly eliminated by estab-
lishing a ratio controller between distillate and reSux
Sow. When the distillate is then increased, the ratio
controller will increase the reSux Sow accordingly.
However, in many cases with a small distillate Sow,
the reSux is used for reSux drum-level control. A sim-
ilar situation holds for the bottom of the column. If
the bottom draw-off B is very small, level control
using the liquid draw-off might not work well. In that
case one could use the steam Sow for base-level con-
trol. However, increasing the steam Sow to the re-
boiler (and accompanying larger vapour Sow in the
column) might temporarily increase the bottom level
in the column. In the long term, however, increased
heat input will result in increased evaporation and
consequently a lower bottom level (Figure 4). This
response is called inverse response or nonminimum
phase response and is not desirable for control
purposes.
For tray columns this effect can easily be quantiRed
using detailed column models. Assuming that any
increase in vapour Sow will propagate through the
1054 II / DISTILLATION / Instrumentation and Control Systems

therefore initially remain at equilibrium, resulting in

no change of liquid content on the trays. However,
for the top tray, N, where the same Sow of liquid
(reSux) is still entering, a mass balance combined
with eqn [4] gives:

l
V

MN" [5]
1#s
*

Figure 4 Bottom-level response to a positive change in steam Substituting eqn [5] into eqn [3] and writing the
flow. result for the entire column yields for the liquid Sow
change for the bottom tray:

column relatively rapidly, linearizing liquid dynamics
L1"[1!Gd(s)]
V [6]

and using deviation variables, the liquid Sow from
tray 1 can be written as (Figure 5):
with:

1 1

Li"
V#
Mi [3] Gd(s)" [7]
L (1#Ls)N

in which: For the column bottom level it can then be written

that:



L M
" , L" [4]
V Mi L V KB

hB" [1!Gd(s)]!1
V [8]
s
where L is the tray hydraulic time constant. Neigh-
bouring trays will have similar parameter values, where KB is the gain of the bottom level in response to
hence the term 
V is the same everywhere, and vapour Sow changes.
therefore initially all changes in liquid Sows will re- From eqn [8] it can be seen that an inverse response
main the same. The mass balance on each tray will can exist, as long as  is not equal to zero. The
magnitude of the inverse response depends on the
magnitude of . For some columns negative values of
 have been found; for others  is positive. It can
be shown that bottom-level control on the steam
(vapour Sow) is strongly delayed by the effect of
 when '0.5. The physical interpretation of the
so-called -effect is that with large tray loads, an
increase in vapour Sow will lead to stagnation of the
liquid Sow and consequently  will be larger than 1.
With small tray loads an increase in vapour Sow will
push more liquid off the tray, thus a larger liquid Sow
to the lower bottom will result. Figure 6 summarizes
all options for control.

Quality Control
The response of the key components to variations in
liquid and vapour Sow can be approximated by the
algebraic sum of the relative Sow variations followed
by a Rrst-order response with a large time constant:



Li#1
Vi 1 Kx

xi" ! \ [9]
Figure 5 Schematic of a stage of a distillation column. Li#1 Vi 1 1# xs
\
 II / DISTILLATION / Instrumentation and Control Systems
Figure 6 Possibilities for the control scheme.
where L and V are the liquid and vapour Sow respec-
tively (see Figure 5), Kx is the gain for concentration
responses and x is a large time constant for concen-
tration responses, usually in the order of hours. This
time constant is approximately proportional to the
square of the number of trays.
For control of the bottom quality, the reSux R or
distillate Sow D are unsuitable candidates, since there
is a large dead time in the dynamic response between
the Sow and the composition. The steam Sow S and
coolant Sow W are acceptable candidates for bottom
quality control, unless, for the same reason as holds
for bottom-level control, the parameter:


Vi L
K " [10]
Li V Mi
is larger than 0.5. The bottom Sow B does not have
a direct impact on the bottom quality; it could have
some impact via a level controller.
For top quality, R, W and S are all suitable candi-
dates for control. If, however, pressure is controlled
by manipulating coolant W, the only viable option
for bottom composition control is the steam Sow S,
after which the only remaining option for control
of the top quality is the reSux R. Not in all
cases, however, is there a dual composition require-
ment, which leaves more options for composition
control.
1055
Location of Sensors
It is important to provide the column with adequate
sensors: Sow measurements and pressure measure-
ments at various locations, for example above the top
tray, at the feed tray and below the bottom tray in
order to be able to calculate pressure differentials for
the purpose of detecting Sooding. Temperature sen-
sors should also be positioned at various locations
along the column. In many cases temperatures can be
used to infer composition. When the temperature
difference between top and bottom of the column is
small, inferring composition from temperature
measurements is generally not feasible, even though
in some cases temperature differences may still pro-
vide a reasonable composition estimate. An advant-
age of using temperature differences is that it is insen-
sitive to pressure changes; unfortunately, the correla-
tion between temperature difference and product
composition is often highly nonlinear. In multicom-
ponent mixtures, the relation of tray temperature to
key component composition is not unique. Further-
more, the tray temperature to key product composi-
tion may also be nonlinear. Therefore, care must be
taken in using temperatures for controlling composi-
tion. For high purity distillation columns sometimes
the logarithm of the temperature (or composition) is
used to linearize the response of the distillation col-
umn.
When the temperature difference between column
top and bottom is large, several temperatures should
be measured at trays above and below the
feed tray, where under normal circumstances the tem-
perature break is located. These temperatures should
then be averaged and could be used in manipulating,
for example, the steam Sow.
An excellent treatment of sensor and valve issues in
distillation control is given by Luyben (see Further
Reading); the location of temperature sensors receives
an especially comprehensive treatment.
For a distillation column Rrst the base control scheme
should be established, i.e. pressure, reSux drum level
and column bottoms level should be controlled. Then
two variables remain available for control of composi-
tion, say the reSux R and steam Sow S.
If temperature is used for control of composition,
the problem of proper temperature sensor location
becomes prime importance.
If only one composition is controlled, a simple
procedure could be followed. By giving a small
change in the reSux R, the sensitivity
Ti/
R can be
determined for each tray. A similar procedure can be
followed for the steam Sow S. A typical plot for
a toluene-o-xylene column with 30 trays is shown in
Figure 7. It can be seen that a 1% change in steam
1056 II / DISTILLATION / Instrumentation and Control Systems
Figure 7 Gain matrix graph for the toluene-o-xylene column.
Sow causes a temperature change of about 103 at tray
24. A 1% change in the reSux Sow causes a max-
imum temperature change on tray 26 of about !43.
Therefore the sensitivity of the temperature to
changes in steam Sow is larger than the sensitivity for
changes in reSux Sow. Where there is dual composi-
tion control, sensor sensitivity should be balanced
against sensor interaction. One tool that has been
used to accomplish this is singular value decomposi-
tion. The sensitivity matrix X, which contains the
sensitivity of the temperature on each tray for
changes in reSux and steam and thus contains two
columns and 30 rows, is now decomposed into
three individual matrices:
[U, S, V ]"SVD(X)"USVT [11]
Luyben gives a comprehensive treatment of the use
of the individual matrices and their physical meaning
and suggests computing a new function (combination
of the principal components) for each tray:
Zi"U1,i !U2,i  [12]
The maximum of Zi is an indication for the best
location of the Rrst sensor, while the minimum of
Zi gives the best location for placement of the second
sensor. For the toluene-o-xylene column this was cal-
culated and the results are shown in Figure 8. The
Figure 8 Sensor placement based on U-vectors.
best location for the Rrst temperature sensor is tray
25, and the best location for the second sensor is tray
28. Note that the location has changed somewhat
compared to the best locations for the single composi-
tion control problem (trays 24 and 26). This is a re-
sult of the process of reducing the interaction between
the vector components at the other trays.
In many columns temperature has been used to
control the separation process. For binary systems at
constant pressure there is a unique relationship be-
tween temperature and composition. For multicom-
ponent mixtures, often a simple relationship may be
found between temperature and composition. If the
column pressure is not controlled tightly, temper-
ature measurement should compensate for pressure
variations:
Tcompensated"Tmeasured#S(Preference!Pmeasured) [13]
where S is the inverse of the slope of the vapour
pressure}temperature curve at normal operating con-
ditions.
No matter how attractive it is to control temper-
ature rather than composition, the ultimate objective
of the separation process is to control composition(s).
This means that an analyser should be used to indi-
cate the true compositions. There are, however,
a number of disadvantages using analysers in control.
First of all, analysers are highly sophisticated instru-
ments and are therefore expensive and require exten-
sive maintenance. In addition, the sampling system of
analysers is prone to malfunctioning and as analysers
are often used for multiple streams the response can
exhibit a large dead time. This means that simple
feedback control using analysers often results in poor
control performance and dead time compen-
sation techniques often have to be used to improve
performance.
Analysers are well suited for use in a cascade con-
trol set-up, where the temperature (or combination of
temperatures) controls one of the Sows and the ana-
lyser controller resets the temperature controller
setpoint. Figure 9 shows one possibility, using the
reSux as manipulated variable.
Control Con\gurations
From Figure 6 it is clear that controlling the top
composition by the reSux (R) and the bottom com-
position by the vapour Sow V (or steam Sow S) is just
one possible option. This control option is called the
RV conRguration or energy balance structure and it is
probably one of the more frequently used options for
dual composition control in many distillation col-
umns. In the literature other options for controlling
 II / DISTILLATION / Instrumentation and Control Systems
Figure 9 Composition control in a cascade structure.
both top and bottom composition are discussed and
compared, such as the DV conRguration and RB
conRguration, which are called the material balance
structures.
Use of Feedforward and Decouplers
In many cases the top and bottom composition con-
trol loops will show interaction and it is advisable to
Figure 10 Example of dual analyser control using DV configuration with decoupling.
1057
design and implement decouplers. Depending on the
required purity of the products and on the selected
control conRguration, decoupling is sometimes difR-
cult to achieve. However, for many industrial distilla-
tion columns, decoupling considerably improves
the performance of the composition control loops
Figure 10.
The main source of disturbances usually enters the
column with the feed. Whereas the feed composition
may vary somewhat, the feed Sow varies considerably
in many cases. In those situations it might be worth-
while applying feedforward control. If the reSux is
controlling the top composition and the steam
Sow the bottom composition, then feedforward af-
fects both these Sows on a column feed change. The
principle for the reSux controller is shown in
Figure 11, which shows a feedforward controller for
feed Sow changes. In a similar manner, a feedforward
controller for feed composition changes could be
implemented. ReSux affects the tray temperature,
say via a model GR. If the feed affects the tray
temperature via a model GF, then the feedforward
controller has the structure GF/GR. One
word of caution is necessary: care should be taken to
identify properly both models, since a poorly de-
signed feedforward controller may cause poorer con-
trol performance than no feedforward controller
at all.
1058 II / DISTILLATION / Instrumentation and Control Systems
Figure 11 Principle of feedforward control.
Ratio Control
From eqn [9] it follows that keeping L/V constant
will reduce variations in the concentration. Figure 12
shows this ratio control implemented. Usually one
cannot directly measure the vapour Sow V, but a rea-
sonable estimate can be made in many ways,
for example from a static heat balance over the
condensor:
Vtop"Fwater cwater(Twater,in!Twater,out)/Hc [14]
Figure 12 Ratio control between vapour flow and reflux flow.
in which Hc is the heat of condensation and cwater is
the speciRc heat of water.
Although this estimate of the vapour Sow is not
dynamically correct, the quality controller can be
tuned such that the control scheme works well.
Multivariable Control
It is possible to consider the 5;5 control problem as
an integrated problem for which an integrated con-
troller should be designed. One technique which tries
to accomplish this is multivariable predictive control.
In this design all the input}output relationships of the
process are identiRed by means of proper plant test-
ing. Based on the models, a controller is designed,
which adjusts all Rve process inputs simultaneously
utilizing all measured process outputs. In its uncon-
strained version the controller is:
u"(A AT# m )\1AT e [15]
in which u is a vector with the changes in process
inputs, A is a matrix with step weight coefRcients
which represent the model dynamics, is a diagonal
matrix with weights representing the relative import-
ance of the process outputs, m is a diagonal matrix
with penalties for the process input changes and e is
a vector of process output error predictions into the
future.
Model predictive control works well for distillation
columns; it becomes more attractive for larger sys-
tems, such as heat-integrated distillation columns.
 II / DISTILLATION / Laboratory Scale Distillation
One of the major beneRts of model predictive control
is its capability in handling constraints. In that case
the control problem is usually solved using quadratic
programming or some other optimization technique.
Commercial software packages are available for
model identiRcation and controller implementation.
Conclusion
Instrumentation and sensor location has been discus-
sed for a distillation column. In addition, a compre-
hensive treatment is given of the various options for
control. It is shown that some understanding of
column dynamics is necessary in order to select the
proper control schemes.
See also: I/Distillation. II/ Distillation: Historical Devel-
opment; Theory of Distillation.
Further Reading
Buckley PS, Luyben WL and Shunta JP (1985) Design of
Distillation Control Systems. Research Triangle Park:
Instrument Society of America.
Laboratory Scale Distillation
R. C. Gillman, Riverside Organics Inc., Lincoln Park,
MI, USA
Copyright ^ 2000 Academic Press
Distillation on a scale ranging from research quantit-
ies as small as ten milligrams to multikilogram lots is
commonly encountered in the laboratory. Based upon
the physical and chemical properties of the substance
to be isolated, in addition to those of the attendant
impurities and the quantity of impure product to be
distilled, a technique can often be chosen which will
result in a product of adequate purity in one opera-
tion. It may be said that planning a synthesis should
include consideration of the difRculty one may en-
counter in separation of the mixture of products re-
sulting therefrom; on the bench several alternate
routes to the desired product may be available, and
the resultant mixtures will differ in ease of separation.
Since the quantities are not large, more expensive
reagents may be chosen, if desirable, than would be
acceptable in a manufacturing process. A major sav-
ing in time and effort can often be thus effected.
1059
Fisher TG (1990) Batch Control Systems, Design Applica-
tion and Implementation. Triangle Research Park:
Instrument Society of America.
Liptak BG (ed.) (1995) Process Control Instrument Engin-
eers’ Handbook, 3rd edn, sections 8.12}8.14. Oxford:
Butterworth-Heinemann.
Luyben WL (1990) Process Modelling, Simulation and
Control for Chemical Engineers, 2nd edn. McGraw-
Hill.
Luyben WL (1992) Practical Distillation Control. Van
Nostrand Reinhold.
McAvoy TJ and Yang YH (1986) Survey of recent dis-
tillation control results. ISA Transactions, 25(1):
5}21.
Roffel B and Chin PA (1981) Computer Control in the
Process Industries. Ann Arbor, Michigan: Ann Arbor
Science.
Roffel B and Rijnsdorp JE (1987) Introduction to Process
Dynamics, Control and Protection. Ann Arbor, Michi-
gan: Ann Arbor Science.
Shinskey FG (1984) Distillation Control for Productivity
and Energy Conservation, 2nd edn. New York:
McGraw-Hill.
Tolliver TL and Waggoner RC (1980) Distillation column
control: a review and prespective from the CPI.
Instrument Society of America, 35: 83}106.
Resort is commonly had to three broad classes of
distillation, steam distillation, Sash or simple distilla-
tion, and fractional distillation. The Rrst, Rnding use
in separation of substances volatile with steam from
those which are not steam-volatile, often uses the
basic equipment of simple distillation described be-
low, with steam being sparged into the distilland;
either the product is collected as part of the conden-
sate of the residual distilland is enriched in the desired
product by removal of steam-volatile impurities. This
technique, when applicable can be a powerful and
convenient method requiring less skill and attention
than fractional distillation.
Simple, or Sash, distillation is in general used to
separate individual compounds from mixtures con-
sisting of substances whose boiling points differ by at
least 403C, and mixtures of volatile and nonvolatile
components. The method consists of simply boiling
the mixture in a vessel equipped with a device, com-
monly referred to as a ‘head’ which conducts the
vapour to a condenser wherefrom the resultant liquid
is collected. The head is usually equipped with
a means by which the vapour temperature may be
1060 II / DISTILLATION / Laboratory Scale Distillation
observed. No forced reSux is used in this method of
rectiRcation; all the vapour reaching the head is con-
densed as efSux. In order to minimize entrainment of
droplets, some form of dephlegmator can be interpo-
sed between the pot and the head. This technique can
be conducted at pressures ranging from atmospheric
to the lowest pressure available; pressures as low as
0.01 torr are not unusual. Very small quantities of
product can be distilled using a device known as
a ‘kugelrohr (ball-tube) apparatus’; it is constructed
of two or more spherical globes connected linearly by
glass tubing to each other and to a vacuum source.
The globe most remote from the vacuum source is
charged with the distilland and vacuum is applied; the
unit is rotated by hand in a horizontal position while
the globe containing the distilland is heated by Same
or oil bath. The vapour condenses in the globe nearer
the vacuum source.
It must be stated that the development of prepara-
tive gas chromatography, a powerful and convenient
method, has in most cases obviated the necessity of
distillation of quantities less than 25 g. Separations
hardly realizable with the most sophisticated and
cumbersome techniques can be accomplished in min-
utes using this method; the fractions taken are usually
of a high degree of purity as isolated.
The remaining method, fractional distillation, is
normally used to purify quantities of crude product
larger than approximately 100 g, although apparatus
such as the Podbielniak column, constructed of
a metal helix Rtted snugly into a small-diameter glass
column, have been utilized successfully to fractionate
amounts less than a tenth as large. The seperation
efRciency of such a column can be very high but
throughput is very low. As stated, preparative gas
chromatography is currently the method of choice for
small quantities. For rectiRcation of larger quantities
a variety of apparatus is available, and the choice of
equipment rests upon the predicted difRculty of the
separation to be accomplished. As mentioned, on the
bench one may be able to choose a synthetic scheme
which results is an easily puriRed crude product;
when this is not feasible recourse is had to a more
efRcient system capable of separating components
whose boiling points lie close together.
As a rule of thumb, if the components of a mixture
differ in boiling point by at least 303, satisfactory
partition can be achieved by use of Vigreux column,
which is constructed of a glass tube 16}25 mm in
diameter 20 cm to 1 m in length having tiers of inden-
tations spaced 1.5}2 cm apart. The fractionating efR-
ciency of the Vigreux column is quite good consider-
ing the ease and economy of construction of the unit.
Due to its relative openness the throughput of such
a column is quite high, and as a result this type of
column has become the workhorse in many prepara-
tive laboratories; its efRciency may be further en-
chanced by operating it under forced reSux.
Throughput of 0.1}1.5 L h\1 are common when us-
ing this column. An additional advantage to use of
this column is the ease of repair; most repairs can be
made in the laboratory.
In cases where the boiling point differentials be-
tween component is less than 303C, use of a more
efRcient column to effect partition will usually be
found advantageous. These units are of three basic
types; the Rrst in which contact between liquid and
vapour in the column is forced by mechanical con-
struction of the column, the second in which the
structured packing is used to present a large surface
area to promote vapour}liquid interaction in the col-
umn (such as a bubble-cap column), and the third,
a column Rlled with glass or metal objects of various
geometries such as metal saddles, glass beads, and
glass helices. The last type is often referred to as
‘dumped packing’.
Units of the Rrst type, whereby the vapour and
liquid are forced into contact by the construction of
the column are fairly efRcient and can be designed for
a reasonably high throughput. They are, however,
very expensive and quite fragile when constructed of
glass; thus, while they have found major application
in industrial processes, where the construction is only
of metal, in the laboratory they have not found use to
the extent to which packed columns are used. Repairs
to glass columns of this type are difRcult and expen-
sive.
The second type, structured column packings, are
usually made up of a fairly closely woven metal mesh,
which is then crimped and rolled into cylindrical
sections which are pushed into a glass or metal col-
umn; the packed length of such a column typically
will be 0.6}3.0 m for laboratory use, with diameters
of 25}100 mm. The packing itself can be fabricated
from many different metals, for instance stainless
steels, monel, or tantalum depending on the chemical
nature of the product to be puriRed. The advantages
of this type of column are relatively high throughput,
low pressure drop and moderate to high fractionation
efRciency. They can be designed to operate at pres-
sures as low as 0.1 mmHg and even greater than
atmospheric pressure. In practice, however, the high
vapour velocities and low vapour densities encoun-
tered at pressures less than 1 mmHg result in signiR-
cant degradation of column efRciency, especially if
a moderate throughput is required. As to size, a col-
umn of this type whose internal diameter is about
25 mm would be suitable for use with a distillation
pot of 5 L capacity, while for use on a 50 L pot
a column of 75}100 mm internal diameter would
 II / DISTILLATION / Laboratory Scale Distillation
sufRce. In addition, this type of column is quite easily
constructed, easily repaired if broken, and can be
used to predict the performance of an industrial unit
of similar construction. In the laboratory, throughput
of 0.1}2.0 L h\1, depending on pressure, can be ex-
pected.
The last group of columns, the so-called ‘dump
packed’ type, are as a group the most efRcient at
fractionation and can be the most tedious to use.
They can be Rlled with a variety of packings, from
Raschig rings (short sections of glass tubing) to col-
umns packed with single turn glass helices dropped
individually into the column. The latter is one of the
most efRcient fractionating columns ever devised, but
as will be seen, is more suitable for distillation of
smaller batches. In practice, columns of this design
intended for laboratory use are 0.6}2.5 m in length,
with internal diameters of 15}60 mm. Packing fab-
ricated of perforated metal should be purchased after
consultation with the manufacturer; a nominal size of
0.4}1.0 cm is usually chosen for bench use. Of the
various packing components, a column Rlled with
glass helices of 4.5}8.0 mm diameter has proven to
have the greatest efRciency per unit column length but
this advantage is offset to some extent by the low
throughput and high pressure drops attendant with
use of these units. Raschig rings are cut to a length
approximating their external diameter from glass
tubing 6}12 mm in size. Columns Rlled with dumped
packing have one advantage other than efRciency in
that they can easily be emptied and reRlled with some
other type of packing should such be desirable; con-
versely, structured packing, once installed can be dif-
Rcult if not impossible to remove from the column
without out ruining it. The disadvantages of dumped
packings are a tendency to form channels through
which vapour can pass without contacting liquid, low
throughput, and, in most cases, high pressure drop, in
addition to being prone to Sooding. As a result these
columns are seldom used at pressures below
10 mmHg; fractionation efRciency and throughput
suffer markedly at lower pressures. In spite of these
disadvantages columns of this construction have
found use in many laboratories because of their abil-
ity to successfully perform separations not possible
with columns of other design. Throughputs of
0.025}0.5 L h\1 can be expected from these columns.
General Comments
The efRciency of any distillation column is dependent
to some extent upon its being operated under
adiabatic conditions. Thus, distillation columns are
usually insulated, enclosed within a silvered vacuum
jacket, or enclosed within a heated jacket the temper-
1061
ature of which can be measured and controlled to
within about 33C of the vapour in the column. Too
much heat will result in superheating of the vapour,
causing insufRcient condensation and reboiling as the
vapours proceed up the column; too little heat can
result in excessive condensation and Sooding in the
column.
Since maximum efRciency also depends upon op-
eration at or near thermal equilibrium it is important
to be able to change from one fraction to another
without interrupting the distillation. A number of
designs are available from scientiRc laboratory glass
supply houses to accomplish this. In this connection,
it will be found convenient to install some form of
manostatic device in the vacuum line to prevent pres-
sure Suctuations during the course of the distillation.
The source of heat input to the pot must be pro-
vided with various zones such that the heat input to
the Sask may be restricted to the area of the Sask
immersed in the distilland as the distillation proceeds.
Failure to control the heat input in this way will result
in serious superheating of the vapour as the distilla-
tion nears its end; driving superheated vapour into the
column will result in signiRcant loss of efRciency.
Finally, allusion was made to use of forced reSux
during fractionation; it is exceedingly important to
the maintenance of column equilibrium and thus frac-
tionation efRciency. Even with a simple column such
as the Vigreux, use of a partial takeoff head will result
in increased ability of the system to furnish fractions
of relatively high purity. In the case of packed col-
umns control of reSux is absolutely essential to
proper column performance; in its absence vapour-
liquid equilibrium is never established resulting in
loss of up to 90% of the fractionating efRciency of the
column. Typically, reSux ratios (reSux ratio is de-
Rned as the ratio of the amount of condensate re-
turned to the column to the amount of condensate
collected as efSux) as high as 50 : 1 are not uncom-
mon during difRcult separations; ratios approaching
1 : 3 are sometimes used during centre cuts or end
fractions. Various means to effect reSux control are
available from laboratory glass suppliers, from simple
devices utilizing manual control by means of a stop-
cock to elaborate units in which a magnetically con-
trolled value is used to divert the condensate stream
either to the column or to the receiver.
Caution
Since a difRcult fractionation may consume days,
during which the pot is at constant reSux, heed must
be paid to considerations of thermal stability in the
distilland, especially since small quantities of acids or
bases can, and do accelerate thermal degradation of
1062 II / DISTILLATION / Modelling and Simulation
some organic chemicals. In fact, some thermal de-
compositions have been shown to be autocatalytic. It
therefore may be well to consider Sash distillation of
the crude product prior to subjecting it to fractional
distillation to remove trace non-volatiles and/or non-
volatiles which would accelerate decomposition or
lead to excessively high pot temperatures. In some
cases one might consider the addition of a stabilizing
agent to the pot to retard decomposition.
Closing Remarks
With all of the above having been stated, fractional
distillation, particularly at reduced pressure, can be
viewed as an opportunity to see physical chemistry at
work. When selecting a system one hopes will result
in satisfactory partition of components it will be
Modelling and Simulation
J. R. Haas, UOP LLC, Des Plaines, Illinois, USA
Copyright ^ 2000 Academic Press
Introduction
Rigorous computer modelling of all types of frac-
tionation columns has become a necessary part of the
development and design process. There are numerous
software products available to do these calculations.
An understanding of the basic mathematics used in
these programmes is helpful to select, use and
troubleshoot a column model. Explained here are the
basic equations, numerical and solution methods
commonly used.
Stage and Column Models
A rigorous method describes a column as a group of
equations and is the mathematical engine to solve and
satisfy these equations to calculate the operating con-
ditions of the column.
Column design and performance calculations pres-
ent the column at steady state, that is, what enters the
column matches what leaves it (material and energy
balances), i.e.:
(molar feed Sow rates)
" (molar product Sow rates)
(mass feed Sow rates)
" (mass product Sow rates)
helpful to consider properties other than the boiling
point. For example, if a mixture of intermolecularly
bound substances is to be separated by distillation,
their partition is likely to be more difRcult than the
differentials between their boiling points would indi-
cate. On the other hand, a mixture of alkanes may
well be more easily separable than comparison of
their boiling points would otherwise indicate. In any
case practice is necessary, both conducting distilla-
tions and selecting systems for distillation. Once ex-
perience has been gained it is satisfying to be able to
rationalize the results of a fractionation in terms of
physico-chemical principles. One positive note: since
distillation does not result in loss of product, in the
worst case one can recombine all the fractions and
redistill using different conditions and, if necessary,
a different system.
(moles of any component in the feeds)
" (moles of the component in the products)
Feed enthalpy#Heat added
"Product enthalpy#Heat removed
Figure 1 shows a complex column with one feed
and one side product. The top stage of the column is
a partial condenser, with a vapour product, D, and
a liquid product, d. The reSux is the liquid, L0, and
the reSux ratio is L0/(D#d). The bottoms product,
B, leaves stage N#1, the reboiler. The stages are
numbered from the top, with the condenser as stage
0, the top tray in the column, stage 1, the bottom tray,
stage N, and the reboiler, as stage N#1.
An ideal or equilibrium stage is where vapour and
liquid entering and leaving the stage are perfectly
mixed and there are no inhibitions to material trans-
fer between the phases. The material and energy Sows
in and out of a simple stage, with no feeds or side
products, is stage j depicted in Figure 2, and i repres-
ents the component number. Components are num-
bered from 1 to the last, C.
The enthalpy terms, Hj and hj, are molar enthalpies
of the vapour and liquid leaving the stage, respective-
ly. These molar enthalpies are multiplied by the total
Sow rates, Vj and Lj, leaving the stage to give the total
energy leaving the stage in each phase.
The feed stage model (stage f in Figure 2) for an
equilibrium stage assumes that the feed liquid mixes

Figure 1 Overall column model with external variables.
with the liquid entering the feed stage while feed
vapour mixes with vapour leaving the stage (though
special consideration is made for the vapour feed at
the bottom of absorber/stripper columns). The distri-
bution is found by an adiabatic Sash of the feed at the
II / DISTILLATION / Modelling and Simulation 1063
feed stage pressure before the feed enters the column.
Regardless of whether the feed is subcooled liquid or
superheated vapour, or if true mixing occurs, the
assumption of an equilibrium stage is maintained in
most rigorous methods.
1064 II / DISTILLATION / Modelling and Simulation
Figure 2 Model of stage variables.
Similar models are drawn for the bottom and top
stages of any column, plus other equipment such as
product withdrawal stages (stage p of Figure 2),
pump-around returns and draws, and inter-reboilers
and inter-condensers. Since a reSux, reboiler vapour,
feeds, or returns are often subcooled, superheated, or
very different in composition from the material on the
stage, the assumption of an equilibrium stage rapidly
becomes invalid.
Equations of Distillation Modelling
The basic equations below fully describe a distillation
column. These equations deRne the overall column
total material balances, energy balances, and product
compositions. Internal to the column, they describe
equilibrium conditions, internal (stage-to-stage) com-
ponent and total material balances, and internal
energy balances. The independent variables of a col-
umn are the product rates and compositions, internal
vapour and liquid rates and compositions, and stage
temperatures. Equilibrium constants, also called
K values, and mixture enthalpies are dependent vari-
ables. Each stage is assumed to be at equilibrium (a
theoretical stage), though an efRciency can be applied
in the equations.
The equations were Rrst referred to as the MESH
equations by Wang and Henke (1966). The MESH
acronym stands for:
Material or Sow rate balance equations, both com-
ponent and total.
Equilibrium equations including the bubble and
dew point equations.
Summation or Stoichiometric equations or com-
position constraints.
Heat or enthalpy or energy balance equations.
The MESH variables are referred to as state variables.
These are:
E Stage temperatures, Tj
E Internal total vapour and liquid rates, Vj and Lj
E Stage compositions, yji and xji, or instead, compon-
ent vapour and liquid rates, vji and lji
 II / DISTILLATION / Modelling and Simulation 1065

The equilibrium equation is:
yji"Kjixji or vji/Vj"Kjilji/Lj
The equilibrium constant or K-value, Kji, can be
a complex function itself, dependent on the composi-
tions, xji and yji
material withdrawn, wpi, is subtracted from the com-
ponent material balance. By convention, material
leaving a tray has a negative value and material enter-
ing a tray has a positive value.
The total material balances are organized in the
same manner as the component balances. The total
material balance for the simple stage of Figure 2 is:

Kji"Kji(Tj, Pj, xji, yji) Vj#1#Lj 1!Vj!Lj"0

\
The dependence of Kji on xji and yji often appears in The same convention applies to feed and product
the MESH equations. The component rates can also trays where the total Sow rate of a feed, Ff, is positive
be expressed in the terms of each other, giving: and the product, Wp, is negative.
The equilibrium equation and the composition
vji"lji(KjiVj/Lj)"ljiSji constraint are combined to get the bubble point equa-
tion:
and 1 C
Kjilji!1"0
C
l * i"1
i"1 ji
lji"vji(Lj/KjiVj)"vjiAji
and the dew point equation:
KjiVj/Lj is termed the stripping factor, Sji, while
Lj/KjiVj is termed the absorption factor, Aji. 1 C
vji
The summation equation or composition con- C * !1"0
v i"1 Kji
i"1 ji
straints simply states that the sum of the mole frac-
tions on each stage is equal to unity. For the liquid These, or some variation, are important in some
phase: methods to Rnd the stage temperature, especially for
C C
more narrow boiling mixtures.
xji!1"0 or lji/Lj!1"0 or The energy balance equations are required in any
i"1 i"1 rigorous method. In narrow-boiling mixtures, they
C inSuence the internal total Sow rates. In wide-boiling
yji/Kji!1"0 mixtures and in columns where there are great
i"1
heat effects (e.g. oil reRnery fractionators) they also
strongly inSuence stage temperatures. The overall
and for the vapour phase:
energy balance for a column with one feed and side
C C
product is:
yji!1"0 or vji/Vj!1"0 or
i"1 i"1 FHF!DHD!BhB!WHW#QR!QC"0
C
Kji xji!1"0 The enthalpy terms, H and h, are per mole of mixture.
i"1
Note that the enthalpies of the top and side products
For a simple column (single feed, no side products), are written so that a vapour or liquid enthalpy can be
the overall component balance equation is: substituted, depending on the phase of the product.
The energy balance for the simple stage, j, of Figure 2
fi!di!bi"0 is:

The component balance for the simple stage (no feed vj#1Hj#1#Lj h !VjHj!Ljhj"0
\1 j\1
or side product), j, of Figure 2, is:
The enthalpies (energy per mole) for each phase
vji#1#lji 1!vji!lji"0 are functions of temperature, pressure and com-
\ position:
The component balance for feed stage, f, of Figure 2
will add the liquid portion of the feed, lFi, while the Hj"Hj(Tj, Pj, yji)
vapour portion, vFi, is added to the component bal-
ance for stage f!1. For the product stage, p, the hj"hj(Tj, Pj, xji)
1066 II / DISTILLATION / Modelling and Simulation
For feed stages, side product stages, and stages with
inter-condensers or inter-reboilers, additional terms
are included in the energy balance equations. The
energy balance for the reboiler is:
LNhN!VN#1HN#1!BhN#1#QR"0
and for a partial condenser with both vapour and
liquid products:
V1H1!L0h0!dh0!DH0!QC"0
Subcooling is accounted for in h0 (the enthalpy of the
reSux, L0, and the liquid distillate, d).
Most computer simulations work with ideal stages
but to characterize a stage for the deviation from
ideality or equilibrium, stage efRciencies are often
used in some software. Commonly, a Murphree
vapour efRciency is used for each component, given
as:
yji!yji 1
EMVji" \ The rigorous methods can be divided into four basic
yHji !yji 1
\ classes. These are:
where yHji is what the vapour composition would be if
the vapour were in equilibrium with the actual liquid
on the stage and yji and yji 1 are actual vapour com-
\ E
positions. If the absorption factor is used, the vapour
efRciency can be expressed in terms of variables al-
ready presented:
vji!vji#1(Vj/Vj#1)
EMVji"
(KjiVj/Lj)lji!vji#1(Vj/Vj#1)
A vaporization efRciency, Eji, based on the Murphree
efRciency is deRned as:
yji#1
Eji"EMVji#(1!EMVji)
Kjixji
This can be used in the MESH equations to account
for stage nonideality. This vaporization efRciency is
applied to the equilibrium constant, Kji, and appears
as the product EjiKji. The vaporization efRciency does
solve a computational problem in placing an efRcien-
cy in the MESH equations. A major disadvantage of
the vaporization efRciency is that it does vary with
composition. Near the top of a high purity column, as
yji#1 and xji approach unity, Eji also approaches
unity, and so a vaporization efRciency does not truly
reSect stage nonidealities.
Another efRciency method is the bypass method
where some of the vapour Sow of a component enter-
ing the stage is sent to the next stage to account for its
inefRciency in separation. The bypass method cannot
be used on trays that have material leaving or enter-
ing from outside the column such as a feed tray,
product draw tray, pump-around return or draw tray,
or side-stripper return or draw tray. The bypass
method will cause one of these trays to be out of mass
balance. Some of the trays adjacent to these trays are
also affected by these actions. In some columns, this
eliminates a large number of trays and makes results
difRcult to apply.
Caution then should be used in any choice of efR-
ciency. More often, it is usually best to perform the
rigorous calculation using ideal stages and then apply
an overall column efRciency based on sound engineer-
ing judgement and experience to account for stage
nonideality, and calculate the number of actual trays
or packing height.
Rigorous Computational Methods
Classi\cation of the Methods
E The bubble point methods (BP)
E The sum-rates methods (SR)
The 2N Newton methods
E The global Newton or simultaneous correction
(SC) methods.
The BP methods get their name because the stage
temperatures are found by directly solving the bubble
point equation. The BP methods generally work best
for narrow-boiling, ideal or nearly ideal systems;
where composition has a greater effect on temper-
ature than the latent heat of vaporization.
The sum-rates (SR) method is suitable for model-
ling absorbers and strippers with extremely wide-
boiling systems, especially those with non-conden-
sables. In these columns, temperatures are the domi-
nant variables and are found by a solution of the stage
energy balances. Compositions do not have as great
an inSuence in calculating the temperatures as do
heat effects or latent heats of vaporization.
The 2N Newton methods calculate temperatures
and total Sow rates together but compositions are
still calculated in a separate, dependent step. The
name 2N Newton means that there are two equations
per stage for a total of 2;N functions and variables
per column solved simultaneously by a Newton}
Raphson method. The 2N Newton methods have
been shown to work well for wide-boiling mixtures
including reRnery fractionators, absorber-stripper
columns and reboiled absorbers.

The Rrst three classes are referred to as equation
tearing or decoupling methods because the MESH
equations are divided and grouped or partitioned and
paired with MESH variables to be solved in a series of
steps. The SC methods attempt to solve all of the
MESH equations and variables together. Additional
classes are:
E Inside-out methods
E Relaxation methods
E Homotopy}continuation methods
E Nonequilibrium models.
The relaxation, inside-out and homotopy}con-
tinuation methods are extensions of whole or part of
the Rrst four methods in order to expand the range of
columns, and to solve difRcult systems or columns.
The nonequilibrium models are rate-based or trans-
port phenomena-based methods that do away alto-
gether with the ideal stage concept and eliminate any
use of efRciencies. They are best suited for columns
where a theoretical stage is difRcult to deRne and
efRciencies are difRcult to predict or apply by any
means.
Numerical Methods ^ The Newton^Raphson
Technique
The MESH equations form a large system of inter-
related, nonlinear, algebraic equations. The mathe-
matical method used to solve all or part of these
equations as a group is the Newton}Raphson
method. An understanding of the numerical method
is needed to understand the performance of all
column methods. Detailed discussion of the Newton}
Raphson method and its variations can be found in
Holland’s (1981) text.
The Newton}Raphson is an approximation tech-
nique. It assumes in the derivatives that the MESH
equations are linear over short distances and the
slopes will point towards the answers. The MESH
equations can be far from linear and the predictions
can take the next trial well off the curves, and move
away from the solution. In some rigorous methods
based on Newton}Raphson, a poor set of starting
values can cause the calculation never to approach
a solution. Also, the calculation can oscillate, with
values swinging to either side of the solution. The
independent variables calculated in a trial need to
move the column to a solution. The software should
include means to prevent or detect these problems
and improve stability, e.g. by damping or limiting the
change to the next set of variables. A Newton}Raph-
son method will normally take even steps toward the
solution.
II / DISTILLATION / Modelling and Simulation 1067
Global Newton Methods
One group of methods that is very popular is the
global Newton methods, also called the simultaneous
correction (SC) methods. A common one is that of
Naphtali and Sandholm (1971), but there are numer-
ous applications in the literature and global Newton
methods have been extended to include additional
equations and variables for solving three-phase and
reactive distillation columns.
In the global Newton methods, all of the equations
are solved together in a Newton}Raphson technique.
The methods vary in their choice of variables and
MESH equations for the Newton}Raphson calcu-
lation but none of the MESH equations are solved in
any separate step. In the BP, SR and 2N Newton
methods, the component balances and compositions
lag the other MESH calculations (since K values and
enthalpies are generated using the compositions from
the previous trial) and compositions of each compon-
ent are calculated independently of the others MESH
variables. These are major disadvantages with highly
nonideal systems, where K values (especially activity
coefRcients
ji) and enthalpies are highly composition
dependent and where the composition of one com-
ponent cannot be readily decoupled from those of
others. The global Newton method includes the com-
ponent balances among the Newton}Raphson inde-
pendent functions and compositions join other
MESH variables as independent variables.
The global Newton methods are the most sensitive
of the rigorous methods to the quality of the initial
values and often require initial values near the
answer. This, and applying the methods to a full
range of column equipment and speciRcations, is
their greatest problem. Variations on global Newton
methods are used in the inside-out, relaxation,
homotopy and nonequilibrium methods, where their
power and reliability is extended.
Inside-out Methods
The inside-out algorithm has become one of the most
popular methods because of its robustness and its
ability to be applied to the solution of a wide variety
of columns. The inside-out concept was developed by
Boston (1980). Russell (1983) presented an inside-out
method that works well for many reRnery frac-
tionators. The inside-out methods are now the
methods of choice for mainstream column simulation
and have displaced other methods.
In older methods, the MESH variables of temper-
atures, total Sow rates and component Sow rates are
the primary solution variables and are used to gener-
ate the K values and enthalpies from complex correla-
tions. These methods update the MESH variables in
1068 II / DISTILLATION / Modelling and Simulation
an outer loop with the K values and enthalpies
updated whenever the MESH variables change. The
inside-out concept reverses this by using the complex
K value and enthalpy correlations to generate para-
meters for simple K value and enthalpy models. These
parameters are unique for each stage and become
the variables for the outside loop. The inside loop
consists of the MESH equations and is a variation on
other methods. In every step through the outside
loop, the simple models are updated using MESH
variables from the inside loop. This sets up the next
pass through the inside loop. Since the K values and
enthalpies are simple, the inside loop works well for
a wide range of mixtures and is little affected by the
nonideality of mixtures or the quality of the initial
values.
The outer loop K value model is based on a simple
composition-independent K method:
ln Kbj"Aj#Bj(1/Tj!1/TH)
where TH is a reference temperature for the K value
correlation. Outer loop variables, Aj and Bj, are gen-
erated for each stage from a reference KbjRef of a com-
posite component:
C The Boston method includes a middle loop to allow
ln KbjRef" wi ln Kji(actual)
i"1
where the wi are weight factors. The temperatures
and compositions used to get the Kji(actual) are the
latest from the inside loop. Simple relative volatil-
ities are among the outside loop variables, and
are used in the Kb method to calculate the temper-
atures and whenever K values are needed in the inside
loop:
ji"Kji(actual)/KbjRef
These simple relative volatilities change little over
the range of temperatures that is seen on a given stage
and greatly simplify temperature and composition
calculations in the inside loop. For nonideal mixtures,
an activity coefRcient for each component accounts
for composition effects in the inside loop. This activ-
ity coefRcient has a simple model, similar to the
Kb model:
ln
Hji "aji#bjixji
where the new outer loop variables, aji and bji, for
each component are determined from the actual ac-
tivity coefRcient model at the current stage temper-
ature and stage composition.
The simple K values used in the inside loop are
easily determined from:
Kji(simple)"Kbjji
Hji
Simple models for the enthalpy of a phase are also
used to reduce effects such as that caused by compo-
nents moving past their critical conditions. Thus, the
outside loop calculation consists of updating the
terms of the simple K value, activity and enthalpy
models which are updated after each inside loop
solution using the latest temperatures and composi-
tions from the inside loop.
The inside loop consists of the actual calculation of
the MESH variables using the simple K value and
enthalpy models. Boston initially used an inside loop
solution method similar to a bubble point method
and from that it may appear that the Boston method
is most appropriate for narrow-boiling mixtures.
However, the forcing style of the method also allows
it to work well for wide-boiling mixtures. The Boston
method works well for tall, high purity (superfrac-
tionator) type columns, but has been extended to
absorbers, to three-phase distillation, and to reactive
distillation by using other arrangements of the MESH
equations.
for column speciRcations and constraints. The ar-
rangement of equations in the inner loop, where the
solution of the MESH variables occur, may allow for
only a few control or speciRed variables, such as Rxed
reSux ratio and product rates. The middle loop ad-
justs the control variables to meet the speciRcations.
The middle loop can be built as an optimization
method with process speciRcation equations and eco-
nomic objectives and constraints.
Russell’s (1983) method differs from Boston’s in
the inside loop by a solution method of the MESH
equations that includes speciRcations for product
quality, stage temperatures, internal Sow rates, etc.,
without the use of a middle loop to solve these. Here,
for each heat exchanger in the column, plus each
additional side product, an additional speciRcation
and operating variable is added to the problem. Rus-
sell’s method has been found to work well for reRnery
fractionators with side strippers and other similar
columns.
Relaxation Methods
A relaxation method Rnds a steady-state solution of
a column as if it were an operating column changing
with time. The column is initialized using some realis-
tic condition and then makes steps to the steady-state
conditions by successive approximations of the

unsteady-state distillation equations. These unsteady-
state equations are modiRcations to the MESH equa-
tions to include changes in the MESH variables with
respect to time. This mimics the physical start-up
of the column, but the objective is not to follow
the dynamic operation but to seek the steady-state
solution.
Homotopy^Continuation Methods
Homotopy or continuation methods are applied to
difRcult-to-solve columns, and are a simple means of
forcing a solution. The MESH equations can be difR-
cult to solve, due either to the nature of the column
(many feeds or side products, side strippers, near
minimum reSux, etc.) or to the nonidealities of the
K values or enthalpies. For three-phase systems, azeo-
tropic systems or systems of columns with two or
more feed/recycle stream combinations, there may be
more than one calculated solution. The method must
be forced to reach the desired solution. Homotopy
methods begin with a known solution of the column
and from there follow a path to the desired solution.
The known solution can be at different conditions or
with much simpler K value and enthalpy methods and
stepped changes are made from there, solving the
column equations at each step, until the Rnal solution
is reached.
Nonequilibrium or Rate-based Methods
Stage efRciency prediction and scale-up from ideal or
equilibrium stages to the actual design can be difRcult
and unreliable for many columns. For highly
nonideal, polar and reactive systems, such as amine
absorbers and strippers, prediction and use of ef-
Rciencies is particularly difRcult. In such mixtures,
mass transfer and not equilibrium often limits the
separation.
Nonequilibrium methods attempt to get around the
difRculty of predicting efRciencies by replacing the
equilibrium stage concept. Instead, they apply
a transport phenomena approach for predicting mass
transfer rates. Here, the bulk vapour and liquid
phases are not at equilibrium with each other, but
there is equilibrium at the interface between phases
with a movement from the bulk phase through the
interface (Figure 3). The net loss or gain of material
and energy at the interface is expressed as transfer
rates. The mass and energy transfer rates are depen-
dent on the mass and energy transfer coefRcients for
each phase which are in turn dependent on composi-
tion and conditions of each bulk phase and at the
interface.
II / DISTILLATION / Modelling and Simulation 1069
The correlations for the mass and heat transfer
coefRcients and interface also take into account pack-
ing or tray geometries for the actual column. The
total mass and energy rates are calculated from inte-
grating the mass and energy Suxes across the total
interface surface.
Krishnamurthy and Taylor (1986) present and test
a nonequilibrium model which includes rate equa-
tions among the traditional MESH equations. These
include individual mass and energy balances in the
vapour and the liquid and across the interface. An
equilibrium equation exists for the interface only. The
solution methods for these equations are the same as
the global Newton methods.
The total mass transfer rates are added to an ex-
panded set of the MESH equations called the
MERQ equations. The new MERQ acronym stands
for:
Material balances for each component } one for the
bulk vapour, one for the bulk liquid and one across
the interface.
Energy balance equations } one for the bulk va-
pour, one for the bulk liquid and one across the
interface.
Rate equations for mass transfer for all but one
component } one from the interface to the bulk
vapour and one from the bulk liquid to the inter-
face, plus one energy transfer rate equation from
the liquid to the vapour.
eQuilibrium equation at the interface only.
Outlook
New rigorous methods for fractionation modelling
may no longer be forthcoming and most enhance-
ments will be driven by greater acceptance of
nonequilibrium methods, and to other methods by
their application to more complex fractionators and
difRcult systems of components. Teaching concepts of
equations and solution may be limited to what is
necessary to understand a programme’s options, di-
agnostics and why a programme acts in a certain
manner. There should be greater emphasis on know-
ledge of the physical reality of a column and where
the actual process is sensitive, to help set up
a problem. Software improvements are needed more
in analysis and troubleshooting thought processes,
tools and reports. Some of these tools may be a return
to use of pre-computer tools such as x-y,
McCabe}Thiele, and Hengstebeck diagrams and
shortcut methods. While computers continue to be-
come more common, faster and easier to use, they
should never be a substitute for sound engineering
experience and judgement.
1070 II / DISTILLATION / Modelling and Simulation
Figure 3 Model of a nonequilibrium separation and mass transfer.
See also: II/Distillation: Historical Development; Theory
of Distillation; Vapour-Liquid Equilibrium: Correlation and
Prediction; Vapour-Liquid Equilibrium: Theory.
Further Reading
Boston JF (1980) Inside-out algorithms for multicompo-
nent separation process calculations. American Chem-
ical Society Symposium Series No. 124: 135.
Brierley RJP and Smith RI (1979) DISTPACK } Using a
combination of algorithms to solve difRcult distillation
and absorption problems. Chemical Engineering Sym-
posium Series No. 56: 89.
Friday JR and Smith BD (1964) An analysis of the equilib-
rium stage separations problem}formulation and con-
vergence. American Institute of Chemical Engineers
Journal 10: 689.
Holland CD (1981) Fundamentals of Multicomponent Dis-
tillation. New York: McGraw-Hill.
Ketchum RG (1979) A combined relaxation}Newton
method as a new global approach to the computation of
thermal separation processes. Chemical Engineering
Science 34: 387.
Kister HZ (1992) Distillation Design. New York:
McGraw-Hill.
Kister HZ (1995) Troubleshooting distillation simulation.
Chemical Engineering Progress 16(6): 63.
Krishnamurthy R and Taylor R (1986) Multicomponent
mass transfer theory and applications. In Cheremisinoff
NP (ed.) Handbook of Heat and Mass Transfer. Gulf
Publishing Company.
Lockett MJ (1986) Distillation Tray Fundamentals. Cam-
bridge, UK: Cambridge University Press.
Naphtali L and Sandholm DS (1971) Multicomponent sep-
arations calculations by linearization. American Insti-
tute of Chemical Engineers Journal 17: 148.
Russell RA (1983) A Sexible and reliable method solves
single-tower and crude-distillation-column problems.
Chemical Engineering 90(20): 53.
Taylor R, Wayburn TL and Vickery DJ (1987) The Devel-
opment of Homotopy methods for the solution of se-
paration process problems. International Chemical
Engineering Symposium Series No. 104: B305.
Wang JC and Henke GE (1996) Tridiagonal matrix
for distillation. Hydrocarbon Processing 45(8):
155.
 II / DISTILLATION / Multicomponent Distillation 1071

Multicomponent Distillation
V. Rico-RamOH rez and U. Diwekar,
Carnegie Mellon University, Pittsburgh, PA, USA
Copyright ^ 2000 Academic Press
Introduction
Distillation is the oldest separation process and the
most widely used unit operation in industry. It in-
volves the separation of a mixture based on the differ-
ence in the boiling point (or volatility) of its compo-
nents. The reason for the wide acceptance of distilla-
tion is that, from both kinetic and thermodynamic
points of view, distillation offers advantages over
other existing processes for the separation of Suid
mixtures:
1. Distillation has the potential for high mass trans-
fer rates because, in general, in distillation there
are no inert materials or solids present.
2. The thermodynamic efRciency for distillation is
higher than the efRciency of most other available
processes in the chemical industry.
Designing a distillation column involves: (1) selecting
the type of column, mostly based on heuristics; (2)
obtaining the vapour}liquid equilibrium data using
thermodynamics; and (3) Rnding the design variables
such as number of equilibrium stages and operating
conditions required to obtain the desired separation
based on mass and energy balances.
When the mixture to be separated contains two
components, the design of a column can be accomp-
lished by using graphical methods. However, for
multicomponent systems the design methods are
more difRcult and are the focus of this article.
still and xF (mole fraction) be the composition of
component A of the mixture. Let B be the number of
moles of material remaining in the still, xB the mole
fraction of component A in the still, xD the mole
fraction of component A in the vapour dB produced
during an inRnitesimal time interval dt. The differen-
tial material balance for component A can be written
as:

 
xB
B dxB
ln " [1]
F xF D!xB
x
Complex mass and heat transfer processes occur in
distillation processes and it is generally assumed that
the vapour formed is in thermodynamic equilibrium
with the liquid. Hence, the vapour composition (xD)
is related to the liquid composition (xB) by an equilib-
rium relation of the functional form xD"f (xB).
Note that, because of the unsteady nature of simple
distillation, the equilibrium relationship between
xD and xB holds only for each inRnitesimal time
interval dt.
The exact equilibrium relationship for a particular
mixture may be obtained from a thermodynamic
analysis and is also dependent upon temperature and
pressure.
Thermodynamics and Equilibrium Data
Accurate and reliable thermodynamic data for
vapour}liquid equilibrium is essential to distillation

Fundamentals
Simple Distillation
Distillation began as a simple still. In such an opera-
tion, a liquid mixture is heated (see Figure 1). As
a result, a vapour stream richer in the more volatile
components comes off, while the liquid, richer in the
less volatile components, remains in the still. The
vapour stream is condensed and collected in the con-
denser.
The analysis of simple distillation for a binary
mixture presented in 1902 by Lord Rayleigh marks
the earliest theoretical work on distillation. Consider
Figure 1. Let F (moles) be the initial feed to the Figure 1 Simple distillation } a still.
1072 II / DISTILLATION / Multicomponent Distillation
design. For binary mixtures, these data are generally
presented in the form of tables containing the liquid
and vapour equilibrium compositions over a range of
temperatures for a Rxed pressure. The same informa-
tion can also be plotted in what is called an x}y
diagram. For multicomponent mixtures, however,
vapour liquid equilibrium data are difRcult to repres-
ent in graphical or tabular form. In such case,
K values are used instead.
K value and relative volatility The K value of a
component i is a measure of the tendency of such
component to vaporize. A K value is deRned by:
yi
Ki"
xi
where yi is the equilibrium composition of the vapour
phase for a composition xi of the liquid phase.
K values are a function of temperature, pressure and
composition, and they are widely reported for binary
and multicomponent mixtures. An associated concept
is the relative volatility,
i,j, which is a measure of the
ease of separation of components i and j by distilla-
tion:
Ki

i,j"
Kj
Ideal and nonideal systems An ideal system is one in
which the liquid phase obeys Raoult’s Law and the
vapour phase obeys the ideal gas law. For such sys-
tems, the K value is given by:
yi p0i
Ki" "
xi P
where p0i is the vapour pressure of pure component
i and P is the pressure of the system. Note that p0i is
a function of temperature.
For a nonideal system, the K values can also depend
upon the composition of the mixture and are ex-
pressed in terms of fugacity coefRcients, where Vi is
the vapour phase fugacity coefRcient and Li is the
liquid phase activity coefRcient, as given below:
Li p0i
Ki" V )
i P
Azeotropic systems represent examples of nonideal
mixtures for which eqn [5] has to be used.
Classi\cation of Distillation Processes
There are many criteria under which one can classify
distillation: type of accessories, operating mode,
design calculation assumptions, etc. Distillation
can either be binary or multicomponent. According
to the type of accessories used to increase the mass
transfer in the separation process, a distillation col-
umn can be packed (use of packing) or staged (use of
plates). It can be batch or continuous. Also, according
to the assumptions made and accuracy expected in
a distillation design calculation, a calculation
technique can either be a shortcut method or a
rigorous method.
Packed columns and staged columns Although
simple distillation in a still historically represents the
start of the distillation process, a complete separation
of the components of the mixture using this process is
[2]
not possible. Therefore, the application of these stills
is restricted to laboratory-scale distillation, where
high purities are not required or when the mixture is
easily separable.
One can look at simple distillation as consisting of
one equilibrium stage where a liquid and a vapour are
in contact with one another and mass and heat trans-
fers take place between the two phases. If N such
stages are stacked one above the other, and are al-
lowed to have successive vaporization and condensa-
tion, that results in a substantially richer vapour and
[3] weaker liquid (in terms of the more volatile compon-
ent) in the condenser and the reboiler, respectively.
This multistage arrangement is representative of
a distillation column, where the vapour from the
reboiler rises to the top and the liquid from the
condenser is reSuxed downwards (see Figure 2). The
contact between the liquid and the vapour phase is
established through accessories such as packing or
[4] plates. When the accessory is a stack of plates, then
the result is a column of trays. Similarly, if the acces-
sory is packing, the result is a packed column.
Continuous distillation and batch distillation The
basic difference between a batch column and a
continuous column is that in continuous distillation
the feed is continuously entering the column, while in
batch distillation the reboiler is normally fed at the
beginning of the operation. Also, while the top prod-
ucts are removed continuously in both batch and
continuous operations, there is no bottom product in
a conventional batch distillation. Since in a continu-
[5]
ous operation the total product Sow equals that of
incoming feed or feeds, the process reaches a steady
state. In batch distillation, on the other hand, the
reboiler becomes depleted over time, so the process
is unsteady. Such differences are illustrated in
Figure 3.
Batch distillation is a direct extension of the simple
distillation still, where the Rayleigh equation
 II / DISTILLATION / Multicomponent Distillation
Figure 2 Equilibrium processes. (A) Single stage; (B) multi-
stage.
(eqn [1]) is applicable. However, in both batch and
continuous distillation, multistage mass transfer and
thermodynamic equilibrium stage calculations are
used for obtaining the steady-state relationship be-
Figure 3 Batch distillation versus continuous distillation.
1073
tween the product composition (instantaneous in case
of batch) and feed composition.
Multicomponent Multistage
Equilibrium Calculations
This section is divided in two parts. In the Rrst we
discuss approximate methods (or shortcut methods);
the second part corresponds to rigorous methods. The
approaches are different depending upon the opera-
tion mode of the column, that is, a continuous opera-
tion or a batch operation.
In this section, our attention is focused on the
approaches to the design of continuous columns. The
reader can refer to the book by Diwekar (1995) for
batch distillation calculations.
Shortcut Methods
Approximate methods constitute a useful for the syn-
thesis, analysis and design of distillation separations.
The main advantage of shortcut methods is that they
can provide the feasible region of operation. They
also provide large saving in computer time, and some-
times, they are sufRciently accurate that more expen-
sive rigorous methods are not justiRed.
Concept of Nmin and Rmin Minimum number of
plates, Nmin, and minimum reSux, Rmin, are very im-
portant concepts in the design of distillation pro-
cesses, as they are considered to be the limiting condi-
tions in the operation of a distillation column.
1074 II / DISTILLATION / Multicomponent Distillation
Nmin corresponds to the number of trays required
for separation in a situation in which the external
reSux ratio R (ratio of the liquid reSuxed to the
distillate rate) of the column is inRnite. This corres-
ponds to total reSux operation.
Rmin corresponds to the minimum value of the ex-
ternal reSux ratio required to achieve the speciRed
separation in a situation in which the number of trays
of the column is inRnite.
Fenske}Underwood}Gilliland method The most
popular of these shortcut methods is the Fenske-
Underwood-Gilliland method (FUG). The basic as-
sumptions of such a method are:
1. The system is ideal.
2. Constant molar overSow (as in the McCabe Thiele
method for binary mixtures).
3. The separation is essentially taking place between
the light key component and the heavy key com-
ponent. The light key (lk) is the lightest compon-
ent appearing in the bottom and the heavy key
(hk) is the heaviest component appearing in the
top.
In the FUG method:
1. Fenske’s equation is used to calculate the min-
imum number of trays, Nmin.
2. Underwood’s equation is used to estimate the min-
imum reSux, Rmin.
3. Gilliland’s correlation is used to calculate the ac-
tual number of trays, N (for any R given), or the
reSux ratio, R, (for any N given) in terms of
previous limiting values Nmin and Rmin.
The Fenske equation is:



xDlk xBhk
log ) methods are: (1) Thiele}Geddes; (2) tridiagonal
xBlk xDhk
Nmin" [6]
log(
lk,hk)
where
lk,hk is the relative volatility between the light
key component and the heavy key component. Since
it can be expected that the value of
changes for each
tray of the column, the geometric average of this
value is generally used:

N"
N )
N 12
1 [7]
\ each stage n in a distillation column (and for each
The Underwood equation can be written as:

i ) xi,D
"Rmin#1 [8]
i
i!
where  is a root of the equation:

i ) xi,F
"1!q [9]
i
i!
such that
hk44
lk.
hk and
lk are the relative
volatilities of the key components (light and heavy) in
the calculation. As stated earlier, such components
are the ones that the designer uses as the basis for the
separation.
Finally, the Gilliland correlation is given by:



N!Nmin 1#54.4G G!1
"1!exp )
N#1 11#117.2G G0.5
[10]
where
R!Rmin
G" [11]
R#1
The main assumptions of the Underwood equation
are the assumption of constant molar Sow rates and
an ideal system. Such assumptions constitute the
main limitation of the algorithm.
Rigorous Methods
Recent developments in computer hardware and soft-
ware have made it possible to use rigorous methods
for the design of distillation processes. In these
methods, the assumption of constant molar Sow rates
is no longer considered. The implication of removing
such an assumption is that rigorous methods not only
consider mass balances, but also enthalpy balances
for each of the trays of the column. Thus, rigorous
methods require simultaneous convergence of mass
and energy equations. Depending on the calculation
sequence, there are several rigorous methods reported
in the literature. The most important of these
methods; (3) Naphtali}Sandholm; (4) inside-out algo-
rithms; (5)  convergence methods; and (6) 2N New-
ton methods. The method of Naphtali}Sandholm and
the inside-out algorithm, which are commonly used
nowadays, are discussed in this work to give an idea
of the scope and applications of rigorous methods.
MESH equations Most rigorous methods involve
the solution of the so-called MESH equations. For
component i in a mixture of C components), the
equations representing mass balance (M), equilib-
rium relationships (E), summation of compositions
(S) and energy balance (H), constitute the MESH
equations. In addition, both K values and enthalpies
 II / DISTILLATION / Multicomponent Distillation 1075

are generally given as functions of temperatures, pres- where N is the number of stages and
sures and compositions. The generalized form of the
MESH equations for the equilibrium stage shown in XM n"[vn,1, vn,2,2, vn,C, Tn, ln,1, ln,2,2, ln,C]T [13]
Figure 4 and the expressions for K values and enthal-
pies are present in Table 1.
Equations:
Naphtali}Sandholm method In the Naphtali}
Sandholm method, the number of variables of the FM "[FM 1, FM 2,2, FM n,2, FM N] [14]
MESH equations is reduced by the introduction of
component Sow rates and side streams. Furthermore, where
the summation of compositions are eliminated. Those
modiRcations result in the equations presented in FM n"[HK n, Mn,1, Mn,2,2, Mn,C, En,1, En,2,2, En,C]T
Table 2.
To solve the system of MESH equations given in [15]
Table 2, the vectors of variables and equations are
ordered as follows. Variables: The solution process is iterative, using one of the
several variations of the Newton method. Thus, cor-
XM "[XM 1, XM 2,2, XM n,2, XM N] [12] rections at each iteration k are obtained from

Figure 4 Equilibrium stage. Derivation of MESH equations.

1076 II / DISTILLATION / Multicomponent Distillation

Table 1 MESH equations

Relationship Equation

Mass balance Ln#1 ) xn#1,i#Vn 1 ) yn 1,i#Fn ) zn,i!(Ln#Un) ) xn,i!(Vn#Wn) ) yn,i"0

\ \
Equilibrium yn,i"Kn,i ) xn,i
Summation of compositions yn,i!1"0
i

H energy balance Ln#1 ) hn#1#Vn 1 ) Hn 1#Fn ) HFn!(Ln#Un) ) hn!(Vn#Wn) ) Hn!Qn"0

\ \

K values and enthalpies Kn,i"Kn,i (Tn, Pn, xnyn)

Hn,i"Hn,i (Tn , Pn , yn)
hn,i"hn,i (Tn , Pn , xn )

(classical Newton}Raphson equations): are used to generate the K values and enthalpies from
complex correlations. Hence, such a method updates


 
FM \1 (k) the primary variables in an outer loop, with the
FM (k)"! ) FM (k) [16] K values and enthalpies updated in an inner loop
XM
whenever the primary variables change.
In inside-out algorithms, the previous situation is
XM (k#1)"XM (k)#t ) XM (k) [17]
reversed. These methods use complex K values and
enthalpy correlations to generate parameters for
where t is such that 04t41. t is the factor that simple K values and enthalpy models. Hence, these
ensures progress toward the solution of the system at parameters become the variables for the outside loop.
equations of each iteration. The inside loop then consists of the MESH equations.
In every step through the outside loop, the simple
Inside-out algorithm In the Naphtali}Sandholm K values and enthalpy models are updated by using
method, the temperatures and component Sowrates the MESH variables from the inside loop. This sets up
are the primary solution variables (see eqn [13]) and the next pass through the inside loop. The book by
Kister (1992) provides detailed guidelines for the use
of the various inside-out methods.
Table 2 MEH equations for method of Naphtali and Sandholm

Relationship Equation Special Separations

Component flow rates vn,i"yn,i ) Vn
When the components of a mixture have low relative
and side streams volatilities, or when the mixture contains a large
ln,i"xn,i ) Ln number of components, separation by distillation be-
fn,i"zn,i ) Fn comes difRcult and expensive because a large number
of trays or a large number of columns are required for
sn"Un /Ln
the separation. Furthermore, some systems may show
Sn"Wn/Vn nonideal behaviour such as the formation of azeo-
M Mn,i"ln,i ) (1#sn)#vn,i ) (1#Sn) tropes or a reversal of the relative volatility with the
!ln#1,i!vn 1,i!fn,i
change in pressure from top to bottom in a column.
\ Complex systems which have these characteristics are
E
H
En,i"Kn,i ) ln,i )

k
vn,k/
k 
ln,k !vn,i"0 common in the pharmaceutical and synthetic chem-
ical industry.
HK n"hn ) (1#sj) ) ln,i This section presents a brief review of separations
i
in which the traditional distillation process is altered,
#Hn ) (1#Sn) ) vn,i but the general principles of multicomponent distilla-
i
!hn#1 ) ln#1,i!Hn 1 ) vn 1,i tion still apply. Three broad categories of such special
\ \
i i
separations exist: azeotropic distillation, extractive
!HFn ) fn,i!Qn"0 distillation and reactive distillation. Petroleum distil-
i
lation will also be discussed since it represents a case
 II / DISTILLATION / Multicomponent Distillation
in which the complexity of the mixture (petroleum)
requires special considerations for the separation.
Azeotropic Distillation
Highly nonideal systems, with components having
close boiling points among them, often produce azeo-
tropes. Azeotropes can be identiRed by using an x}y
diagram. When an azeotrope is present, the equilib-
rium curve crosses the line x"y (453 line), as shown
in Figure 5.
Azeotropes limit the separation that can be
achieved by conventional distillation. Sometimes it is
possible to shift the equilibrium by changing the pres-
sure of the system sufRciently to move the azeotrope
away from the region where the separation must be
made. Other cases, however, require the addition of
a new material in order to achieve separation.
In azeotropic distillation, the equilibrium behav-
iour of the mixture is modiRed by adding a new
material (called the solvent or entrainer). The added
entrainer forms a minimum boiling point azeotrope
with one or more components and distils overhead.
The distillate is generally heterogeneous, that is, it is
composed of two immiscible liquids when condensed.
Such a heterogeneous nature facilitates the separation
of the product from the entrainer.
Extractive Distillation
Extractive distillation also involves the addition of
the third component to the mixture (solvent or en-
trainer). However, in the case of extractive distilla-
tion, the solvent is a relatively high boiling point
material, which is present at high concentration on
each stage and exits at the bottom. To improve the
efRciency of the process, the entrainer has to be added
Figure 5 Azeotropic behaviour.
1077
at the top of the column, so that its concentration on
each stage will be enough to produce the desired
effect in the equilibrium of the original mixture. Fi-
nally, the entrainer is separated from the bottoms
product in another distillation column.
Reactive Distillation
The idea of combining reaction and separation in
a single apparatus has been extensively investigated.
Doherty and Buzad (1992) present a survey of the
available design techniques for reactive distillation.
Reactive distillation is particularly attractive when-
ever a chemical reaction provides the favourable ef-
fect of reacting away azeotropic mixtures so that the
behaviour of the liquid phase is simpliRed. In addi-
tion, it has been shown that reactive distillation
has the potential of eliminating recycle costs when
a liquid reaction involves a large excess of one
reactant.
In general, the current trend in reactive distillation
design is using experimental results from bench-scale
problems in the initial stages of the design, and then
using computer-aided simulation tools for scale-up
and operability issues.
Possible proRtable applications of reactive distilla-
tion processes are numerous. However, an incom-
plete understanding of the interactions of the many
nonlinear phenomena such as chemical reaction,
phase equilibrium, mass transfer and countercurrent
Sow has prevented the widespread use of such pro-
cesses. Considerable research effort in the area is
currently being conducted.
Petroleum Distillation
Petroleum distillation is particularly difRcult because
of the large number of components of the mixture
and large scale of the processes. This type of distilla-
tion involves products that are not easily identiRable
components. Instead, separation is achieved in terms
of pseudo-components, which are generally charac-
terized in terms of their true boiling point ranges
(TBP), an average relative molecular mass and an API
gravity. TBP data are widely available and are gener-
ally presented in form of curves.
There are two main approaches to the design of
petroleum distillation columns. The Rrst consists of
the solution of mass and energy balances based on
empirical correlations, and is basically a calculation
by hand. This approach was developed by Packie.
In the second approach, each pseudo-component is
characterized for properties (such as vapour pressure
and enthalpy) by using homologous-series ap-
proaches. Thus, rigorous mass and energy balances
can then be applied to determine the separation in
1078 II / DISTILLATION / Multicomponent Distillation
terms of the reSux ratio. Several efRcient computer
programs following this approach have been de-
veloped.
Packed Columns
Several approaches exist for the design of packed
columns. These are based on the concepts of number
of transfer units (NTU), height of transfer units
(HTU) and height equivalent to a theoretical plate
(HETP). The last of these concepts is the most widely
used.
Since methods for the design of staged distillation
columns are well developed, a common approach is
to calculate the number of trays N using such ap-
proaches and then to Rnd the height of the packed
column, h, by the relation:
h"N ) HETP [18]
There exist various correlations for predicting the
value of the HETP. One of most commonly used is
the Sherwood correlation. It can be expected that
HETP will change with respect to the operating con-
ditions, physical properties of the liquid, etc., so, it is
calculated in terms of correlations containing many
factors.
Nonequilibrium Distillation
All the mathematical methods (binary, rigorous,
shortcut) presented earlier assume that each stage in
the column is an equilibrium stage. In reality, how-
ever, this assumption is rarely satisRed.
Stage Ef\ciency
An approach to nonequilibrium calculations is the
use of the concept of stage efRciency. The most com-
mon approach is to modify the rigorous methods with
the introduction of the so-called Murphree
efRciency in the calculations. The Murphree efRcien-
cy in a stage calculation can be deRned as:
xout,i!xin,i
ELMi" [19]
xi !xin,i
for the liquid and
yout,i!xin,i
EVMi" [20]
yi !yin,i
for the vapour. xi are the compositions of the liquid
that would be in equilibrium with the outlet composi-
tion of the vapour. yi are the compositions of the
liquid that would be in equilibrium with the outlet
composition of the liquid.
Mass Transfer Rates
It has been shown that stage efRciency prediction and
scale-up are difRcult and unreliable. For highly
nonideal, polar and reactive systems, a transport phe-
nomena approach for predicting mass transfer rates is
preferred. Such mass transfer rates are calculated
continuously along the column similarly to the HETP
calculation for packed columns.
Nonequilibrium models for the calculation of mass
transfer rates assume that, while the bulk vapour and
liquid phase are not in equilibrium with each other,
there is an equilibrium at the interface. Hence,
the net loss or gain for a component at the interface
is expressed in a rate form. For instance, the net
gain by the vapour because of the transfer at the
interface is:
NVij "Nvij ) daj
0
[21]
where NVij is the vapour Sux of the component at some
point through the interface and daj is the interface
area through which the Sux passes. The mass transfer
rates for liquid and vapour, NVij and NLij , are dependent
on the mass transfer coefRcients for each phase. There
exist several correlations for the heat and mass trans-
fer coefRcients and these are dependent on the com-
positions in the bulk phase, the temperatures in the
bulk phase and interface, and on the packing or tray
geometries.
Industrial Applications
Distillation is by far the most widely used separation
technique in the petroleum, natural gas and chemical
industries so, applications of multicomponent distil-
lation are numerous. A couple of industrial applica-
tions are described in this section.
Primary Distillation of Crude Oil
A typical conRguration for the distillation of a crude
oil unit includes two main columns, an atmospheric
tower and a vacuum tower (see Figure 6). In the
atmospheric tower, crude oil is rectiRed (at a pressure
no greater than 275.8 kPa (40 psi); to yield a distillate
product containing light hydrocarbon gas, light and
heavy naphtha, kerosene, diesel oil, and a bottom
product of heavier components (TBP greater than
4203C). Each of the side streams of the atmospheric
tower are sent to side strippers that have a partial
reboiler or steam stripper. The side stream strippers
serve to remove the light components. Stripping by
 II / DISTILLATION / Multicomponent Distillation 1079

Figure 6 Crude oil distillation unit.

steam is also frequently used in the bottom of the
tower.
The bottom product of the atmospheric tower is
further separated by rectiRcation in the vacuum
tower. The feed-tray pressure of a vacuum tower is
usually 6 kPa (45 Torr). Vacuum towers are mainly
designed to obtain heavy distillates such as gas oil,
lubricating oils and bunker fuels with asphalt as the
bottom product.
The pump-around systems shown in both of the
towers serve to make much larger liquid Sows on the
intermediate stages and produce a net increase in
liquid Sow. This serves as a point of control to keep
the plates from running dry.
1080 II / DISTILLATION / Multicomponent Distillation
Highly developed procedures for the preliminary
design of fractionators that process petroleum are
commercially available through computer programs.
The program ‘REFINE’ of the ChemShare Corpora-
tion and the ‘PROCESS’ (now PRO-II) program of
Simulation Sciences Inc. are two examples.
Ethylene and Propylene Production
The manufacture of ethylene and propylene is one of
the most important operations of the petrochemical
industry. In that process, ethylene and propylene are
formed from the thermal cracking of other hydrocar-
bons, such as ethane, propane and naphtha. The
mixture resulting from the thermal cracking is very
complex. Hence, the mixture has to be separated into
relatively pure ethylene and propylene, ethane and
propane to be used as a recycle, methane and hydro-
gen to be used as fuel, and heavier products to be used
for gasoline. A typical reRnery gas feed to the separ-
ation system of this process contains hydrogen,
ethylene, methane, ethane, propane, propylene
and lower compositions of other heavy hydro-
carbons. The distillation sequence most commonly
used for the separation of the mixture is shown in
Figure 7.
In a high pressure plant (no refrigeration is needed
for condensation of products), the distillation se-
quence consists of Rve distillation columns:
Figure 7 Separation of products of the manufacture of ethylene and propylene.
1. Demethanizer
2. Deethanizer
3. Ethylene/ethane separator
4. Depropanizer
5. Propylene/propane separator.
Both the propylene/propane and the ethylene/
ethane separator require high towers with large dia-
meters because such mixtures contain components
with very close relative volatilities. A plant that uses
the conRguration described here was built by Pullman
Kellogg Inc., Houston, Texas.
In the case of a lower pressure plant, the deethan-
izer precedes the demethanizer because refrigeration
is required for the feed of the demethanizer. So, by
placing the deethanizer Rrst, important utility savings
are obtained.
Future Work
Enormous progress has been made on the application
and design of distillation technology. However, chal-
lenges still exist in some areas, which lead to the
following ongoing research:
1. Improvement of mass transfer coefRcients in
packed distillation columns. Great effort is being
made on the design of efRcient packings and accu-
rate correlation of their performance.
 II / DISTILLATION / Packed Columns: Design and Performance
2. The simulation, synthesis and design of reactive
and azeotropic distillation. Such topics still consti-
tute a gap in the knowledge of distillation tech-
nology.
3. Investigation of complex conRgurations for batch
distillation processes.
4. Use of optimization methods for obtaining opti-
mal conRguration and design of batch and con-
tinuous distillation processes.
5. Online optimization and control of columns.
See also: II/Distillation: Batch Distillation; Theory of Dis-
tillation; Vapour-Liquid Equilibrium: Correlation and Pre-
diction; Vapour-Liquid Equilibrium: Theory.
Further Reading
Diwekar UM (1995) Batch Distillation: Simulation, Opti-
mal Design and Control. Series in Chemical and Mech-
anical Engineering. Washington, DC: Taylor & Francis.
Packed Columns: Design and Performance
L. Klemas, Bogota, Colombia
J. A. Bonilla, Ellicott City, MD, USA
Copyright ^ 2000 Academic Press
Use of Packing in Distillation
Use of packing in mass transfer has its origins in the
early 1800s for simple applications such as alcohol
distillation, and in sulfuric acid plant absorbers. Glass
balls, coke or even stones were used as packing ma-
terials. Nevertheless packings for distillation were not
established until the 1930s with the use of regular
shape materials such as ceramic Raschig rings and
Berl saddles, as well as the availability of distillation
calculations such as the McCabe}Thiele and Pon-
chon}Savarit methods. Early in the second half of the
century, the use of packing for distillation went
through a transformation, producing the second-
generation packings (see Table 1). Regular and im-
proved shape of packings, such as pall rings, became
available with larger open areas that permitted a sub-
stantial increase both in capacity and column efRcien-
cy. In the 1960s Sulzer introduced the wire-mesh
packings with very high efRciency (low height equiva-
lent to a theoretical plate, HETP), resulting in a new
transformation in the use of packings. In the 1970s
1081
Doherty MF and Buzad G (1992) Reactive distillation by
design. Transactions of the Institution of Chemical En-
gineers 70: part A.
Gmehling J and Onken U (1977) Vapor}Liquid Equilib-
rium Data Collections, DECHEMA Chemistry Data
series, vol. 1. Frankfurt:
Henley EJ and Seader JD (1981) Equilibrium-Stage Separ-
ation Operations in Chemical Engineering. New York:
Wiley.
Holland CD (1981) Fundamentals of Multicomponent Dis-
tillation. New York: McGraw-Hill.
King CJ (1980) Separation Processes, 2nd edn. New York:
McGraw-Hill.
Kister HZ (1992) Distillation Design. New York:
McGraw-Hill.
Perry RH, Green DW and Maloney JO (1984) Perry’s
Chemical Engineers’ Handbook, 6th edn. New York:
McGraw-Hill.
Schweitzer PA (1979) Handbook of Separation Techniques
for Chemical Engineers. New York: McGraw-Hill, The
Kingsport Press.
Treybal RE (1980) Mass Transfer Operations, 3rd edn.
New York: McGraw-Hill.
and 1980s all major mass-transfer equipment manu-
facturers developed structured packings. Compared
to the traditional tray columns spectacular improve-
ments in plant capacity were achieved, but also some
projects were pitfalls, when the expected beneRts did
not materialize. Manufacturers started realizing that
liquid distributors had to be improved, but there was
no coherent understanding, nor correlations, that
could lead to a safe distributor-column system design.
Many manufacturers returned to trays, producing
new improved designs, using the area under the
downcomer for vapour Sow: these trays are offered
with new names that indicate their increased vapour
Sow capacity (MaxySow, Superfrack, etc.). The need
for good distribution and its effect on the column
efRciency are now well understood, allowing safe
design and efRcient applications for random and
structured packings in large industrial columns.
General Concepts
Distillation separation is based in relative volatility
that makes it possible to concentrate the more volatile
components in the vapour phase while the less vol-
atile ones remain in the liquid phase. Distillation
columns are countercurrent vapour}liquid mass-
transfer devices, where the required separation and
puriRcation of components is achieved.
1082 II / DISTILLATION / Packed Columns: Design and Performance

Table 1 Evolution of packing

First generation, Second generation, Third generation, after

before 1950 1950}1970 1970

Random packings Rashing rings Intalox威 (Norton) IMTP威 (Norton)

Lessing rings Pall Ringsa CMR威 (Koch Glitsch)
Saddles Chempak威b
Fleximax威 (Koch Glitsch)
Nutter Ring威 (Nutter)

Grids C-Grid (Koch Glitsch)c

EF-25 (Koch Glitsch)c

Structured packing Wire-mesh typed Sulzer BX and CY

Mellapack威 (Sulzer)
Flexipack威 (Koch Glitsch)
Gempack威 (Koch Glitsch)
Intalox威 (Norton)
Montz packing (Montz)

a
Developed by BASF, still marketed (or variations of it) by most packing manufacturers.
b
Developed by Leva, marketed by Nutter.
c
Variations of these grids are now offered by most packing manufacturers.
d
Developed by Sulzer, they are now offered by other manufacturers.

The main variable inSuencing the column design
requirements is the relative volatility,
. Figure 1
illustrates the effect of
on the column perfor-
mance:
E As
increases, the number of theoretical stages
(NTS) required to achieve a Rxed product quality
decreases, since NTS is proportional to 1/ln(
). As

decreases and approaches 1, the number of stages
required increases approaching inRnity. At any
given
, the minimum number of stages required to
achieve a given separation corresponds to a total
reSux operation. At total reSux all overhead va-
pours are condensed and returned to the column as
reSux, so that there is no net product. The min-
imum reSux sets the limiting slope of the operating
line, required to achieve a given separation.
E At constant
, the NTS increases as the product
purity increases. The increase is proportional to the
logarithm of the key components purity ratio.
It can be also demonstrated that:
E At constant product purity, the minimum reSux
decreases as
increases.
E At constant product purity, the minimum number
of stages decrease as
increases.
E At constant
, the minimum reSux decreases as the
product purity decreases.

Figure 1 Number of stages required vs. relative volatility at several product purities.
 II / DISTILLATION / Packed Columns: Design and Performance 1083

E At constant
, the minimum number of stages in-

creases as the product purity increases.
All these statement say that
deRnes the separation
difRculty. For values around 1.1 and lower, separ-
ation by distillation becomes very difRcult, requiring
very large and expensive columns. For
"1 the mix-
ture is azeotropic and would require the addition of
selective entrainers if azeotropic or extractive distilla-
tion is to be applied.

Packed Column Description

Figure 2 illustrates a tower with structured packing.
In addition to the packing itself, packed columns
require other internals to assure the performance of
the packing. These internals are:
E Liquid feed pipes to deliver the Suid to the liquid
distributors, as seen at the top of the tower and at
the intermediate distributor.
E Liquid collection and mixing as shown below the
top bed.
E Liquid draw-off sump and pipe as shown below the
top bed.
E Liquid redistributors, as presented between the two
beds.
E Vapour feed pipes as shown at the vapour inlet
nozzle, at the bottom of the tower.
E Packing support plates resting on beams and level-
led rings welded to the vessel.
E Hold-down plates.
Incorrect design or incorrect installation of any of
these elements can lead to tower failure. One of the
most critical element, and often the culprit of tower
failures, is the liquid distributor.

Packing Selection
Figures 3 and 4 illustrates random and structured
packings. There are many parameters to be con-
sidered in the selection of packings; in some cases,
there are one or two considerations that dictate the
selection, such as capacity for a revamp, which could
favour structured packing. There are also some con-
siderations or applications, such as high-pressure
distillation, that could make structured packing a
questionable choice. Table 2 gives some general guid-
lines on packing selection.

Pressure Drop in Packed Beds

The dry-bed pressure gradient is given by the follow-
ing equation:
Figure 2 Packed tower illustration. (Photo courtesy of Sulzer
Pd"C1gu2g [1] Chemtech.)
1084 II / DISTILLATION / Packed Columns: Design and Performance

Figure 3 Random packings: (A) IMTP威. (Photo courtesy of Norton Chemical Process Products Corporation.) (B) Nutter Ring威.
(Photo courtesy of Sulzer Chemtech.) (C) Cascade Mini-Rings威 (CMR2+) and Fleximax威. (Courtesy of Koch}Glitsch Inc.) (D) Pall
Rings metal and plastic. (Courtesy of Koch}Glitsch Inc.)

Leva extended the correlation to irrigated beds: where:

Pi"C110@  u .
u1 2
g g Gf"G(0.075/g)0.5(Fp/20)0.5100.024 Mg
Robbins developed the following set of general pres- (for pressures over 1 atm)H
sure-drop correlations:
P"C2G2f 10C3Lf HNote: in this correlation the original term 100.3Mg was replaced by
100.024Mg since the original correlation predicts too high a pressure
#0.4(Lf/20 000)0.1(C2G2f10C3Lf)4 [2] drop.
 II / DISTILLATION / Packed Columns: Design and Performance 1085

Figure 4 Structured packings: (A) Wire gauze structured packing. Close view, packing and wiper bands. (Photo courtesy of
Koch}Glitsch Inc.) (B) Two structured packing layers rotated 903. (Photo courtesy of Koch}Glitsch Inc.) (C) One structured packing
element for small towers. (Photo courtesy of Sulzer Chemtech.) (D) Structured packed bed for a small tower. (Photo courtesy of
Koch}Glitsch Inc.) (E) Packed bed for a large tower built in sections. (Photo courtesy of Norton Chemical Process Products Corp.)
1086 II / DISTILLATION / Packed Columns: Design and Performance

Figure 4 Continued
Figure 5 Bed P vs. rates. (Permission from Gulf Publishing
Company.)

Lf"L(62.4/1)(Fp/20)0.50.1
l (for Fp over 15)
dry-column line (which is a function of the drag
C2"7.4;10\8 and C3"2.7;10\5
only). Equation [3] allows calculation of the packing
For the case of dry packing Lf"0, the pressure- factor, Fp, by measuring the slope of the dry-packing
drop equation reduces to: pressure-drop data. As the vapour rate increases, the
slope of the constant liquid rate lines increase; this
P"C2G2f"C2(0.075/20)FpG2/g. [3] increase is also proportional to the liquid rate. The
initial departure from the dry-line slope indicates
Figure 5 presents a family of pressure drop-lines at interaction between the vapour and liquid, and rep-
constant liquid Sow as a function of the vapour Sow. resents a loading point. EfRcient mass-transfer
The constant liquid rate lines start parallel to the operations can be achieved only above the loading

Table 2 Packing selection guidelines (trays included as a reference)

Application in distillation Random packing Structured packing Traditional trays High-capacity trays

Pressure drop/theoretical stage 2 1 3 3

Maximum capacitya
Efficiency at high pressure
Efficiency at low pressure
Efficiency at low liquid ratec
Efficiency at high liquid rated
Low residence time
High residence time
Heat transfer
Foaming systems
Non-metallic servicesb
Fouling systems
Efficiency in high  systems
Inspection and maintenance
Low cost
2 1 3 2
2 4 2 1
2 1 2 3
2 1 3 4
3 4 2 1
2 1 4 4
3 4 1 1
2 1 2 2
2 2 3 3
1 2 4 4
4f 2f 1e 1e
2 4 1 1
3 4 1 1
2 4 1 3

Application rating: 1, best; 2, good; 3, fair; 4, poor.

a
Efficiency may be reduced at high capacities.
b
As may be required based on corrosion protection considerations, such as ceramic.
c
Systems below 5 gallons min\1 ft\2.
d
Systems over 15 gallons min\1 ft\2.
e
Applies to sieve trays, specially dual-flow, not to valve trays.
f
It would require a fouling-resistant distributor, which may result in reduced efficiency.
 II / DISTILLATION / Packed Columns: Design and Performance
point. For any given liquid rate, as the vapour rate
further increases, the pressure-drop line slope in-
creases rapidly until the line becomes near vertical. At
this point the Sow and P are unstable, and the bed is
Sooded; the vapour Sow does not allow the liquid to
Sow down the bed and there is massive entrainment
of liquid in the vapour phase and mass transfer is no
longer viable.
For most packings, bed Sooding occurs between
1 and 2 inches of water-pressure drop per foot of
packing. Pressure drop at Sooding seems to be a func-
tion of the packing size. Kister cited Zenz and later
Strigle and Rukovena observations indicating that
Sooding (Pfl) is higher for smaller size packings, and
proposed a correlation to determine the pressure drop
at Sooding as a function of the packing factor.
Pfl"0.115(Fp)0.7 [4A]
We also obtained by regression from data published
by Strigle:
Pfl"0.146Fp0.75 inch liquid ft\1 or [4B]
P"0.146SgFp0.75 inch H2O ft\1 [5]
Pressure drop at incipient loading may be estimated:
Pl"0.072SgFp0.75 [6]
and pressure drop at maximum efRciency loading
may be estimated by:
Pe"0.082SgFp0.75 [7]
All the above correlations have been regressed for
metallic random packings (Pall Rings and IMPT威).
For column design, it is well-accepted practice to
assume Sooding at 1 inch of water per foot of packing
pressure drop and design the packing for an operation
at 80% Sood. However, when reliable packing-factor
information is available, the use of the calculated
Pfl, using one of the eqns [4A], [4B] and [5], is
a more accurate approach.
Caution: Presence of foam, even incipient foam, has
a great impact on a packing column pressure drop
and performance and should be avoided. Amines,
insoluble Rne solids (such as corrosion products),
high-viscosity organic liquid (0.5}1 cP or higher) and
immiscible liquids are known to foam. For these
systems, or other systems known to be prone to foam,
continuous or intermittent dosing of antifoam
agents may be required to maintain an efRcient
packed-column operation. Nevertheless, uncontrol-
1087
led antifoam injection is known to aggravate foaming
problems. Filtration of liquids and adsorption of
contaminants on activated carbon has proven valu-
able to control foaming in some systems such as
amines.
Flooding Correlations
Several generalized Sooding and pressure-drop cor-
relations have been proposed for commercial pack-
ings. Sherwood, Shipley and Holloway presented
the Rrst correlation between a ‘Sow parameter’ X
deRned as:
X"(L/G) (g/l)0.5 [8]
and a ‘Sooding parameter’ Y deRned as:
Yf"(u2g/gc)(a/3)(g/l)0.2"(G2f/gc)(a/3)0.2/(gl).
[9]
Sherwood and co-workers correlated dumped and
stacked random packing data and found that Yf is
around Rve times higher for stacked than for dumped
packing, which means that mass velocity at Sood is
over two times higher for stacked packing. This was
the precursor idea for the later development of ‘struc-
tured’ packings.
Lobo and Friend presented a similar correlation of
Y and X with indication of pressure-drop lines and
Sooding line.
Leva proposed a similar correlation with the same
Sow parameter given by eqn [8] and modiRed the
Sooding parameter Yf"(G2f/gc)(a/3)0.2 (w/l)2/l.
According to this correlation, minimum loading
Ym occurs at about one-third of Yf which means that
loading starts at 50% of the mass Sow rates corre-
sponding to the Sooding point.
Eckert observed that the packing geometrical pro-
perties factor (a/3) did not represent correctly the
packing in the Sooding correlations. He introduced
a packing factor, Fp. The value of Fp is determined
experimentally from pressure-drop data. The new
Sooding parameter became:
Yf"(G2f/gc)Fp0.2(w/l)2/(gl) [10]
and is correlated to the same Sow parameter
X"(L/G) (g/l)0.5.
The most recent proposed correlation was present-
ed by Strigle (see Figure 6):
Y"CsFp0.5(/Sg) 0.05"CsFp0.50.05 [11A]
1088 II / DISTILLATION / Packed Columns: Design and Performance

Figure 6 Striegle pressure drop chart. (Permission from Gulf Publishing Company.)

Y is the vapour Sow parameter and is a function of
vapour capacity factor Cs"ug(g/(l!g))0.5, the
packing factor and the kinematic viscosity "/Sg.
Note that at Sooding Y"Yf. Y is plotted in a linear
ordinate as a function of the Sow parameter X in
a logarithmic abscissa and a family of constant P
lines. No Sooding line is shown. The advantage of the
linear ordinate is that it is easier to interpolate than
the older log}log charts.
Figure 7 presents the Sooding lines of packings as
a function of the packing factor Fp and the Sow
parameter X. The ordinate is the modiRed Sooding
parameter Y fH, deRned as follows:
Y Hf "Yf/Fp0.5"Cs0.05 [11B]
Y Hf is plotted as a function of the Sow parameter X,
eqn [8], at constant packing factors.

Figure 7 Modified flooding parameter as a function of the flow parameter.

 II / DISTILLATION / Packed Columns: Design and Performance 1089

Comparing Packed Column vs. Tray ations that occur in actual operations and for
Tower Capacity process control requirements.
6. Determine the column diameter Dc"12(4Ac/)0.5.
Table 5 presents packing capacities, calculated from
the above relations, compared to tray Sooding capac- Turndown and Minimum Wetting Flow
ities at several tray spacings.
In general, the turndown of a packed tower is limited
to the turndown of the liquid distributor, which is its
Packed Tower Diameter ability to reduce liquid load and still maintain a
homogeneous distribution. Most standard liquid
Figures 6 or 7 can be used to determine the column distributors can operate efRciently at 50% of its de-
diameter. Using Figure 7 the procedure is as follows: sign liquid load; turndown as low as 25% can be
achieved.
1. Determine the value of the abscissa X"
To operate efRciently as mass-transfer devices,
L/G(g/l) .
0.5
packing should be homogeneously wetted to assure
2. Obtain from the manufacturer the selected pack-
use of the total surface. Minimum recommended
ing Fp value, or from Tables 3 or 4.
values of liquid irrigation depend on the packing
3. Determine the ordinate YHf "Cs 0.05
from Fig-
material and surface wettability, as follows:
ure 7.
4. Calculate the capacity factor at Sood Cs from Random packing
the YfH value, the gas velocity at Sood ug" Ceramic 0.2 gallons min\1 ft\2
Cs(l!g) /g and the Sooding gas mass velocity
0.5
Surface-treated or 0.5 gallons min\1 ft\2
Gfl"ugg. rusted metals
5. Determine the column cross-sectional area Glass, glassed ceramic 1.0 gallons min\1 ft\2
Ac"V/(0.8Gfl), based on 80% of the G Sooding and stainless steel
rate. This is standard design practice for new Plastics 1.5}2.9 gallons
column sizing, and allows for normal Sow Suctu- min\1 ft\2

Table 3 Random packing design parameters

Packing metal Nominal size Packing Specific surface Void ft 3 ft\3 () Bulk density
factor (Fp) ft 2 ft\3 (a) (lb ft\3)

Pall Rings 0.625 81 103 0.918 39.9

1 56 61 0.953 23.1
1.5 40 39 0.971 14.3
2 27 30 0.969 14.1
3.5 18 18 0.972 13.9
CMR威 0 60 103 0.957 20.96
1 38 76 0.968 15.51
1.5 33 57 0.961 18.66
2 26 44 0.970 14.29
2.5 21 38 0.974 12.54
3 14 32 0.979 10.22
4 12 23 0.985 7.36
5 8 15 0.989 5.46
IMTP威 No 15 51 88.7 0.961 17.9
No 25 41 69.8 0.970 14.1
No 40 24 46.9 0.969 14.6
No 50 18 31.2 0.981 9.3
No 60 16 25.3 0.982 8.7
No 70 12 17.5 0.984 8.1
Nutter Rings威 0.7 N/A 69 0.978 11.0
1.0 30 51 0.978 11.1
1.5 24 38 0.978 11.3
2.0 18 29 0.979 10.8
2.5 16 25 0.982 9.0
3.5 13 20 0.984 8.3
1090 II / DISTILLATION / Packed Columns: Design and Performance

Table 4 Structured packing design parameters

Packing 453 Size Packing factor Specific surface Void fraction () Bulk density
Crimp angle ft2 ft\3 (a) (lb ft\3)

Mellapack威 (Sulzer) 125Y 10 35 0.989 5.09

250Y 20 78 0.987 5.61
350Y 23 107 0.983 7.8
500Y 34 155 0.975 10.92
Sulzer BX (Gauze) BX 21 150

Gempack威 (Koch 4A 55 138.1 0.942 17

Glitsch) 3A 23 91.4 0.962 9.9
2A 15 67 0.972 6.3
1A 9 35 0.977 4.7
Intalox威 (Norton) 1T 28.0 95.2 0.980 10.14
2T 20.0 65.3 0.984 8.23
3T 15.0 51.9 0.987 6.55
4T 13.5 40.6 0.986 6.75
5T 12.0 27.0 0.991 4.5
Montz B1-100 30
B1-200 20 61 0.94
B1-250 76
B1-300 33 91

Structure packings especially for high-purity separations, only gravity

Surface-treated metals 0.2 gallons min\1 ft\2 distributions are used. Table 6 illustrates the main
Plain surface metals 0.5 gallons min\1 ft\2 type of distributors and the main factors to be con-
sidered for selection:

E Pipe oriTce headers (POH) (Figure 8) consist of

Type of Liquid Distributors a pipe ladder arrangement with calibrated oriRces
Liquid distributors can be gravity or pressure fed drilled in the pipe laterals in a uniform layout.
depending on how the liquid is introduced to the POH can be pressure or gravity fed.
distributor. Pressure distributors are limited to heat E Pan distributors (PAN) (Figure 9) consist of a Sat
transfer and some simple mass-transfer operations, horizontal plate (tray) with uniformly spaced calib-
mainly in stripping or absorption. For distillation, rated oriRces that allow the passage of liquid to the
packing below. Round or rectangular risers (chim-
neys), located within the oriRce pattern, distribute
Table 5 Relative capacity of packing and traysa the vapour to the packing above. The riser layout
should be uniform and should not interfere with
Tray spacing Ratio of packing to tray capacity according to the uniformity of the oriRce layout. PAN distribu-
packing factor (Fp) tors are always gravity fed.
10 20 30 40 50 60 E Narrow trough distributors (NTD) (Figure 10A
and 10B). This distributor is composed of a series
36 inches 1.15 0.96 0.87 0.81 0.76 0.73 of narrow (3}4 inches) parallel troughs fed by one
24 inches 1.45 1.22 1.10 1.03 0.97 0.93 or more larger troughs (parting boxes) oriented at
18 inches 1.90 1.60 1.44 1.35 1.27 1.22
903 from the narrow troughs. The narrow troughs
12 inches 2.41 2.03 1.84 1.71 1.62 1.55
distribute the liquid to the packing below, through
a
Tray capacity based on the column full cross-sectional area, calibrated oriRces drilled at the bottom or at the
without discounting any area for downcomers (which implies high- wall. NTD are always fed by gravity.
capacity trays). For conventional trays the ratio of packing capa- E Spray nozzle header (SNH) (Figure 11). They are
city/tray capacity will be higher. Tray capacity taken from the
similar to POH but spray nozzles are used instead
generalized correlation of tray flooding proposed by Fair JR and
Matthews RL (Petroleum Refiner 37(4): 153). The packing capa- of oriRces. The density of nozzles in the SNH is
city taken from the generalized correlations presented by RF lower than the density of oriRces in the POH. The
Striegle Jr and Figure 6). SNH relies on the liquid cone leaving the nozzle for
 II / DISTILLATION / Packed Columns: Design and Performance 1091

Table 6 Guidelines for distributor selection

Gravity-fed distributors Pressure-fed distributors

POH PAN NTD POH SNH

Uniformity 1 1 1 2 3
High-purity fractionation 1 1 1 3 3
Maximum drip points per area 2 1 1 2 2
For large diameter towers (over 10 ft) 1 3 1 1 1
Leakage potential C Ha C C C
For high liquid rates 2 1 2 2 1
For high vapour rates 1 3 1 1 1
Residence time C A B C C
Solids handling 3 3 2b 2 1
Turndown 1 1 1 1 3
Easy installation and levelling 1 3 2 1 1
Cost B A A C C

1, Good; 2, fair; 3, poor; A, high; B, medium; C, low.

a
Unless it is seal-welded.
b
Very good if a V-notch is provided at the top of the trough wall for liquid flow. Nevertheless, the quality and turndown of the distributor
are affected.

further spreading. This results in either an overlap
or a gap of the cone projection over the packed
bed, and deteriorates the uniformity of the distri-
bution. SNHs can handle very large liquid rates
and are very efRcient for heat transfer.
Liquid Mixing, Redistribution and
Maximum Bed Height
Initial liquid distribution is essential to achieve good
packed tower efRciency. Hoek suggested that at
a given Sow rate, each packing has its natural distri-
bution determined by its radial spreading coefRcient.
Although this effect does spread the initial liquid
distribution, this effect is not sufRcient to correct
poor initial distribution. Radial concentration gradi-
ents already established at the top of the bed cannot
be compensated by additional packing. The result is
permanent efRciency loss.
In general, if a good distribution is established at
the top of the bed, the packing will develop its natural
distribution and maintain it for bed depths of 10 NTS
or more. Columns requiring more than 10 NTS per
section should be subdivided into several packing
beds to maintain coefRcient HETP values. Liquid
redistribution, and often mixing, are required be-
tween these bed sections.
Distributor Design Parameters
Distributor Liquid Level and Hole Diameter
The basic distributor design equation relates the total
oriRce open area, the liquid head and the volumetric
Sow:
Q"Cona0(h!hd)0.5 [12]

Figure 8 POH distributor. (Courtesy of Norton Chemical Pro-

cess Products Corp.) Figure 9 PAN distributor.
1092 II / DISTILLATION / Packed Columns: Design and Performance
Figure 10 NTD distributor: (A) Photo courtesy of Norton Chemical Process Products Corp. (B) Photo courtesy of Koch}Glitsch Inc.
where Q is the volumetric Sow rate, Co the oriRce
Sow coefRcient, n the number of oriRces, a0 the open
area of one oriRce, h the liquid head over
the oriRce, and hd the vapour-pressure drop across
the distributor given in head of liquid. The value of
Co varies between 0.5 to 0.8 and is near 0.6 for most
commercial distributors. Using this value, eqn [12]
becomes:
Q"4.0nd 2(h!hd)0.5
and:
n"0.25Q/d 2(h!hd)0.5 [13]
The minimum recommended oriRce diameter, to pre-
vent plugging, is 3/8 inch for carbon steel and 1/8
inch for stainless steel. The minimum recommended
liquid level at minimum Sow is 2 inches. If a 50%
turndown is speciRed, the required liquid level at
normal liquid load becomes 8 inches.
Figure 11 SNH distributor. (Courtesy of Norton Chemical Pro-
cess Products Corp.)
Uniformity of the Drip Point Layout
Density of liquid drip points is not enough to assure
a good distributor quality. The distribution must be
homogeneous; the same amount of liquid should irri-
gate the packing at any fraction of the tower cross-
sectional area. Areas near the tower wall should
receive the same amount of liquid as areas near the
centre.
Other Considerations
A number of factors need to be considered when
selecting and designing packing and distributors.
Ratio Tower to Packing Size
The minimum recommended ratio of the tower dia-
meter to the packing size is 8. In the case of structured
packings, this ratio applies to the ratio of the tower
diameter to the crimp size.
Fouling
Some solids are usually present even in ‘clean systems’
because of corrosion products, especially after main-
tenance shutdowns, when rust and debris can remain
in the tower. The tower shell metallurgy should be
adequate to prevent formation of scale or corrosion
products that can plug distributors. Distributors with
small oriRces should be protected with Rlters in all
liquid lines entering the tower. In other cases, solids
are expected to be present because of the process
itself. In these cases the distributor should be designed
to handle the solids. A NTD distributor with V-
notches for liquid overSow is adequate to handle
 II / DISTILLATION / Packed Columns: Design and Performance
some slurries. SNHs can also handle slurries but their
application is limited to heat transfer.
Vapour Distribution Requirements
Vapours entering the tower have a kinetic energy
proportional to their velocity, which is converted into
pressure as the vapour turns to start Sowing upward
in the tower. The resulting radial pressure proRle is
not uniform; areas of higher pressure would allow
higher vapour up-Sow. This is especially critical for
low-pressure drop packings such as structured pack-
ings. Vapour radial velocity proRles are corrected by
pressure drop and by diffusion devices. The following
is the recommended practice for vapour distribution:
E Low vapour inlet velocity (velocity head below 0.5
inches of water): no inlet distributor required, pro-
vide as minimum 112 column diameters, or 36
inches between the top of the vapour inlet nozzle
and the bottom of the bed.
E Intermediate vapour inlet velocity (velocity head
between 0.5 and 1.5 inches of water): provide an
inlet vapour diffuser directing vapour Sow down
the tower. This type of device can be a horizontal
pipe with the bottom half cut as shown at the
bottom of the column in Figure 2. Vertical bafSes
can be provided for better vapour distribution. The
purpose of these bafSes is to stop the horizontal
velocity component of the vapour.
E High vapour velocity (velocity heads above 1.5
inches of water): provide an inlet vapour diffuser,
as described above, plus a small riser chimney tray
with a pressure drop of a minimum of 2 inches of
water. The pressure drop can be created by oriRces
at the bottom of the risers. A vapour distributor, as
shown in Figure 12 is a good alternative to the
vapour diffuser in critical systems.
Distributor Testing
Water test of assembled distributors at the manufac-
turer’s workshop is always a good practice for all
high-efRciency distributors. The test should deter-
mine liquid rate gradients under the distributor,
liquid level in the distributor itself at design and
Figure 12 Vapour distributor. (Courtesy of Sulzer Chemtech.)
1093
turndown liquid rate, liquid level gradient in the
trough and uniformity of the drip point layout. These
parameters should be compared to the distributor
design parameters and adjustments made to the dis-
tributor if necessary. Figure 13 shows a distributor
testing facility.
Packing Performance in Distillation
Factors to Consider in Determining the Column
Design HETP
The height equivalent to a theoretical plate (HETP) is
determined by the following main three factors:
Intrinsic geometric shape and size of the packing
This factor determines the surface per unit of volume,
and the packing capacity of establishing effective va-
pour liquid interfacial surface. It is a well-known fact
that, for any packing, the smaller its particle size, the
larger its surface : volume ratio, and the lower the
HETP value. All the other factors being equal, numer-
ous available data tend to indicate that the expected
reference packing HETPo may be correlated as fol-
lows:
HETPo"Kp/Fpf [14]
where Fp is the packing factor. The constants Kp and
f for different types of commercial packing are corre-
lated as in Table 7 for a reference system.
Figure 13 Distributor testing facilities. (Photo courtesy of
Koch}Glitsch Inc.)
1094 II / DISTILLATION / Packed Columns: Design and Performance

Table 7 HETP correlation factors for a reference system, re- properties:

gressed by the authors for eqn [14]
HETP"HETPo(
/Sg)n /(
/Sg )n0 [15]
Packing Constant Kp Exponent f

Structured packings when the n exponent best Rt is between 0.15 and

Sulzer Mellapak威 126 0.73 0.21. Replacing HETPo from eqn [14] into eqn [15],
Koch Flexipac威 100 0.69 and using the reference system we obtain:
Koch}Glitsch Gempak威 120 0.76
Average of above structured 106 0.70
packings HETP"(2.0Kp/F fp) (
/Sg)0.2 [16]
Random metallic packings
Koch}Glitsch CMR威 73 0.43
In addition, theoretical considerations suggest that
Norton IMPT威 198 0.69 the HETP is related to "m/(L/V), the ratio of the
Pall Rings 250 0.69 slopes of the equilibrium line and operating line, by
Average above random packings 110 0.50 the correlation:

HETP" ln( )/( !1)HTU

System properties Numerous investigations have
tried to correlate experimental HETP data with the
distillation system Suid physical properties. The best
and most consistent correlations tend to conRrm
that the HETP is proportional to reference HETPo
and a factor proportional to the system physical
where HTU is the height of a transfer unit. Then:
HETP"(2.0Kp/F fp) (
/Sg)0.2 f ( ) [17]
Packing loading Figure 14 shows the pilot plant
performance of Sulzer/Nutter ring No. 2.5 in

Figure 14 HETP vs. loads. (Courtesy of Sulzer Chemtech.)

 II / DISTILLATION / Packed Columns: Design and Performance 1095

isobutane}n-butane separation. Note that although

only the vapour rate appears in the abscissa, actually
both the vapour and the liquid rates increase in the
same proportion since the chart was developed at
total reSux. All packings present similar curves in
small size experimental columns. The initial HETP is
high (low efRciency) owing to the low loads that
result in liquid maldistribution, poor packing wetting
and little interaction between the vapour and the
liquid (this left section of the curve is not shown in
Figure 14). Nevertheless, the HETP continuously de-
creases as the loads increase. At a point, correspond-
ing to the loading point of the packing, the HETP
becomes constant over a range of loads. This range
represents the operating range of the packing. As the
loads continue to increase, the HETP shows a dip
corresponding to high interaction between the Suids,
followed by a rapid increase in the HETP caused by Figure 15 Effect of number of drip points and liquid irrigation
recirculation of liquid within the bed. This corres- rate on maldistribution. (Permission from Chemical Engineering
Progress.)
ponds to the initial Sooding of the bed.

Maldistribution
Liquid maldistribution has a very large effect on col-
umn distillation performance. Liquid maldistribution
is originated by uneven liquid Sow from the distribu-
tor to the top section of the packing. Some degree of
maldistribution cannot be avoided and it is related to
the following factors.
Drip points density (total drip points/column cross-
section area) In principle, a smaller number of drip
points equates to a higher initial maldistribution. This
could be solved by constructing distributors with
a high number of drip points. However, there are
physical and mechanical limits that make it difRcult
to build distributors with more than 20 drip
points ft\2. It has also been demonstrated that if the
distributor deck is not levelled, the resulting maldis-
tribution effect may increase as the number of drip
points is increased above an optimal number. The
optimal number of drip points is related to the liquid
irrigation Sow as follows (Figure 15):
Liquid irrigation 0.25 0.5 1.0 2.0 4.0
g ) m\1 ) ft\2
Optimum number drip 5 8 13 21 32
points per square foot
Furthermore, the drip points themselves may create
additional maldistribution if they are not evenly dis-
tributed across the entire column cross-sectional area.
Poor construction making holes of variable diameters
or unlevelled installation of the distributor will also
induce additional maldistribution. Operational prob-
lems such as plugging of the distributor deck areas
will cause large sectors to be dry, thus producing
a macroscopic or sectorial maldistribution.
Maldistribution and spreading factor Initial maldis-
tribution produces a condition of uneven liquid/
vapour Sow ratio across the column cross-sectional
area. Some areas or spots are underirrigated and some
are overirrigated. The column packing does spread
the liquid resulting in some correction or attenuation
of the initial maldistribution. The overall weighted
maldistribution is attenuated better in small diameter
columns than in larger columns. This is determined
by the nondimensional number (Zb/CD2c), where Zb is
the bed height in feet, Dc the column diameter in
inches, and C is the spreading factor in ft ) in\2 units
(see Figure 16). The spreading factor is related to the
packing particle size and the liquid irrigation.
The lost column efRciency is proportional to the
liquid maldistribution, and this effect is ampliRed by
the number of theoretical stages required to achieve
the separation. Figure 17 presents a useful correlation
for the calculation of the column efRciency in packed
distillation columns.
Liquid distributor quality The liquid distributor in-
trinsic maldistribution, Md (related to its design and
manufacture) should be measured at the factory by
a water test measuring the liquid Sow under each
subsection of the column cross-section. The smaller
and more numerous the test area subdivisions,
the more precise will be the maldistribution
1096 II / DISTILLATION / Packed Columns: Design and Performance
Figure 16 M vs. Zb/CD 2c. (Permission from Chemical Engin-
eering Progress.)
measurement. The mathematical expression of the
maldistribution is:
Md"100[ ((Li/Lav!1)2)/n]0.5
for each point area subdivision from i"1 to i"n.
The following is the correlation between distributor
quality and its maldistribution:
Qd (%)"100/[1#(Md/100)2]
Figure 17 Efficiency vs. NTS.
If the distributor quality is 90%, the actual measured
maldistribution should not exceed 33%. A 95% qual-
ity implied a maximum measured maldistribution
of 23%.
Total maldistribution Additional maldistribution
can originate from operational factors related to
levelness and obstructions. The total initial maldis-
tribution, Mo, can be calculated by:
Mo"( M2)0.5 [20]
Assuming a maximum operational maldistribution
Mop"15%, and using a 90% distributor quality
(Md"33%), the total effective operating maldis-
tribution at the top of the packing is
Mo"(332#152)0.5"36.2.
The effective bed attenuated maldistribution is cal-
culated by the following equation:
Mbz"Mo/[1#0.16Mo(Z/CD2c)] [21]
With this Mbz value, the bed efRciency Ez may be
obtained from Figure 17. The calculated bed efRcien-
[18] cy should be used to correct the packing HETP and
obtain the bed operating HETPop:
HETPop"HETP Ez/100 [22]
HETPop"[(2.0 Kp/Ffp)(
/Sg)0.2 f ( )]Ez/100.
[19] [23]
 II / DISTILLATION / Packed Columns: Design and Performance 1097

The bed effective number of theoretical stages can be h Liquid head over inches of
calculated by: distributor oriRce liquid
hd Vapour pressure drop inches of
NTS"Z/HETPop. [24] across liquid
liquid distributor
For a column requiring more than 10 NTS, it is in G Vapour Sow mass velocity lb ft\2 h\1
general advantageous to subdivide the packing in two Gf Vapour Sow mass velocity lb ft\2
or more beds and limit the NTS per bed to around 10. (in Robbins equation)
The lower the NTS per bed, the higher the resulting V Vapour Sow lb h\1
bed efRciency. The limiting factor of subdividing the L Liquid mass Sow lb ft\2 h\1
column into a large number of redistributed beds, is Lf Liquid mass Sow lb ft\2 h\1
the extra column height (or the effective packed height Li Liquid mass Sow at point i gallons
loss for column revamps) necessary to accommodate min\1 ft\2
each redistributor, and the resulting increased cost. For Lav Liquid average mass Sow gallons
new columns, the optimal number of beds is the one min\1 ft\2
that results in the required performance at minimum Fp Packing factor
cost, for revamps, it is often the one that results in the Kp HETP correlation factor
maximum available overall NTS. The best choice in n Number of measured points
each case is determined by an optimization. (in distributor testing)
HETP Height equivalent of a in
theoretical plate
Future Developments HTU High of a transfer unit in
With the ability to accurately design and predict the l Liquid viscosity cP
performance of packings in distillation, it is expected Sg Liquid speciRc gravity
that the use of packings in distillation will become X Flow factor"
better accepted, not only for plant revamps but also (L/G) (g /l)0.5
for grass roots applications. The design and evalu- Y Vapour Sow parameter. At
ation of liquid distributors needs to be better under- Sooding Y"Yf
stood by users and equipment manufacturers; stan- Yf,YHf Flooding parameters,
dard methods for distributor quality rating should be deRned by eqns [10] and
implemented based on the basic concepts presented in [11]
this contribution. Readers interested in further ex- a Packing surface area ft2 ft\3
ploring the column design methods outlined in a0 Open area of one drip point
this article may download a free demo of BDSIM
Relative volatility
at url http://www.geocities.com/&combusem/  Void fraction
BDSIM.HTM  Surface tension dynes cm\1
gc Gravitational constant 32.2 ft s\2
NTS Number of theoretical
Nomenclature stages
Ac Column cross-sectional ft2 R ReSux ratio
area Rm Minimum reSux ratio
P Packing pressure drop in ft\1 Md Maldistribution originated %
C Packing spreading factor ft in\2 by the distributor design
Co OriRce Sow coefRcient Mo Total initial maldistribution%
C1 , C2 , C3 Constants in pressure drop Mbz Effective bed %
correlations maldistribution
d OriRce diameter in Q Liquid Sow gallons
Dc Column diameter in min\1
Ez Bed efRciency % Qd Distributor quality %
Cs Vapour capacity factor, ft s\1 Zb Bed height ft
deRned by  Kinematic viscosity
Cs"ug(g /(l!g))0.5 constant"l/Sg
g Gas density lb f\3 Ratio of the equilibrium
l Liquid density lb f\3 curve slope to the operating
ug Vapour velocity ft s\1 line slope
1098 II / DISTILLATION / Pilot Plant Batch Distillation
See also: I/Distillation: Historical Development; Model-
ling and Simulation; Theory of Distillation; Tray Columns:
Performance; Tray Columns: Performance; Vapour-Liquid
Equilibrium; Correlation and Prediction; Vapour-Liquid
Equilibrium: Theory.
Further Reading
Bonilla J (1993) Don’t neglect liquid distributors. Chemical
Engineering Progress 83(3): 47.
Eckert JS (1961) Chemical Engineering Progress 57(9): 54.
Fair JR and Matthews RL (1958) Petroleum ReTner 37(4):
153.
Klemas L and Bonilla J (1995) Accurately assess packed-
column efRciency. Chemical Engineering Progress
91(7): 27.
Pilot Plant Batch Distillation
M. A. P. de Carvalho and
W. R. Curtis, The Pennsylvania State University,
PA, USA
Introduction
Laboratory distillation encompasses an operating
range from millilitres in bench-top devices to pilot
units with the capacity for producing several hundred
kilograms of product per day. While the design of
bench-top assemblies is generally geared towards the
achievement of a speciRed purity grade of the desired
product, quantitative predictions are not usually feas-
ible for such equipment and their construction relies
a great deal on ingenuity and craftsmanship. For
dedicated applications, glassware companies offer
off-the-shelf equipment. This article will therefore
focus on the pilot-scale units, where the analytical
principles of mass and heat transfer can be applied
to the operation, design and optimization of the
equipment.
The section on theory presents analytical descrip-
tions of batch distillation for three different ap-
proaches in order of decreasing complexity. It starts
with a comprehensive model for a nonadiabatic, non-
zero hold-up, nonconstant molar overSow, nonideal
multicomponent column. The second model present-
ed neglects stage hold-ups and assumes adiabatic
stages and constant molar overSows to arrive at a set
of equations describing the transient behaviour of the
equipment, which can be solved for a binary system
using a simple spreadsheet. If constant relative vola-
tility and operation at minimum reSux are further
Kister HZ (1992) Distillation Design. New York:
McGraw-Hill.
Leva M (1954) Chemical Engineering Progress 50(10):
51.
Lobo WE et al. (1945) Transaction of the American Insti-
tute of Chemical Engineers 41: 693.
Robbins LA (1991) Chemical Engineering Progress, May,
p. 87.
Sherwood TK, Shipley GH and Holloway FA (1938) Indus-
trial and Engineering Chemistry 30.
Strigle RF Jr (1994) Packed Tower Design and Applica-
tions. Houston: Gulf Publishing.
Strigle RF Jr and Rukovena F (1979) Chemical Engineering
Progress 75(3): 86.
Zenz FA (1953) Chemical Engineering, August, p. 176.
assumed, the derivation of a third model is possible,
where the transient states within the equipment are
given by direct analytical expressions.
The design of a batch column can be a challenging
task because batch distillation presents unique con-
siderations that are not addressed in most of the
available literature, which is concerned with continu-
ous operation. The section on design is a collection
of advice and criteria for the design of batch
columns. SpeciRc information is given about equip-
ment for batch distillation and accompanying instru-
mentation and safety circuitry. Details are drawn
from a pilot-scale column that is installed in Penn
State University’s Department of Chemical Engineer-
ing. The section on column operation extends the
scope of the two preceding sections by providing
information on establishing operating strategies and
operating protocols for batch runs. Much of this
information is based on hands-on experience ac-
quired with the column described in the subsection on
equipment.
The last section is a synopsis of numerical tech-
niques that have been developed in recent years to
facilitate the optimization of the operation and design
of batch columns. Inherent difRculties associated
with the implementation of these numerical tech-
niques into computer codes prevents their widespread
use in equipment operation and design. However, it is
likely that these techniques will be integrated into
commercial simulators in the near future and be read-
ily available to users with little knowledge of pro-
gramming. The aim here is to introduce the reader to
the topic, rather than to offer extensive coverage,
 II / DISTILLATION / Pilot Plant Batch Distillation 1099

providing references for those interested in further

reading.

Theory
The theory of batch distillation permits design, opera-
tion and optimization calculations by integrating the
concepts of thermodynamic equilibrium, mass and
heat transfer, energy and material balances to solve
the problem of predicting the compositions and Sow
rates of process streams. The traditional approach to
the development of governing equations is to model
the equipment as a stack of equilibrium stages. De-
partures from ideality are taken into account by in-
troducing the concept of stage efRciencies. Modelling
in terms of equilibrium stages is convenient because
of the availability of extensive equilibrium data for
multicomponent systems and the associated predic-
tive thermodynamic models. Packed columns, which
do not possess physical mass transfer stages, can be
translated into this technical framework by the use of
the transfer unit concept (see, for example, McCabe
et al., 1993).
Relinquishing or including levels of complexity and
interdependence in the fundamental equation and
process variables will result in more or less rigorous
treatments, with gains in accuracy usually being ac-
companied by substantial drawbacks in complexity
and computational difRculty. Hereunder, three differ-
ent sets of model equations are presented in decreas- Figure 1 Batch distillation column schematic for a rectifying
ing levels of complexity. A rigorous approach to the configuration.
problem involves the solution of a set of time-depen-
dent differential and algebraic equations for the ma-
terial and heat balances and for the equilibrium rela- Energy balance
tions. Batch columns are usually constructed so that


NC
dHLi,j dT
there is only one section, either above or below the hLtot,i xi,j "Li 1xi 1,jHLi 1,j
dT dt \ \ \
feed stage. The typical column design depicted in x"1

Figure 1 represents the rectifying section. From this #Vi#1yi#1,jHVi#1,j

simpliRcation it is possible to write the following
equations to deRne the problem completely in each !Lixi,jHLi,j!Viyi,jHVi,j#Qi
stage except the top stage and the feed drum (re-
boiler). i"1, Nstage, j"1, NC [3]

Vapour liquid equilibrium relation

Total material balance
yi,j"f (xi,j) [4]
dhLtot,i
"Li 1#Vi#1!Li!Vi [1]
dt \ Summation of liquid mole fractions
NC
Component material balances xi,j"1 [5]
j"1

dxi,j
hLtot,i "Li 1xi 1,j#Vi#1yi#1,j!Lixi,j!Viyi,j Summation of vapour mole fractions
dt \ \
NC
yi,j"1 [6]
i"1, Nstage, j"1, NC [2] j"1
1100 II / DISTILLATION / Pilot Plant Batch Distillation

Liquid total hold-up constraint Component material balances

dxR,j


NC
hLtot,i xi,jvj "Vol [7] hLtot, R "LN stage xN stage,j!VRyR,j [12]
dt
j"1

Energy balance
Eqns [1]}[7] ignore the tray hydraulic behaviour

and assume identical compositions for the liquid
hold-up within a stage and the liquid outSow out of
the stage. Other major features and assumptions of
the model are nonadiabatic stages, negligible vapour
hold-up and constant volumetric liquid hold-up. The
assumptions concerning the hold-up are very reason-
able and the equations can be readily translated to
real stages by the introduction of Murphree tray ef-
Rciencies as correction factors for either the liquid or
the vapour compositions.
For the top stage, the liquid inSow is related to the
distillate outSow by the reSux ratio, and for the feed
drum, the depletion of material should be taken into
account. There is also no liquid outSow for the re-
boiler, thereby decreasing the number of necessary
equations by one (eqn [7]). The following equations
are the modiRed set for the situation in the top stage
and reboiler.
Top stage: total material balance
dhLtot,1
"DRD#V2!L1!V1
dt
Component material balances
dx1,j
hLtot,1 "DRDxD,j#V2y2,j!L1x1,j!V1y1,j
dt
Energy balance


NC
dHLR,j dT
hLtot,R xR,jHLR,j
x"1 dT dt
"LN stagexN stage,jHLN stage,j!VRyR,jHVR,j#QR [13]
The set of equations written for all the stages forms
a system of nonlinear differential algebraic equations,
with initial conditions given by the original charge in
the feed drum, tray hold-ups and tray composition
proRles and internal Sow rates. The vector of initial
conditions represents a pseudo steady-state solution
for an initial feed whose composition is equal to the
vapour in equilibrium with the liquid charge of the
feed vessel. The transient behaviour is obtained by the
simultaneous solution of eqns [1]}[13], which re-
quires linearization and a combination of matrix in-
version and integration techniques.
Despite the large range of computational complex-
ity, simulations of batch distillation show that, in
most cases, short-cut and rigorous models agree very
well. A distinct advantage of the simpliRed models is
they can be implemented in a spreadsheet. In these
[8] models the stage hold-up is considered negligible,
except for the feed drum (reboiler) where the
following equations hold for the total and
volatile component material balances in a column
operating at constant distillate composition and vari-
able reSux.
[9]
Total cumulative material balance
DM "VM !LM [14]



LM


NC
dHL1,j dT R
hLtot,1 x1,jHL1,j dW"!dDM " 1! dVM " 1! dVM
x"1 dT dt VM R#1

"(1!S) dVM [15]

"DRDxD,jHLD,j#V2y2,jHV2,j!L1xi,jHL1,j
Cumulative component balance
!V1y1,jHV1,j#Q1 [10]
Wixwi"Wxw#(Wi!W) xD
[16]
For the feed drum (reboiler) the following equa-
tions are modiRed. By differentiating and rearranging one gets:

Wi(xwi!xD
)
Total material balance W" [17]
(xw!xD
)

dhLtot,R Wi(xD
!xwi) dxw

"LN stage!VR [11] !dW" "dDM [18]
dt (xD
!xw)2
 II / DISTILLATION / Pilot Plant Batch Distillation 1101

If eqn [18] is substituted in [15] one gets: The remaining amount of charge in the still is then
calculated by combining the above equation with the
Wi(xD
!xwi) dxw total differential balance dW"!dDM and sub-
(1!S) dVM " [19]
(xD
!xw)2 sequently integrating the resultant expression:

The total amount of vapour produced will then be: dW dxw

" [28]
W xD
!xw

  
xwf
Wi(xD
!xwi) dxw
VM " [20]
(1!S) (xD
!xw)2

xwi xwf
Wf dxw
ln " [29]
Since the cumulative vapour produced is VM "V, Wi xwi xD
!xw

the time  necessary for a run is calculated from the

above equation as: Eqns [20]}[22] provide an efRcient way of calcu-
lating the total amounts of vapour and distillate pro-

  
xwf
Wi(xD
!xwi) dxw duced and the time necessary for the separation with-
" [21] out having to solve the system of equations comprised
xwi V (1!S) (xD
!xw)2
by eqns [1]}[13]. This calculation provides an eco-
The total amount of distillate produced can be nomic benchmark since it deRnes the optimum time
found by integration of eqn [18]: for a run, based on the recovered product value and
the operating costs. As the run time increases, the

 
D
f xwf
Wi(xD
!xwi) dxw cumulative revenues given by the total amount of
dDM " [22] recovered product multiplied by its value will Rrst
0 xwi (xD
!xw)2
increase but then approach an asymptotic value. The
Finally, rearrangement of eqn [16] yields: decreased economic beneRt results either because the
amount of distillate decreases (as in the case of con-
Wixwi!(Wi!W)xD
"Wxw [23] stant composition distillate) or because the product
stream becomes progressively less pure (as in the case
Wixwi!DM xD
"Wxw [24] of constant reSux ratio operation). The operating
costs on the other hand increase steadily with time
The above can be differentiated to give the follow- and the proRt function, which combines these two
ing result: costs, undergoes a maximum, after which the proRt
will decrease, as illustrated by Figure 2.
d(Wixwi!DM xD
)"d(Wxw) [25] An analytical solution can be found for a limiting
case which assumes an inRnite number of stages (cor-
!d(DM xD
)"d(Wxw) [26] responding to minimum reSux) and constant relative
volatility. At minimum reSux the operating line ends
!xD
dDM "xwdW#Wdxw [27] at a pinch zone and xw is located in the equilibrium

Figure 2 Optimum profit profile: operating costs versus length of time. Recovered product increases rapidly at first but then levels off.
1102 II / DISTILLATION / Pilot Plant Batch Distillation

line. The slope of the operating line is then given by: of eqn [31] into [30] to yield:

dLM xD
!yw R dLM xD
#xD
xw(!1)!xw

S" " " [30] " [36]
dVM xD
!xw R#1 dVM (xD
!xw) [1#xw(!1)]

Under the constraint of constant relative volatility Substitution of [36] into the general expression for
the equilibrium relation becomes: the vapour requirement (eqn [20]) and integration
leads to the vapour requirement equation when the
column is operated at constant distillate composition
xw and variable reSux:
yw" [31]
1!xw(#1)
Wi(xD
!xwi)
If one takes advantage of eqns [30] and [31], ana- VM "
(1!xD
) (xD
) (!1)
lytical forms for the cumulative distillate production
and remaining charge left in the feed still can be
 


xD
!xwf xwi
derived for the cases of either constant reSux ratio or ; (1!xD
) ln #
xD
!xwi xwf
constant distillate composition operation.
For constant reSux ratio, eqn [15] can be inte-



xD
!xwf 1!xwi
grated to the expression for the total vapour require- #xD
 ln [37]
ment: xD
!xwi 1!xwf

Eqns [34] and [37] were developed by Bauerle and

VM "(R#1)DM "(R#1) (Wi!Wf) [32]
Combination of eqns [15], [30] and [31] yields the
functional dependence of distillate composition,
xD
on recycle ratio, R, composition of the remaining
feed, xw in the reboiler and relative volatility, :
Sandall, assuming the ideal pinched columns operat-
ing at minimum reSux. None the less, their applica-
tion to real columns yields good approximated results
if the equipment operates in a near-pinched zone at
the bottom. This is often the case for columns with
Rve or more theoretical stages.

(R#1)xw!Rxw!Rx2w(!1)
xD
" [33] Design
1#xw(!1)
The operation of a batch distillation column, even
Substitution of eqn [33] into the general mass bal- pilot-scale equipment, is often as technically involved
ance expression, eqn [29] and subsequent integration as operation of an industrial-scale column, and the
produces an analytical form for the mass balance same amount of care in start-up and safety proced-
given by: ures should be taken. Whether designing a new col-
umn or revamping an existing one, the necessary
safety and physical properties data such as Sash and




Wf 1 1!xwi xwf
ln " ln ignition points, Sammability and toxicity must be
Wi (R#1) (!1) 1!xwf xwi compiled for each component in the mixture. Predic-
tive equations or experimental values for the vapour



1 1!xwf pressures of all components and binary equilibrium
# ln [34]
R#1 1!xwi data should be compiled together with parameters of
equations of state or activity coefRcient models when-
If the column is operated at constant distillate com- ever available. Other physical properties to be in-
position, direct integration of eqn [29] produces the cluded are liquid and vapour heat capacities, heats of
expression for the mass balance: vaporization and viscosities.
Once the physical property data bank has been put
together, preliminary design calculations can be per-


Wf xD
!xwi formed. For a multicomponent distillation column
" [35]
Wi xD
!xwf a light-key and a heavy-key component should be
chosen in order to reduce the preliminary design to
The integration of eqn [20], however, requires the a pseudo-binary system. At this point it is possible to
development of an expression for the time-dependent use graphical methods like McCabe}Thiele or even
operating line, which is accomplished by substitution something more involved like Ponchon}Savarit to
 II / DISTILLATION / Pilot Plant Batch Distillation
carry our a case study to Rnd out the system response
in terms of required number of theoretical stages for
a speciRed purity at different reSux ratios. To accom-
plish this, the optimization techniques described later
can be useful, but since they are also hard to imple-
ment. The alternative approach of using a simpliRed
calculation method such as presented in the section
on column operation might be more desirable. This
initial set (reSux ratio!number of theoretical stages)
will permit the preliminary design.
Depending on the intended purpose of the laborat-
ory-scale column, these initial calculations are likely
to be sufRcient for specifying the details of column
design. Reboiler and condenser heat loads permit
sizing of steam and condensation coils. Environ-
mental concerns have introduced complexities in de-
sign which were not previously an issue for pilot-scale
distillation. Current regulations at our site require
condensate return to steam generation facilities to
recover waste heat. Even a moderate condenser heat
load can require prohibitively large quantities of cold
tap water and the condenser heat load for even
a small distillation column will typically be substan-
tially larger than can be handled by laboratory-scale
recirculated chillers. Much of the Rnal decisions on
absolute sizing will be dependent upon available fa-
cilities and the anticipated intensity of column use. It
is important to get to a reasonably accurate prelimi-
nary design early in the design process, so that such
practical constraints can be considered.
In many situations, a laboratory-scale distillation
column will be used for multiple separations, or as
a testing ground for additional full-scale design data.
Under these circumstances, design for Sexibility is
a primary concern. Instead of focusing on detailed
physical property information, the data collection
should focus on obtaining ranges of anticipated phys-
ical properties as well as ranges in batch size. The
actual design should then reSect the appropriate
bounds of properties and separations that may be
encountered. It should be kept in mind that there is
a practical minimum volume that can be handled, due
to tray hold-up, while larger volumes can be handled
with multiple batches. Undersizing either reboiler or
condenser heat transfer capacity may render the col-
umn useless for a speciRc separation.
In most batch distillation operations, the lighter
component is the desired product and the actual col-
umn is the rectifying section of a continuous tower.
The preceding discussion in this section as well as in
the next section implicitly assume this situation.
Nevertheless, there might arise design situations
where the economical interest lies in the heavier com-
pounds. In more complex operations the designer
might even be faced with the task of devising a separ-
1103
ation sequence involving two or more columns. For
the case where it is desired to recover the heavy
component, the calculations for the number of stages
should be performed as a stripping column instead.
Since the principles and computational basis of
distillation are quite advanced, additional assump-
tions allow the derivation of simple expressions for
the distillate composition and Sow rate and the
amount of material left in the feed drum. These as-
sumptions render the evaluation of columns with
recycle amenable to straightforward solutions.
The remainder of this section includes a description
of a versatile laboratory distillation column and its
instrumentation and safety systems.
Equipment
Batch distillation equipment can be custom-made to
meet particular design speciRcations or be directly
purchased by catalogue selection if no stringent con-
struction features or materials are required. Ordering
can be greatly facilitated by a previous search of the
manufacturers or suppliers in the worldwide web.
Equipment intended to be used for research or educa-
tional purposes should be made of glass whenever
possible, given the easy observation of the internal
Sow regimes and their change with the internal Sow
rates. An existing batch glass column is described
here as an example.
Figure 3 shows a distillation column which has
a simple conceptual design but is versatile enough to
be used for research or teaching applications. The
column is atmospheric and functions as the rectifying
section of a regular distillation column. The feed
drum doubles as a kettle vessel where the feed is
vaporized by a coil heater having steam as the heating
medium. Instrumentation is reduced to the essentials:
the distillate and reSux Sow rates are controlled by
varying the rotation speed of the distillate and reSex
pumps, and the feed Sow rate can be controlled by
varying the steam pressure in the coil. The safety
system consists of a relay actuated by the occurrence
of any of the failure conditions in the column, which
are pressurization within the equipment, zero Sow of
condenser cooling water or loss of power to the
ventilation system. When any of these conditions
occurs, steam admission to the feed drum is switched
off.
The pumps are actuated in the remote mode by
a driver board that receives signals in the range be-
tween 4 and 20 mA from an analog output board
installed in a PC. The connection between the pump
driver board and the analog output board consists of
a screw terminal connector. The variation of
the output signal to the pumps is accomplished via
1104 II / DISTILLATION / Pilot Plant Batch Distillation
Figure 3 Rectifying batch distillation column. The abbreviations are as follows: A/D board"analog to digital interface board,
CW"cooling water, D. Pump"distillate pump, F. Pump"feed pump, FIC"flow indicator control, P. chiller"product chiller, P.
Tank"product tank, PI"pressure indicator, PIC"pressure indicator control, R. Disk"rupture disk, S. Valve"safety valve and
TI"temperature indicator.
a software utility provided with the board that emu-
lates a control panel, where each of the instruments
hooked to the board is assigned a channel number
displayed in the panel screen. The user varies the
pump Sow rate by changing the output current at the
computer screen. Manual local control of each pump
is also provided in case of failure of the computer-
interfaced control. The stepper motor of the steam
valve for the feed drum steam coil is also interfaced in
a comparable manner.
The column is also provided with thermocouples
for each stage, including condenser and reboiler. The
thermocouples are wired to a screw terminal connec-
tor, that provides the interface to an analog/digital
I/O board installed in the PC. When there is a signiR-
cant difference between the boiling points of the two
components of a binary system, the stage temperature
is an efRcient and straightforward way of evaluating
compositions. Under these circumstances, the real-
time composition proRle within the column can be
updated to the computer screen. Sample ports for the
liquid phase are installed in every stage, including the
condenser and reboiler to corroborate thermal
measurements under circumstances where thermal
gradients are not sufRciently steep to provide accurate
composition correlation. The composition analysis
can be performed by a variety of methods. If there is
a signiRcant density difference between components
being separated, composition can be deduced from
a density}concentration curve. Density can be deter-
mined gravimetrically, or equipment is available for
online density measurement. A particularly versatile
online implementation of density measurement is in
the condensate stream of an off-set condenser as
described in more detail below. If one of the compo-
nents is an organic acid, sample analysis can be car-
ried out either by titration or by the measurement of
any other property related to the dissociation state
(such as pH), provided the metering apparatus is
sufRciently accurate to discriminate stage-to-stage
differences. For organic mixtures, other properties
such as refractive index may also be used as analytical
method.
The feed drum in the described pilot-scale column
is a large glass bulb equipped with a steam heating
coil to vaporize the feed. A pressure relief rupture
disk provides a mechanical fail-safe against reboiler
pressurization. Another pressure gauge is installed at
the steam inlet to the coil. The steam outlet is pro-
vided with a trap to ensure the total condensation of
the steam, and therefore the use of its latent heat.
A useful energy balance is achieved by cooling the
condensate as it is discharged to the drain. The heat
load to the column can be crudely calculated by
measuring the discharge Sow rate and multiplying
this valve by the heat of vaporization of the steam at
the inlet pressure.
The distillate is collected in a separate vessel whose
volume equals approximately half that of the feed
drum. The product collection vessel is Rtted to permit
 II / DISTILLATION / Pilot Plant Batch Distillation
charging of material to the feed drum. The remaining
material in the drum after a batch processing can be
discharged by a valve in the bottom. The subsequent
batch charge can also be combined with the remain-
ing heavy ends of the previous operation.
The off-set condenser depicted in Figure 3 provides
for direct measurement of condensate Sow rate. This
eliminates the need to calculate condensate from the
condenser energy balance. This is particularly impor-
tant for pilot-scale units where complete condensa-
tion may not be achieved at high boil-up rates. The
condensed top vapours drip down and accumulate in
the bottom part of the vessel, from which they are
removed either to the distillate tank or back to the
column as reSux. A match between the condensation
rate and sum of product and liquid Sow rate returned
to the column can be assured by visual monitoring of
the condenser liquid level or computer monitoring of
the liquid head in the bottom of the condenser with
a pressure transducer. The gas entrance to the con-
denser doubles as a liquid overSow in the event of
excessive condensate accumulation.
Instrumentation and Safety Circuits
Although a batch distillation column can be run man-
ually by an attentive and experienced technician, the
dynamic nature of operation requires extensive in-
strumentation for all but the simplest mode of opera-
tion. To gain the Sexibility necessary for the opera-
tion at constant distillate composition, Sow rates of
the reSux and distillate must be independently con-
trolled. In small units devoted to research, Sow con-
trol can be easily accomplished by varying the speed
of a gear pump through a control panel displayed on
a computer screen. The connection of the pump to the
computer consists of a driver board wired to a screw
terminal connector. The latter is attached to an ana-
log output board. The output signal to the driver
board can be varied based on the results of calcu-
lations performed by external application programs.
If the top composition is to be kept constant, the
update on the reSux rate can be calculated by a user
routine with the aid of operating charts such as those
described later. The value of the reSux rate translated
to Sow rates (typically current values) controls the
distillate and reSux pumps. The product and reSux
Sow rates must be constrained to balance the rate of
condensation. This is conveniently accomplished by
monitoring the condenser level as an indicator of the
difference between boil-up and distillate and product
Sow rates. The direct use of the heat load output to
control the condenser level is not recommended due
to the large dead-time between a change in the re-
boiler conditions and the resulting effect in the liquid
1105
level. The result of the reSux calculation is stored in
a data buffer from which it can be retrieved by an-
other application and used to update the correspond-
ing analog output channel.
Analog and digital I/O boards can also be installed
to retrieve information such as cooling water temper-
ature and Sow rate, steam pressure, Sow rate to the
reboiler and the temperature proRle of the column.
The stage temperature is a direct indicator of stage
composition; however, it is only useful as a control
variable when the temperature variation between suc-
cessive stages is signiRcant. The greatest variation in
temperature during the batch run will take place in
the reboiler which is an excellent means of monitor-
ing overall progress of the separation.
Variation of the heat load to the column can be
actuated remotely by Rtting the steam valve with
a stepper motor. A variable heat load adds operation
Sexibility and can be used in conjunction with other
strategies to maximize recovery and purity of a de-
sired component at a lower energy cost. For example,
the feed drum contains the highest fraction of the
volatile component at the beginning of the
process. Therefore, the column can be started
up at a lower boil-up rate, which will be gradually
increased to match the enrichment of the charge in
the heavier component. Figure 3 shows typical
column instrumentation, and data acquisition and
control.
Three operating conditions are monitored con-
stantly and may independently activate the safety
system, stopping steam delivery to the reboiler by
closing a steam safety valve that precedes the steam
controller valve. These monitored conditions are the
column pressure, cooling water Sow and ventilation
fans. The pressure transducers, Sow transmitter and
power to the fans are set up as a logic relay where loss
of any one of the direct current voltages is sufRcient to
actuate the steam safety valve. It is important to
choose the logic of these circuits such that electrical
and mechanical failures will default to termination of
the batch run. The safety system circuitry is depicted
in Figure 4.
Column Operation
While dedicated laboratory-scale distillation can be
used for solvent recovery, experimental laboratory-
scale columns are used for collection of design data.
The operational objective is usually to maximize the
recovery of a component under the constraint of
a desired purity level. The feature of the batch process
which distinguishes it from the more familiar con-
tinuous counterpart is its inherently transient nature.
The continuously changing feed composition must be
1106 II / DISTILLATION / Pilot Plant Batch Distillation
Figure 4 Safety system circuitry.
accounted for in the calculation theory. A variety of
operation and completion criteria may be used de-
pending on the process economics, equipment charac-
teristics and product value. Several different basic
operational modes are possible:
1. total reSux, with periodic dumping of the accumu-
lated material from the condenser to the distillate
tank;
2. constant reSux, with continuous variation of the
instantaneous distillate composition, starting
above and Rnishing below the desired product
speciRcation;
3. constant composition, with variable reSux ratio in
order to keep the instantaneous distillate composi-
tion constant.
Operation strategies can take advantage of all three
of these operational modes.
Initially the column is operated at total reSux with
subsequent product collection at a purity higher than
the Rnal product speciRcation. After this initial cut is
withdrawn to the product tank, operation is switched
to the constant composition mode and the equipment
is run until the reSux ratio becomes so high that
product collection is minimal. At this point the
 II / DISTILLATION / Pilot Plant Batch Distillation
operation is switched to the constant reSux mode and
continued until the average composition of the distil-
late drops to the desired level, when the equipment is
shut down. Variations on this approach are often
required due to equipment limitations. If the
column is to be operated manually, then it might be
difRcult to maintain constant distillate composition.
Also, if the column consists of less than Rve ideal
stages, it is more advantageous to operate at constant
reSux.
Prior to the initiation of any operational procedure,
it is necessary to elaborate an operation schedule that
can be used as a guide throughout the run. Although
the procedures discussed here can be implemented for
columns with a very low level of automation, they
can also be used in application programs that run as
a part of an automated control loop. One can use
simpliRed calculation methods (e.g. McCabe}Thiele)
or resort to more extensive numerical computations if
the effect of some variables such as the hold-up is to
be taken into account. Independent of the calcu-
lational basis, better predictions of the composition
proRles can be obtained if the stage efRciencies
or at least the overall efRciency is known. EfRciencies
depend on the physical properties of the system, par-
ticularly the viscosity and the relative volatility, but
also on the geometric characteristics of the equip-
ment. Determination of the overall and stage efRcien-
cies can be easily accomplished by running the col-
umn at total reSux. When the column has reached
Figure 5 Representation of stage efficiencies.
1107
steady state, the composition proRles within the col-
umn must be determined. The overall efRciency is
easily calculated by stepping off theoretical stages in
the McCabe}Thiele diagram between the equilibrium
and the operating line, which at total reSux coincides
with y"x. The number of theoretical stages is deter-
mined when the bottom composition is crossed. The
Murphree efRciency for stage n receiving liquid from
stage n!1 and vapour from stage n#1 is deRned as:
yn!yn 1
M" eq \ [38]
yn !yn 1
\
Eqn [38] is a measure of the degree of separation
achieved in the vapour going from stage n#1 to
stage n and can be visualized in the McCabe}Thiele
diagram as a segment ratio, as shown in Figure 5. The
maximum degree of separation is represented by the
difference in the denominator, where the vapour leav-
ing the stage is in equilibrium with the liquid phase of
the same stage.
Determination of the vapour-phase composition is
often more difRcult than liquid and sample ports are
generally only provided for the liquid phase. It is
therefore useful to deRne Murphree efRciency of the
liquid compositions as:
xn#1!xn
M" [39]
xn#1!xeq
n
1108 II / DISTILLATION / Pilot Plant Batch Distillation
Once the efRciencies are determined, an operation
schedule at constant reSux should be prepared. The
operation schedule consists of a family of curves
where the reSux ratio is plotted as a function of
reboiler composition, holding the distillate composi-
tion constant as depicted in Figure 6. The curves can
easily be generated in a spreadsheet if the equilibrium
curve for the system can be regressed as an analytical
form, x"f(y). A distillate composition (xD) is Rxed as
the fulcrum, around which all the operating lines
pivot, as deRned for different reSux ratios (RD), as
shown in Figure 7. Since the number of ideal stages is
known, the bottoms composition (xB) is found for
each operating line by stepping off these stages be-
tween the equilibrium curve and the operating lines.
Once xB is found for a particular operating line, the
reSux ratio is changed, thereby deRning another oper-
ating line, and the procedure is repeated to Rnd the
corresponding xB. In this way, a set of data points
(RD, xB) corresponding to a Rxed distillate composi-
tion xD is determined. The next set is determined by
the same procedure, changing the value of xD. These
plots of RD versus xB represent the reSux ratio re-
quired to achieve a speciRed distillate composition at
a given composition within the reboiler.
The RD versus xB curves can either be used in
manual operation or integrated into an automated
control strategy, where information about the feed
drum composition is used to calculate the new re-
quired reSux ratio to keep xD constant. The choice of
xD will depend on the minimum acceptable purity for
the product. Sometimes, even when a higher purity is
desired, the operating xD may be imposed by equip-
ment restrictions. The Sow rate range of the pumps,
for example, might restrict operation to a certain
range of the reSux ratio. In this case, switching to
Figure 6 Operation schedule at constant composition. Reflux schedule (based on 45%  overall). Xb is the reboiler composition,
Xd is the distillate composition and Rd is the reflex ratio.
a lower distillate composition will allow longer runs,
thus increasing the total amount of product.
Stopping criteria for an industrial distillation is
generally dictated by economics. Operating costs ac-
cumulate continuously with time, because of energy
and labour costs, as discussed above.
Optimization Techniques
Optimization of batch distillation operations is not
addressed with the same frequency as the continuous-
case counterpart. The likely reason for the scarcity of
publications in this area lies in the transient nature of
the problem, which introduces a system of differential
equations to describe the dynamics. The optimization
problem therefore consists of a target functional (see
below) to be minimized and a set of constraints em-
bodied by the differential equations for the time-
dependent behaviour of material and energy balances
and algebraic equations for phase equilibrium and
column hydraulics plus additional constraints such as
bounds on certain variables. Optimization problems
with nonlinear algebraic model equations and con-
straints can be solved in a straightforward way by
nonlinear programming strategies. On the other
hand, unconstrained problems with differential equa-
tion models can be handled through the calculus of
variations. Models that combine both of these fea-
tures are currently optimized by imposing some level
of approximation to the problem. The problems usu-
ally reported in the literature for batch distillation can
be classiRed as:
1. Maximum distillate problem: to maximize the
amount of distillate of a speciRed purity for a spe-
ciRed time.
 II / DISTILLATION / Pilot Plant Batch Distillation
Figure 7 Construction of the operation schedule.
2. Minimum time problem: to minimize the batch
time needed to produce a prescribed amount of
distillate of a speciRed purity.
3. Maximum proRt problem: to maximize a proRt
function for a speciRed purity of distillate.
The maximum proRt problem for a column oper-
ated at constant distillate composition involves the
evaluation of the net proRt of the column along its
batch run time. The net proRt function behaviour has
already been discussed. The proRt curve displays an
extrema and the proRt optimization problem there-
fore seeks the value of the batch run time (a number)
that will maximize the net proRt function. It is amen-
able to a simple graphic solution that can be obtained
from a spreadsheet as long as a simple zero hold-up
model is employed to describe the column operation.
On the other hand, the solution of the maximum
distillate problem is given by a time-dependent func-
tion (the distillate Sow rate) that will maximize the
cumulative distillate production, a function of the
distillate Sow rate. This latter function of another
function is called a functional.
In this section some of the recently developed tech-
niques to extremize functionals are brieSy reviewed.
The objective is to offer the reader an introduction to
the theme, and provide useful references for further
information. The optimization problem belonging to
one of the above categories can be posed in terms of
an objective function subjected to constraints such as:

b
Min "(z(b), p)# G(z(t), u(t), p)dt [40]
u(t),z(t),p a unconstrained one, the augmented problem is
1109
subject to:
zR (t)"F(z(t), u(t), p) [41]
g(u(t), z(t))40 [42]
gf (z(b))40 [43]
z(a)"z0 [44]
z(t)L4z(t)4z(t)U [45]
u(t)L4u(t)4u(t)U [46]
In the above set of equations the integral part of the
objective function to be minimized can be viewed as
the total amount of distillate, withdrawn as top prod-
uct, whereas the function (z(b), p) may account for
the Rnal hold-up within the equipment, which can be
incorporated into the product at the end of a batch
run. The vector z(t) represents the state variables of
the system, such as composition, internal Sow rates
and temperatures, and the vector p represents con-
stant parameters. The vector u(t) carries the control
proRles, i.e. the variables used to manipulate eqn [40]
and achieve the minimization goal. Distillate of a spe-
ciRed purity can be maximized for instance by chang-
ing the reSux ratio. The time-dependent reSux ratio
would be then the control variable u(t) for such an
optimization problem. The constraints represented by
eqn [41] embody the material and energy balances,
which are written in their transient form for the batch
problem. Algebraic constraints included in eqn [42]
may represent the equilibrium relations and the sum-
mation of the liquid mole fractions. Inequality con-
straints with lower and upper bounds (eqns [45] and
[46]) may represent either purity requirements in the
product or physical constraints in the maximum and
minimum attainable values of the control variables.
Initial and Rnal states of the system are also written as
constraints, as represented by eqns [44] and [43] re-
spectively.
The optimal control problem posed above seeks the
time-dependent control proRle (control function) that
minimizes the objective functional (i.e. a function of
functions represented by eqn [40]). It can be solved in
a variety of ways, depending on how one chooses
to handle the differential equation constraints.
Methods based on the calculus of variations use Lag-
range multipliers and slack variables to restate the
constrained problem of eqns [40]}[46] as an uncon-
strained one. Since the minimum of the constrained
problem is equivalent to the minimum of the
1110 II / DISTILLATION / Pilot Plant Batch Distillation

represented by: Pontryagin’s maximum principle

Min 1"(z(b), p)#Tgf(z(b)) H(xH, uH , H, t)4H(xH, uH# u, H, t) [55]

u(t),z(t),p

Boundary condition


b
# [G(z(t), u(t), p)
a
( z#Tz! )T bdz(b)#( t#Tt#H) bdt(b)"0
# T(t)(F(z(t), u(t), p)!z(t)) [56]

#MT(t)(g(u(t), z(t))#s2)]dt [47] For problems with constraints in the control vari-
ables like those given by eqn [46], the stationary
By analogy to Hamilton’s equation of motion, one condition must be modiRed to include Pontryagin’s
can deRne a Hamiltonian as: maximum principle which establishes that the solu-
tion values for the constrained control variables must
H(z(t), u(t), p)"G(z(t), u(t), p)# TF(z(t), u(t), p) lie along an optimal path. That is, any variation in the
optimal control proRle uH(t) at time t, while keeping
#MT(g(u(t), z(t))#s2) [48]
the state and co-state variables z(t), (t) and M(t) at
their optimal values, will force an increase in the
And therefore the problem in [47] becomes:
value of the Hamiltonian. This replaces the uncon-
strained minimum condition of eqn [52] and is stated
Min 1"(z(b), p)#Tgf(z(b)) mathematically in eqn [55]. Also, the second term of
u(t),z(t),p
the boundary condition in eqn [56] vanishes for
Rxed-time problems.


b
# [H(z(t), u(t), p)! (t)z(t)]dt [49] Solution of the optimization problem of
a eqns [40]}[46] in its variational formulation requires
integration of two sets of differential equations given
The minimum of the latter is found in the usual by [50] and [51] to get the state variables z and
way by taking the derivatives of the augmented prob- adjoint variables for the ordinary differential equa-
lem with respect to all the independent variables (z, , tion (ODE). Since these equations are also a function
, s, u, t). Setting those to zero and taking into of the control proRle u(t), their integration is Rrst
account the conditions for a Rxed initial condition performed with guessed values of this vector.
problem (dt t0"0, dz(t0"0), one obtains the follow- Eqns [53] and [54] are used to Rnd the second set of
ing variational formulation. adjoint variables (M) and the slack variables (s2)
associated with the constraint on g(u(t), z(t)). Finally,
State equations eqn [52] or [55] provides the updated values for u(t),
the control proRle, whereas the adjoint variables for
H the boundary conditions are calculated from
zR " [50]
eqn [56]. The whole procedure involves successive
iterations of the control vector and can be computa-
Co-state equations tionally intensive, especially for problems with many
constraints.
H An alternative solution can be formulated to over-
!Q" [51]
z come the difRculty posed by the differential con-
straints. Eqns [40]}[46] are discretized using Rnite
Stationary condition elements. Within each element, function approxima-
tion is expressed in terms of orthogonal polynomials
H and the resulting problem is amenable to a mathemat-
"0 [52]
u ical treatment intended to minimization problems
involving only algebraic equations.
H Discretization of the optimal control problem leads
"0 [53]
M to the nonlinear problem model given below. This
formulation consists of the discretized objective func-
H tion of the original problem, the continuity equations
"0 [54]
s for state variables and inequality constraints in the
 II / DISTILLATION / Pilot Plant Batch Distillation 1111

original formulation. ratic programming problem as follows (see Logsdon

and Biegler, in the Further Reading section):
NE K
Equation [58] is solved in each element for the
Min "(zf, p)# wijG(zij, uij, p, i)
values of zij in the interior collocation points, starting
D
uij,zij,p, i i"1 j"1
with the Rrst element, from the initial values of the
[57]
state variables and guessed element lengths (i) and
subject to: control proRles (uij). The rightmost (exterior) collo-
cation point for the state and control proRles in eqn
irij"zR K#1ij!iF(zij, ui, p)"0 [58] [58] is calculated from the values at the interior collo-
cation points by:
g(uij, zij, i)40 [59]
k

gf(zf)40 [60] zi,k#1" zij j [67]

j"0

z10!z0"0 [61] k
ui,k#1" zijj [68]
\1 (i)"0 i"2,2, NE
zi0!ziK#1 [62] j"1

zf!zNE
K#1(NE#1)"0 [63]
zLij4zij4zUij [64]
uLij4uij4uUij [65]
NE
i"Total [66]
i"1
Problem discretization introduces the time element
lengths (i) as additional variables. Thus, variables
in [57]}[66] include: i, the Rnite element lengths for
i"1,2, NE; zf, the value of the state at the Rnal
time; zij and uij, the collocation coefRcients for the
state and control proRles where i refers to the element
and j to the collocation point within each element;
and p, any additional design parameters (such as
boil-up rate and Rnal time). In addition, wij are quad-
rature weights from the integral in [40]. Lagrange
polynomials are applied for the orthogonal collo-
cation within the Rnite elements. The order of the
collocation method should be equal to the index of
the system of the state variable differential constraints
and algebraic equations. The index is equal to the
number of times the algebraic equations must be
derived in order to recover the standard form of
a Rrst-order ODE.
The solution of eqns [57]}[66] looks for the values
of the coefRcients zij and uij of the polynomial approx-
imation for the state variables and control proRles
respectively. In addition, the discretized problem also
includes the length of the discretization interval i.
The problem variables are partitioned into a set of
state variables (zij) and optimization variables (uij and
i), which provides a solution strategy where the
state variables are calculated separately using the
state equations, whereas the control proRle and ele-
ment lengths are obtained via the solution of a quad-
where j and j in the above equation are Lagendre’s
orthogonal polynomials.
Continuity for the state variables is ensured by
eqn [62] which establishes the equality between the
state variables of the rightmost collocation point of
element i!1 and their initial value in element i. The
initial value problem presented in eqn [58] is thus
integrated element-by-element using a marching tech-
nique with collocation within each element.
After a new set of state variables is generated by the
technique described above, the control proRles (uij)
and element lengths (i) are updated using a success-
ive quadratic programming algorithm that solves the
following:
Min TZ#12T(ZTBZ) [69]
K
subject to:
g#gTZ40 [70]
In the above problem, the variables u and  were
included in the vector  and the inequality con-
straint is the same as that of the original problem
formulation. (ZTBZ is the Hessian matrix of the ob-
jective function  and it is also updated during the
quadratic programming step using the BFGS
(BroydendFletcherdGoldfarbdShanno) formula). The
reduced gradients for the objective and constraint
functions appearing in eqns [69] and [70] above are
calculated in the iteration t during the integration step
according to the formula:
zf,k#1  zc,k#1 gn
ZTj" ZTgnj"
j zf j zc
[71]
1112 II / DISTILLATION / Pilot Plant Batch Distillation

In the equation above, the partial derivatives of the Subscripts

state variables at the rightmost exterior collocation i, j" tray number, component
point (zk#1), in relation to the optimized vector , is i" initial
calculated, via chainruling, by the formula: f" Rnal
R" reboiler
zi,k#1 zi,k#1 zi 1,k#1 zj,k#1 w" material in the still
" \ 2 [72] tot" total
j zi 1,k#1 zi 2,k#1 j
\ \
Section 5 ^ Optimization Techniques
The new set of control variables and element
Variables
lengths calculated in the optimization step replaces
a" initial condition for the optimization problem
the old one and the integration step is performed once
b" Rnal condition for the optimization problem
again. The Kuhn}Tucker conditions, which deter-
G" component of the objective function due to
mine the attainment of constrained minimum, are
the integral state
then checked and the calculations are stopped if these
H" Hamiltonian function
conditions have been reached. Otherwise, the optim-
M" Lagrange multipliers for the algebriac in-
ization step is performed again and the whole proced-
equality constraints
ure is repeated.
p" vector of design parameters
The preceding development typiRes the complexity
t" time
involved in rigorous optimization of batch distilla-
s2" vector of slack variables for the inequality
tion. The gains of reduced costs or shorter process
constraints
times that can be achieved by such an optimization
u" vector of control variables
would not probably be worth the effort for routine
z" vector of state variables
operation. None the less, it is possible to utilize soph-
isticated laboratory-scale distillation units to test al-
Superscripts
ternative control strategies indicated by computa-
U, L"upper and lower limits of the constrained
tional approaches.
variables

Subscripts
List of Variables z" derivative with respect to z
b" evaluate at point b
Section 2 ^ Theory

Variables Greek alphabet

h" tray hold-up " vector of time-Rnite element lengths
H" molar enthalpy " Lagrange polynomial approximation for the
DM " cumulative distillate production, moles control variables
L" liquid molar internal Sow rate " Lagrange multipliers for the differential
LM " cumulative amount of reSux, moles equality constraints
MW"molecular weight " Lagrange multipliers for the inequality con-
Q" heat load straints at Rnal conditions
R" reSux ratio " Lagrange polynomial approximation for the
S" R/(R#1) state variables
T" temperature " objective function
t" time " term of the objective function evaluated at
V" vapour molar internal Sow rate Rnal conditions
v" liquid molar volume
Vol" volume of the stage See also: II/Distillation: Historical Development; Instru-
mentation and Control Systems; Theory of Distillation.
VM " cumulative vapour production, moles
W" moles of material left in the still
x" liquid molar fraction Further Reading
y" vapour molar fraction
Al-Tuwaim MS and Luyben WL (1991) Multicomponent
batch distillation. 3. Shortcut design of batch distillation
Superscripts columns. Industrial and Engineering Chemistry Re-
L, V"liquid and vapour phases search 30: 507.

Bauerle GI and Sandall OC (1987) Batch distillation of
binary mixtures at minimum reSux. AIChE Journal 33:
1034.
Block B (1961) Batch distillation of binary mixtures pro-
vides versatile process operations. Chemical Engineering
68: 87.
Chiotti OJ and Iribarren OA (1991) SimpliRed models for
binary batch distillation. Computers and Chemical En-
gineering 15: 1.
Diwekar UM and Madhavan KP (1991) Batch-Dist: a com-
prehensive package for simulation, design, optimization
and optimal control of multicomponent, multifraction
batch distillation columns. Computers and Chemical
Engineering 15: 833.
Diwekar UM (1992) UniRed approach to solving optimal
design-control problems in batch distillation. AIChE
Journal 38: 1551.
Kumana JD (1990) Run batch distillation processes with
spreadsheet software. Chemical Engineering Progress
6: 53.
Lewis F (1986) Optimal Control. New York: John
Wiley.
Sublimation
J. D. Green, BP Amoco Chemicals, Hull, UK
This article is reproduced from Encyclopedia of Analyti-
cal Science Copyright ^ 1995 Academic Press
Introduction
Sublimation is not a procedure that is generally re-
garded as an analytical technique. It is a process,
however, by which compounds can be puriRed or
mixtures separated and as such can be of value as
a single step or as an integral part of a more complex
analytical method. It is applicable to a range of solids
of inorganic or organic origin in a variety of different
matrices and can be particularly useful when heat-
labile materials are involved.
As a method of sample puriRcation sublimation has
been used to produce high-purity materials as analyti-
cal standards. A speciRc and common example of
sublimation used as a means of puriRcation is the
removal of water from heat-labile materials in the
process known as freeze-drying. The technique is
described more fully below.
As a separation technique fractional sublimation
has been used either to purify samples for analysis by
removing undesirable components of the matrix or to
remove the analyte from the matrix for subsequent
analysis.
II / DISTILLATION / Sublimation 1113
Logsdon JS and Biegler LT (1990) On the simultaneous
optimal design and operation of batch distillation
columns. Transactions of the Institution of Chemical
Engineers, Part A 68: 434.
Logsdon JS and Biegler LT (1992) Decomposition strat-
egies for large-scale dynamic optimization problems.
Industrial and Engineering Chemistry Research 32:
692.
Logsdon JS and Biegler LT (1993) Accurate determination
of optimal reSux policies for the maximum distillate
problem in batch distillation. Chemical Engineering
Science 47: 851.
McCabe WL, Smith JC and Harriott P (1993) Unit Opera-
tions of Chemical Engineering, 5th edn. New York:
McGraw-Hill.
McCausland I (1985) Introduction to Optimal Control.
Malabar, FL: Robert Krieger.
Rao S (1996) Engineering Optimization. New York: John
Wiley.
Van Dongen DB and Doherty MF (1985) On the dynamics
of distillation processes d VI. Batch distillation. Chem-
ical Engineering Science 40: 2087.
Principles
Sublimation is the direct conversion of a solid to a gas
or vapour:
solid#heatNgas or vapour (heat"Hsubl)
The heat supplied in this endothermic process is
termed the heat of sublimation (Hsubl). The condi-
tions under which sublimation occurs may be pre-
dicted for a given substance from its phase diagram,
but in practice it is more common to use typical
experimental parameters to determine the optimized
procedure.
The heat of sublimation is a crucial parameter in
deciding upon the applicability of sublimation to
a particular substance, or indeed on the possibility of
separating two components in a mixture.
An empirical approach to determining the appro-
priate temperature and pressure for sublimation can
be used based upon previously determined data. The
temperature (T, 3K) and pressure (P) of sublimation
can be related by an expression of the form:
log10P (mmHg)"A!(B/T )
in which the constants A and B for compounds
of interest are available from published tables. The
1114 II / DISTILLATION / Sublimation

result of the sublimation process can be seen in

a freezer where ice sublimes and resolidiRes as crys-
tals. Iodine is a common substance that sublimes at
room temperature and pressure; the result of this can
be observed in a reagent bottle of the element.
The theory and mechanism of sublimation is of less
practical importance to analytical procedures than it
is in some other specialized areas of chemical science.
Knowledge of sublimation characteristics can aid im-
provements in the stability of materials used at high
temperatures or low pressures. For analytical pur-
poses it should be sufRcient to recognize that rates of
sublimation depend upon the topology of the vaporiz-
ing surface (dislocations, atomic steps and ledge con-
centrations) and upon any atomic rearrangement that
occurs during the sublimation process.
Experimentally, to effect sublimation, a number of
criteria need to be satisRed. Firstly, the sample in
question must be maintained at a temperature that
ensures a sufRciently high vapour pressure for subli-
mation to occur, whilst remaining below that point at
which the material either decomposes or melts. Sec-
ondly, a secondary surface must be available on
which the sublimed vapour can condense or solidify.
A number of experimental arrangements have been Figure 1 Simple apparatus for demonstrating the principles of
used that allow these criteria to be established; these sublimation. S, sample; P, sublimate (product); FP, perforated
filter paper.
are described in a later section of this article. It is also
possible to enhance the sublimation process by
changing the physical parameters under which the
process is carried out: surface and an appropriately placed sealing ring im-
proves the performance (Figure 2). Coils through
E The sample may be heated in order to increase its which coolant is circulated can promote the sublima-
vapour pressure. tion process.
E The application of a vacuum to the apparatus en- An early form of sublimation apparatus of which
courages vaporization and enhances the sublima- the above arrangements are derivatives (Figure 3)
tion.
E Selectively cooling part of the apparatus increases
the efRciency of the condensation process.
E Using an entraining gas can improve the mass
transport in the system and thereby increase the
overall efRciency of the sublimation process.

Apparatus
The simplest form of sublimation apparatus consists
of a beaker or porcelain dish on top of which is placed
an upturned watch-glass. The beaker contains the
solid to be sublimed and the underside of the watch-
glass provides the surface upon which the sublimed
components condense (Figure 1). A perforated Rlter
paper is commonly placed between the beaker and
Figure 2 Apparatus for simple sublimation at atmospheric
the watch-glass to prevent sublimate falling back into
pressures. Watch-glass, W, with sample, S, surmounted by filter
the sample. funnel, FF, with cooling coils, C, glass wool, G, and collected
A variant upon the above system uses an upturned sublimed product, P. A sealing ring, R, is included between the
funnel instead of a watch-glass as the condensing watch-glass and filter funnel.
 II / DISTILLATION / Sublimation 1115

materials such as enzymes and the process has been

Figure 3 Early form of sublimation apparatus. The heated cru-
cible, CR, rests in the cooling device. The sublimate product, P,
collects on the underside of the watch-glass, W. S, sample.
termed lyophilization.
Sublimation of metallic elements from rock or ore
samples requires high temperatures. The equipment
used is based upon silica furnace tubes in order to
withstand the necessary conditions. The silica tube is
heated in a furnace and the sublimate condenses
either on a cool part of the tube or on a cooled surface
immediately after leaving the tube.
The conditions of sublimation must be chosen ac-
cording to the requirements of the application. For
simple puriRcation the sample temperature is raised
slowly, under reduced pressure if necessary, until
sublimate is observed on the condensing surface.
These established conditions should then be main-
tained until no further sublimation appears to be
occurring, at which point the sample temperature can
be raised again if other components of the sample can
be further removed. At any point in this cycle the
apparatus can be dismantled and the sublimate re-
moved. This process allows selective separation or
fractional sublimation to be carried out.

included a means of cooling the surface on which

the sublimate condenses. Cooling can be achieved
Applications
in a number of ways, for example, using Rlter Sublimation is applicable to a wide range of organic
papers moistened with cool water, or in the case and inorganic compounds in an equally wide range
of the upturned funnel, a suitably shaped coil of
circulating coolant liquid can be placed around the
surface.
Sublimation under reduced pressure uses a modi-
Red form of apparatus in which a sealed enclosure
allows a vacuum to be applied. The cooled surface is
orientated with respect to the sample so as to maxi-
mize the condensation once sublimation has occurred
(Figure 4). Reducing the distance that the sublimed
substance(s) must travel is beneRcial provided the
necessary temperature gradient between sample and
condensing surface can be maintained. An alternative
arrangement for sublimation applications is shown in
Figure 5.
Freeze-drying, a special application of the sublima-
tion principle, uses apparatus of a different kind (Fig-
ure 6). The sample is dispersed around the walls of
a round-bottomed Sask whilst it is frozen by immer-
sion in a suitable freezing mixture, for example dry
ice}acetone. The Sask is then attached to the evacuat-
ing system which usually comprises an oil vacuum
pump protected from the ice sublimate by a train of
condenser traps. Over a period of typically several
hours the ice sublimes from the sample and condenses
in the traps. Air is then admitted to the apparatus and
the dried sample can be removed whilst the sublimed Figure 4 Apparatus for sublimation at reduced pressure. Cool-
ice is drained off as water through the drain tap. ant is circulated through the cold finger, CF, whilst a vacuum is
Frequently this system is used to dry heat-sensitive applied to the sample chamber.
1116 II / DISTILLATION / Sublimation

mode } the most easily sublimed compounds are

deposited Rrst whilst those requiring a higher temper-
ature are deposited later. The TLC plate is then de-
veloped in the normal way to give what has been
termed a ‘thermofractogram’ in which the substances
are separated as a function of their heats of sublima-
tion along one axis and as a function of their
chromatographic characteristics along the other axis.
This approach has been applied to a wide range of
substances including pharmaceutical preparations,
plant components and foodstuffs.
Only a very limited number of standard methods
have been reported in which sublimation is an impor-
tant aspect. These comprise an ASTM standard for
measurement of sublimation from thermionic emitters,
and two standards from Germany and Japan testing
the stability of dyes and printing inks to sublimation.
The Rrst covers the determination of the quantity, rate,
and identity of sublimed, evaporated, or sputtered
materials, whilst the latter two are concerned with
textile materials and semi-manufactured products.
Sublimation is often the mechanism by which pre-
concentration of an analyte is effected, although this
fact is frequently not appreciated. Dynamic head-
space concentration from solid samples such as plant
materials, foodstuffs or polymeric materials
occurs by sublimation of the volatile components.
Indeed, given appropriate apparatus that can be
operated at different temperatures for dynamic

Figure 5 Improved sublimation apparatus proposed by Eisen-

braun et al. (1978). Sample, S, sublimes from the lower to the
upper chamber where condensation takes place on the cooled
surface. A vacuum is applied to the apparatus and cooling coils,
C, improve the condensation process. A specially formed spatula,
SP, can be used to help remove the sublimate after the upper part
of the apparatus is removed.

of different matrices. Sublimable substances include

ice, iodine, arsenic(III) oxide, cadmium sulRde, am-
monium chloride and a large number of organic com-
pounds. Common matrices from which substances
are sublimed include biological Suids, plant mater-
ials, carbonaceous materials, samples of crude or-
ganic solids and samples of rocks and ores.
Sublimation as a method of applying substances to
thin-layer chromatography (TLC) plates involves the Figure 6 Typical freeze-drying apparatus. The frozen sample,
sample being sublimed and the vapour produced be- S, is attached to the condenser assembly and a vacuum is
applied. A cold trap, CT, protects the pump as ice sublimes from
ing directed by means of a drawn capillary onto the
the sample and subsequently condenses in the refrigerant con-
surface of a TLC plate which is slowly moved in one densers, RC. On completion of the drying the sample is removed
dimension. This results in the sublimed materials be- and the collected ice melts and drains from the system through
ing deposited upon the TLC plate in a differential the stopcock, SC.

headspace concentration, the heat of sublimation can
be determined for various compounds.
Derivatization procedures carried out on crude
samples can produce materials with improved subli-
mation characteristics. This technique has been used
to produce volatile compounds of lanthanides and
actinides which have then been sublimed prior to
analytical determinations. Derivatives have been
made using
-diketones (hexaSuoroacetylacetone or
acetylacetone), benzoyltriSuoroacetone and thenoyl-
triSuoroacetonates.
Low-temperature sublimation, which in some cir-
cumstances is termed freeze-drying, has been used to
separate water, as ice, from biological Suids such as
serum, urine or saliva. The technique has been parti-
cularly useful in paediatric cases where sample volumes
are extremely low. Determinations have then been ac-
complished using IR spectroscopy or mass spectro-
metry. Preparation of physiological samples for deter-
mination of deuterium oxide has included sublimation
techniques prior to spectrophotometric determinations.
Low-temperature sublimation has been used to pre-
pare samples for cryo-scanning electron microscopy
(SEM) analysis in order to examine herbicide particles
in a water suspension. The sublimation of herbicide-
containing frozen water droplets provides a suitable
etching of the surface for the SEM technique.
High-temperature sublimations are often the
methods of choice in sample preparations from min-
eral ores, particularly in the case of trace enrichment
of noble metals and the actinides and lanthanides
prior to activation methods. Temperatures of 800}
12003C are typical. The procedure is carried out in
silica tubes with entrainment gases, for example air or
argon, being used to increase the sublimation process.
Polycyclic aromatic compounds have been separ-
ated using sublimation techniques from a variety of
samples including coal, solids derived from oil, coal
and petroleum processing, and residues (soots) result-
ing from the use of such fossil fuels.
Theory of Distillation
I. J. Halvorsen and S. Skogestad, Norwegian
University of Science and Technology (NTNU),
Trondheim, Norway
Copyright ^ 2000 Academic Press
Introduction
Distillation is a very old separation technology for
separating liquid mixtures that can be traced back to
II / DISTILLATION / Theory of Distillation 1117
A variety of miscellaneous applications have
been developed for separation from difRcult matrices
and puriRcation of speciRc materials. These include:
E Mercury separated from impurities by conversion
to its iodide followed by sublimation.
E Isolation of proazulene and chamomile from the
Sower heads of plants.
E Isolation of aroma compounds from wheat and rye
samples prior to determination using isotope dilu-
tion methods.
E Determination of tin in casserite.
E Selective sublimation of molybdenum and tung-
sten.
Sublimation is used in some procedures for the
preparation of samples for SEM in which gold is
sublimed in vacuum from a heated tungsten Rlament
to the sample being examined.
See also: II/Distillation: Freeze-Drying.
Further Reading
Chickos JS (1987) Heats of sublimation. In: Liebman JF
and Greenberg A (eds) Molecular Structure and Ener-
getics, vol. 2. Weinheim, Germany: VCH Publishers.
Davies M (compiler) (1991}1992) Sublimation pressure for
organic compounds. In: Lide DR (ed.) CRC Handbook
of Physics and Chemistry, 72nd edn, pp. 5}91. Boca
Raton: CRC Press.
Eisenbraun EJ, Moyer CJ and Vuppalapaty P (1978) Subli-
mation apparatus: design improvement. Chemistry and
Industry 229}230.
Holden CA and Bryant HS (1969) PuriRcation by sublima-
tion. Separation Science 4(1): 1}13.
Somorjai GA (1968) Mechanism of sublimation. Science
162: 755}760.
Stahl E (1976) Advances in the Reld of thermal procedures
in direct combination with thin-layer chromatography.
Accounts of Chemical Research 9: 75}80.
the chemists in Alexandria in the Rrst century AD.
Today distillation is the most important industrial
separation technology. It is particularly well suited
for high purity separations since any degree of
separation can be obtained with a Rxed energy con-
sumption by increasing the number of equilibrium
stages.
To describe the degree of separation between two
components in a column or in a column section, we
1118 II / DISTILLATION / Theory of Distillation
introduce the separation factor:
(xL /xH)T
S" [1]
(xL /xH)B
where x denotes mole fraction of a component, sub-
script L denotes light component, H heavy component,
T denotes the top of the section, and B the bottom.
It is relatively straightforward to derive models of
distillation columns based on almost any degree
of detail, and also to use such models to simulate
the behaviour on a computer. However, such simula-
tions may be time-consuming and often provide
limited insight. The objective of this article is to
provide analytical expressions that are useful for
understanding the fundamentals of distillation and
which may be used to guide and check more detailed
simulations:
1. minimum energy requirement and corresponding
internal Sow requirements
2. minimum number of stages
3. simple expressions for the separation factor
The derivation of analytical expressions requires the
assumptions of:
1. equilibrium stages
2. constant relative volatility
3. constant molar Sows
These assumptions may seem restrictive, but they are
actually satisRed for many real systems, and in any
case the resulting expressions yield invaluable in-
sights, also for systems where the approximations do
not hold.
Figure 1 Equilibrium-stage concept.
Fundamentals
The Equilibrium-Stage Concept
The equilibrium (theoretical)-stage concept (Fig-
ure 1) is central in distillation. Here we assume va-
pour}liquid equilibrium (VLE) on each stage and that
the liquid is sent to the stage below and the vapour to
the stage above. For some trayed columns this may be
a reasonable description of the actual physics, but it is
certainly not for a packed column. Nevertheless, it is
established that calculations based on the equilib-
rium-stage concept (with the number of stages ad-
justed appropriately) Rts data from most real columns
very well, even packed columns.
One many reRne the equilibrium stage concept, e.g.
by introducing back-mixing or a Murphee efRciency
factor for the equilibrium, but these ‘Rxes’ often have
relatively little theoretical justiRcation, and are not
used in this article.
For practical calculations, the critical step is usu-
ally not the modelling of the stages, but to obtain
a good description of the VLE. In this area there has
been signiRcant advances in the last 25 years, espe-
cially after the introduction of equations of state for
VLE prediction. However, here we will use simpler
VLE models (constant relative volatility) which apply
to relatively ideal mixtures.
Vapour+Liquid Equilibrium
In a two-phase system (PH"2) with Nc nonreacting
components, the state is completely determined by
Nc degrees of freedom ( f ), according to Gibb’s phase
rule:
f"Nc#2!PH [2]

If the pressure (P) and Nc!1 liquid compositions or
mole fractions (x) are used as degree of freedom, then
the mole fractions (y) in the vapour phase and the
temperature (T) are determined, provided that two
phases are present. The general VLE relation can then
be written:
[y1, y22 yNc 1, T]"f (P, x1, x22 xNc 1)
\ \
[3]
[y, T]"f (P, x)
Here we have introduced the mole fractions x and y in
the liquid and vapour phases respectively, and we
trivially have ni"1 xi"1 and ni"1 yi"1.
In ideal mixtures, the VLE can be derived from
Raoult’s law which states that the partial pressure
pi of a component (i) in the vapour phase is propor-
tional to the vapour pressure (poi) of the pure compon-
ent (which is a function of temperature only:
poi"poi(T)) and the liquid mole fraction (xi):
pi"xi poi(T) [4]
For an ideal gas, according to Dalton’s law, the par-
tial pressure of a component is proportional to the
mole fraction pi"yiP and since the total pressure
P"p1#p2#2#pNc" i pi" i xi poi(T) we derive:
poi xi poi(T)
yi"xi " [5]
P xi poi(T)
i (light component) and x2"1!x (heavy compon-
The following empirical formula is frequently used to
compute the pure component vapour pressure:
b
ln po(T)+a# #d ln (T)#eT f [6]
c#T
The coefRcients are listed in component property
databases. The case with d"e"0 is called the
Antoine equation.
K-values and Relative Volatility
The K-value for a component i is deRned as:
Ki"yi/xi. The K-value is sometimes called the equi-
librium constant, but this is misleading as it depends
strongly on temperature and pressure (or composi-
tion).
The relative volatility between components i and
j is deRned as:
(yi/xi) Ki
ij" " [7]
(yj/xj) Kj
II / DISTILLATION / Theory of Distillation 1119
For ideal mixtures that satisfy Raoult’s law we have:
(yi/xi) Ki poi(T)
ij" " " [8]
(yj/xj) Kj poj(T)
Here poi(T) depends on temperature so the K-values
will actually be constant only close to the column
ends where the temperature is relatively constant. On
the other hand, the ratio poi(T)/poj(T) is much less
dependent on temperature, which makes the relative
volatility very attractive for computations. For ideal
mixtures, a geometric average of the relative volatil-
ities for the highest and lowest temperature in the
column usually gives sufRcient accuracy in the com-
putations: ij"(ij,top ) ij,bottom.
We usually select a common reference component
r (usually the least volatile or heavy component), and
deRne:
i"ir"poi(T)/por(T) [9]
The VLE relationship (eqn [5]) then becomes:
ixi
yi" [10]
ixi
i
For a binary mixture we usually omit the component
index for the light component, i.e. we write x"x1
ent). Then the VLE relationship becomes:
x
y" [11]
1#(!1)x
This equilibrium curve is illustrated in Figure 2.
The difference y!x determines the amount of
separation that can be achieved on a stage. Large
relative volatilities imply large differences in boiling
points and easy separation. Close boiling points im-
ply relative volatility closer to unity, as shown below
quantitatively.
Estimating the Relative Volatility from
Boiling Point Data
The Clapeyron equation relates the vapour pressure
temperature dependency to the speciRc heat of vapor-
ization (Hvap) and volume change between liquid
and vapour phase (Vvap):
dpo(T) Hvap(T)
" [12]
dT TVvap(T)
1120 II / DISTILLATION / Theory of Distillation

Figure 2 VLE for ideal binary mixture: y"x /1#(!1)x.

If we assume an ideal gas phase, and that the gas Here we may use the geometric average also for the
volume is much larger than the liquid volume, then heat of vaporization:
Vvap+RT/P, and integration of Clapeyron’s equa-
tion from temperature Tbi (boiling point at pressure HM vap"(Hvap
i (Tbi) ) Hj (Tbj)
vap

Pref) to temperature T (at pressure poi) gives, when

Hvap
i is assumed constant:
This results in a rough estimate of the relative volatil-
ity ij, based on the boiling points only:


H vap
i
!

 
H vap
1 R
ln poi+
i
#ln Pref # [13] HM vap
R Tbi T ij+e@(Tbj\Tbi)/TM b where
" [16]
RTM B
This gives us the Antoine coefRcients:
ai"(Hvap /R) "dH If we do not know HM vap, a typical value
+13 can
i /R, ci"0.
vap
i (1/T bi)#ln P ref, bi
In most cases Pref"1 atm. For an ideal mixture that be used for many cases.
satisRes Raoult’s law we have ij"poi(T)/poj(T) and
we derive: Example For methanol (L) and n-propanol (H),
we have TBL"337.8 K and TBH"370.4 K and
Hvap
i 1 Hvap
j 1 Hvap
j !Hi
vap the heats of vaporization at their boiling
ln ij" ! # [14] points are 35.3 and 41.8 kJ mol\1 respectively.
R Tbi R Tbj RT
Thus TM B"(337.8 ) 370.4"354 K and HM vap"
We see that the temperature dependency of the rela- (35.3 ) 41.8"38.4. This gives
"HM vap/RTM B"
tive volatility arises from different speciRc heats of 38.4/(8.83 ) 354)"13.1 and +e13.1  32.6/354+3.34
i +Hj ), the
vaporization. For similar values (Hvap vap which is a bit lower than the experimental value.
expression simpliRes to:
HM vap Tbj!Tbi Material Balance on a Distillation Stage
ln ij+ where TM b"(TbiTbj [15]
RTM b TM b
GHI Based on the equilibrium-stage concept, a distillation

column section is modelled as shown in Figure 3.

Figure 3 Distillation column section modelled as a set of con-
nected equilibrium stages.
Note that we choose to number the stages starting
from the bottom of the column. We denote Ln and
Vn as the total liquid and vapour molar Sow rates
leaving stage n (and entering stages n!1 and n#1,
respectively). We assume perfect mixing in both
phases inside a stage. The mole fraction of species i in
the vapour leaving the stage with Vn is yi,n, and the
mole fraction in Ln is xi,n.
The material balance for component i at stage
n then becomes (in mol s\1):
dNi,n
"(Ln#1xi,n#1!Vnyi,n)!(Lnxi,n!Vn 1yi,n 1)
dt \ \
where Ni,n is the number of moles of component i on
stage n. In the following we will consider steady-state
operation, i.e: dNi,n/dt"0.
It is convenient to deRne the net material Sow
(wi) of component i upwards from stage n to n#1
(mol s\1):
wi,n"Vn yi,n!Ln#1 xi,n#1
At steady state, this net Sow has to be the same
through all stages in a column section, i.e. wi,n"
wi,n#1"wi.
The material Sow equation is usually rewritten to
relate the vapour composition (yn) on one stage to the
II / DISTILLATION / Theory of Distillation 1121
liquid composition on the stage above (xn#1):
Ln#1 1
yi,n" xi,n#1# wi [19]
Vn Vn
The resulting curve is known as the operating line.
Combined with the VLE relationship (equilibrium
line), this enables us to compute all the stage com-
positions when we known the Sows in the system.
This is illustrated in Figure 4, and forms the basis of
McCabe}Thiele approach.
Assumption about Constant Molar Flows
In a column section, we may very often use the as-
sumption about constant molar Sows. That is, we
assume Ln"Ln#1"L (mol s\1) and Vn 1"Vn"
\
V (mol s\1). This assumption is reasonable for
ideal mixtures when the components have similar
molar heats of vaporization. An important
implication is that the operating line is then a straight
line for a given section, i.e. yi,n"(L/V)xi,n#1#wi/V.
This makes computations much simpler since the
internal Sows (L and V) do not depend on composi-
tions.
The Continuous Distillation Column
We study here the simple two-product continuous
distillation column in Figure 5. We will Rrst limit
ourselves to a binary feed mixture, and the compon-
ent index is omitted, so the mole fractions (x, y, z)
refer to the light component. The column has N equi-
librium stages, with the reboiler as stage number 1.
The feed with total molar Sow rate F (mol s\1) and
mole fraction z enters at stage NF.
The section above the feed stage is denoted the
[17] rectifying section, or just the top section, and the
most volatile component is enriched upwards to-
wards the distillate product outlet (D). The strip-
ping section, or the bottom section, is below the
feed, in which the least volatile component is
enriched towards the bottoms product outlet (B).
(The least volatile component is stripped out.) Heat is
supplied in the reboiler and removed in the conden-
ser, and we do not consider any heat loss along the
column.
[18]
The feed liquid fraction q describes the changes in
liquid and vapour Sow rates at the feed stage:
LF"qF
[20]
VF"(1!q) F
1122 II / DISTILLATION / Theory of Distillation

Figure 4 Combining the VLE and the operating line to compute mole fractions in a section of equilibrium stages.

The liquid fraction is related to the feed enthalpy (hF) does not apply since it relates the thermodynamic
as follows: degree of freedom inside a single equilibrium stage.


This implies that if we know, for example, the
'1 Subcooled liquid reSux (LT) and vapour (VB) Sow rate into the column,
"1 Saturated liquid all states on all stages and in both products are com-
hV,sat!hF pletely determined.
q" " 0(q(1 Liquid and vapour
Hvap
"0 Saturated vapour External and Internal Flows

(0 Superheated vapour The overall mass balance and component mass bal-
ance is given by:
[21]

When we assume constant molar Sows in each sec- F"D#B

tion, we get the following relationships for the Sows:
VT"VB#(1!q)F
LB"LT#qF
[22]
D"VT!LT
B"LB!VB
Degrees of Freedom in Operation of a Distillation
Column
With a given feed (F, z and q), and column pressure
(P), we have only 2 degrees of freedom in operation of
the two-product column in Figure 5, independent of
the number of components in the feed. This may be
a bit confusing if we think about degrees of freedom
as in Gibb’s phase rule, but in this context Gibb’s rule
[23]
Fz"DxD#BxB
Here z is the mole fraction of light component in the
feed, and xD and xB are the product compositions. For
sharp splits with xD+1 and xB+0 we then have that
D"zF. In words, we must adjust the product split
D/F such that the distillate Sow equals the amount of
light component in the feed. Any deviation from this
value will result in large changes in product composi-
tion. This is a very important insight for practical
operation.
Example Consider a column with z"0.5, xD"
0.99, xB"0.01 (all these refer to the mole fraction of
light component) and D/F"B/F"0.5. To simplify
the discussion set F"1 (mol s\1). Now consider
a 20% increase in the distillate D from 0.51 to

Figure 5 An ordinary continuous two-product distillation
column.
0.6 (mol s\1). This will have a drastic effect on com-
position. Since the total amount of light component
available in the feed is z"0.5 (mol s\1), at least 0.1
(mol s\1) of the distillate must now be heavy com-
ponent, so the amount mole faction of light compon-
ent is now at its best 0.5/0.6"0.833. In other words,
the amount of heavy component in the distillate will
increase at least by a factor of 16.7 (from 1% to
16.7%).
Thus, we generally have that a change in external
Sows (D/F and B/F) has a large effect on composi-
tion, at least for sharp splits, because any signiRcant
deviation in D/F from z implies large changes in
composition. On the other hand, the effects of
changes in the internal Sows (L and V) are much
smaller.
McCabe+Thiele Diagram (Constant Molar Flows,
but any VLE)
The McCabe}Thiele diagram where y is plotted as
a function x along the column provides an insightful
graphical solution to the combined mass balance (op-
eration line) and VLE (equilibrium line) equations. It
is mainly used for binary mixtures. It is often used to
II / DISTILLATION / Theory of Distillation 1123
Rnd the number of theoretical stages for mixtures
with constant molar Sows. The equilibrium relation-
ship yn"f (xn) (y as a function of x at the stages) may
be nonideal. With constant molar Sow, L and V are
constant within each section and the operating lines
(y as a function of x between the stages) are straight.
In the top section the net transport of light compon-
ent w"xDD. Inserted into the material balance
eqn [19] we obtain the operating line for the top
section, and we have a similar expression for the
bottom section:


L
Top: yn" (xn#1!xD)#xD
V T
[24]


L
Bottom: yn" (xn#1!xB)#xB
V B
A typical McCabe}Thiele diagram is shown in
Figure 6.
The optimal feed stage is at the intersection of the
two operating lines and the feed-stage composition
(xF, yF) is then equal to the composition of the Sashed
feed mixture. We have that z"qxF#(1!q)yF. For
q"1 (liquid feed) we Rnd xF"z and for q"0 (va-
pour feed) we Rnd yF"z (in the other cases we must
solve the equation together with the VLE). The pinch,
which occurs at one side of the feed stage if the feed is
not optimally located, is easily understood from the
McCabe}Thiele diagram, as shown in Figure 8 (see
below).
Typical Column Pro\les + Pinch
An example of a column composition proRle is shown
in Figure 7 for a column with z"0.5, "1.5,
N"40, NF"21 (counted from the bottom, includ-
ing the reboiler), yD"0.90, xB"0.002. This is a case
where the feed stage is not optimally located, as seen
from the presence of a pinch zone (a zone of constant
composition) above the stage. The corresponding
McCabe}Thiele diagram is shown in Figure 8. We see
that the feed stage is not located at the intersection of
the two operating lines, and that there is a pinch zone
above the feed, but not below.
Simple Design Equations
Minimum Number of Stages + In\nite Energy
The minimum number of stages for a given separ-
ation (or equivalently, the maximum separation for
a given number of stages) is obtained with inRnite
internal Sows (inRnite energy) per unit feed. (This
always holds for single-feed columns and ideal
1124 II / DISTILLATION / Theory of Distillation

Figure 6 McCabeIThiele diagram with an optimally located feed.

mixtures, but may not hold, for example, for extrac- we have:
tive distillation with two feed streams.)
 
yL,n xL,n xL,n#1 xL,n
With inRnite internal Sows (total reSux) Ln/F"R " " [25]
yH,n xH,n xH,n#1 xH,n
and Vn/F" R, a material balance across any part of
the column gives Vn"Ln#1, and similarly a material
balance for any component gives Vnyn"Ln#1xn#1. For a column or column section with N stages, re-
Thus, yn"xn#1, and with constant relative volatility peated use of this relation gives directly Fenske’s

Figure 7 Composition profile (xL, xH ) for case with nonoptimal feed location. Continuous line, light key; dashed line, heavy key.

Figure 8 McCabe-Thiele diagram for the same example as shown in Figure 7. Observe that the feed stage location is not optimal.
formula for the overall separation factor:

 

xL xL
S" "N
xH T xH B
For a column with a given separation, this yields
Fenske’s formula for the minimum number of stages:
ln S
Nmin"
ln 
These Fenske expressions do not assume constant
molar Sows and apply to the separation between any
two components with constant relative volatility.
Note that although a high purity separation (large S)
requires a larger number of stages, the increase is only
proportional to the logarithm of the separation fac-
tor. For example, increasing the purity level in a prod-
uct by a factor of 10 (e.g. by reducing xH,D from
0.01 to 0.001) increases Nmin by about a factor of
ln 10"2.3.
A common rule of thumb is to select the actual
number of stages N"2Nmin (or even larger).
Minimum Energy Usage + In\nite Number of Stages
For a given separation, an increase in the number of
stages will yield a reduction in the reSux (or equiva-
II / DISTILLATION / Theory of Distillation 1125
lently in the boil-up). However, as the number of
stages approaches inRnity, a pinch zone develops
somewhere in the column, and the reSux cannot be
[26]
reduced further. For a binary separation the pinch
usually occurs at the feed stage (where the material
balance line and the equilibrium line will meet), and
we can easily derive an expression for the minimum
reSux with N"R. For a saturated liquid feed
(q"1) (King’s formula):
[27]
rL,D!rH,D
LTmin" F [28]
!1
where rL,D"xDD/zF is the recovery fraction of light
component, and rH,D of heavy component, both in the
distillate. The value depends relatively weakly on the
product purity, and for sharp separations (where
rL,D"1 and rH,D"0), we have LTmin"F/(!1). Ac-
tually, eqn [28] applies without stipulating constant
molar Sows or constant , but then LTmin is the liquid
Sow entering the feed stage from above, and  is the
relative volatility at feed conditions. A similar expres-
sion, but in terms of VBmin entering the feed stage from
below, applies for a saturated vapour feed (q"0)
(King’s formula):
rH,B!rL,B
VBmin" F [29]
!1
1126 II / DISTILLATION / Theory of Distillation
For sharp separations we get VBmin"F/(!1). In
summary, for a binary mixture with constant molar
Sows and constant relative volatility, the minimum
boil-up for sharp separations is:
1 yH"HHxH heavy component; xHP0)
Feed liquid, q"1: VBmin" F#D
!1
[30]
1 relative volatility, Henry’s constant in the bottom is
Feed vapour, q"0: VBmin" F HL" and in the top is HH"1/.) Thus, with con-
!1
Note that minimum boil-up is independent of the
product purity for sharp separations. From this we
establish one of the key properties of distillation: we
can achieve any product purity (even inRnite separ-
ation factor) with a constant Rnite energy (as long as
it is higher than the minimum) by increasing the
number of stages.
Obviously, this statement does not apply to azeo-
tropic mixtures, for which "1 for some composi-
tions but we can get arbitrarily close to the azeotropic
composition, and useful results may be obtained in
some cases by treating the azeotrope as a pseudo-
component and using  for this pseudo-separation.
Finite Number of Stages and Finite Re]ux
Fenske’s formula S"N applies to inRnite reSux
(inRnite energy). To extend this expression to real
columns with Rnite reSux we will assume constant
molar Sows and consider three approaches:
1. Assume constant K-values and derive the Kremser
formulas (exact close to the column end for a high
purity separation.)
2. Assume constant relative volatility and derive the
following extended Fenske formula (approximate
formula for case with optimal feed-stage location):
(LT/VT)NT
S+N
(LB/VB)NB
Here NT is the number of stages in the top section
and NB in the bottom section.
3. Assume constant relative volatility and derive
exact expressions. The most used are the Under-
wood formulas which are particularly useful for
computing the minimum reSux (with inRnite
stages).
Constant K-Values + Kremser Formulas
For high purity separations most of the stages are
located in the corner parts of the McCabe}Thiele
diagram where, according to Henry’s law, we may
approximate the VLE relationship, even for nonideal
mixtures, by straight lines:
Bottom of column:
yL"HLxL light component; xLP0)
Top of column:
where H is Henry’s constant. (For the case of constant
stant molar Sows, both the equilibrium and
mass}balance relationships are linear, and the result-
ing difference equations are easily solved analytically.
For example, at the bottom of the column we derive
for the light component:
xL,n#1"(VB/LB)HLxL,n#(B/LB)xL,B
"sxL,n#(1!VB/LB)xL,B [31]
where s"(VB/LB)HL'1 is the stripping factor. Re-
peated use of this equation gives the Kremser formula
for stage NB from the bottom (the reboiler would here
be stage zero):
xL,NB"sNBxL,B[1#(1!VB/LB)(1!s\NB/(s!1)]
[32]
(assuming we are in the region where s is constant, i.e.
xL+0). At the top of the column we have for the
heavy component:
yH,n 1"(LT/VT)(1/HH)yH,n#(D/VT)xH,D
\
"ayH,n#(1!LT/VT)xH,D [33]
where a"(LT/VT)/HH'1 is the absorption factor.
The corresponding Kremser formula for the heavy
component in the vapour phase at stage NT counted
from the top of the column (the accumulator in stage
zero) is then:
yH,NT"aNTxH,D[1#(1!LT/VT)(1!a\NT)/(a!1)]
[34]
(assuming we are in the region where a is constant,
i.e. xH+0).
For hand calculations one may use the
McCabe}Thiele diagram for the intermediate com-
position region, and the Kremser formulas at the
column ends where the use of the McCabe}Thiele
diagram is inaccurate.
Example We consider a column with N"40,
NF"21, "1.5, zL"0.5, F"1, D"0.5,
VB"3.206. The feed is saturated liquid and exact

calculations give the product compositions
xH,D"xL,B"0.01. We now want to have a bottom
product with only 1 p.p.m. heavy product, i.e.
xL,B"1.e!6. We can use the Kremser formulas to
estimate easily the additional stages needed when we
have the same energy usage, VB"3.206. (Note that
with the increased purity in the bottom we actually
get D"0.505.) At the bottom of the column
HL""1.5 and the stripping factor is
s"(VB/LB)HL"(3.206/3.711)1.5"1.296.
With xL,B"1.e!6 (new purity) and xL,NB"0.01
(old purity) we Rnd by solving the Kremser equation
[31] for the top with respect to NB that NB"34.1,
and we conclude that we need about 34 additional
stages in the bottom (this is not quite enough since the
operating line is slightly moved and thus affects the
rest of the column; using 36 rather than 34 additional
stages compensates for this).
The above Kremser formulas are valid at the col-
umn ends, but the linear approximation resulting
from the Henry’s law approximation lies above the
real VLE curve (it is optimistic), and thus gives too
few stages in the middle of the column. However, if
there is no pinch at the feed stage (i.e. the feed is
optimally located), then most of the states in the
column will be located at the column ends where the
above Kremser formulas apply.
Approximate Formula with Constant Relative
Volatility
We will now use the Kremser formulas to derive an
approximation for the separation factor S. First note
that for cases with high purity products we have
S+1/(xL,BxH,D). That is, the separation factor is
the inverse of the product of the key component
product impurities.
We now assume that the feed stage is optimally
located such that the composition at the feed stage is
the same as that in the feed, i.e. yH,NT"yH,F and
xL,NB"xL,F. Assuming constant relative volatility and
using HL", HB"1/, "(yLF/xLF)/(yHF/xHF) and
N"NT#NB#1 (including total reboiler) then
gives:
(LT/VT)NT c
S+N
(LB/VB)NB (xHFyLF)
where:

 
VB (1!s\NB)
c" 1# 1!
LB (s!1)

 
LT (1!a\NT)
; 1# 1!
VT (a!1)
II / DISTILLATION / Theory of Distillation 1127
We know that S predicted by this expression is some-
what too large because of the linearized VLE. How-
ever, we may correct it such that it satisRes the exact
relationship S"N at inRnite reSux (where
LB/VB"VT/LT"1 and c"1) by dropping the factor
1/(xHFyLF) (which as expected is always larger than 1).
At Rnite reSux, there are even more stages in the feed
region and the formula will further overestimate the
value of S. However, since c'1 at Rnite reSux, we
may partly counteract this by setting c"1. Thus, we
delete the term c and arrive at the Rnal extended
Fenske formula, where the main assumptions are that
we have constant relative volatility, constant molar
Sows and that there is no pinch zone around the feed
(i.e. the feed is optimally located):
(LT/VT)NT
S+N [35]
(LB/VB)NB
where N"NT#NB#1. Together with the material
balance, FzF"DxD#BxB, this approximate formula
can be used to estimate the number of stages for
column design (instead of e.g. the Gilliand plots), and
also to estimate the effect of changes of internal Sows
during column operation. However, its main value is
the insight it provides:
1. We see that the best way to increase the separation
S is to increase the number of stages.
2. During operation where N is Rxed, the formula
provides us with the important insight that the
separation factor S is increased by increasing the
internal Sows L and V, thereby making L/V closer
to 1. However, the effect of increasing the internal
Sows (energy) is limited since the maximum separ-
ation with inRnite Sows is S"N.
3. We see that the separation factor S depends mainly
on the internal Sows (energy usage) and only
weakly on the split D/F. This means that if we
change D/F then S will remain approximately con-
stant (Shinskey’s rule), that is, we will get a shift in
impurity from one product to the other such that
the product of the impurities remains constant.
This insight is very useful.
Example Consider a column with xD,H"0.01 (1%
heavy in top) and xB,L"0.01 (1% light in bottom).
The separation factor is then approximately
S"0.99;0.99/(0.01;0.01)"9801. Assume we
slightly increase D from 0.50 to 0.51. If we assume
constant separation factor (Shinskey’s rule), then we
Rnd that xD,H changes from 0.01 to 0.0236 (heavy
impurity in the top product increases by a factor 2.4),
whereas xB,L changes from 0.01 to 0.0042 (light
1128 II / DISTILLATION / Theory of Distillation

 
 
NT
impurity in the bottom product decreases by a factor LT LT (1!a\NT)
1# 1!
2.4). Exact calculations with column data: N"40, VT VT (a!1)
NF"21, "1.5, zL"0.5, F"1, D"0.5, ;

 
 
NB
VB VB (1!s\NB)
LT/F"3.206, given that xD,H changes from 0.01 to 1# 1!
LB LB (s!1)
0.0241 and xB,L changes from 0.01 to 0.0046
(separation factor changes from S"9801 to The last big term is close to 1 in most cases and can be
8706). Thus, Shinskey’s rule gives very accurate pre- neglected. Rewriting the expression in terms of the
dictions. light component then gives Skogestad’s short-cut for-
However, the simple extended Fenske formula also mula for the feed stage location:
has shortcomings. First, it is somewhat misleading

  
since it suggests that the separation may always be (1!yF) xB
ln
improved by transferring stages from the bottom to xF (1!xD)
the top section if (LT/VT)'(VB/LB). This is not gener- NT!NB" [36]
ln 
ally true (and is not really ‘allowed’ as it violates the
assumption of optimal feed location). Second, al- where yF and xF at the feed stage are obtained as
though the formula gives the correct limiting value explained above. The optimal feed-stage location
S"N for inRnite reSux, it overestimates the value of counted from the bottom is then:
S at lower reSux rates. This is not surprising since at
low reSux rates a pinch zone develops around the [N#1!(NT!NB)]
NF"NB#1" [37]
feed. 2

where N is the total number of stages in the column.

Example Consider again the column with N"40,
NF"21, "1.5, zL"0.5, F"1, D"0.5;
LT"2.706. Exact calculations based on these data
give xHD"xLB"0.01 and S"9801. On the other
hand, the extended Fenske formula with NT"20 and
NB"20 yields:
(2.7606/3.206)20 0.34
S"1.541; 20 "16 586 000;
(3.706/3.206) 18.48
"30 774
corresponding to xHD"xLB"0.0057. The error may
seem large, but it is actually quite good for such
a simple formula.
Optimal Feed Location
The optimal feed-stage location is at the intersection
of the two operating lines in the McCabe}Thiele
diagram. The corresponding optimal feed-stage com-
position (xF, yF) can be obtained by solving the
following two equations: z"qxF#(1!q)yF and
yF"xF/(1#(!1)xF). For q"1 (liquid feed) we
Rnd xF"z and for q"0 (vapour feed) we Rnd yF"z
(in the other cases we must solve a second-order
equation).
There exists several simple short-cut formulas to
estimate the feed point location. One may be derived
from the Kremser equations given above. Divide
the Kremser equation for the top by the one for the
bottom and assume that the feed is optimally located
to derive:
yH,F xH,D (NT NB)
"  \
xL,F xL,B
Summary for Continuous Binary Columns
With the help of a few of the above formulas it is
possible to perform a column design in a matter of
minutes by hand calculations. We will illustrate this
with a simple example.
We want to design a column for seperating
a saturated vapour mixture of 80% nitrogen (L) and
20% oxygen (H) into a distillate product with 99%
nitrogen and a bottoms product with 99.998% oxy-
gen (mole fractions).
Component data Normal boiling points (at 1 atm):
TbL"77.4 K, TbH"90.2 K, heat of vaporization at
normal boiling points: 5.57 kJ mol\1 (L) and
6.82 kJ mol\1 (H).
The calculation procedure when applying the
simple methods presented in this article can be done
as shown in the following steps:
1. Relative volatility: The mixture is relatively
ideal and we will assume constant relative vola-
tility. The estimated relative volatility at 1 atm
based on the boiling points is ln +
(HM vap/RTM b) [(TbH!TbL)/TM b] where HM vap"
(5.57 ) 6.82"6.16 kJ mol\1, TM b"(TbHTbL"
83.6 K and TH!TL"90.2!77.7"18.8. This
gives (HM vap)/(RTM b)"8.87 and we Rnd +3.89
(however, it is generally recommended to obtain
 from experimental VLE data).
2. Product split: From the overall material balance
we get D/F"(z!xB)/(xD!xB)"(0.8!0.00002)/
(0.99!0.00002)"0.808.
 II / DISTILLATION / Theory of Distillation 1129

3. Number of stages: The separation factor is
S"(0.99;0.99998)/(0.01;0.00002)"4 950 000,
i.e. ln S"15.4. The minimum number of stages
required for the separation is Nmin"
ln S/ln "11.35 and we select the actual number
of stages as N"23 (+2Nmin).
4. Feed-stage location: With an optimal feed location
we have at the feed stage (q"0) that yF"zF"
0.8 and xF"yF/(!(!1)yF)"0.507. Skoges-
tad’s approximate formula for the feed-stage loca-
tion gives:
The result is a slightly lower vapour Sow due to
a higher relative volatility ( in the range 3.99}4.26
with an average of 4.14). More precisely, a simula-
tion with N"23, NF"15 gave V/F"0.291, which
is about 11% higher than the minimum value
Vmin"0.263 found with a very large number of
stages (increasing N'60 did not give any signiRcant
energy reduction below Vmin). The optimal feed stage
(with N"23) was indeed found to be NF"15.
Thus, the results from HYSYS conRrm that a col-
umn design based on the very simple short-cut
methods is very close to results from much more

  
(1!yF) xB
NT!NB"ln (ln ) rigorous computations.
xF (1!xD)

   
0.2 0.00002
"ln
0.507
;
0.01
1.358 Multicomponent Distillation +
Underwood’s Methods
"!5.27
We present here the Underwood equations for multi-
corresponding to the feed stage NF"[N#1! component distillation with constant relative volatil-
(NT!NB)]/2"(23#1#5.27)/2"14.6. ity and constant molar Sows. The analysis is based on
5. Energy usage: The minimum energy usage for considering a two-product column with a single feed,
a vapour feed (assuming sharp separation) is but the usage can be extended to all kinds of column
Vmin/F"1/(!1)"1/2.89"0.346. With the section interconnections.
choice N"2Nmin, the actual energy usage (V) is It is important to note that adding more compo-
then typically about 10% above the minimum nents does not give any additional degrees of freedom
(Vmin), i.e. V/F is about 0.38. in operation. This implies that for an ordinary two-
product distillation column we still have only two
This concludes the simple hand calculations. Note
degrees of freedom, and thus we will only be able to
again that the number of stages depends directly on
specify two variables, e.g. one property for each prod-
the product purity (although only logarithmically),
uct. Typically, we specify the purity (or recovery) of
whereas for well-designed columns (with a sufRcient
the light key in the top and of the heavy key in the
number of stages) the energy usage is only weakly
bottom (the key components are deRned as the com-
dependent on the product purity.
ponents between which we are performing the split).
Remark 1 The actual minimum energy usage is The recoveries for all other components and the inter-
slightly lower since we do not have sharp separations. nal Sows (L and V) will then be completely deter-
The recovery of the two components in the bottom mined.
product is rL"(xL,BB)/(zFLF)"0.9596 and rH" For a binary mixture with given products, as we
(xH,BB)/(zFHF)+0, so from the formulas given earlier increase the number of stages, there develops a pinch
the exact value for nonsharp separations is zone on both sides of the feed stage. For a multicom-
Vmin/F"(0.9596!0.0;3.89)/(3.89!1)"0.332. ponent mixture, a feed region pinch zone only devel-
ops when all components distribute to both products,
Remark 2 For a liquid feed we would have to use and the minimum energy operation is found for a par-
more energy, and for a sharp separation: ticular set of product recoveries, sometimes denoted
as the preferred split. If all components do not distrib-
Vmin/F"1/(!1)#D/F"0.346#0.808"1.154 ute, the pinch zones will develop away from the feed
stage. Underwood’s methods can be used in all these
Remark 3 We can check the results with exact cases, and are especially useful for the case of inRnite
stage-by-stage calculations. With N"23, NF"15 number of stages.
and "3.89 (constant), we Rnd V/F"0.374, which
is about 13% higher than Vmin"0.332. The Basic Underwood Equations
Remark 4 A simulation with more rigorous VLE The net material transport (wi) of component i up-
computations, using the Soave}Redlich}Kwong wards through a stage n is:
(SRK) equation of state, has been carried out using
the HYSYS (Hypnotech Ltd.) simulation package. wi"Vn yi,n!Ln#1xi,n#1 [38]
1130 II / DISTILLATION / Theory of Distillation

Note that wi is always constant in each column divide an equation for k with the one for , the
j
section. We will assume constant molar Sows following expression appears:
(L"Ln " Ln 1 and V"Vn"Vn#1) and, assuming
\
ixi,n#m ixi,n



constant relative volatility, the VLE relationship is:


(i! k) k
m
(i! k)
ixi (yi/xi)
i
" i
[45]
yi" where i" [39] ixi,n#m j ixi,n
ixi (yr/xr) i (i! j) i (i! j)
i

We divide eqn [38] by V, multiply with the factor and we note the similarities with the Fenske and
i/(i! ), and take the sum over all components: Kremser equations derived earlier. This relates the
composition on a stage (n) to a composition on an-
2ixi,n other stage (n#m). The number of independent
equations of this kind equals the number of Under-
1 iwi (i! ) L ixi,n#1
" i
! [40] wood roots minus 1 (since the number of equations of
V i (i! ) ixi,n V i (i! ) the type as in eqn [44] equals the number of Under-
i
wood roots), but in addition we also have xi"1.
The parameter is free to choose, and the Under- Together, this is a linear equation system for comput-
wood roots are deRned as the values of which make ing xi,n#m when xi,n is known and the Underwood
the left-hand side of eqn [40] unity, i.e.: roots are computed from eqn [41].
Note that so far we have not discussed minimum
iwi reSux (or vapour Sow rate), thus these equations hold
V" [41]
i (i! ) for any vapour and reSux Sow rates, provided that the
roots are computed from the deRnition in eqn [41].
The number of values satisfying this equation is
equal to the number of components. Some Properties of the Underwood Roots
Most authors usually use a product composition or Underwood showed a series of important properties
component recovery (r) in this deRnition, e.g. for the of these roots for a two-product column with a re-
top (subscript T) section or the distillate product boiler and condenser. In this case all components
(subscript D): Sows upwards in the top section (wi,T50), and
downwards in the bottom section (wi,B40). The
wi"wi,T"wi,D"Dxi,D"ri,DziF [42] mass balance yields: wi,B"wi,T!wi,F where
wi,F"Fzi. Underwood showed that in the top section
but we prefer to use w since it is more general. Note (with Nc components) the roots ( ) obey:
that use of the recovery is equivalent to using net
component Sow, but use of the product composition 1' 1'2' '3'2'N '
3 N .
is only applicable when a single product stream is  

leaving the column. If we apply the product recovery, And in the bottom section (where wi,n"wi,B40) in
or the product composition, the deRning equation for general we have a different set of roots denoted
the top section becomes: () computed from VB" i [iwi,B/(i!)]"
i [i(!ri,B)zi/(i!)]" i [i(!(1!ri,D))zi/(i!)]
iri,Dzi ixi,D which obey:
VT" F" D [43]
i (i! ) i (i! )
1'1'2'2'3'3'2'N 'N
Stage-to-Stage Calculations  

With the deRnition of from eqn [41], eqn [40] can Note that the smallest root in the top section is
be simpliRed to: smaller than the smallest relative volatility, and the
largest root in the bottom section is larger than the
ixi,n largest volatility. It is easy to see from the deRning
L ixi,n#1 (i! ) equations that VTPRN iPi and similarly
" i
[44] VBPRNiPi.
V i (i! ) ixi,n
i When the vapour Sow is reduced, the roots in the
top section will decrease, while the roots in the bot-
This equation will be valid for any of the Underwood tom section will increase, but interestingly Under-
roots, and if we assume constant molar Sows and wood showed that i5i#1. A very important result
 II / DISTILLATION / Theory of Distillation 1131

by Underwood is that for an inRnite number of stages rj#1,D"0). The procedure is then simply:
VPVminN iPi#1.
Then, at minimum reSux, the Underwood roots for 1. Compute the common root (j) for which
the top ( ) and bottom () sections coincide. Thus, if j'j'j#1 from the feed equation: (1!q)"
we denote the common roots (), and recall that i [aizi/(ai!)]
VT!VB"(1!q)F, we obtain the following equa- 2. Compute the minimum energy by applying j
tion for the common roots () by subtracting the to the deRnition equation:
deRning equations for the top and bottom sections:
VTmin/F" j
a z /(ai!j).
i"1 i i

izi Note that the recoveries

(1!q)" [46]
i (i!)

We call this expression the feed equation since only
the feed properties (q and z) appear. Note that this is
not the equation which deRnes the Underwood roots
and the solutions () apply as roots of the deRning
equations only for minimum reSux conditions
(N"R). The feed equation has Nc roots (but one of
these is not a common root) and the Nc!1 common
roots obey: 1'1'2'2'2'N!1'N .

Solution of the feed equation gives us the possible
common roots, but all pairs of roots ( i and i#1) for
the top and bottom section do not necessarily co-
incide for an arbitrary operating condition. We illus-
trate this with the following example.
Example Assume we start with a given product split
(D/F) and a large vapour Sow (V/F). Then only one
component i (with relative volatility i) can be distri-
buted to both products. No roots are common. Then
we gradually reduce V/F until a second component
j (this has to be a component j"i#1 or j"i!1)
becomes distributed, e.g. for j"i#1 one set of roots
will coincide: i"i#1"i, while the others do not.
As we reduce V/F further, more components become
distributed and the corresponding roots will coincide,
until all components are distributed to both products,
and then all the Nc!1 roots from the feed equation
also are roots for the top and bottom sections.
An important property of the Underwood roots is
that the value of a pair of roots which coincide
(e.g. when i"i#1"i) will not change, even if
only one, two or all pairs coincide. Thus all the
possible common roots are found by solving the feed
equation once.

1 for i4j
ri,D"
0 for i'j
For example we can derive King’s expressions for
minimum reSux for a binary feed (zL"z,
zH"(1!z), L", H"1, and liquid feed (q"1)).
Consider the case with liquid feed (q"1). We Rnd
the single common root from the feed equation:
"/(1#(!1)z), (observe 551, as ex-
pected). The minimum reSux expression appears as
we use the deRning equation with the common root:
LTmin VTmin D ri,Dzi rL,Dz rH,D(1!z)
" ! " " #
F F F i (i!) ! 1!
and when we substitute for  and simplify, we obtain
King’s expression: LTmin/F"(rL,D!rH,D)/(!1).
Another interesting case is minimum energy opera-
tion when we consider sharp split only between the
most heavy and most light components, while all the
intermediates are distributed to both products. This
case is also denoted the preferred split, and in this
case there will be a pinch region on both sides of the
feed stage. The procedure is:
1. Compute all the Nc!1 common roots () from
the feed equation.
2. Set r1,D"1 and rNc ,D"0 and solve the following
linear equation set (Nc!1 equations) with respect
to [VT, r2,D, r3,D2rNc 1] (Nc!1 variables):
\
Nc
airi,Dzi
VT"
i"1 (ai!1)

Minimum Energy + In\nite Number of Stages :

When we go to the limiting case of inRnite number of
Nc
stages, Underwood’s equations become very useful. airi,Dzi
VT" [47]
The equations can be used to compute the minimum i"1 (ai!Nc!1)
energy requirement for any feasible multicomponent
separation. Note that, in this case, when we regard the most
Let us consider two cases: Rrst we want to compute heavy and light components as the keys, and all the
the minimum energy for a sharp split between two intermediates are distributed to both products, King’s
adjacent key components j and j#1 (rj,D"1 and very simple expression will also give the correct
1132 II / DISTILLATION / Theory of Distillation
minimum reSux for a multicomponent mixture (for
q"1 or q"0). The reason is that the pinch then
occurs at the feed stage. In general, the values com-
puted by King’s expression give a (conservative) up-
per bound when applied directly to multicomponent
mixtures. An interesting result which can be seen
from King’s formula is that the minimum reSux at the
preferred split (for q"1) is independent of the feed
composition and also independent of the relative
volatilities of the intermediates.
However, with the Underwood method, we
also obtain the distribution of the intermediates,
and it is easy to handle any liquid fraction (q) in the
feed.
The procedure for an arbitrary feasible product
recovery speciRcation is similar to the preferred split
case, but then we must only apply the Underwood
roots (and corresponding equations) with values be-
tween the relative volatilities of the distributing com-
ponents at the limit of being distributed. In cases
where not all components distribute, King’s min-
imum reSux expression cannot be trusted directly,
but it gives a (conservative) upper bound.
Figure 9 shows an example of how the components
are distributed to the products for a ternary (ABC)
mixture. We chose the overhead vapour Sow
(V"VT) and the distillate product Sow (D"V!L)
as the two degrees of freedom. The straight lines,
Figure 9 Regions of distributing feed components as a function of V and D for a feed mixture with three components: ABC.
Pij represents minimum energy for sharp split between components i and j. For large vapour flow (above the top sawtooth), only one
component distributes. In the triangle below PAC, all components distribute.
which are at the boundaries when a component
is at the limit of appearing/disappearing (distri-
bute/not distribute) in one of the products, can be
computed directly by Underwood’s method. Note
that the two peaks (PAB and PBC) give us the minimum
vapour Sow for a sharp split between A/B and B/C.
The point PAC, however, is at the minimum vapour
Sow for a sharp A/C split and this occurs for a speciRc
distribution of the intermediate B, known as the pre-
ferred split.
King’s minimum reSux expression is only valid in
the triangle below the preferred split, while the
Underwood equations can reveal all component re-
coveries for all possible operating points. (The shaded
area is not feasible since reSux and vapour Sow rates
have to be positive (V'D, V'(1!q) F.)
Further Discussion of Speci\c Issues
The Energy Balance and the Assumption of
Constant Molar Flows
All the calculations in this article are based on the
assumption of constant molar Sows in a section, i.e.
Vn"Vn 1"V and Ln"Ln#1"L. This is a very
\
common simpliRcation in distillation computations,
and we shall use the energy balance to see when we
can justify it. The energy balance is similar to the
 II / DISTILLATION / Theory of Distillation 1133

mass balance, but now we use the molar enthalpy (h) At Rrst glance, these assumptions may seem re-
strictive, but the assumption of constant molar Sows
of the streams instead of composition. The enthalpy is
computed for the actual mixture and will be a func-actually holds well for many industrial mixtures.
tion of composition in addition to temperature (or In a binary column where the last assumption
about equal Hvap
pressure). At steady state the energy balance around bpi is not fulRlled, a good estimate of
stage n becomes: the change in molar Sows from the bottom (stage 1)
to the top (stage N), due to this effect for a case with
LnhL,n!Vn 1hV,n 1"Ln#1hL,n#1!Vn hV,n [48] saturated liquid feed (q"1) and close to pure prod-
\ \
ucts, is given by: VN/V1+Hvap H /HL , where the
vap

Combining this energy balance with the overall molar heat of vaporization is taken at the boiling
material balance on a stage (Vn 1!Ln" point temperatures for the heavy (H) and light (L)
\
Vn!Ln#1"W where W is the net total molar Sow components respectively.
through a section, i.e. W"D in the top section and Recall that the temperature dependency of the rela-
W"B in the bottom section) yields: tive volatility is related to different heat of vaporiza-
tion also, thus the assumptions of constant molar
hV,n 1!hL,n hL,n!hL,n#1
Vn"Vn 1 \ #W [49] Sows and constant relative volatility are closely
\ hV,n!hL,n#1 hV,n!hL,n#1 related.

From this expression we observe how the vapour Sow
will vary though a section due to variations in heat of
vaporization and molar enthalpy from stage to stage.
We will now show one way of deriving the con-
stant molar Sow assumption:
1. Choose the reference state (where h"0) for each
pure component as saturated liquid at a reference
pressure. (This means that each component has
a different reference temperature, namely its
boiling point (Tbpi) at the reference pressure.)
2. Assume that the column pressure is constant and
equal to the reference pressure.
Calculation of Temperature when Using Relative
Volatilities
It may seem that we have lost the pressure and tem-
perature in the equilibrium equation when we intro-
duced the relative volatility. However, this is not the
case since the vapour pressure for every pure compon-
ent is a direct function of temperature, thus it is also the
relative volatility. From the relationship P" pi"
xi poi(T) we derive:
P"por(T) xii [50]
i

3. Neglect any heat of mixing such that

Remember that only one of P or T can be speciRed
hL,n" i xi,ncPLi(Tn!Tbpi). when the mole fractions are speciRed. If composition
4. Assume that all components have the same molar and pressure are known, a rigorous solution of the
heat capacity, cPL. temperature is found by solving the nonlinear equa-
5. Assume that the stage temperature can be approxi- tion:
mated by Tn" i xi,nTbpi. These assumptions give
hL,n"0 on all stages and eqn [49] for change in P" xipoi(T) [51]
boil-up is reduced to:
However, if we use the pure component boiling
hV,n 1 points (Tbi), a crude and simple estimate can be com-
Vn"Vn 1 \ puted as:
\ hV,n

6. The molar enthalpy in the vapour phase is T+ xiTbi [52]

given as: hV,n" i xi,nHvap bpi # i xi,ncPVi(Tn!Tbpi)
where Hvapbpi is the heat of vaporization for the
For ideal mixtures, this usually gives an estimate
pure component at its reference boiling temper- which is a bit higher than the real temperature; how-
ature (Tbpi). ever, a similar approximation may be done by using
7. We assume that cPV is equal for all components, the vapour composition (y), which will usually give
and then the second summation term above will a lower temperature estimate. This leads to a good
become zero, and we have: hV,n" i xi,nHvap bpi .
estimate when we use the average of x and y, i.e.:
8. Then if Hvap
bpi "H
vap
is equal for all components


we get hV,n"hV,n 1"Hvap, and thereby con- xi#yi
\ T+ Tbi [53]
stant molar Sows: Vn"Vn 1 and also Ln"Ln#1. 2
\
1134 II / DISTILLATION / Theory of Distillation
Figure 10 Temperature profile for the example shown in Figure 7 (continuous line) compared with various linear boiling point
approximations.
Alternatively, if we are using relative volatilities we
may Rnd the temperature via the vapour pressure of
the reference component. If we use the Antoine equa-
tion, then we have an explicit equation:
Br
T+ #Cr where por"P/ xii [54]
log por!Ar i
This last expression is a very good approximation to
a solution of the nonlinear eqn [51]. An illustration of
how the different approximations behave is
shown in Figure 10. For that particular case (a fairly
ideal mixture), eqns [53] and [54] almost coincide.
In a rigorous simulation of a distillation column,
the mass and energy balances and the VLE have to be
solved simultaneously for all stages. The temperature
is then often used as an iteration parameter in order
to compute the vapour pressures in VLE computa-
tions and in the enthalpy computations of the energy
balance.
Discussion and Caution
Most of the methods presented in this article are
based on ideal mixtures and simplifying assumptions
about constant molar Sows and constant relative
volatility. Thus there are many separation cases for
nonideal systems where these methods cannot be ap-
plied directly.
However, if we are aware of the most important
shortcomings, we may still use these simple methods
for short-cut calculations, for example, to gain insight
or check more detailed calculations.
See also: II/Distillation: Historical Development; Instru-
mentation and Control Systems; Modelling and Simula-
tion; Vapour-Liquid Equilibrium: Theory.
Further Reading
Franklin NL and Forsyth JS (1953) The interpretation of
minimum reSux conditions in multicomponent distilla-
tion. Chemical Engineering Research and Design 31.
Reprinted in volume 75, December 1997, pp. 56}81.
King CJ (1980) Separation Processes, 2nd edn. New York:
McGraw-Hill.
Kister HZ (1992) Distillation Design. New York:
McGraw-Hill.
McCabe WL, Smith JC and Harriot P (1993) Unit Opera-
tions of Chemical Engineering, New York: McGraw-Hill.
Shinskey FG (1984) Distillation Control } for Productivity
and Energy Conservation. New York: McGraw-Hill.
Skogestad S (1997) Dynamics and control of distillation
columns } a tutorial introduction. Chemical Engineering
Research and Design 75: 539}562.
Stichlmair J and James RF (1998) Distillation: Principles
and Practice. New York: Wiley.
Underwood AJV (1948) Fractional distillation of multi-
component mixtures. Chemical Engineering Progress
44: 603}614.
 II / DISTILLATION / Tray Columns: Design 1135

Tray Columns: Design

K. T. Chuang and K. Nandakumar, top to the bottom of the column occurs mainly via
University of Alberta, Edmonton, Alberta, Canada downcomers.
Copyright ^ 2000 Academic Press There are three types of cross-Sow trays: (1) sieve,
(2) valve and (3) bubble cap. Among them, sieve trays
offer high capacity and efRciency, low pressure drop,
Distillation has remained an important separation ease of cleaning, and low capital cost, but smaller
technology for the chemical process industries. In turndown ratio. Although the design procedure is
1997 it was reported in the journal Chemical Engin- similar for all three types of trays, only sieve tray
eering that about 95% of all worldwide separation performance data are readily available in the public
processes use this technology. In the USA alone, some domain. The valve and bubble cap designs are often
40 000 distillation columns represent a capital invest- protected by patents, and thus the performance data
ment of about US $8 billion. They consume the en- are supplied by the vendors. This article describes the
ergy equivalent of approximately 1 billion barrels of procedure for designing an optimum sieve tray.
crude oil per day. Such columns are used in reRneries, A similar procedure can be applied in principle to the
petrochemical plants, gas processing plants and or- valve and bubble cap trays, provided critical perfor-
ganic chemical plants to purify natural gas, improve mance data are available.
gasoline, produce petrochemicals and organic prod- The cost of a tray column is determined by two
ucts, recover pollulant species, etc. factors:
Distillation can be carried out in a tray or a packed
column. The major considerations involved in the 1. column diameter, which determines the through-
choice of the column type are operating pressure and put;
design reliability. As pressure increases, tray coulmns
become more efRcient for mass transfer and can often
tolerate the pressure drop across the trays. The design
procedure for the large diameter tray column is also
more reliable than that for the packed column. Thus,
trays are usually selected for large pressurized column
applications.
Distillation trays can be classiRed as:
1. cross-Sow trays with downcomers (see Figure 1A);
2. countercurrent trays without downcomers (also
known as dual-Sow trays) (see Figure 1B).
The-dual Sow tray allows the gas and liquid to pass
through the same tray openings. This results in a lim-
ited operating range because the dispersion height is
very sensitive to the gas/liquid Sow rates. In general,
dual-Sow trays are employed only in cases where high
capacity or high resistance to fouling are required.
However, because of its narrow operating range, the
market share is small and such trays will not be
discussed further.
The cross-Sow tray utilizes a weir on the down-
comer to control the spray height on the tray, and
thus provides a stable gas}liquid dispersion over
a wide range of gas/liquid Sows. A tray is the combi-
nation of a tray deck, where froth is generated to
provide vapour}liquid contact, and a downcomer,
where the vapour}liquid mixture is separated. The
bulk of the vapour rises from the aerated liquid
through the vapour disengagement space to the tray Figure 1 (A) Sieve ray with downcomer in a 30 cm diameter
above. However, the passage of the liquid from the column. (B) Dual-flow tray in a 30 cm diameter column.
1136 II / DISTILLATION / Tray Columns: Design
2. column height, which delivers the number of equi-
librium stages required for the separation.
The minimum cost is generally achieved when the
column volume is minimized. The Rnal selection of
the tray design is based on the combined cost of the
column shell, internals and installation.
It should be noted that the fraction of the cross-
sectional area available for vapour}liquid disengage-
ment decreases when the downcomer area is in-
creased. Thus, optimum design of the tray involves
a balance between the tray area and the downcomer
area (i.e. the capacity for the tray deck should match
the capacity of the downcomer). The correlations for
sizing trays are implicit in column diameter, tray
spacing and tray geometry, thus requiring trial-and-
error calculations to arrive at the Rnal selection.
Characteristics of Tray Operation
Typical tray layout is shown in Figure 2, and tray
operation is shown in Figure 3. High speed photogra-
phy of a large operating tray indicates that the vapour
erupts through the liquid sporadically. The holes that
are not erupting do not weep appreciably at a vapour
rate above the weep point, although the supporting of
the liquid by the vapour is not absolutely complete.
The interaction of vapour and liquid on a properly
designed tray results in a highly turbulent two-phase
mixture of a high speciRc interfacial area with net
Figure 2 Tray layouts.
liquid movement in a crossSow direction to the rising
vapour stream. The aerated liquid may be either
liquid-continuous (froth) at relatively low vapour vel-
ocities or vapour-continuous (spray) at high vapour
velocities.
The maximum capacity of a sieve tray is reached
when the tray is Sooded. This may be due to excessive
spraying (entrainment) taking place in the intertray
space or the froth in the downcomer backing-up to
reach the top of the outlet weir. The onset of Sooding
is accompanied by a sharp increase in tray pressure
and a sharp decrease in tray efRciency.
As vapour rates decrease to the point that the
vapour Sow cannot totally support the liquid on the
tray, some liquid will weep through the holes. If
the weepage is so severe that no liquid Sows over the
outlet weir, the tray cannot operate stably under these
dumping conditions. The minimum capacity of the
tray is normally reached when moderate weepage is
encountered. Ideally, a sieve tray should operate in
the shaded area shown in Figure 4 to ensure proper
operation.
Tray efRciency can be divided into two compo-
nents:
1. point efRciency as determined by the vertical Sow
of vapour through the froth;
2. tray efRciency enhancement by the crossSow of
liquid.
The physical properties of the vapour}liquid mix-
ture determine the point efRciency, although froth
height, which inSuences the gas residence time, also
has a signiRcant effect, especially for low efRciency
systems. Liquid Sow pathlength determines the liquid
residence time and the extent of crossSow tray efR-
ciency enhancement. Entrainment and weeping de-
press tray efRciency by disrupting the concentration
proRle in the column. The froth height and the liquid
Sow path are two parameters that are optimized to
give maximum tray efRciency. Other geometric vari-
ables, such as open hole area, hole diameter and
downcomer arrangement, also affect tray hydraulics
and efRciency. The goal for a tray design is to reach
maximum tray efRciency without compromising hy-
draulic stability.
The steps required for tray column design are
shown in Figure 5; a detailed discussion of each step
is given below.
Input Data
Once sieve trays are selected for a given application,
the input data that are required in the design
calculations include density, viscosity, surface ten-
sion, diffusivity and Sow rate of the liquid stream, as
 II / DISTILLATION / Tray Columns: Design 1137

Figure 3 Tray operation schematic diagram.

well as density, diffusivity and Sow rate of the vapour
stream. This information can be obtained by per-
forming tray-to-tray distillation calculations; several
commercial computer packages are available for this
purpose (e.g. PRO II, ASPEN PLUS, HYSIM).
As the physical properties and the vapour and
liquid Sow rates vary throughout a given column, it is
difRcult to provide a single design for the entire col-
umn. Instead, the column is divided into a number of
sections. Within each section, trays are designed with
the same layout. Normally the section is a set
of trays bounded by two column penetrations (feed
and/or drawoff). Tray design calculations
should be performed to ensure that trays at the top
and bottom of the section meet the design require-
ments.
Preliminary Speci\cations
Tray Spacing
Tray spacing is set by maintenance requirements, and
also by support structure design in large-diameter
columns. SufRcient crawl space must be provided for
tray cleaning and repair. From these considerations,
the minimum tray spacing is about 12 in (30 cm) for
column diameter less than 5 ft, and (150cm) and
18 in (45 cm) for a column diameter greater than
10 ft (300 cm). In general, it is best to keep tray
spacing to a minimum, which is often the most econ-
omical.
the stability of the froth in the downcomer and deter-
mined by the residence time required for achieving
the separation of the two-phase mixture. For non-
foaming systems, such as lower alcohols, a residence
time of 3 s is sufRcient, whereas for extremely high
foaming systems such as caustic regenerators, 9 s is
required.
To prevent the liquid coming off the bubbling area
from splashing against the column wall, the minimum
downcomer width is 5 in (12.7 cm). Also, the min-
imum side chord length should be 60% of the column
diameter. This is required to maintain good liquid
distribution on the tray.
Since the separation of the vapour}liquid mixture
is complete at the bottom of the downcomer, a sloped
downcomer can be used to maximize the active tray
area. In this case, the downcomer area at the bottom
should be about 60% of that at the top.

Downcomer Area
The downcomer area at the top is sized such that the
velocity of the ascending vapour bubbles exceeds the
downSow velocity of the liquid. The size is related to Figure 4 Sieve tray performance diagram.
1138 II / DISTILLATION / Tray Columns: Design
Figure 5 Sieve tray design procedure.
It should be noted that the downcomer area occu-
pies only a small fraction of the cross-sectional area.
Thus, a small overdesign does not result in a signiR-
cant economic penalty.
Column Diameter
The column diameter can be calculated once the tray
spacing and downcomer area have been speciRed.
The Fair correlation, based on the Souders and Brown
criterion, is recommended by most designers. The
vapour Sooding velocity can be calculated from
eqn [1].




L!
V 0.5
L 0.2
UN,f"CSB [1]

V 20
In eqn [1] CSB is the Souders}Brown coefRcient,

L and L (dyne cm\1) are liquid density and surface
tension, respectively, and
V is the vapour density in
the same units as
L . UN,f is based on the net area,
AN(ft2), which is the active area plus one downcomer
area. The unit for UN,f is ft s\1. The most popular
empirical formula for calculating CSB is given in
eqn [2].
CSB(ft s\1)"0.04232#0.1674TS
#(0.0063!0.2686TS)FlV
#(0.1448TS!0.008)F 2lV [2]
In this equation FlV"(L/V)(
V/
L)0.5, TS is tray spac-
ing in feet, and L and V are mass Sow rates of the
liquid and vapour. The CSB is valid for trays with
a fractional hole area greater than 10%. For areas of
8% and 6%, CSB should be multiplied by 0.9 and 0.8,
respectively.
Knowing UN,f and the total vapour Sow rate, the
column diameter can be calculated by assuming that
the column will be operated at a lower vapour velo-
city, say 80% of the Sood point.
Number of Flow Passes
The number of Sow passes is set to allow the tray to
operate at a weir loading that does not result in
excessive weir crest. The weir loading can be cal-
culated once the column diameter and the down-
comer area are determined. The optimum weir load-
ing is 4}6 US gallons per minute and the maximum
loading is about 20. Downcomer choking, which
causes liquid build-up on the tray, may occur if the
maximum value is exceeded. Increasing the number
of Sow passes provides a solution to this problem (see
Figure 2). However, shorter liquid Sow path and pos-
sible maldistribution of liquid and vapour streams in
multipass trays may result in lower tray efRciency.

As a rule of thumb, the liquid and vapour handling
capacity are a direct function of weir loading and
column area, respectively. Since weir length and col-
umn area are proportional to column diameter
and diameter squared, respectively, the use of
multipass trays is often necessary for large-diameter
columns.
Tray Geometry
Tray geometry should be chosen so that hydraulic
and efRciency calculations can be performed to arrive
at the optimum design. The following parameters
must be speciRed for tray design calculations.
Tray Thickness
The choice of material for the fabrication of trays is
dependent mainly on the corrosion properties of the
process Suids. In general, tray thickness is about
gauge 10 (0.134 in; 3.40 mm) for carbon steel and
gauge 12 (0.109 in; 2.77 mm) for stainless steel. For
economic reasons the holes are punched, which dic-
tates that the thickness must be less than the hole
diameter.
Hole Diameter
3
Small holes with a diameter in the range of 16 to 14 in
(4.76}6.35 mm) give better hydraulic and mass trans-
fer performance than the large ones in the range of 12
to 34 in (12.7}19.0 mm). However, large-hole trays
are cheaper and show more resistance to fouling.
Choose the hole size according to design require-
ments.
Hole Area
The hole area is normally in the range of 5}16% of
the bubbling area. Lower hole area allows the tray to
operate at higher efRciency and turndown ratio, but
at the expense of higher pressure drop. Since the
operating pressure of the column dictates the max-
imum allowable pressure drop, the hole area is se-
lected according to the type of service. Recommended
values are 5}10% for pressure and 10}16% for vac-
uum operations.
Hole areas below 5% are not used because the
distance between holes becomes excessive and liquid
channelling may occur. However, the distance can
also be adjusted by changing the hole diameter. In
general, the hole pitch should not be larger than
2.5 in (6.35 cm). On the other hand, if the hole areas
are greater than 16%, signiRcant weeping and en-
trainment may coexist and the design equations may
not apply under these conditions.
II / DISTILLATION / Tray Columns: Design 1139
Weir Design
Outlet weirs are used to control the froth height on
the tray. For most trays, the outlet weir height is
about 1}4 in (2.5}10 cm) and the downcomer clear-
ance, where the liquid is discharged from the bottom
of the downcomer onto the tray below, should be
0.5 in (1.25 cm) smaller than the outlet weir height to
ensure a positive downcomer seal.
From the above discussion, it may be concluded
that the object of tray design is to obtain the optimum
combination of the following parameters:
1. column diameter
2. tray spacing
3. top and bottom downcomer area
4. hole diameter and hole area
5. outlet weir height and downcomer clearance.
Design Criteria
The trays should be designed for maximum through-
put. However, owing to inaccuracies in the design
equations and Suctuation of process conditions (e.g.
Sow rates, temperature and pressure), safety factors
are needed to ensure stable column operation at all
times (see Figure 4).
Jet Flood Safety Factor
The jet Sood safety factor (JFSF) is deRned as the ratio
of vapour velocity required to entrain the entire liquid
Sow (Umax) to the operating velocity (Uop). It is a use-
ful measure of entrainment and hydraulic stability.
The typical JFSF value is 1.2.
Turndown Ratio
For various reasons, the column may be operated at
a reduced throughput. Weeping is encountered if the
vapour velocity can no longer support the liquid on
the tray. Although Sow dynamics permit stable op-
eration as long as dumping is avoided, tray efRciency
suffers because weeping reduces the vapour}liquid
contact. The turndown ratio is the ratio of the design
vapour Sow rate to the Sow rate that permits some
weeping without seriously affecting the tray efRcien-
cy. Recommended weepages at turndown conditions
for vacuum and pressure operations are 3% and 7%,
respectively.
Downcomer Area Safety Factor (DCASF)
and Downcomer Backup Safety Factor (DCBUSF)
The liquid handling capacity of a tray is determined
by downcomer design and tray spacing. The DCASF
determines the approach of the top downcomer area
1140 II / DISTILLATION / Tray Columns: Performance
to the minimum area to the minimum area required
for vapour}liquid disengagement. The DCBUSF de-
termines the approach of the downcomer froth height
to the downcomer depth ("tray spacing#outlet
weir height). Safety factors in the range of 1.5}2.0 are
recommended.
Pressure Drop
The pressure drop across an operating tray should be
speciRed if it affects the number of equilibrium stage
requirements for the separation. This is often the case
for vacuum applications. Stable operation can be
obtained at a pressure drop of 1}3 in (2.5}7.6 cm) of
liquid per tray for vacuum and 2}5 in (5.1}12.7 cm)
for pressure operations.
Design Calculations
Tray Hydraulics
The hydraulic performance of a sieve tray for a given
layout may be calculated using the methods presented
in ‘Distillation/Tray Columns: Performance’.
Tray Ef\ciency
Tray efRciency is a strong function of the physical
properties of the vapour and liquid streams. It is also
affected, to a lesser extent, by the Sow rates and tray
layout. In the latter case, only hole diameter, hole
area and weir height have a small inSuence on the
tray efRciency. The optimum design, which gives the
maximum number of equilibrium stages in a column,
is often obtained at minimum tray spacing and min-
imum number of Sow paths that satisfy the hydraulic
design criteria.
Conclusions
A well-designed tray should be economical while
meeting all process design requirements. Economic
Tray Columns: Performance
K. Nandakumar and K. T. Chuang, University of Alberta,
Edmonton, Alberta, Canada
Copyright ^ 2000 Academic Press
Introduction
As pointed out in the article entitled distillation tray
columns: design, a sieve tray is designed with a num-
considerations suggest that it is best to use the
smallest column diameter and height that satisfy the
process requirements within reasonable safety allow-
ances. Process requirements include accommodation
of the expected liquid and vapour Sow ranges and the
optimization of tray efRciency.
See also: II/Distillation: Packed Columns: Design and
Performance; Theory of Distillation; Tray Columns: Perfor-
mance.
Further Reading
Billet R (1979) Distillation Engineering. New York: Chem-
ical Publishing Co.
Fair JR (1963) In: Smith BD (ed.) Design of Equilibrium
Stage Processes. New York: McGraw-Hill.
Fair JR (1987) In: Rousseau RW (ed.) Handbook of
Separation Process Technology, ch. 5. New York: John
Wiley.
Fair JR, Steinmeyer DE, Peuney WR and Crocker BB
(1997). In: Perry RH and Green D (eds) Perry’s Chem-
ical Engineers’ Handbook, 7th edn, sect. 14. New York:
McGraw-Hill.
Humphrey JL and Keller GE II (1997) Separation Process
Technology. New York: McGraw-Hill.
Kister HZ, (1992) Distillation Design. New York
McGraw-Hill.
Lockett MJ (1986) Distillation Tray Fundamentals. Cam-
bridge: Cambridge University Press.
Lygeros AI and Magoulas KG (1986) Column Sooding
and entrainment. Hydrocarbon Processing 65:
43}44.
McCabe WL, Smith JC and Harriott P (1993) Unit Opera-
tions of Chemical Engineering, 5th edn. New York:
McGraw-Hill.
Ogboja O and Kuye A (1991) A procedure for the design
and optimization of sieve trays. Transactions of the
Institution of Chemical Engineers 445.
Rose LM (1985) Distillation Design in Practice. Amster-
dam: Elsevier.
ber of objectives in mind. They include: (i) achieving
high efRciency of contact between the liquid and the
vapour so that the phases leaving a tray are as close to
equilibrium conditions as possible; (ii) balancing the
tray deck area provided for vapour/liquid contact
with the downcomer area provided for disengage-
ment of the two phases so that neither limits the
capacity of the column to process large amounts
of feed; and (iii) avoiding detrimental operating
 II / DISTILLATION / Tray Columns: Performance 1141

conditions in the column such as weeping, Uooding or

high vapour entrainment.
Numerous geometrical factors have to be selected
by the designer such as: (i) column diameter; (ii) tray
spacing; (iii) top and bottom downcomer area; (iv)
number of Sow passes; (v) hole diameter and density;
(vi) tray thickness; and (vii) weir design. This is
a highly empirical process which depends on empiri-
cal design equations that describe the tray hydraulics
and rule-of-thumb guidelines that have evolved over
several decades of operating experience. Thus, the
design of sieve tray columns has remained an art,
although commercial process simulation software
packages such as ASPEN, PRO II, HYSIM, etc., are
trying to codify these procedures into their design
packages. The conceptual steps in the design proced-
ure together with the rule-of-thumb guidelines have
been presented in the Tray Columns: Design article.
Since frequent reference will be made to that article,
we will henceforth refer to it simply as Part I.
In contrast, the performance analysis problem is
relatively more scientiRc, in the sense that a series
of well-deRned steps leads to the estimation of the
Murphree tray efRciency, the column efTciency and
the actual number of trays. The overall column efR-
ciency, Eo, is deRned as:
Nequilibrium
Eo" [1]
Nactual
where Nequilibrium is obtained from stagewise equilib-
rium design calculations. Performance evaluation
boils down to estimating Eo so that the actual number
of trays, Nactual, can be determined. The overall col-
umn efRciency, Eo, is related to the Murphree tray
efRciency, EMV, through the Lewis relationship (as-
suming constant slopes of equilibrium and operating
lines), given by:
Figure 1 (A) Murphree tray efficiency. (B) Head in the down-
comer.
Figure 1A illustrates various compositions. yn is the
actual composition of the vapour stream leaving tray
n, while yHn is the composition that is in equilibrium
with the exit liquid stream.
These two compositions would be the same, if and
only if the condition of ideal equilibrium tray is satis-
Red. Since it is never satisRed in practice, it is impor-
tant to be able to predict the tray efRciency. In fact,
the compositions are not even uniform across the tray
deck. Hence the above deRnition is applied at a local
point on the tray and the point efRciency is integrated
with the variations in Uow conditions to predict a tray
efRciency. The relationship between the inputs and
the sequence of calculations is shown in Figure 2.
In Figure 2 the point efTciency is a function of local
Sow conditions such as local mass transfer coefR-
cients in the liquid and vapour phases. The dry Mur-
phree tray efRciency incorporates the effects of liquid
and vapour distribution on the point efTciency, while
the wet Murphree tray efRciency incorporates the
additional effects of entrainment and weeping.

ln [1#EMV(
!1)]
Eo"
ln

where
"mG/L is the separation factor, m is the

slope of the equilibrium line, and (G, L) are the
vapour and liquid Sow rates in kmol s\1. Thus the
Murphree tray efRciency, EMV, must be estimated in
order to determine the column efRciency. The urphree
tray efRciency is deRned to provide a measure of
departure from the assumption of ideal equilibrium
tray that is used to determine the number of ideal
stages required to achieve a given separation. It is
deRned as:

yn!yn 1
EMV" \ [2]
yHn !yn 1 Figure 2 Steps in performance evaluation.
\
1142 II / DISTILLATION / Tray Columns: Performance
The tray efRciency, EMV, clearly depends on: (i) the
geometrical design parameters chosen as outlined in
Part I; (ii) the physical properties of the system such as
density, viscosity, surface tension, etc.; and (iii) the
operating conditions like the vapour/liquid Sow
rates.
Having selected the design parameters identiRed in
Part I, the objective of the performance analysis step
is to predict: (i) the tray hydraulics (including the
pressure drop, the Sow regime, the froth density, the
entrainment and weeping factors); (ii) the point efR-
ciency; (iii) the Murphree tray efRciency; and (iv) the
column efRciency. In the initial stages of designing
a new tray column, there is feedback between the
design and performance analysis steps to arrive at
a set of optimal design parameters, as outlined in the
Sow chart (Figure 5 of part I). But the performance
analysis steps to be outlined in this part, are also
useful in analysing the performance of an existing
tray column, although the opportunity to pick opti-
mal design conditions is not present as one is forced
to deal with an existing tray.
An excellent summary of the equations used to
study the performance of a sieve tray column can be
found in Zuiderweg (1982), Lockett (1986) and Kis-
ter (1992) (see Further Reading). The detailed steps
involved in the performance analysis include: (i) the
pressure drop prediction; (ii) froth height and density
calculations; (iii) point efRciency prediction; and (iv)
tray efRciency prediction. The inputs required are: (i)
tray geometry; (ii) physical properties; and
(iii) Sow conditions.
Steps in Performance Analysis
Pressure Drop Calculation in Vapour Phase
The pressure drop in the vapour phase across a sieve
tray is modelled as (Zuiderweg, 1982):
P"Pdry#l ghl [3]
where the dry pressure drop is given by:


2
1 ug,h
Pdry" G [4]
2 CD
Here, g is the acceleration due to gravity, hl is the
liquid height or hold-up in metres, ug,h is the vapour
velocity in the hole in metres per second, CD is the
drag coefRcient, and (G, l) are the densities of the
vapour and liquid, respectively. The second term in
eqn [3] represents the static head due to the liquid
hold-up on the tray. Hence the liquid height, hl must
be predicted from correlations that depend on the
weir geometry. One such equation that predicts the
liquid height is given by:
hl"0.6H0.5
w p
0.25
(FP/b)0.25 25 mm(Hw(100 mm
[5]
Here, Hw is the weir height in metres, p is the pitch
of the holes in the sieve plate in metres,
FP"(ul/ug)((l/g) is the Sow parameter, b is the
weir length per unit bubbling area in metres\1.
The discharge coefRcient, CD in eqn [4] is a
function of the Sow conditions near a hole. It is in fact
dependent on the liquid present on the tray. It is
correlated by:

 
ghll 2/3
CD"0.7 1!0.14 [6]
u2g,hg
All of the quantities appearing on the right-hand side
have been deRned previously.
Pressure Drop Calculation in the Liquid Phase
The liquid is transported down through the down-
comer. The capacity of the downcomer should be
sufRcient to handle the liquid load without becoming
the rate limiting factor, i.e. without the liquid backing
up the downcomer to a signiRcant extent. Figure 1B
shows the pressure differential components making
up the total head differential on the liquid side as the
liquid backs up the downcomer to a height of hdc. The
extent of liquid back-up can be estimated from:
hdc"ht#hda#hL [7]
where ht is the pressure difference between points
a and b in the vapour phase that is necessary to keep
the vapour Sowing upwards, hL refers to the effective
clear liquid height on the tray deck that must be
overcome by the liquid in the downcomer, and
hda refers to the pressure loss due to the liquid Sow
under the downcomer apron. Note that ht is necessary
to keep the upward Sow of vapour, but acts as a pres-
sure differential that works against the natural liquid
Sow in the downcomer. If this pressure differential is
large, the liquid will back up more in the downcomer.
This points out the coupling between the pressure
loss in the vapour phase through the tray deck area
and the liquid Sow in the downcomer. An optimal
design must balance these two factors carefully.
hL and ht can be estimated from the correlations
provided in the previous section. hda can be estimated
from:
hda"165.2U2da
 II / DISTILLATION / Tray Columns: Performance
where hda is in millimetres of liquid and Uda is the
velocity under the downcomer apron in metres per
second.
Froth Height and Density Calculation
The froth density (or the two-phase density) has been
measured using gamma ray techniques. The average
liquid volume fraction on a sieve tray, deRned as
N l"hl/hb, is correlated by:


1 ug G 0.5 n
!1"c1 [8]
N l (ghl) 0.5
L
Here, hb is the froth or bed height in metres and ug is
the vapour velocity on bubbling area in metres per
second. The constants c1 and n depend on the type of
Sow regimes. In the spray regime, they take on the
values of c1"265 and n"1.7, while in the
mixed/emulsion regime, they are 40 and 0.8. This
requires one to estimate the Sow regime to be ex-
pected under a given set of operating conditions. In
Figure 3 of Part I, we identiRed the limits of operation
to lie between the weeping and Sooding conditions
as the vapour rate is increased. Even within this
permissible range of operation, the Sow condition has
been observed to change from spray to froth to emul-
sion to bubble Uow regimes. The transition into
the spray regime is given by the capacity factor de-
Rned as:


g 0.5
g h )F
0.5 1.5
l
CF"ug "0.85
l dh
Here, CF is the capacity factor deRned as ug((G/L)
in metres per second, ug is the vapour velocity in the
bubbling area in metres per second, F is the fractional
hole area per unit bubbling area and dh is the hole
diameter in metres. The transition from the
spray/froth to emulsion/bubble Sow regime is con-
trolled by the ratio of horizontal liquid momentum to
vertical vapour momentum and is given by:


ul l 0.5
FP
" '3.0
ug G b ) hl
where ul is the horizontal liquid velocity, ug is the
vapour velocity on bubbling area in metres per second,
and FP is the Sow parameter deRned in eqn [5], b is the
weir length per unit bubbling area in metres\1, hl is the
liquid height or hold-up in metres.
Point Ef\ciency Calculation
There are many empirical correlations for predicting
the mass transfer efRciencies on sieve trays. The most
recent one is that proposed by Chen and Chuang
(1993). It is based on data from industrial sized col-
umns of Fractionation Research Inc. The point efR-
1143
ciency is related to the overall number of transfer
units by:
EOG"1!e\NOG [9]
Chen and Chuang present the following correlation for
NOG using data free of weeping and entrainment. But
the data set spans both the froth and spray regimes:
 
1 LF2s 1/3
11 0.1 0.14 (DGtG)0.5

2
NOG" [10]



11 DGG 0.5 MGL

#1
14 DLL MLG
Here
"mG/L is the separation factor, Fs"us(G
is the superRcial F-factor in kg0.5/m0.5s, tG"hf/us is
the vapour-phase contact time in seconds, and hf is
the froth height in metres. Note that this correlation
combines the geometrical parameters such as
, the
fractional perforated area, Ab the bubbling area, the
system properties such as densities (L, G), diffusivi-
ties (DL, DG) viscosity (), the interfacial tension ()
in newtons per metre, the molecular weights
(ML, MG), and operating conditions such as (L, G),
Sow rates. This correlation appears to predict the
point efRciencies to within 5% of experimental data
over a wide range of pressures.
Murphree Tray Ef\ciency Calculation
The point efRciency model presented above is based
on a detailed examination of mass transfer at the
vapour/liquid interface. The ideal equilibrium tray
assumption used in the McCabe}Thiele method as-
serts that the Sow condition on a tray is homogeneous
everywhere. If that were true, the point efRciency
would be the same everywhere on the tray. But there
is strong evidence that the Sow is not homogeneous,
the degree of inhomogeneity being larger in large
diameter columns. Several researchers have tried to
measure the velocity proRles across a sieve tray and
increasingly computational Suid dynamics is being
used as a tool to predict such Sow Relds. (See for
example Solari and Bell (1986) and Mehta et al.
(1998)). This information on Sow proRle must be
integrated with the point efRciency calculations in
order to predict a Murphree tray efRciency. One such
method is given below as an illustration. This model
considers only the effect of longitudinal mixing.
A measure of the effective diffusivity, DE is needed in
this model. Models of other Sow conRguration are
discussed in Lockett (1986):
EMV 1!e\(E#Pe)
"
EOG (#Pe)1#[(#Pe)/]
e\E!1
# [11]
1#[/(#Pe)]
1144 II / DISTILLATION / Tray Columns: Performance

where: sieve plate, instead of the downcomer, which is the

preferred path for the liquid. If weeping is signiRcant,


 
1/2
Pe 4
EOG
" 1# !1 then it results in mixing of liquid streams between
2 Pe two neighbouring trays, thus degrading the perfor-
and Pe is the Peclet number, deRned as Pe"Z2l/DEtl. mance of the column. The need to avoid weeping
Here Zl is the length of liquid travel, or the distance places a limit on the minimum vapour velocity.
between the two weirs and tl is the liquid residence Zuiderweg presents the following correlations to pre-
time. The effective diffusivity is given by: dict the minimum operating limit.

(DE"0.0124#0.017uG#0.0025L#0.0150W Mixed/free bubbling regime

 
[12] FP
CFw"F(ghl 1!0.15
where DE is in square feet per second, uG is superRcial bhl
gas velocity, expressed as cubic feet per second
divided by the active bubbling area in square feet. As
the Peclet number becomes large, this model predicts Emulsion Uow regime
efRciency enhancement much large than unity. In
large diameter columns (large Zl) the Peclet number CFw"0.45F(ghl
can tend to take a large value which would suggest
signiRcant efRciency enhancements. But it should be
remembered that the above model considers only the where CFw"ug(G/(L!G) is the capacity factor
longitudinal mixing process. In large diameter col- at the weep point in metres per second, and F is the
umns, the liquid Sow structure can be much more fractional hole area per unit bubbling area. Correla-
complicated as documented by Solari and Bell (1986). tions to estimate the type of Sow regime are given by
Hence, predicted values of EMV/EOG greater than 1.2 Zuiderweg. Note that weeping will seldom occur in
by the longitudinal mixing model should be viewed the spray regime as vapour velocities are sufRciently
with caution, as they may not be realized in the Reld. large under design conditions. The effect of weeping
on the tray efRciency calculation has been studied by
Effect of Entrainment on Murphree Kageyama (1969).
Tray Ef\ciency
The effect of entrainment on the Murphree tray efR- Extensions to Multicomponent
ciency is estimated from: Systems

 
1 The methods outlined above have been developed
EMV,entrain"EMV
1#EMV /(1! )
where:
e absolute entrainment
" "
L#e total liquid flow rate
where e is the entrained liquid in moles per hour.
Zuiderweg presents the following empirical equation
to predict the liquid entrainment in the spray regime:
[13]
largely using experimental data for binary, two-phase
systems. The question of whether they can be applied
to multicomponent systems can be examined as fol-
lows. Tray hydraulics factors such as pressure drops,
Sow regimes, froth densities, etc., depend only on the
Suid mechanics of the two-phase mixture on sieve
trays; hence one can expect the correlations to be
useful for multicomponent mixtures as long as mix-
ture properties for densities, viscosities, interfacial
tensions, etc., are used. On the other hand, the point



3
hb ug,h hb efRciency (and hence the Murphree tray efRciency)
"1.0;10\8 for 0.3( (0.9
Hs ul Hs depends on the mass transfer resistance of each
component species in each phase. Since the diffusivi-
Here Hs is the tray spacing in metres, hb is the bed ties and the equilibrium ratios (or the slope of the
height as deRned in eqn [8], ug,h is the vapour velocity equilibrium curve, m) could vary for each species, the
in the hole in metres per second and ul is the horizon- point efRciency will be different for each species. The
tal liquid velocity. correlation given in eqn [10] is based on binary mass
transfer data.
Weeping Point Determination
In the pseudo binary method of calculation (see
When the vapour velocity is too small, the liquid on Kister, 1992) two components are identiRed as
a tray deck can Sow down through the holes on the the light key and heavy key components and the
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction
Murphree tray efRciency is determined for such a bi-
nary pair. One then has the option of either using the
efRciency so calculated for all of the remaining com-
ponents or repeating the procedure for all possible
binary pairs. Such detailed estimates of com-
ponent efRciencies are then used as inputs to ad-
vanced process simulators such as ASPEN.
Issues Relating to Scale-up
of Ef\ciency Data
Since the point efRciency data and correlations (like
eqn [10] are (or should be) based on local conditions,
they should, in principle, remain valid on all scales.
They are then integrated with Sow conditions to
predict the overall tray efRciency. Correlations such
as eqn [11], which provide this function of integrat-
ing the point efRciency to provide tray efRciency, do
not remain valid at all scales. It has been well
documented that the liquid Sow patterns change
quite dramatically depending on the diameter of the
column and the location of the weirs near the down-
comer. In future one can expect computational Uuid
dynamics to provide detailed Sow information using
models that remain scale invariant over a wide range
of diameters.
Concluding Remarks
A series of correlations taken from the literature are
presented. They permit the evaluation of the perfor-
mance of a sieve tray, once a set of design parameters
has been chosen as outlined in Part I. At the design
stage of a new sieve tray column, one can embed this
design and performance analysis steps into an optim-
ization procedure, in such a way that the design
parameters may be altered until a speciRed objective
function is satisRed. The objective function could be
Vapour+Liquid Equilibrium: Correlation and Prediction
B. C.-Y. Lu, University of Ottawa, Ottawa,
Ontario, Canada,
D.-Y. Peng, University of Saskatchewan,
Saskatoon, Saskatchewan, Canada
Copyright ^ 2000 Academic Press
Introduction
Distillation is a process used to separate liquid mix-
tures into two or more streams, each of which has
a composition that is different from that of the
1145
a cost function that includes the capital cost of the
equipment (which determines the column diameter,
tray spacing, etc.) and operating costs (which deter-
mine the reSux and reboil rates and the number of
ideal stages).
See also: II/Distillation: Historical Development; Instru-
mentation and Control Systems; Theory of Distillation;
Tray Columns: Design; Packed Columns: Design and Per-
formance; Vapour-Liquid Equilibrium: Correlation and Pre-
diction; Vapour-Liquid Equilibrium: Theory.
Further Reading
Chen GX and Chuang KT (1993) Prediction of Point EfT-
ciency for Sieve Trays in Distillation. I & EC Research,
vol. 32, p. 701.
Fair JR et al. (1997) In: Perry RH and Green D (eds.),
Perry’s Chemical Engineers’ Handbook } Section 14,
7th edn. New York: McGraw-Hill.
Kageyama, O. Plate efRciency in distillation towers with
weeping and entrainment, I. Chem. E. Symposium Series
No. 32.
Kister HZ (1992) Distillation Design. New York:
McGraw-Hill.
Lockett MJ (1986) Distillation Tray Fundamentals. Cam-
bridge University Press.
Mehta B, Chuang KT and Nandakumar K (1998) Model
for liquid phase Sow on sieve trays. Transactions of the
Institute of Chemical Engineers, part A (in press).
Rose LM (1985) Distillation Design in Practice. Amster-
dam: Elsevier.
Rousseau RW (1987) Handbook of Separation Process
Technology, New York: John Wiley & Sons.
Solari RB and Bell RL (1986) Fluid Sow patterns and
velocity distributions on commercial scale sieve trays.
American Institute of Chemical Engineers Journal 32:
640.
Zuiderweg FJ (1982) Sieve trays: A view of the state of the
art. Chemical Engineering Science 37: 1441.
original mixture. The process involves both the va-
porization of the original liquid in order to
generate the vapours and the subsequent condensa-
tion of the vapours to form the desired liquid
products. It is evident that vapour}liquid equilibria
(VLE) are essential to this separation process.
Typical temperature}composition (T}x}y) diagrams,
pressure}composition (P}x}y) diagrams, and va-
pour}liquid composition (x}y) diagrams for com-
pletely miscible binary systems are depicted in
Figure 1.
1146 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction

Figure 1 Four types of binary P}x}y, T}x}y and x}y equilibrium phase diagrams. Correlation curves are for: (A) System
acetone#benzene. (B) System 1-butanol#acrylic acid. (C) System cyclohexane#2-propanol. (D) System acetone#chloroform.
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction
Consider an N-component closed system in which
a liquid mixture at temperature T and pressure P is in
equilibrium with a vapour mixture at the same tem-
perature and pressure. The condition of thermodyn-
amic equilibrium is that the chemical potentials of the
components in both phases must satisfy the equality
relation:

iV"
iL (i"1, 2, 2, N) (constant T) [1]
The introduction of the quantity called fugacity by
G.N. Lewis has facilitated the application of this
condition. The fugacity of a system, f, at constant
temperature is deRned in the following two equa-
tions:
dG"RT d ln f [2]
and:
lim ( f/P)"1 as P approaches zero [3]
where, in eqn [2], G is the molar Gibbs energy.
The fugacity of a component i in a solution is
related to the chemical potential of the same compon-
ent by the equation:
d
i"RT d ln fK i [4]
A practical expression for vapour}liquid equilib-
rium consideration thus takes the form:
fK iV"fK iL [5]
where the subscripts V and L indicate fugacity in the
vapour phase and liquid phase, respectively. Correla-
tion and prediction of vapour}liquid equilibria must
satisfy the equal fugacity condition. The main con-
cern is to relate these fugacities to T, P, and the
compositions of the liquid and vapour phases.
Design of distillation operations requires reliable
experimental vapour}liquid equilibrium values for
conditions corresponding to the desired operation.
Available data may frequently be either fragmentary
or for conditions different from the desired operating
conditions. On many occasions, the needed experi-
mental values are not available at all. In order to
make suitable interpolation and extrapolation of the
available data, and to make acceptable estimates of
unavailable data, it is necessary to take advantage of
the limited data available and apply prediction
methods developed on the basis of reasonable as-
sumptions. In this article, discussion is limited to
correlation and prediction of the vapour}liquid equi-
librium values for organic and nonelectrolyte mix-
tures at conditions such that Raoult’s law cannot be
1147
used to represent the behaviour of all components
over the complete concentration range. The emphasis
is placed on the equilibrium T}P}x}y.
Correlation of Vapour+Liquid
Equilibria (VLE)
The quality of the available experimental data is the
Rrst concern of any VLE correlation. It is known that
a considerable proportion of experimental values are
not of good quality owing to the impurity of the
chemicals and poor equilibrium stills used, equipment
set-ups, and operator errors. In many instances there
are quantitative discrepancies among the experi-
mental results for the same system investigated under
the same thermodynamic conditions by different
authors. Therefore, it is desirable to determine
whether the available experimental values are ther-
modynamically consistent prior to correlation. Con-
sistency tests can be applied whenever the measured
properties are more than those that are needed, on the
basis of the phase rule, to deRne the intensive proper-
ties of the system under consideration. Although the
thermodynamic consistency of experimental data
does not guarantee their correctness, inconsistent
data are deRnitely not acceptable.
There are two frequently used methods in correlat-
ing vapour}liquid equilibria. One is through the
gamma}phi approach, and the other is by means of
an equation of state. A comparison of the two
methods is presented in Table 1.
Thermodynamic Consistency Test of Data
The Gibbs}Duhem equation is a differential equation
that represents the interrelationship among the cha-
nges of T, P and composition (in terms of chemical
potentials) of an equilibrium system. The equation
has the form:
!ns dT#nv dP! ni d
i"0 [6]
When the Gibbs}Duhem equation is applied to the
experimental results for systems under isothermal
conditions and at low to moderate pressures, eqn [6]
is reduced to ni d
i"0. In terms of liquid activity
coefRcient, which is deRned by:
i"fK iL/xi f sat
i [7]
the simplest expression for testing thermodynamic
consistency has the form:
xi d ln i"0 [8]
1148 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction

Table 1 Comparison of vapour}liquid equilibrium calculation methods

Method Gamma}phi approach Equation-of-state approach

Advantage Applicable to a wide variety of mixtures, including
polar systems, electrolytes, and polymers.
Simple solution models suffice for the correlation
of vapour}liquid equilibrium data.
Disadvantage Difficult to apply to systems involving supercritical
components.
Additional correlations must be used to represent
the volumetric behaviour and thermal properties.
Applicable to a given system over wide ranges of
temperature and pressure, including the supercritical
region
Thermodynamic properties, such as the enthalpy and
entropy, can be consistently calculated from the same
equation of state.
A single equation cannot represent the properties of all
components precisely at the same time.
Conventional mixing and combining rules are not
applicable to systems containing polar components,
polymer molecules, or electrolytes.

In eqn [7], the standard-state fugacity, fi sat, is the Gamma+Phi approach

fugacity of component i at the system temperature.
In the gamma}phi approach, the dimensionless fugac-
The standard state is usually taken to be the pure
ity coefRcients and the activity coefRcients 
liquid at the T and P of the system. As data for binary
are used to describe the vapour phase and the liquid
systems are the basis for further correlation, it is
phase:
desirable to have their consistency tested Rrst. The
simplest method is the ‘visual test’, which is indepen- fK iV" K iyiP [9]
dent of the models of equations used for expressing
fK iL"i xi f sat
i [10]
the excess Gibbs energy. A brief description of the
visual test method is presented in Table 2 with exam- The fugacity coefRcients can be calculated from the
ples depicted in Figure 2. More precise testing vapour phase PTv composition data by means of an
methods, such as the point-by-point test and the area equation of state, such as the virial equation. When
test, which take into consideration the effect of tem- the system pressure is low, either of the volume-
perature change on the data for systems at isobaric explicit virial equation and the pressure-explicit virial
conditions or the effect of pressure change on the data equation may be truncated after the second term and
for systems at isothermal conditions, are available. used for the calculation. The resulting expressions are
However, excess enthalpies or volume changes owing respectively:
to the mixing of components may be required. When-
ever the liquid activity coefRcients obtained from
experimental data can be represented by an integ-
and:
ln K i"(P/RT) 2

yiBij!B
 [11]

rated Gibbs}Duhem equation (an appropriate model-

ling equation for ), the data are considered thermo- ln K i"(2/v) yiBij!ln Z [12]
dynamically consistent. For high pressure VLE, test-
ing methods such as those developed by Won and In these two equations, the expression:
Prausnitz in 1973 and Christiansen and Fredenslund
in 1975 may be applied. B" yiyjBij [13]

Table 2 Visual test of consistency of activity coefficients for binary systems

1. log 1 evaluated at x1"0.25 should be approximately equal to log 2 evaluated at x1"0.75.

2. Let  be the value of log 1 evaluated at x1"0 and  be the value of log 2 evaluated at x1"1.
log 1 evaluated at x1"0.5 should be approximately equal to 0.25.
log 2 evaluated at x1"0.5 should be approximately equal to 0.25.
3. If  is greater than or equal to , then the value of log 1 evaluated at x1"0.5 should be less than or equal to the value of log 2
evaluated at x1"0.5.
4. If  is less than , then the value of log 1 evaluated at x1"0.5 should be greater than the value of log 2 evaluated at x1"0.5.
5. Both log i versus xi curves show horizontal tangency as xi approaches unity and log i approaches zero.
6. If there is a maximum (or minimum) on one of the log i versus xi curves, there is a corresponding minimum (or maximum) on the
other curve at the same x1.
7. If there is neither a maximum nor a minimum on the curves, both curves should be on the same side of the horizontal line that is
representing log i"0.
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction 1149

Figure 2 Visual consistency test of binary vapour}liquid equilibrium of (A) the acetone#benzene system at 318.15 K (Brown I and
Smith F (1957) Liquid}vapor equilibrium VIII. The system acetone#benzene and acetone#carbon tetrachloride at 453C. Australian
Journal of Chemistry 10: 423I428), (B) the heptane#3-pentanone system at 368.15 K (Geiseler G and Koehler H (1968) Thermodyn-
amic behavior of the binary systems methyl ethyl ketoxime/n-heptane, diethyl ketone/n-heptane, and methyl ethyl ketoxime/diethyl
ketone. Berichte, Bunsengesellschaft fuel Physikalische Chemie. 72: 697I706), (C) the toluene#n-octane system at 101.3 kPa.
(Bromiley EC and Quiggle, D (1933) Vapor}liquid equilibria of hydrocarbon mixtures. Industrial and Engineering Chemistry 25:
1136}1138).

is used for the second virial coefRcient of the mixture
under consideration. When iOj, the cross second
virial coefRcient Bij represents the interaction between
molecule i and molecule j. Z is the compressibility
factor of the mixture. The i values obtained from
eqns [11] and [12] are practically identical. A general-
ized method proposed by Hayden and O’Connell in
1975 may be used for predicting the second virial
coefRcients for pure components and the cross second
virial coefRcients. For an ideal gas mixture, K i is
unity.
Expressions are available for representing the de-
pendence of i on the composition of the solution. The
activity coefRcients are related to the excess Gibbs
energy function by the equation
GEi "RT ln i
where G "(nG /ni)T,P,nj. Some of these expres-
E
i
E
sions, such as the two-parameter equations of van
Laar, those of Margules, and the multiparameter
equations of Redlich and Kister, are listed in Table 3.
Additional expressions involving higher-order terms,
such as those based on the local composition concept
proposed by Wilson in 1964, the nonrandom two-
liquid (NRTL) model by Renon and Prausnitz in
1968, and the universal quasi-chemical theory
(UNIQUAC) equation by Abrams and Pransnitz in
1975, are frequently applied to activity coefRcient
calculations for binary and multicomponent systems.
According to eqn [8], the relation between ln 1 and
ln 2 for a binary system at constant temperature and
at low to moderate pressure, is given by:
xi(d ln 1/dx1)#x2(d ln 2/dx1)"0
which indicates that the two activity coefRcients are
not independent of each other. These activity coefR-
cients may be evaluated by means of the equations
listed in Table 4. If the vapour phase can be con-
sidered ideal at low pressures, the calculation of  is
much simpliRed. Pure-component fugacities may be
substituted directly by the pure-component vapour
pressures:
pi,yiP"i xi psat
i [16]
where pi is the partial pressure of component i, P the
system pressure, and psat
i the vapour pressure of pure
component i.
The effect of temperature on ln i is a concern in
data correlation; a suitable representation of the tem-
[14] perature effect permits the determination of data for
isobaric conditions from data for isothermal condi-
tions, and vice versa. However, a consistency test for
isothermal data is much easier than that for isobaric
data because the pressure effect is generally much
smaller than the temperature effect. The effect of
temperature on ln i is related to the partial molar
enthalpy. Lu in 1959 considered the variation of
excess enthalpies with temperature for binary systems
and suggested that the variation of ln i at constant
liquid composition be represented by an expression
involving three terms for data interpolation and ex-
trapolation:
ln i"a#(b/T)#c ln T (constant composition)
[17]
In the absence of excess enthalpy data, isothermal
[15] data determined at three conditions sufRce for the
1150 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction

Table 3 Selected acitivity coefficient models

Name G E /(RT ) ln i

Margules G E /(RT )"x1x2 (Ax1 #Bx 2) ln 1"x 22 [B #2(A !B )x 1]

ln 2"x 21 [A#2(B !A )x 2]
van Laar G E /(RT )"ABx 1x 2/(Ax 1#Bx 2) ln 1"A [1#Ax 1/(Bx 2)]\2
ln 2"B [1#Bx 2/(Ax 1)]\2
Redlich}Kister G E /(RT )"x 1x 2 [A #B (x 1!x 2)#C (x 1!x 2)2 ln 1"a 1x 22#b 1x 32#c 1x 32#d 1x 42#2
#D (x 1!x 2)3#2] ln 2"a 2x 21#b 2x 31#c 2x 31#d 2x 41#2
where: a 1"A #3B #5C #7D #2
b 1"!4(B #4C #9D )#2
c 1"12(C #5D )#2
d 1"!32D #2
a 2"A !3B #5C !7D #2
b 2"4(B !4C #9D )#2
c 2"12(C !5D )#2
d 2"32D #2
Wilson G E /(RT )"! i xi ln ( j xj ij ) ln i "1!ln( j xj ij )! k [xk ki /( j xj kj )]


NRTL G E
21G 21 12G 12


 
"x 1x 2 # G 21 2
12G 12
RT x 1#x 2G 21 x 2#x 1G 12 ln 1"x 22 21 #
x 1#x 2G 21 (x 2#x 1G 12)2
g 12 g 21
where: 12" , 21"
RT RT


 
G 12 2
21G 21
ln G 12"!a 1212, ln G 21"!a 1221 ln 2"x 21 12 #
x 2#x 1G 12 (x 1#x 2G 21)2


i i ri
UNIQUAC G E"G E (combinatorial)#G E (residual) ln i "ln #5qi ln # j li ! lj
xi i rj


G E (combinatorial) 1 2 ji ij
"x 1 ln #x 2 ln !qi ln ( i # j ji )# jqi !
RT x1 x2 i # 
j ji j # i ij

#5 q1x1 ln
1

1
#q 2x 2 ln
2

2  where: i"1, j"2 or i"2, j"1

GE (residual)
"!q1x1 ln ( 1# 221)!q2x2 ln ( 2# 112)
RT

x 1r 1 x 1q 1 z
"
1 , 1" li " (ri !qi )!(ri !1)
x 1r 1#x 2r 2 x 1q 1#x 2q 2 2
u21 u 12 z
ln 21"! , ln 12"! lj " (rj !q j )!(rj !1)
RT RT 2

determination of isobaric data within a reasonable Equation-of-State Approach

range of temperatures. Similarly, if isobaric vapour}
liquid equilibrium data are available at three condi-
tions, isothermal data can be obtained by the same
approach and then tested for consistency. The num-
ber of sets of vapour}liquid equilibrium data required
can be reduced when excess enthalpies are available,
but generally one set of experimental values should
be used in the correlation. In the absence of the
required data for the determination of parameters in
eqn [17], RT ln i at a given composition may be
assumed to be constant as an approximation. The
correlated results can also be used for the prediction
purposes.
Fugacities of both phases are represented in this ap-
proach by the same equation of state, which provides
a relationship between the intensive thermodynamic
variables T, P, v and composition. Such an equation
may be explicit in P or v. The pressure-explicit equa-
tions in the form of :
P"P(T, v, x1, x2, 2 , xn 1) [18]
\
are more useful for solving phase-equilibrium prob-
lems. In terms of the fugacity coefRcients,
K iV ("fK iV/yiP) and K iL ("fK i L/xiP), formulation of
vapour}liquid equilibria is based on the equilibrium
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction 1151

Table 4 Barker’s method for the determination of activity coefficients from experimental
data

At equilibrium:
fK i l "fK i v
fK vi
K vi , (by definition)
yi P
yi P K vi "xi i f li
fi l f sat
i
K li " , K sat
i " sat
P pi
Therefore:
yi P K vi "xi i l
i P
ln (yiP K )"ln (xi i )#ln
v
i
l
i #ln P
v il
"ln (xi i )#ln sat
i # (P !p sat sat
i )#ln pi
RT
Hence:
yi P v li (P !p sat
i )
ln i "ln #ln K vi !ln sat
i !
xi psat
i RT
For a binary system:
B "y 21B 11#2y 1y 2B 12#y 22B 22
Let:
12 "2B 12!B 11!B 22
Then:
B "y 1B 11#y 2B 22#y 1y 2 12

At low pressure:
PvI BP
Z " "1#
RT RT


B 11#y 22 12
ln K v1 " P
RT
For pure component i :


P
dP Bii P
ln i " (Zi !1) , Zi "1#
0 P RT
sat
p i dP Bii sat
ln sat
i " (Zi !1) " pi
P RT
0


y 1P B 11#y 22 12 B 11 sat v l1(P !p sat
1 )
ln 1"ln sat
# P! p1 !
x 1p 1 RT RT RT
y 1P (B 11!v )(P !p sat
1 ) y 22 12P l
1
"ln sat
# #
x 1p 1 RT RT
Similarly:
y 2P (B 22!v l2)(P !p sat
2 ) y 21 12P
ln 2"ln # #
x 2p sat
2 RT RT

Barker JA (1953) Determination of activity coefficients from total-pressure measurements.

Australian Journal of Chemistry (1953) 6: 207}210.


equations: 
RT ln K i" [(P/ni)T,V,nj!RT/V] dV!RT ln Z
V
yi K iV"xi K iL (i"1, 2,2 , N) [19] [20]

with both the K iV and K iL calculated from the equa- The advantage of this approach is that it is applic-
tions: able to calculations of VLE at high pressures and it
1152 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction
can also be used to obtain other conRgurational prop-
erties such as enthalpy, entropy and volume changes
of mixing, which are useful in the design of distilla-
tion columns.
The Rrst equation of state with a theoretical foun-
dation was proposed by van der Waals in 1873,
several decades after the ideal gas equation of state
had been formulated. This equation not only yields
qualitatively correct representation of the phase be-
haviour of a real Suid, but also provides the basis of
the principle of corresponding states. Hundreds of
equations of state have been developed since the pub-
lication of the van der Waals equation. They may be
theoretical, semi-theoretical or empirical. However,
most of the modiRcations are generally limited to
a speciRc purpose.
In order to apply an equation of state to va-
pour}liquid equilibrium calculations for pure compo-
nents, a suitable equation should satisfy the three
conditions at a given saturation temperature:
vV,calc."vV, vL,calc."vL, fV,calc."fL,calc. [21]
Mixing and combining rules for the equation para-
meters are required for extending its application to
mixtures. However, most of the practical equations
available at present have their inherent advantages
and disadvantages and may not satisfy both of the
volumetric conditions.
The equations of state expressed in terms of poly-
nomials in volume are of practical importance. For
VLE calculations, especially when the properties un-
der consideration are limited to T, P and composi-
tions, the simplest and frequently used form is that
which is cubic in v. In spite of their shortcomings,
these cubic equations are the most frequently used in
practice at present. Currently, the most popular two-
parameter cubic equations of state include the
Soave}Redlich}Kwong equation (1972):
P"RT/(v!b)!a/[v(v#b)] [22]
and the Peng}Robinson equation (1976):
P"RT/(v!b)!a/[v(v#b)#b(v!b)] [23]
Both equations can be obtained from a general form
of a four-constant cubic equation of the van der
Waals type:
P"RT/(v!b)!a/[(v#c1b)(v#c2b)] [24]
Additional multiparameter cubic equations, which
are of the form represented by eqn [24] but developed
for improving the representations of pure-component
vapour pressures, saturated liquid volumes, the criti-
cal compressibility factors, and phase behaviour of
polar}nonpolar mixtures, appear continuously in the
literature. The maximum number of parameters in
a cubic equation is Rve. A list of some selected cubic
equations of state is presented in Table 5. In some
of the cubic equations, different repulsion terms (the
Rrst term on the right-hand side of the equations
listed in the table) have been adopted. The forms of
these equations are frequently inSuenced by the desire
to improve the theoretical basis of the equation, and
that of Rtting the volumetric properties. It should also
be mentioned that one of the inherent limitations of
a two-parameter equation is that the critical com-
pressibility factor is a constant for all components.
The ability of a cubic equation in VLE representation
is controlled by the selection of an adequate temper-
ature function for the parameter ‘a’ for vapour pres-
sures of pure components, and a set of suitable mix-
ing and combining rules for all the parameters of the
equation for mixtures.
Temperature function for ‘a ’ The importance of
using a proper temperature function to represent the
parameter ‘a’ cannot be overemphasized. In the
1960s, Wilson began the consideration of the temper-
ature effect on the parameter ‘a’ of the
Redlich}Kwong equation. The expression which has
gained wider acceptance was developed by Soave for
the same equation in 1972. The parameter ‘a’ was
expressed by:
a"ac [25]
with  expressed by a function involving the reduced
temperature Tr ("T/Tc) in the form:
"[1#m(1!T1/2
r )]
2
[26]
where the subscript c refers to the critical-point con-
dition, and m represents a quadratic function of the
acentric factor of Pitzer. This form and its variations
have been adopted subsequently in many cubic equa-
tions. A selected set of temperature functions for the
parameter ‘a’ is listed in Table 6.
Mixing and combining rules To extend the applica-
tion of his equation of state to representing the behav-
iour of mixtures, van der Waals proposed that the
constants ‘a’ and ‘b’ be expressed by:
a" xi xj aij [27]
a" xi xj bij [28]
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction 1153

Table 5 Selected cubic equations of state and the corresponding fugacity coefficient expressions

Equation of state Fugacity coefficient for pure component i Fugacity coefficient for component i in mixture a


A B Bi
Soave}Redlich}Kwong (1972) ln "Z!1!ln(Z!B )! ln 1# ln " (Z!1)!ln(Z!B )
i
B Z B
RT a (T )

  
P" ! aP bP Pv A 2 j yj aji Bi B
v!b v (v#b) where: A" 2 2 , B" , Z" ! ! ln 1#
R T RT RT B a B Z

Bi
ln " (Z!1)!ln(Z!B)
Peng}Robinson (1976) ln "Z!1!ln(Z!B) i
B

 
  
RT a (T ) A Z#(1#(2)B A 2 j yj aji Bi Z#(1#(2)B
P" ! ! ln ! ! ln
v!b v (v#b)#b(v!b) 2(2B Z#(1!(2)B 2(2B a B Z#(1!(2)B
aP bP Pv
where: A" 2 2 , B" , Z"
RT RT RT


bi
Patel}Teja (1982) ln "Z!1!ln(Z!B) RT ln i "!RT ln (Z!B)#RT
v!b

 
RT a (T ) a Z#M


P" ! # ln xj aij Q#d a(bi #ci )
v!b v (v#b)#c(v!b) 2RTN Z#Q ! ln #
d Q!d 2(Q 2!d 2)
bP Pv a
where: B" , Z" # 3 ci (3b#c)#bi (3c#b)
RT RT 8d


b#c P

 
M" !N Q#d 2Qd
2 RT ; ln # 2
Q!d Q !d 2
where:

 
(b#c)2 \1/2
N" bc#

2 b#c (b#c)2
Q"v# , d" bc#
2 4


b#c P
Q" #N
2 RT

 
b1i a b2i b3i
Adachi}Lu}Sugie (1983) ln "Z!1!ln (Z!B1) ln "
i ! #
v!b1 RT(b2#b3) v!b2 v#b3


RT a (T ) v!b2


a a 2 j yj aji b2i#b3i
P" ! # ln # !
v!b1 (v!b2)#(v#b3) RT(b2#b3) v#b3 RT (b2#b3) a b2#b3

 
v!b2
;ln !ln(Z!B1)
v#b3

bj"
i
xibji


  
v bi v a
Iwai}Margerum}Lu (1988) ln "Z!1!ln Z#ln ln "
i #ln !ln Z# (v 2!cv#cb)
v!b v!b v!b RTA
RT a (T )

 
P" ! 1 a ;[(bi c#ci b)(2v!c)#ci (2bc!cv )]
v!b v 2#ub(v!b) !
RT (c2!4bc)0.5

 
a

 
2v!c#(c2!4bc)0.5 # 2 xj aij# 2(bi c#ci b)!cci
;ln j A
2v!c!(c2!4bc)0.5
1 2v!c!(A
where: ; ln
Pv RT(A 2v!c#(A
Z" , c"!bu
RT where:
A"c 2!4bc
a
The mixing rule: a" i j xi xj aij , b" i xi bi, c" i xi ci; aij "(ai aj )1/2(1!kij ). References: Adachi YB, Lu BC-Y and Sugie H (1983)
A four-parameter equation of state. Fluid Phase Equilibria 11: 29}48. Iwai Y, Margerum R and Lu BC-Y (1988) A new three-pararmeter
cubic equation of state for polar fluids and fluid mixtures. Fluid Phase Equilibria 42: 21}41. Patel NC and Teja AS (1982) A new cubic
equation of state for fluids and fluid mixtures. Chemical Engineering Science 37: 463}473. Peng D-Y and Robinson DB (1976) A new
two-constant equation of state. Industrial and Engineering Chemistry Fundamentals 15: 59}64. Soave G (1972) Equilibrium constants
from a modified Redlich}Kwong equation of state. Chemical Engineering Science 27: 1197}1203.
1154 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction

Table 6 Some different forms of the  function proved theoretical considerations have appeared in
the literature. A list of some mixing and combining
Form Reference rules is presented in Table 7.
"1#m(1!Tr) 1 In general, vapour}liquid equilibrium of a great
"[1#m(1!(Tr)]2 2 variety of mixtures, including polar}nonpolar mix-
"[1#m1(1!(Tr)#m2(1/Tr!1)]2 3 tures, can be well represented. For a given mixture,
"[1#m1(1!(Tr)#m2(1!Tr)(0.7!Tr)]2 4 the equation-of-state mixing rules with one set of
"1#m1(1!Tr)#m2(1/Tr!1) 5
parameters can frequently represent the data over
"10[m(1\Tr)] 6
" 1#[m1#m2(1#(Tr)(0.7!Tr)](1!(Tr)2 7 wide ranges of temperature and pressure.
"10f (Tr), f (Tr)"m3(m0#m1Tr#m2T 2r)(1!Tr) 8 Examples of binary data representation by means
"exp[m1(1!Tr)#m2(1!(Tr)2] 9 of the two approaches are depicted in Figure 3.
"T (rm2\1)m3 exp [m1(1!T m
r H )]
2 m3
10

References: 1. Wilson GM (1964) VaporIliquid equilibriums, cor- Prediction of Vapour}Liquid Equilibria

relation by means of a modified Redlich}Kwong equation of state.
Advances in Cryogenic Engineering. 9: 168}176. 2. Soave G Although vapour}liquid equilibria have been investi-
(1972) Equilibrium constants from a modified Redlich}Kwong gated for more than 10 000 systems, values resulting
equation of state. Chemical Engineering Science 27: from various combinations are still unknown. It
1197}1203. 3. Harmens A and Knapp H (1980) Three-parameter
would be impractical to determine experimentally all
cubic equation of state for normal substances. Industrial and
Engineering Chemistry Fundundamentals 19: 291}294. 4. the systems needed individually.
Mathias PM (1983) A versatile phase equilibrium equation of In principle, experimental values of some thermo-
state. Industrial and Engineering Chemistry. Process Design and dynamic properties can be used to estimate other
Development 22: 385}391. 5. Soave G (1984) Improvement of properties. For examples, binary vapour}liquid equi-
the van der Waals equation of state. Chemical Engineering
librium can be estimated from the liquid activity
Science 39: 357}369. 6. Adachi Y and Lu BC-Y (1984) Simplest
equation of state for vapor}liquid equilibrium calculations: a modi- coefRcients calculated from mutual solubility data for
fication of the van der Waals equation. American Institute of the same mixture, and the inRnite-dilution activity
Chemical Engineers Journal 30: 991}993. 7. Stryjek R and Vera JH coefRcients measured from gas}liquid chromatogra-
(1986) PRSV: An improved Peng}Robinson equation of state for phy can be used to predict the vapour}liquid equilib-
pure compounds and mixtures. Canadian Journal of Chemical
ria over the complete concentration range. Some pre-
Engineering 64: 323}333. 8. Yu JM and Lu BC-Y (1987) A three-
parameter cubic equation of state for asymmetric mixture density diction methods are brieSy described below with em-
calculations. Fluid Phase Equilibria 34: 1}19. 9. Melhem GA, phasis placed on binary mixtures. Extending the
Saini R and Goodwin BM (1989) A modified Peng}Robinson
equation of state. Fluid Phase Equilibria 47: 189}237. 10. Twu CH,
Bluck D, Cunnigham JR and Coon JE (1991) A cubic equation of
Table 7 Some mixing and combining rules for cubic equations
state with a new alpha function and a new mixing rule. Fluid
of state
Phase Equilibria 69: 33}50.
van der Waals/Berthelot
aij"(ai aj )1/2
The simplest combining rules for ‘aij’ and ‘bij’ are
obtained by using the geometric mean for aij and the a" yi yj aij , b" yi bi
i j i
arithmetic mean for bij, i.e:
Modified van der Waals/Berthelot
aij"(aiaj) 1/2
[29] a" yi yj aij b" yi yj bij
aij"(ai aj ) (1!kij )
1/2
bij"(bi#bj)/2 [30]
bij"12(bi#bj )(1!cij )
A binary interaction parameter kij is frequently intro- Wong}Sandler
duced in eqn [29] to correct the discrepancy gener-


a aij
ated by the geometric mean: b! " bij!
RT i j RT

aij"(aiaj)1/2(1!kij) [31] bi#bj

bij" (1!kij )
Occasionally, a binary interaction parameter lij is 2
ai#aj
introduced in eqn [30] to yield improved bij values: aij" (1!kij )
2
a xi ai GE
bij"(bi#bj)(1!lij)/2 [32] " !
b bi CRT
where: G E is a selected excess Gibbs energy model
More recently, new mixing rules, such as the one C is characteristic of the equation of state
proposed by Wong and Sandler in 1992, with im-
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction 1155

Figure 3 (A) Correlating the phase behaviour of the (ethanol#benzene) system at 323.15 K by means of the gamma}phi approach.
The lines represent the values calculated by using the Margules equations and the points represent the experimental values reported
by ND. Litvinov (1952) Isothermal equilibrium of vapor and liquid in systems of three fully miscible liquids. Zhurnal Fizicheskoi Khimii
26: 1405}1412. (B) Predicting the phase behaviour of the (0.2654 mole fraction ethane#0.7346 mole fraction n-heptane) mixture by
means of the equation-of-state approach. The smooth curve represents the values calculated by using the Peng}Robinson equation,
and the points represent the experimental values reported by WB Kay (1938) Liquid}vapor phase equilibrium relations in the
ethane-n-heptane system. Industrial and Engineering Chemistry 30: 459}464.

application to multicomponent mixtures is feasible
once good correlation of the vapour}liquid equilibria
of its constituent binary systems becomes available.
Prediction from Pure Component Properties
Application of the regular solution theory For mix-
tures containing nonpolar components that are not
much different in size and shape, the regular solution
theory of Hildebrand leads to a semi-quantitative
prediction of k values of all components in a mixture.
In terms of the solubility parameter, the activity co-
efRcients of the components in a regular solution can
be calculated from the equation:
RT ln k"vk( k! )2
where the volume average solubility parameter is
given by:
where UVk and HVk are, respectively, the molar en-
ergy and enthalpy of vaporization of pure liquid k at
temperature T. The assumption involved here is that
T is well below the critical temperature in order to
make the approximation valid. The calculated liquid
activity coefRcients can then be used to obtain
the desired vapour}liquid equilibrium values. For a
binary mixture:
RT ln 1"v1 2
( 1! 2)2 [37]
2
RT ln 2"v2 2
1 ( 1! 2)2 [38]
Should a binary interaction parameter be required
to improve the data representation, an extension of
[33] the approach to the prediction of multicomponent
vapour}liquid equilibrium may not be practical;
attempts made to correlate the binary parameters
have not been successful.

Liquid activity coefVcients at inVnite dilution


" [34]
i i
Values of  are particularly useful for obtaining the
parameters of any of the two-constant equations for
and the volume fraction i is deRned by: the excess Gibbs energy; the  values for a binary
system are the parameter values. For example,
i "xivi/ xjvj [35] 
1 "A and  2 "B for the van Laar and Margules
equations presented in Table 3. If a three-parameter
The solubility parameter for substance k, k , is de- equation is used, the third parameter must be deter-
Rned by: mined by an independent approach.
The modiRed separation of cohesive energy density
k"(UVk /vVk )1/2"[(HVk !RT)/vk]1/2 [36] (MOSCED) method proposed by Thomas and Eckert
1156 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction
in 1984 may be used to predict  values from pure
component parameters. This method is based on
a modiRed regular solution theory and the assump-
tion that the forces contributing to the cohesive
energy are additive. It has been reported that the
average error of 3357  values predicted by this
method was 9.1%.
In general, calculated equilibrium vapour composi-
tions are relatively insensitive to moderate errors in
the  used in the calculation.
Prediction of Binary
Values Using Azeotropic or
Mutual Solubility Data
Prediction from azeotropic data Many binary sys-
tems exhibit azeotropic behaviour. At an azeotropic
condition, the compositions of the liquid and vapour
phases are identical. At low pressures, the liquid ac-
tivity coefRcients can be simply calculated by:
1"P/psat and 2"P/psat [39]
1 2
The parameters of any two-parameter expression of
the excess Gibbs energy can then be obtained and
used for extrapolating vapour}liquid equilibrium
over the complete concentration range.
Prediction of  values from mutual solubility data
The thermodynamic consideration applicable to a bi-
nary system at vapour}liquid equilibrium is also ap-
plicable to a partially miscible binary liquid mixture
at equilibrium. Hence, the activity coefRcients of the
two liquids at the temperature at which the solubili-
ties were experimentally determined can be expressed
by:
1x1"1x1 and 2x2"2 x2 [40]
where the two superscripts refer to the two liquid
phases. Applying these relationships to any two-para-
meter expression of the excess Gibbs energy leads to
the determination of the parameter values, which
permit vapour}liquid equilibrium estimation of the
mixture.
Prediction of
from Group Contribution Methods
In group contribution methods, the calculation of
thermodynamic properties of pure Suids is based on
the assumption that each molecule is an aggregate of
functional groups. Langmuir in 1925 extended the
concept to mixtures. Redlich, Derr, Pierotti and
Papadopoulos developed a group interaction model
for heats of solution in 1959. Adopting the concepts
presented by these authors, Wilson and Deal sugges-
ted in 1962 a solution of the groups approach by
which liquid activity coefRcients can be estimated on
the basis of group contributions. In this approach, the
logarithm of the activity coefRcient of a component is
assumed to be the sum of two contributions: the
conRgurational contribution, which accounts for the
differences in molecular size, and the group interac-
tion contribution, which accounts for the inter-
molecular forces originating from the different func-
tional groups.
The group contribution approach to calculating  is
attractive because through this approach it is possible
to estimate vapour}liquid equilibria of nonideal mix-
tures without experimentation. Although a large
number of mixtures can result from pure compounds,
the functional groups, such as CH2, OH, CO, COO
and COOH, that constitute these compounds are
limited. If the activity coefRcients of the mixture
components could be obtained from a knowledge of
the interactions of these groups, and with the assump-
tion that the contribution to  by one group within
a molecule is independent of that made by any other
group in that molecule, a relatively small number of
parameters would sufRce for the prediction of the
activity coefRcients for mixtures containing the same
groups. This assumption implies that the contribution
of the group interaction is independent of the nature
of the molecule.
Two of these approaches are mentioned here.
Analytical solutions of groups (ASOG) method
Following the concept of the solutions of groups of
Wilson and Deal, the analytical solutions of groups
(ASOG) method was Rrst presented by Derr and Deal
in 1969. Basically, the practical application of the
solutions of groups concept involves the reduction of
liquid activity coefRcients obtained from experi-
mental data for vapourdliquid equilibria into a num-
ber of binary group interaction parameters. The
working equations of the ASOG method are present-
ed in Table 8.
Kojima and Tochigi in 1979 calculated the group
interaction parameters for 31 groups and used the
method to predict the vapour}liquid equilibria for
936 binary systems, 103 ternary systems, Rve quater-
nary systems, and two quinary systems at low pres-
sures. They reported that the average absolute devi-
ation of the predicted vapour compositions was 1.2%.
Universal functional group activity coefVcient
(UNIFAC) method The universal functional group
activity coefRcient (UNIFAC) method proposed by
Fredenslund, Gmehling and Rasmussen and the
ASOG method are based on the same principle of
group contributions. The main difference between
these two methods is in the equations used for
 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction 1157

Table 8 The analytical solution of groups (ASOG) method a wide range of temperature. Experience indicates
that equations that could meet the equal fugacity
1. ln i"ln Si#ln Gi
condition as well as vV or vL would be suitable for the
2. ln Si"1!ri!ln ri
where: vi intended vapour}liquid equilibrium calculations.
ri" Should a situation arise such that the saturated liquid
xj vj
xj"mole fraction of molecule j molar volume is required in the estimation, cubic
vj"number of atoms other than hydrogen in mol- equations containing more parameters may be se-
ecule j lected for the representation. Adachi and Lu in 1990
suggested that it is possible to assign different two-
3. ln Gi" vki (ln k!ln  (ki ))
k parameter or three-parameter cubic equations to dif-
where: vki"number of atoms other than hydrogen in ferent components of the binary mixture under
group k in molecule i
consideration, and then use a four-parameter cubic
ln k"1!Ck!ln Dk
ln  (ki )"1!C (ki )!ln D (ki ) equation to combine the equations in the
vapour}liquid equilibrium calculations for mixtures.
Xj Ajk X (j i )Ajk Special case should be applied when equations of
Ck" , C (ki )"
j Dj Dj state are used to represent the experimental data
Dk" Xj Akj , D (ki )" X (j i )Akj measured at or near theoretical points of the Suids.
1 vLi The selection of mixing and combining rules plays
XL" xi vLi, X (Li )" a crucial role in the application of equations of state
S vki
k to the correlation and prediction of vapourdliquid


nkL equilibria. The importance of selecting an appropri-
AkL"exp mkL#
T ate expression for the excess Gibbs energy cannot be
S" xi vki overemphasized.
i k The prediction of binary vapour}liquid equilibria
from pure-component properties by means of the
MOSCED method is attractive. However, poor re-
representing the excess Gibbs energy. The Wilson sults are obtained for systems where steric consider-
equation is used in the ASOG method, whereas the ations are signiRcant. The general applicability of this
two-parameter universal quasi-chemical (UNIQUAC) model is limited due to the difRculties involved in
equation of Abrams and Prausnitz is used in the
UNIFAC method. The working equations of the Table 9 The UNIFAC method
UNIFAC method are presented in Table 9. There are
50 main groups together with their subgroups for the 1. ln i "ln ci#ln Ri


determination of the parameters involved in the cal- 2.
Ji
ln ci "1!Ji #ln Ji !5qi 1! #ln
Ji
culation. For  calculations for multicomponent sys- Li Li
tems, the adjustable binary parameters are evaluated where: ri qi
from binary vapour}liquid equilibrium data. Ji " , Li "
xr
j j j
xj qj
ri " v (ki )Rk , qi " v (ki )Qk
Prediction using Equations of State k k
Rk "volume parameter for group k
The equations of state successfully used for correlation Qk "surface area parameter for group k
of binary vapour}liquid equilibria can be used for the v (ki ) "number of subgroups of type k in molecule
purposes of predicting multicomponent vapour}liquid of species i



equilibria, provided that the binary interaction para- k ik ik
3. ln Ri"qi 1! !ki ln
meters of all the constituent binary systems are avail- k Sk Sk
able. All the parameters should be obtained by regres- where: v Qk (i )
k
sion of the binary data using the same mixing and ki "
qi
combining rules. Interpolated and estimated values of ik " m mi mk
these parameters are available for some systems, but
xi qi ki
their values are subject to frequent revision. k " i

j xj qj

Sk " m

m mk

Additional Comments on Applications


mk
mk "exp !
T
There is no simple equation of state that can represent mk "group interaction parameter
satisfactorily the three conditions of eqn [21] over
1158 II / DISTILLATION / Vapour}Liquid Equilibrium: Correlation and Prediction

Figure 4 Algorithms for correlating and predicting vapour}liquid equilibrium values.

the determination of pure component parameters. Fur-
thermore, the approach cannot be used for aqueous
mixtures, nor for systems with very large  values.
In general, the accuracy of prediction depends on
the availability of some reliable binary data for either
the system of interest or another one that is closely
related to it.
All the group-contribution methods are approxim-
ate. The fundamental assumption involved in the
group solution approach is additivity, and the esti-
mated  values are necessarily approximate.
Finally, computer software packages for vapour}
liquid equilibrium calculations are available commer-
cially from a number of process engineering software
development companies. Typical algorithms for data
correlation and prediction are depicted in Figure 4.
See also: II/Distillation: Modelling and Simulation;
Multicomponent Distillation; Theory of Distillation; Vapour-
Liquid Equilibrium: Theory. III/Physico-Chemical Meas-
urements: Gas Chromatography.
Further Reading
Abrams DS and Prausnitz JM (1975) Statistical thermo-
dynamics of liquid mixtures. New equation for the ex-
cess Gibbs energy of partly or completely miscible sys-
tems. American Institute of Chemical Engineering Jour-
nal 21: 116}128.
 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory
Adachi Y and Lu BC-Y (1990) Taking advantage of avail-
able cubic equations of state. Canadian Journal of
Chemical Engineering 68: 639}644.
Christiansen LJ and Fredenslund A (1975) Thermodynamic
consistency using orthogonal collocation or composition
of equilibrium vapor compositions at high pressures.
American Institute of Chemical Engineers Journal 21:
49}57.
Derr EL and Deal CH Jr (1969) Analytical solutions of
groups. Correlation of activity coefRcients through
structural group parameters. Proceedings of Interna-
tional Symposium of Distillation 3: 40}51.
Denbigh K (1981) The Principles of Chemical Equilibrium,
4th edn. Cambridge: Cambridge University Press.
Fredenslund A, Gmehling J and Rasmussen P (1977) Va-
pour}Liquid Equilibria Using UNIFAC. Amsterdam:
Elsevier.
Gmehling J, Onken U and Arlt W (1974}1990) Vapour}
Liquid Equilibrium Data Collection; Dechema Chem-
istry Data Series, vol. I, parts 1}8. Frankfurt: Dechema.
Hala E, Pick J, Fried V and Vilim O (1967) Vapour}Liquid
Equilibrium, 2nd edn. Oxford: Pergamon Press.
Hayden JG and O’Connell JP (1975) Generalized method
for predicting second virial coefRcients. Industrial and
Engineering Chemistry. Process Design and Develop-
ment 14: 209}216.
Knapp H, Doring R, Oellrich L, Plocker U and Prausnitz
JM (1982) In: Behrens D and Eckerman R (eds) Chem-
istry Data Series, Vol. VI: VLE for Mixtures of Low
Boiling Substances. Frankfurt: Dechema.
Kojima K and Tochigi T (1979) Prediction of Vapour}
Liquid Equilibria by the ASOG Method. New York:
Elsevier.
Lewis, GN and Randall M (1923) Thermodynamics and
the Free Energy of Chemical Substances. New York:
McGraw-Hill.
Lu BC-Y (1959) Heats of mixing and vapor}liquid equilib-
rium calculations. Canadian Journal of Chemical Engine-
ering 37: 193}199.
Lu BC-Y (1962) Binary vapor}liquid equilibrium data:
Thermodynamic consistency tests. Canadian Journal of
Chemical Engineering 40: 16}24.
Malanowski S and Anderko A (1992) Modelling Phase
Equilibria, Thermodynamic Background and Practical
Tools. New York: John Wiley.
Vapour}Liquid Equilibrium: Theory
A. S. Teja and L. J. Holm, Georgia Institute of
Technology, Atlanta, GA, USA
Copyright ^ 2000 Academic Press
Introduction
The concept of an equilibrium stage in distillation is
based on the assumption that the vapour leaving the
1159
Papadopoulos MN and Derr EL (1959) Group interaction.
II. A test of the group model on binary solutions of
hydrocarbons. Journal of American Chemical Society
81: 2285}2289.
Prausnitz JM, Lichtenthaler RN and de Azevedo EG
(1999) Molecular Thermodynamics of Fluid-Phase
Equilibria, 3rd edn. Englewood Cliffs, NJ: Prentice-
Hall.
Raal JD and Muhlbauer AL (1998) Phase Equilibria,
Measurement and Computation. Washington, DC:
Taylor & Francis.
Redlich O, Derr EL and Pierotti GJ (1959) Group interac-
tion. I. A model for interaction in solutions. Journal of
American Chemical Society 81: 2283}2285.
Reid RC, Prausnitz JM and Poling BE (1987) The Proper-
ties of Gases and Liquids, 4th edn. New York: McGraw-
Hill.
Renon H and Prausnitz JM (1968) Local compositions in
thermodynamic excess functions for liquid mixtures.
American Institute of Chemical Engineers Journal 14:
135}144.
Starling KE (1977) Fluid Properties for Light Petroleum
Systems. Houston, TX: Gulf Publishing Co.
Thomas ER and Eckert CA (1984) Prediction of limiting
activity of coefRcients by a modiRed separation of cohe-
sive energy density model and UNIFAC. Industrial and
Engineering Chemistry. Process Design and Develop-
ment 23: 194}209.
Walas SM (1985) Phase Equilibria in Chemical Engineer-
ing. Boston: Butterworth.
Wilson GM (1964) Vapor}liquid equilibrium. XI. A new
expression for the excess Gibbs energy of mixing. Jour-
nal of American Chemical Society 86: 127}130.
Wilson GM and Deal CH (1962) Activity coefRcients and
molecular structure } activity coefRcients in changing en-
vironments } solutions of groups. Industrial and Engin-
eering Chemistry Fundamentals 1: 20}23.
Won KW and Prausnitz JM (1973) High-pressure vapor}
liquid equilibriums. Calculation of partial pressures
from total pressure data. Thermodynamic consistency.
Industrial and Engineering Chemistry Fundamentals 12:
459}463.
Wong DSH and Sandler SI (1972) A theoretically correct
mixing rule for cubic equations of state. American Insti-
tute of Chemical Engineers Journal 38: 671}680.
stage is in equilibrium with the liquid leaving the
same stage. The use of this concept in the design of
distillation columns requires a description of how the
components of a multicomponent mixture distribute
between the two phases in equilibrium. This description
is provided by phase equilibrium thermodynamics.
The equilibrium relationship for any component
i in an equilibrium stage is deRned in terms of the
1160 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory

distribution coefRcient or K value: relative volatilities) is provided by thermodynamics

yi
Ki"
xj
where yi is the mole fraction of component i in the
vapour phase and xi is the mole fraction of i in the
liquid phase. The more volatile components of a mix-
ture will have higher K values, and vice versa. In
distillation, the efRciency of separation of two com-
ponents is often compared via a quantity called the
relative volatility
ij:
Ki yi/xi

ij" "
Kj yj/xj
A relative volatility close to unity means that the
separation of the two components is likely to be
difRcult, whereas a relative volatility much greater or
much less than unity means that few equilibrium
stages are likely to be needed for separation. For
binary system, eqn [2] can be rearranged to give:
and is discussed below. A more detailed discussion
may be found in textbooks of thermodynamics (see
[1]
Further Reading).
Thermodynamic Framework
Vapour}liquid equlibria (VLE) can be modelled using
the thermodynamic relationship for the equality of
fugacities of a component i in the vapour and liquid
phases. Thus:
fK iL"fK iV (i"1 to m) [4]
[2]
where m is the total number of components in the
system, and L and V represent the liquid phase and
the vapour phase, respectively. In order to use this
relationship, the fugacities must Rrst be related to the
compositions in the two phases in equilibrium. This is
done using the following thermodynamic relationship
in terms of the variables T, P and ni:

  
 
fiK P
V RT

ijxi RT ln " ! dP [5]
yi" [3] ziP ni P
1#(
ij!1)xi 0 T,P,nj

or the equivalent relationship in terms of T, V and ni:

Eqn [3] is plotted in Figure 1 for various (constant)
values of the relative volatility. Note that an increase

  
 
fKi  P RT
in the relative volatility leads to an increase in the RT ln " ! dP
concentration of the more volatile component in the ziP V ni T,V,nj V
vapour phase. When the relative volatility has a value


of 1, yi"xi and separation is no longer feasible. PV
!RT ln [6]
A relative volatility of 1 also signiRes the existence of RT
an azeotrope or a critical point. A framework for the
correlation and prediction of K values (and hence In the above equations, zi is either xi or yi depend-
ing on the phase being considered, and ni is the
number of moles of component i in that phase. The
quantity (fiK /ziP) is called the fugacity coefRcient K i of
component i in the mixture.

Ideal Systems
In the case of an ideal gas mixture, substitution of the
ideal gas equation PV"nRT into eqn [5] leads to:

fK Vi "yiP [7]
Similarly, substitution of the volume additivity re-
lation for ideal liquid mixtures V"nivLi (where vLi is
the molar volume of component i at the temperature
and pressure of the solution) leads to:

fK Li"xi f Li [8]

where f Li is the fugacity of pure liquid i at the pressure

Figure 1 The y}x behaviour of a binary mixture at constant and temperature of the solution. At constant temper-
temperature for various values of the relative volatility
ij. ature, the effect of pressure on the pure liquid
 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory 1161

fugacity can be obtained from:

v Li
d ln f Li" dP [9]
RT

Integration of eqn [9] from the saturation pressure

to the system pressure leads to:


P
vLi
f Li"f sat
i exp dP [10]
Psat
i
RT

where f sat
i (" sat sat
i Pi ) is the fugacity of pure i at
sat
saturation, i is the fugacity coefRcient of pure i at
saturation, and Psat
i is the vapour pressure of pure i.
The liquid molar volume vLi can often be assumed to
be constant with respect to pressure (since liquids are
Figure 2 The P}x}y behaviour at constant temperature of a bi-
incompressible), thus simplifying the exponential nary mixture that obeys Raoult’s law. The dashed line shows the
terms of eqn [10], called the Poynting factor, to: P}x behaviour or the bubble curve. The solid line represents the
P}y behaviour or the dew point curve.

 

i !P)


P
vLi vLi(Psat
exp ! dP +exp [11]
Psat
i
RT RT
only at low pressures ((1 MPa). As a consequence,
At low pressures (P(1 MPa), the Poynting terms K values can be predicted from pure component data
and sat
i both approach unity and eqn [10] reduces to:
only for such mixtures. The majority of real systems
are nonideal and their thermodynamic description is
f Li"Psat
i [12] discussed below.

Nonideal Systems
Thus, for the simplest case of an ideal gas mixture in
equilibrium with an ideal liquid solution at low pres- Figure 3A, B shows the P}x}y behaviour of systems
sures, eqns [4]}[12] lead to Raoult’s law: that exhibit small negative and positive deviations
from Raoult’s law, whereas Figure 3C and D show
i "yiP
xiPsat [13] systems that exhibit large negative and positive devi-
ations from Raoult’s law, respectively. Large devi-
and, therefore: ations from Raoult’s law often lead to the formation
of minimum pressure (maximum boiling) or max-
yi Psat
i imum pressure (minimum boiling) azeotropes, as
Ki" " [14]
xi P shown in Figure 3C and D. The relative volatility has
and: a value of unity at the azeotropic composition.
Nonideal behaviour depicted in these Rgures can be
Psat
i
described using two approaches } the activity coefRc-

ij" sat [15]
Pj ient approach and the equation of state approach.

The relative volatility of a system that obeys
Raoult’s law is thus a ratio of two vapour pressures
and is a function only of the temperature. The
yi versus xi behaviour at constant temperature is
therefore as shown in Figure 1, and the P}x}y behav-
iour of such a system is shown in Figure 2. A feature
of this system is that the P}x behaviour (or the bubble
curve) is linear and given by:
i #x1(P1 !P2 )
P"Psat sat sat
[16]
Only a small number of systems containing chemic-
ally similar components obey Raoult’s law, and then
Activity coefVcient approach In this approach, va-
pour-phase fugacities are written in terms of the va-
pour-phase composition as follows:
fK iV" K ViyiP [17]
where K Vi is the vapour-phase fugacity coefRcient of
component i, yi is the mole fraction of i in the vapour
phase and P is the total system pressure. In addition,
liquid-phase fugacities are written in terms of the
liquid-phase composition as follows:
fK iL" i x i f i0 [18]
1162 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory

Figure 3 (A) The P}x}y behaviour of a system that exhibits small negative deviations from Raoult’s law. The P}x behaviour of
a system that follows Raoult’s law is shown by the broken line. (B) The P}x}y behaviour of a system that exhibits small postive
deviations from Raoult’s law. Raoult’s law behaviour is shown by the broken line. (C) The P}x}y behaviour of a system that exhibits
significant negative deviations from Raoult’s law leading to the formation of a minimum pressure (maximum boiling) azeotrope. (D) The
P}x}y behaviour of a system that exhibits significant positive deviations from Raoult’s law leading to the formation of a maximum
pressure (minimum boiling) azeotrope.

where i is the activity coefRcient of component i in Hence:

the liquid phase, xi is the mole fraction of i in the
ixiPsat
i "yiP [21]
liquid phase and f 0i is the fugacity of pure liquid i at
the pressure and temperature of the system. Combin- and:
ing eqns [17] and [18], we obtain: Ki"iPsat
i /P [22]
i xi f " K y P
0
i
V
i i [19] Vapour pressures at subcritical temperatures may
be obtained from experimental data using equations
At low pressures ((1 MPa), eqn [19] can be simpli-
such as the Antoine equation. Activity coefRcients
Red further to yield:
may be obtained from excess Gibbs energy gE models,
i xiPsat
i " K i yiP
V
[20] as described below.

as discussed in the previous section. Often K Vi&1.0 Equation of state approach The equation of state
for vapour phases at moderate pressures. approach uses eqn [17] for the vapour phase and an
 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory 1163

analogous expression for the liquid phase. Thus, for The truncated virial eqn [28] is only applicable to
the liquid phase: gases at densities that are less than about half the
critical density. One of the advantages of the virial
fK Li" K LixiP [23] equation, however, is that virial coefRcients can be
calculated from intermolecular potential functions.
Substituting these relationships into eqn [4] results
Also, statistical mechanics provides rigorous expres-
in:
sions for the composition dependence of the virial
K Viyi" K Lixi [24] coefRcients. Thus, the mixture second virial coefRc-
ient is given by:
and:
B" yiyjBij [29]
Ki" K Li/ K V
i [25] i j

The calculation of K values is therefore reduced where Bii is the second virial coefRcient of component
to the calculation of fugacity coefRcients in the equa- i and Bij is a cross second virial coefRcient. Substitu-
tion of state approach and, at moderate pressures, to tion of eqns [28] and [29] into eqn [5] leads to:
the calculation of activity coefRcients in the activity
coefRcient approach. P
ln K Vi" [Bii# ykyl(2ki!kl)] [30]
RT 

Calculation of Fugacity where kl"2Bkl!Bkk!Bll. The fugacity coefRcient

Coef\cients of any component in the vapour phase can thus
Calculation of the fugacity coefRcient using eqn [5] or be calculated if the second virial coefRcients of the
eqn [6] requires knowledge of the P}V}T}x behav- pure components and the cross second virial
iour of the system. This information is obtained from coefRcients are available. Since the truncated virial
an equation of state. Two representative types of equation is only applicable to gases at low to moder-
equation of state will be discussed below } a volume- ate pressures, fugacity coefRcients calculated using
explicit virial equation and a pressure-explicit cubic eqn [30] are generally only employed when eqn [20]
equation. A more detailed discussion of types of is used to calculate VLE.
equation of state is available elsewhere (see Further Pressure-explicit Cubic Equation of State
Reading).
Cubic equations of state express the pressure as a cu-
Volume-explicit Virial Equation bic function of the molar volume, and their origin
Virial equations of state are inRnite-series expansions stems from the van der Waals equation, which was
of the compressibility Z as a function either of the the Rrst cubic equation of state to represent qualita-
density or pressure. The pressure series may be writ- tively both vapour and liquid phases. Several hundred
ten as: modiRcation of the van der Waals equation have been
reported in the literature. An example of a recent
PV modiRcation that is better able to represent P}V}T}x
Z" "1#BP#CP2#2 data for both vapour and liquid mixtures is the equa-
RT
tion of state proposed by Patel and Teja in 1982. This
BP (C!B2) 2 equation may be written as:
"1# # P #2 [26]
RT (RT2)
RT a

where B is the second virial coefRcient, C the third P" ! [31]
v!b v2#bv#cv!bc
virial coefRcient, and so on. Typically, the two-terms
truncated form of the virial equation is used for gases where:
at low pressures:
R2T 2c
a"
a [32]
BP Pc
Z"1#BP"1# [27]
RT RTc
b"b [33]
which can be rearranged in the volume-explicit form: Pc

RT RTc
V" #B [28] c"c [34]
P Pc
1164 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory



"[1#m(1!(T/Tc)]2 [35] (2v#b#c)#[4bc#(b#c)]
;ln
(2v#b#c)![4bc#(b#c)]
a"32c#3(1!2c)b#2b#1!3c [36]
2a
(bi#ci)
c"1!3c [37] #
RT ((2v#b#c)![4bc#(b#c)])
c"Pcvc/RTc [38]
a

#
and b is the smallest positive root of: RT[4bc#(b#c)]

3b#(2!3c)2b#32cb!3c"0 [39] ;[ci (3b#c)#bi (3c#b)]



In the above equations, the subscript c denotes (2v#b#c)#[4bc#(b#c)]
a value at the critical point. Note that by setting the ; ln
(2v#b#c)![4bc#(b#c)]
parameter c"0 in eqn [31], the Patel}Teja equation
reduces to the Redlich-Kwong-Soave equation of

2(2v#b#c)[4bc#(b#c)]
state; and by setting c"b, it reduces to the Peng} ! [45]
(2v#b#c)![4bc#(b#c)]
Robinson equation of state. Both the Redlich}
Kwong}Soave and the Peng}Robinson equations are
widely used in process design calculations. For non- Eqn [45] can be used to calculate both the vapour-
polar Suids, c and m are calculated from the follow- and liquid-phase fugacity coefRcients. In the case of
ing relationships in terms of the acentric factor : vapour phase, zi"yi and v is the vapour molar vol-
ume, whereas for the liquid phase, zi"xi and v is the
c"0.329032!0.076799 #0.0211947 2
[40] liquid molar volume. The ratio of the two fugacity
coefRcients yields the K value under the conditions of
m"0.452413#1.30982 !0.295937 2
[41] interest.

A knowledge of Pc, Tc and is therefore sufRcient

to calculated the parameters of the equation of state. Calculation of Activity Coef\cients
Alternatively, c and m may be calculated from ex- Activity coefRcients, i are generally calculated by
perimental values of the vapour pressure and liquid differentiation of the excess Gibbs energy gE:
density of the substance. Several other forms of


eqn [35] suitable for complex molecules have been ngE
RT ln i" [46]
proposed. ni T,P,nj
The parameters a, b and c for a mixture can be
calculated using the following mixing rules: A number of expressions have been proposed for
gE as a function of composition. Some of the more
a
" zizj(a
)ij [42] popular of these are outlined below.
i j

Margules Equation
b" zibi [43]
i The Margules equation is one of the simplest expres-
sions for the molar excess Gibbs energy. For a binary
c" zici [44] solution:
i
gE
where zi can be xi or yi and (a
)ij"(1!kij ((a
)i(a
)j. "x1x2(A21x1#A12x2) [47]
RT
kij is a binary interaction parameter that is usually
obtained by Rtting experimental VLE data. where A12 and A21 are binary parameters dependent
The fugacity coefRcient can be obtained by substi- on temperature, but not on the composition. The
tuting eqns [31]}[44] into eqn [6] leading to: Margules activity coefRcients in a binary mixture are
obtained by differentiation of eqn [47] and are given

  
fK i bi P by:
ln K i"ln " !ln (v!b)
zi P v!b RT
ln 1"[A12#2(A21!A12)x1]x22 [48]
2 xj
ij aij
!
RT [4bc#(b#c)] ln 2"[A21#2(A12!A21)x2]x21 [49]
 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory 1165

A12 and A21 are generally obtained by Rtting VLE
data. Note that the value of the activity coefRcient of
each component tends to unity as the mole fraction of
that component goes to unity. This behaviour is in-
the molar liquid volume vi of the pure component i,
and the energies of interaction ij between the molecu-
les i and j as follows:
ij!

 
herent in all gE models. The Margules equation works vj ii
ij " exp ! [54]
well for binary systems in which the two components vi RT
are very similar in size, shape and chemical nature.
Margules parameters for a large number of systems The expression for the liquid activity coefRcients
are tabulated in DECHEMA books on VLE data. are:
Van Laar Equation ln 1"!ln(x1# x)
12 2

The van Laar equation for the excess Gibbs energy

may be written as: #x2

x1#
12

12x2
!
x2#
21

x
21 1  [55]

gE 2a12x1x2q1q2
" [50] ln 2"!ln(x2# x)
RT x1q1#x2q2 21 1

where q1 and q2 are the effective volumes of the two

molecules and a12 is an interaction parameter. Differ-
#x1

x2#
21

21x1
!
x1#
12

x
12 2  [56]

entiation according to eqn [46] leads to the following

expressions for the activity coefRcients: The Wilson equation has two parameters 12 and
21 (or equivalently, 12! 11 and 21! 22) and is
A12 able to correlate VLE data for a wide variety of
ln 1" [51]

 
A12 x1 2 miscible systems, including those containing polar or
1# associating components in nonpolar solvents. How-
A21 x2
ever, the equation is incapable of predicting
A21 liquid}liquid immiscibility in a system.
ln 2" [52]

 
A21 x2 2 For multicomponent mixtures, the Wilson equa-
1# tion is written as follows:
A12 x1

 
m m
where A12"2q1a12 and A21"q2a12. As in the case of xi ik
ln k"!ln xj kj #1! m
[57]
the Margules equation, the two parameters A12 and j"1 i"1
A21 are obtained by Rtting VLE data. The van Laar xj ij
j"1
equations have been shown to work well for a num-
ber of binary systems where the size, shape and chem- Note that only binary parameters ij are required
ical nature of the components are dissimilar, and to evaluate activity coefRcients in multicomponent
parameters for many binary systems have been systems. These parameters are obtained by Rtting
tabulated in the DECHEMA data books. VLE data for the binary pairs, and many of these
parameters have been tabulated in the DECHEMA data
Wilson Equation
books. Moreover, because a temperature dependence
The Margules equation is based on the assumption is included in eqn [54], the same binary parameters
that the ratio of species 1 to species 2 molecules may be used over a range of temperatures (although
surrounding any molecule is the same as the ratio of no more that about 50 K).
the mole fractions of species 1 and 2. A different class
of gE models has been proposed based on the assump- NRTL Equation
tion that, around each molecule, there is a The NRTL (non-tandom two-liquid theory) equation
local composition that is different from the bulk com- is also based on a local composition model for the
position. The Wilson equation is such a local com- excess Gibbs energy. However, it is applicable to
position model and the Wilson excess Gibbs miscible as well as partially miscible systems due to
energy has the following form for a binary system: the inclusion of a third binary parameter in the
model. The expression for the molar excess Gibbs
gE
"x1 ln(x1# x )!x2 ln(x2#
12 2 x ) [53]
21 1 energy is:
RT

 
where 12 and 21 are parameters speciRc to the gE 21G21 12G12
"x1x2 # [58]
binary pair. These parameters are deRned in terms of RT x1#x2G21 x2#x1G12
1166 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory

where ij and Gij are deRned as: gE(combinatorial) 1 2

"x1 ln #x2 ln
RT x1 x2
Gij"exp(!ij ij) [59]


z 1 2
gij!gjj # q1x1 ln #q2x2 ln
ij" [60] 2 1 2
RT [65]
gij describes the energy of interaction between com-
gE(residual)
ponent i and j and ij ("ji) is a nonrandomness "!q1x1 ln(1#2 )
21
parameter which is often set equal to 0.3. Thus, only RT
two parameters ij and ji (or, equivalently, gij!gjj !q2x2 ln(2#1 )
12 [66]
and gji!gii) are required per binary pair.
The activity coefRcients expressions are as follows: where:
xiri
" [67]


 
2 i
G21 12G12 xjrj
ln 1"x 2
2 21 # [61]
x1#G21x2 (x2#G12x1)2 j
xiqi
i" [68]
xjqj

 
2
G12 G21 21
ln 2"x 2
1 12 # [62] j
x2#G12x1 (x1#G21x2)2


aji
ji "exp ! [69]
RT
A major advantage of the NRTL equation lies in its
ability to represent highly nonideal systems, par- In eqn [65] z is a coordination number ("10 usu-
ticularly partially miscible systems. ally), i are volume fractions, and i are surface area
For multicomponent mixtures, the liquid-phase ac- fractions for component i. The volume and surface
tivity coefRcients are expressed as: area parameters ri and qi can be evaluated from pure
component molecular structure information and are


m m tabulated in the DECHEMA data books. Thus, there are
Gjixj
ji m
xr rjGrj two binary parameters aij and aji in the UNIQUAC model
xjGij
ln i"j"1
m
# m ij!
r"1
m and these are found by Rtting binary VLE data. The
j"1
Glixl Gljxl Gljxl activity coefRcient expressions become:
l"1 l"1 l"1



[63] 1 z 1 r1
ln 1"ln # q1 ln # 2 l1! l2
x1 2 1 r2
As with the Wilson equation, only binary data are
needed to calculate activity coefRcients in multicom- !q1 ln(1#2 21 )
ponent systems, and these parameters have been


tabulated in the DECHEMA data books for many sys- 21 12
#2q1 ! [70]
tems. Furthermore, because of the inclusion of the 1#2 21 2#1 12
temperature in eqn [60] the parameters obtained by



Rtting VLE data at one temperature may be used to 2 z 2 r2
calculated VLE at other temperatures (within a range ln 2"ln # q2 ln # 1 l2! l1
x2 2 2 r1
of about 50 K).

UNIQUAC Equation !q2 ln(2#1 12 )


The UNIQUAC (universal quasi-chemical theory) equa- 12 21
#1q2 ! [71]
tion expresses the molar excess Gibbs energy as a sum 2#1 12 1#2 21
of a combinatorial part and residual part.
where:
gE"gE(combinatorial)#gE(residual) [64]


z
li" (ri!qi)!(ri!1) [72]
The combinatorial part accounts for differences in 2
the size and shape of the molecules, whereas the
residual contribution accounts for energetic interac- The UNIQUAC equation is applicable to a wide range
tions. of systems, including partially miscible systems.
 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory 1167

For multicomponent systems, the UNIQUAC equation XmQm

m" [83]
becomes:
XnQn
zi i i
m n
ln i"ln # qi ln #li! xjlj


xi 2 i xi j"1 amn
mn"exp ! [84]
T


m m
j ij
!qi ln j ji #qi!qi [73]
m xivmi
j"1 j"1
k kj
Xm" i
[85]

j"1
Once again, only pure component and binary
data are needed to calculate the parameters. UNIQUAC
parameters for over 6000 binary systems have been
tabulated in the DECHEMA data series on VLE.
UNIFAC Group Contribution Method
When values of Margules, van Laar, Wilson, NRTL
or UNIQUAC parameters are not available in the litera-
ture, or when no VLE data for the system of interest
have been measured, the UNIFAC (UNIQUAC functional
group activity coefRcients) method may be used to
estimate activity coefRcients. The UNIFAC method is
a group contribution technique for the estimation of
the parameters amn of the excess Gibbs energy model.
The method expresses the molar excess Gibbs energy
as a sum of a combinatorial part and a residual part
and uses the same combinatorial part as the UNIQUAC
equation. In terms of the activity coefRcient:
ln i"ln Ci#ln Ri


xi vki
i k
Since the group volume parameters Rk and the
group area parameters Qk are known, the only un-
knowns in the UNIFAC equations are the group interac-
tion parameters amn and anm. These have been
tabulated for a large number of groups. Moreover,
updated parameters are published regularly in the
literature. The UNIFAC method has been successfully
applied to a wide variety of binary and multicompo-
nent systems.
Examples of Use
Subcritical Vapour+Liquid Equilibria
Figure 4 shows the P}x}y behaviour of the meth-
anol}water system at 313 K calculated using the
activity coefRcient approach. Activity coefRcients
were obtained from the Wilson equation using para-
[74]
meters 12"!449.3 and 21"!835.9 reported
in DECHEMA data books. Fugacity coefRcients in the



i z i ri
ln Ci"ln # qi ln # i li! lj [75] vapour phase were assumed to be equal to 1.0. Note
xi 2 i rj that the Wilson parameters were obtained by Rtting
the data, and therefore reproduce the experimental
ri" vkiRk [76]
k

qi" vkiQk [77]

k

The group Rk, and the group area Qk have been

tabulated for a large number of groups. vki is the
number of groups of k kind in molecule i.
The residual contribution is expressed as follows:

ln Ri" Qk(ln k!ln (i)

k ) [78]
k

ln k"Qk(1!ln Ek!Fk) [79]

ln (k)
k "Qk (1!ln Ek !F k )
(i) (i) (i)
[80]

Ek"11k#22k#33k#2 [81]

1k1 2k2 3k3

Fk" # # #2 [82] Figure 4 The P}x}y behaviour of methanol}water at 313 K
E1 E2 E3 correlated with the Wilson equation.
1168 II / DISTILLATION / Vapour}Liquid Equilibrium: Theory

Figure 5 The P}x}y behaviour of methanol}water at 373 K

predicted using Wilson equation constants obtained at 313 K.
Figure 7 The P}x}y behaviour of methanol}water at 373 K
predicted with Patel}Teja equation using binary parameters ob-
tained at 313 K.
data (open circles) reasonably well in this system.
Figure 5 shows that when the same parameters are
used to calculate VLE data for methanol}water at 373 K using the same values of the kij parameters, as
373 K, good agreement is obtained with experimental shown in Figure 7.
data.
Figure 6 shows the P}x}y behaviour at 313 K of
Supercritical Vapour+Liquid Equilibria
the same system correlated using the Patel}Teja equa-
tion of state. Two kij values were required to correlate Figure 8 shows the P}x}y behaviour of the carbon
the data (k12"!0.0923 and k21"0.0748) and, in dioxide}propane system at 328 K predicted using the
general, excellent agreement was obtained between equation of state approach with the Patel-Teja
calculated and experimental values. Moreover, the equation of state. The two binary interaction
equation of state was successful in predicting data at

Figure 8 The P}x}y behaviour of CO2}propane mixtures at

Figure 6 The P}x}y behaviour of methanol}water at 313 K 328 K predicted using the Patel}Teja equation of state with binary
correlated with the Patel}Teja equation. parameters obtained from data at 244 K.
 II / ELECTROPHORESIS / Agarose Gels 1169

See also: II/Distillation: Historical Development; Model-

ling and Simulation; Multicomponent Distillation; Theory of
Distillation; Vapour-Liquid Equilibrium: Correlation and
Prediction.

Further Reading
Dymond JH and Smith EB (1980). The Virial CoefTcients
of Pure Gases and Mixtures: A Critical Compilation.
New York: Oxford University Press.
Fredenslund A, Gmehling J and Rasmussen P (1977) Va-
pour}liquid Equilibria Using UNIFAC } A Group
Contribution Method. New York: Elsevier Science.
Ghemling J, Onken U and Arlt W (eds) (1978) Vapour}
liquid Equilibrium Data Collection. Frankfurt: DE-
CHEMA.
Hansen HK, Rasmussen P and Fredenslun A et al. (1991)
Vapour}liquid equilibrium by UNIFAC group contribu-
tion. 5. Revision and extension. Industrial & Engineer-
ing Chemistry Research 30: 2352}2355.
Figure 9 The P}x}y behaviour of methane-n-butane mixtures Knapp H, Reichl A and Sandler SI (1998) Analysis of
at 344 K predicted using the Patel}Teja equation of state with
thermodynamic model equations: mixing rules in cubic
binary parameters obtained from data at 186 K.
equations of state. Industrial & Engineering Chemistry
Research 37: 2908}2916.
parameters (k12"0.143 and k21"0.121) were ob- Patel NC (1996) Improvements of the Patel}Teja equation
tained by Rtting data at a much lower temperature of of state. International Journal of Thermophysics 17:
244 K. The predictions are in excellent agreement 673}682.
with experiment, even though the extrapolation is to Patel NC and Teja AS (1982) A new cubic equation of state
a temperature that is above the critical temperature of for Suids and Suids mixtures. Chemical Engineering
carbon dioxide (304 K). A similar extrapolation using Science 37: 463}473.
the Patel}Teja equation of state is shown in Figure 9 Prausnitz JM, Lichtenthaler RN and Gomes de Azevedo
where VLE in the methane-n-butane system at 344 K E (1999) Molecular Thermodynamics of Fluid-phase
Equilibria, 3rd edn. Englewood Cliffs: Prentice
have been predicted using binary parameters (k12"
Hall.
0.021 and k21"0.002) obtained at 186 K. Note that Reid RC, Prausnitz JM and Poling B (1987) The Properties
the extrapolation is carried out to a temperature that of Gases and Liquids, 4th edn. New York: McGraw-
is well above the critical temperature of methane Hill.
(190 K). Finally, it should be added that the activity Smith JM, Abbott MM and Van Ness HC (1996) Introduc-
coefRcient approach described above cannot be used tion to Chemical Engineering Thermodynamics, 5th
to correlate or predict supercritical VLE behaviour. edn. New York: McGraw-Hill.

ELECTROPHORESIS

following Araki’s preparation of it in 1937 as an

Agarose Gels
J. R. Shainoff, Cleveland State University,
Cleveland, Ohio, USA
Copyright ^ 2000 Academic Press
Development
Agarose is a uniquely nonadhesive hydrocolloid that
has found many uses in the separation sciences
apparently sulfate-free component distinct from the
sulfate-rich agaropectin in agar. Agar and many of
the other hydrocolloids derived from certain species
of seaweed had been used mainly in food preparation
dating back to the seventeenth century in Japan. Agar
was introduced as a medium for immunoelectrophor-
esis by Grabar and Williams in 1953, and the original
technique is occasionally used today. Citrated agar
electrophoresis is the current principal method
for identiRcation of haemoglobin variants. Agar was
1170 II / ELECTROPHORESIS / Agarose Gels
applied by Polson in 1961 for chromatographic sep-
arations based on molecular sieving, following which
its use gradually gave way to agarose beginning with
HjerteH n. A word-search of the Medline abstracts indi-
cates that the number of papers per year employing
agarose electrophoresis did not surpass agar elec-
trophoresis until 1976.
The sulfate content of grades of agarose supplied
for laboratory use is generally below 0.12% com-
pared to 4% in agaropectin (see Armisen, 1997). The
ease with which agarose can be cast into gels without
additives or cross-linking agents and its inertness to-
wards interacting with proteins and nucleic acids are
the principal reasons for its popularity as a separation
medium. All that is needed to prepare gels is to
dissolve the agarose powder by careful heating to
boiling or near boiling temperatures, and let the clear
solution cool. Gelation proceeds spontaneously and
completely as the solution cools. In addition to the
simplicity of gel casting with agarose and its inertness
towards interacting with proteins and nucleic acids, it
offers many other advantages as a separation me-
dium, such as easy sample recovery, nontoxicity, and
amenability to cast and use it in an open faced
format. The latter option underlies its unique
applicability to numerous immunoelectrophoretic
procedures. Agarose is very porous compared to
polyacrylamide, and that limits its suitability in
unmodiRed form for sieving-based separations of
proteins with molecular masses below 60 kDa. On
the other hand, its high permeability makes it, again,
superior to polyacrylamide for immunochemical
studies.
Properties of Agarose Gels
Chemically, agarose is a polysaccharide composed of
alternating D- and L-galactose biose (agarobiose)
units in chains that are on the order of 400 agarobiose
units in length, &120 kDa. Rees and Arnott each
characterized the gelation as proceeding initially
through pairing of the 120 kDa chains into
0.8}1.4 nm thickness double helical coils, followed
by lateral coalescence of the helical dimers into thick
protoRbrils with average thickness of the order of
24 nm, although the number of helical dimers can
range from 5 to 5000. By contrast, conventional poly-
acrylamide gels have Rbre thickness of only 0.4 nm.
Gelation occurs as the thick protoRbrils interlink at
their loosely coiled ends into a net-like matrix (Fig-
ure 1). The gels are translucent because of light scat-
tering by the thick Rbrils, but become transparent on
drying. The ability to form thick, strong Rbres enables
agarose to form gels at concentrations down to 0.2%
as utilized in studies by Serwer. The incorporation of
Figure 1 Depiction of structure of the agarose biose unit and
the structure of interlocking aggregates of coiled coil dimers of
agarose chains based on electron microscopy and X-ray diffrac-
tion. (Reproduced with permission from Westermeier, 1997.)
the agarose into thick rather than thin Rbres makes
the gels highly porous at concentrations up to 6%, the
limit to which agarose can be dissolved without re-
sorting to autoclave temperatures. Concentrations up
to 16% can be reached by autoclaving.
The high porosity makes agarose superior to poly-
acrylamide for separation of large proteins, poly-
peptides, complexes with size ranging from 100 kDa
to several megadaltons, and for DNA fragments
ranging from 1000 to 23 000 bp. In the interest of
retaining the gel casting simplicity of agarose while
enhancing its sieving capacity, nonsupercoiling hy-
droxyalkylated agarose derivatives have been de-
veloped which, according to Chrambach, have Rbre
thicknesses of the order of 0.8 nm, approaching poly-
acrylamide, but have gel strengths that are low com-
pared to unmodiRed agarose. Recently, blends of
agarose derivatives have been formulated which have
improved strength.
The sulfate present in most commercial agarose
induces a slight electroendosmosis (EEO) during elec-
trophoresis. Electroendosmosis is the Sow of buffer
through the gel. In effect it arises because the nega-
tively charged sulfate groups in the matrix cannot
move towards the anode which results in a compensa-
tory pumping of buffer to the cathode. The slight
effect on mobility of proteins is usually of little conse-
quence; however, it can have pronounced effects on
mobilities of buffer boundaries in discontinuous buf-
fer systems used for sharpening bands in the applied
samples. It can also induce syneresis and collapse of
gels with discontinuous buffer systems due to unequal
electroendosmosis across the buffer boundaries.
Thus, some of the numerous buffer systems that have
been described in the computer output by Jovin for
polyacrylamide gels do not work well with agarose
gels. Many do, nevertheless. The zero-EEO agarose
which is used for isoelectric focusing has its electro-
endosmotic tendency neutralized by charge-balancing
with electropositive groups. The additives in some

zero-EEO agaroses contribute to background
staining.
The gelation of agarose depends entirely on hydro-
gen bonding, and is inhibited by chaotropic agents
such as concentrated (&20%) glycerine or urea.
Although the high concentrations of these substances
usually used in biochemical procedures can block
the gelation, they do not appreciably affect gels once
solidiRed. As with all electrophoresis, concentrated
ionic chaotropes such as guanidine-HCl and KSCN
which are used as protein solubilizing and denaturing
agents should be removed and replaced with either
urea or SDS prior to electrophoresis. Since the
aggregation of agarose is fully reversible, gels
can melt under high current. Also, agarose gels
are compressible, and tend to drop out of vertical
electrophoresis cells unless supported. The ability
using weights to mechanically compress agarose
gels to within a few per cent of original thickness
can be used to advantage. Gels containing more
than 2.5% agarose will rehydrate to the original
thickness within a fraction of an hour, but gels with
1% or less agarose have a greatly reduced tendency to
rehydrate.
Agarose Derivatives
Numerous agarose derivatives have been described
for afRnity chromatography where addition of
charged groups is not critical. All derivatives pre-
pared for electrophoresis involve modiRcations not
imparting electrical charges. As indicated earlier,
hydroxyethylated (FMC BioProducts) and hydroxy-
methylated agarose (Hispanagar, SA) are low melting
derivatives which form gels with thin Rbres for en-
hanced sieving. These low melting agaroses have been
used to impart sieving during capillary electrophor-
esis. Allylglycidyl agarose, a very low melting deriva-
tive, is frequently used in place of bis-acrylamide
for cross-linking acrylamide into stronger gels. All
of the commercially available derivatives yield gels
with low melting points. That is because the agarose
is derivatized in a molten state. As we learned in
studies on alkylation of agarose with either glycidol
or allylglycidyl ether from preparing glyoxyl
agarose and its analogues, the hydroxyl groups
involved in the hydrogen bonding functioning in
gelation are protected during derivatization of the
agarose in the gel state, and the derivatized agarose
gel retains much of its original melting and gelling
characteristics.
Glyoxyl agarose (oxidized glyceryl agarose) has
acetaldehyde substituent groups which can be utilized
to form Schiff’s base linkages with protein amino
groups. At pH (8.5, these Schiff’s base linkages are
too dissociated to retard electrophoresis of the pro-
II / ELECTROPHORESIS / Agarose Gels 1171
teins. Once separated on the glyoxyl agarose gel, the
proteins can be driven to bind to it, either reversibly
by immersion in pH 10 buffer to suppress dissocia-
tion of the Schiff’s base linkages or by immersion in
buffer containing sodium cyanoborohydride which
rapidly and speciRcally drives alkylation of amino
groups of the protein with the aldehyde groups in the
gel matrix. This functionality enables the gel to be
used sequentially and repetitively as a protein-
separating and immobilizing medium. It can also
Rx small peptides containing at least a single
amino group, but does not Rx nucleotides. By compo-
siting it with a removable polyacrylamide Rller, it can
be used for separations over any molecular weight
range.
Agarose/Polyacrylamide Composites
These media were constructed initially to provide
a supporting agarose matrix for polyacrylamide at
concentrations below 3}4% which do not form Rrm
gels. The combinations can yield exceedingly strong
gels (Figure 2) and provide facility to achieve a wide
range of sieving characteristics. As described by Pea-
cock and Dingman in 1968, strong gels are obtained
only when formulated to allow the agarose to gel
before the acrylamide polymerizes. By using either
non-cross-linked polyacrylamide or cross-linkers
such as alkali or periodate degradable 1,2-dihy-
droxyethylene-bis-acrylamide instead of the usual
bis-acrylamide, the poorly-permeable acrylamide Rl-
ler can be removed leaving a highly permeable
agarose matrix to allow direct immunochemical char-
acterization of the Rxed components of the elec-
tropherogram.
Electrophoresis
Initial applications of agarose electrophoresis focused
heavily on diagnostic separations of plasma proteins,
particularly lipoproteins, with great improvement
over results obtained with paper or with starch which
had been used earlier. Straight analytical separations
were followed by more deRnitive immunoelectro-
phoretic protein identiRcations. These separations
were usually applied to native, undenatured proteins.
Following introduction in 1967 of sodium dodecyl
sulfate (SDS) by Shapiro, Vin
uela and Miazel as a car-
rier enabling polyacrylamide-based separation of pro-
teins according to molecular weight by imposing
overwhelming negative charge over the original
and by denaturing conformational differences be-
tween proteins, the practice was widely adopted as
a rapid means for separating large-sized proteins on
agarose. The converse approach of separating pro-
teins according to differences in isolectric points
1172 II / ELECTROPHORESIS / Agarose Gels
Figure 2 Demonstration of the strength of agarose/polyacrylamide composite gels. (Reproduced with permission from Shainoff,
1993.)
independent of molecular size by establishing a pH
gradient across the gel was demonstrated in 1969 by
Vesterberg, but a decade passed before the develop-
ment of zero-EEO agarose enabled this separation
method to be used with agarose.
Equipment and Buffer Systems
Agarose gels are usually used for electrophoresis on
a horizontal platen on which the gels are placed
between anodic and cathodic electrode chambers
Rlled with buffer and connected to the gel with buf-
fer-Rlled wicks. Except for the usual need to purchase
a platen that can be cooled with circulating water,
this is a simple arrangement that is easily set up,
largely because of the ease with which agarose gels
are cast either open faced or in cassettes consisting of
two spacer-separated glass plates which can be separ-
ated from the gel because the gel is nonadherent. The
use of wicks can be troublesome, because voltage is
lost across them, and this produces much of the heat-
ing and vapour condensation in the chamber. Voltage
drop across the gels must be gauged directly from the
gel rather than the power supply. Condensation of
vapour on the gel is easily prevented by simply laying
a vapour barrier, usually GelBond威 (FMC Bio-
Products), over the gel. For nucleic acid electrophor-
esis the use of wicks has been obviated by a technique
known as ‘submarine’ electrophoresis in which the
gel is laid into a buffer-Rlled platform connecting the
electrode chambers. The submarine mode is usually
not applicable to protein electrophoresis because pro-
teins tend to diffuse into the surrounding buffer,
while nucleic acids which have very low diffusion
constants migrate almost entirely due to the applied
voltage with little diffusion. Running agarose gels in
vertical format with electrode chambers directly
above and below the gel-containing cassette, as gener-
ally used with polyacrylamide gels, is seldom recom-
mended because of the tendency of agarose to
synerese and slip out of the cassette, a problem that
can be prevented by use of etched rather than clear
glass for the cassette. Slipping can also be prevented
by partially immersing the cassette into dilute 0.5%
agarose, and letting the agarose drain and dry before
pouring the running gel.
Sample applications into agarose gels were initially
made by either cutting or punching a slot or hole to
receive the solution. This approach continues to be
used with nucleic acids which form sharp bands as
they migrate out of the nonrestrictive sample solution
into the migration-restrictive gel, but comparatively
little band sharpening occurs with proteins which
easily permeate the gels. Further, the wells induce
band distortions of proteins due: (i) to partial per-
meation into the sides of the wells prior to elec-
trophoresis; and (ii) to uneven voltage drops across
the wells. Because of the ease with which solutions
can be imbibed into agarose following a temporary
compression with blotting paper, samples can be
drawn into the gels in nicely demarcated bands
through a ‘sample application foil’, thin plastic over-
lays with slots through which the samples are drawn
into the gels. Large samples need band-sharpening
which can be achieved either mechanically by tem-
porarily overlaying a dialysis membrane and thick gel
to produce a sharp voltage drop near the origin, or by
use of discontinuous buffer systems.
 II / ELECTROPHORESIS / Agarose Gels 1173

A method for sharpening bands at the beginning of

runs at the outset of electrophoresis was devised by
Ornstein and Davis using a discontinuous buffer sys-
tem, ‘disc electrophoresis’, in which proteins and all
but common electrolytes stack in hypersharp bands at
the boundary between initial buffer ions producing
low conductivity and secondary buffer ions produ-
cing high conductivity with a common counter ion.
The discontinuous buffer systems worked predictably
with over 7000 computer generated buffer systems
based on the pKa values of the leading and trailing
buffering ions, as far as known, with polyacrylamide
gels which produce no EEO, but only a few have been
tested and found to yield operable systems with
agarose, as already noted.
The ability to run agarose gels horizontally in a vir-
tual open-faced format is its real advantage. That
enables the gels to be cut and manipulated post-
electrophoresis to conduct secondary electrophoretic
operations with it. This facility has made it the usual
medium for immunoelectrophoretic analyses.

Staining
With the exception of silver staining, all commonly
used staining methods (Coomassie-based, Suorescent,
chemiluminescent, and colloidal-type) work well
with agarose, colloidal Coomassie being one of the
fastest. An important consideration is to minimize
exposure to acid which hydrolytically weakens the
gel. Unlike polyacrylamide, agarose gels can be dir-
ectly immunostained as illustrated (Figure 3), but
the antibodies should be free of large entrapable
aggregates.
Immunoelectrophoresis
These methods are based on immune-precipitate
(precipitin) formation at antigen/antibody equiva-
lence, analogous to Ouchterlony immunodiffusional
analyses in which antibody and antigen placed in
separate wells in agar(ose) form precipitin arcs in the
gel at position(s) depending on levels of antigen and
antibody in the wells. In the method of Grabar and
Williams the proteins under analysis are separated on
agar gels, then lateral slots are cut approximately
8 mm to each side of the gel to accommodate anti-
body. After a day lateral diffusion of the protein
bands towards the counter-diffusing antibody one or
more precipitin arcs will form in the gel depending on
the antibody and antigen heterogeneity. Quantitation
of antigen/antibody by this method requires consider-
able effort and multiple runs. A simpler approach to
quantitation was devised by Laurell, using a method
known as ‘rocket immunoelectrophoresis’.
Figure 3 (See Colour Plate 39). Direct dual immunostaining of
a polyacrylamide / glyoxyl-agarose composite gel to profile fibrino-
gen -chain (grey), -chain (amber) cross-linking and hybrid  / -
chain (umber) cross-linking by plasma transglutaminase (right
lane), and the chain composition of plasma fibrinogen. The illus-
tration depicts sieving equivalent to a regular polyacrylamide gel,
and subsequent rendering of the gel for antibody permeation by
removing the polyacrylamide. (Reproduced with permission from
Shainoff et al., 1991, Journal of Biological Chemistry 266: 6429.)
In the method of Laurell, the gel is poured with
antibody added to it at 563C, and wells are punched
in the solidiRed gel to accommodate the antigen solu-
tion. As the antigen moves by electrophoresis out of
the well it sweeps soluble immune complexes along
with it in a comet-shaped proRle until the antigen/
antibody levels become equivalent, whereupon the
comet-shaped precipitin arc comes to a virtual stop
because the antibodies themselves do not migrate at
around pH 8.6. The area contained within the arc is
usually directly proportional to the antigen level. Be-
cause of the ease with which agarose gels can be cut
and Rlled with interposed gels open-faced, this tech-
nique is amenable to a myriad of variations, as de-
scribed by Axelsen and associates.
1174 II / ELECTROPHORESIS / Agarose Gels
In one variation of the Laurell method, called
crossed immunoelectrophoresis, the proteins are sub-
jected to pre-electrophoresis, and antibody-contain-
ing agarose is cast around a strip of the gel. Then, in
a crossed electrophoresis the antigens are transferred
out of the primary gel into the antibody where they
form rocket(s) in line with their initial position. The
technique is ‘found’ useful for quantifying multiple
antigens and variant forms of an antigen. If the initial
gel contains SDS, it should be quenched by adding
nonionic detergent such as Lubrol威 to the antibody-
containing gel. In the event that the protein becomes
insoluble when stripped of SDS, initially it can be
Rxed, then probed with charge-enhanced car-
bamylated primary antibody. Retained antibody is
measured by electrophoresis into secondary antibody,
as illustrated in Figure 4. Use of secondary antibodies
is also essential when using monoclonal IgG antibod-
ies as primary probes, because these monoclonal anti-
bodies do not form immunoprecipitates on their
own unless the epitope is multiply expressed in the
antigen.
Direct Immunoprobing
Because of the permeability of agarose gels to anti-
body it is possible to probe electropherograms
directly. Direct immunoprecipitation within the gel
is seldom used because of uncertainties of levels or
antibody required for substantial precipitation of un-
Figure 4 Crossed immunoelectrophoresis to profile plasma fib-
rinogen derivatives in plasma by: (i) probing the electropherogram
with primary antifibrinogen antibodies; and (ii) measuring the
retained antibody by displacing to form rockets in secondary gel
containing anti-IgG antibodies, with standards for the IgG forming
the left- and right-most peaks. This approach was made neces-
sary because insolubility of fibrinogen and its high molecular
mass derivatives, once denatured by SDS, cannot be transferred
out of the primary electropherogram to form rockets directly.
(Reproduced from Dardik et al., 1989, Cleveland Clinical Journal
of Medicine 56: 451.)
known levels of antigen. The proteins are usually
Rxed or immobilized, and probed by imbibing label-
led antibody into and out of the gel. If secondary
antibodies are to be used to report retention of the
primary antibody, the primary antibody should in
turn be Rxed before the secondary probing to avoid
dissociative losses of the primary during the second-
ary probing which takes periods of the order of an
hour. Glyoxyl agarose was developed to enable these
Rxations by chemical immobilization.
Blotting
This widely used procedure involves transfer of com-
ponents out of the gel on to a blotting membrane to
immobilize them for immunostaining or composi-
tional analysis on an open-faced surface. It is an
essential means for probing reactivities of compo-
nents separated on polyacrylamide because of the low
permeability of the gel. Much uncertainty attends this
‘blotting’, because the transfers are incomplete, and
frequently nil with high molecular mass proteins.
These proteins not only transfer slowly, but often
precipitate within the gel as SDS transfers away from
them. Also, very low molecular mass peptides fail to
be retained by the blotting membrane.
While blotting from polyacrylamide gels is usually
effected by crossed electrophoresis, the transfers out
of agarose gels are more simply effected by either
compressing them against the membrane supported
on Rlter paper stacks (a method that yields only
partial transfer), or by light vacuum suctioning of
buffer through the gel on to the membrane supported
on a ‘gel drier’. However, when SDS is present it must
either be quenched by a prior 15 min immersion in
buffer containing 1}2% nonionic detergent, or pre-
cipitated by immersing the gel in 0.1 M KCl.
Af\nity Electrophoresis
These methods use speciRc ligands, either added to
the buffer or immobilized on the gel matrix, to induce
shifts in mobilities or apparent concentrations of tar-
get proteins. Again, ability to work with open-faced
gels to interpose ligand-rich zones makes it an ideal
medium for these procedures, as devised with numer-
ous examples from crossed immunoelectrophoresis.
With glyoxyl agarose a ligand containing at least
a single amino group can be imbibed into the gel
through a mask and Rxed in place.
Conclusion
Equipment and supplies for general electrophoresis
on agarose have undergone little change over the last

two decades. Improved imaging systems and labelling
agents for high sensitivity detection and linear
quantitation by Suorescence and chemiluminescence
offer attractive alternatives to immunostaining and
the use of radioisotopes, and have made rapid auto-
mated nucleic acid sequencing possible. Except for
those involved in nucleic acid work, few laboratories
perform electrophoresis on agarose on a day-to-day
basis; thus, accessories for many techniques such as
immunoelectrophoresis have been dropped by sup-
pliers. There may be a revival because of growing
interest in several proteins which are not separable
in polyacrylamide matrices, von Willebrand factor
multimers for diagnosis of certain types of von
Willebrand’s disease, and high molecular mass
derivatives of Rbrinogen as markers of vascular
disease.
More efRcient cooling and thermal control would
allow the application of much higher voltages for
high performance agarose electrophoresis above the
present limitations imposed by gel meltdown. By us-
ing an apparatus in which the gel is enveloped with
a membranous bladder for cooling, von Willebrand
factor multimers can be separated in 20 min runs
producing resolution superior to that obtained with
the usual overnight runs widely used through the
1990s.
Precast polyacrylamide gels have become popular,
and precast agarose gels are available for submarine
electrophoresis. Compressibility of agarose is the
principal detriment to precasts for other modes of
electrophoresis. Agarose/polyacrylamide composites
would not be subject to that drawback, and would
probably be marketable in precast form because they
offer advantages and are not as easily constructed as
gels of agarose itself.
Summarily, agarose gels are simple to construct.
They are highly porous and ideal for separating high
Autoradiography Electrophoresis
See II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
Blotting
See II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
II / ELECTROPHORESIS / Agarose Gels 1175
molecular mass proteins and nucleic acids, and can be
modiRed or composited to extend their utility.
See Colour Plate 39.
Further Reading
Andrews AT (1989) In Peacock AR and Harrington WF
(eds) Electrophoresis. Theory, Techniques, and Bio-
chemical and Clinical Applications, 2nd edn. New York:
Oxford University Press.
Armisen R (1997) In Imeson A (ed.) Thickening and Gelling
Agents for Food, 2nd edn, pp. 1}21. London: Blackie.
Axelsen NH (1975) Quantitative Immunoelectrophoresis.
New Developments and Applications, Oslo: Universi-
tetsforlaget.
Axelsen NH, Kroll J and Weeke B (1973) A Manual of
Quantitative Immunoelectrophoresis. Scandinavian
Journal of Immunology, Vol. 2, Supplement no. 1, Oslo:
Universitetsforlaget.
BuzaH s Z and Chrambach A (1982) Un-supercoiled agarose
with a degree of molecular sieving similar to that of
crosslinked polyacrylamide. Electrophoresis 3: 130}134.
Chrambach A (1985) The Practice of Quantitative Gel
Electrophoresis. Weinheim: VCH.
Hames BD and Rickwood D (1981) Gel Electrophoresis of
Proteins. Oxford: IRL Press.
Orban L, Hahn E and Chrambach A (1988) Discontinu-
ous buffer systems optimized for the agarose gel elec-
trophoresis of subcellular particles. Electrophoresis 9:
167.
Pharmacia Fine Chemicals (1982) Isoelectric Focusing,
Principles and Methods. Uppsala: Pharmacia.
Serwer P (1980) A technique for electrophoresis in mul-
tiple-concentration agarose gels. Analytical Biochemis-
try 101: 154}159.
Shainoff JR (1993) Electrophoresis and direct immunopro-
bing on glyoxyl agarose and poyacrylamide composites.
Advances in Electrophoresis 6: 61}177.
Westermeier R (1997) Electrophoresis in Practice. Wein-
heim: VCH.
1176 II / ELECTROPHORESIS / Capillary Electrophoresis
Capillary Electrophoresis
S. F. Y. Li and Y. S. Wu, National University of
Singapore, Singapore
Copyright ^ 2000 Academic Press
Introduction
The migration of charged particles under the inSu-
ence of an electric Reld was discovered and character-
ized theoretically more than 100 years ago by
Kolrausch et al. Foreseeing the possibility of separ-
ation of charged species through the application of
a voltage, the term ‘electrophoresis’ was coined soon
after. However, early attempts to use electrophoresis
as an analytical tool were persistently frustrated by
the existence of Joule heating, which acts to discount
the electrophoretic effect. Thus a way of combatting
the thermal effect during the electrophoretic process
was needed. By 1950s, Tiselius et al. found that
a variety of substances such as agarose and polymeric
gels could serve as stabilizing agents in electro-
phoretic analysis owing to their anticonvective
properties. This eventually led to the creation of slab
gel electrophoresis, which has become a fundamental
technique for the study of proteins, DNA fragments
and other biomacromolecules in life sciences and
biotechnology. Notwithstanding its great success,
slab gel electrophoresis has its drawbacks with re-
spect to speed and automation when compared
with contemporary chromatographic techniques
such as high performance liquid chromatography
(HPLC).
A straightforward way to speed up an electro-
phoretic separation process is to apply higher electric
Relds, and this necessitates systems able to release the
heat generated more efRciently. Electrophoresis with
a tube as a separation channel is hence an attractive
choice since the desired surface-to-volume ratio can
be achieved by simply reducing the tube radius. Per-
forming electrophoresis based on the tube format has
an added advantage in that simultaneous detection
may be implemented in a way analogous to HPLC,
thus rendering the entire procedure fast and auto-
matic. Running electrophoresis with a tube conRgura-
tion was initiated by Hjerten as early as the 1960s,
and further attempted by Virtanen et al. and Mikkers
et al. in 1970s. During this period, the adopted inner
diameters of tubes were in the range of 0.2}3 mm,
and thermal effects conRned the applied voltage to
around 1000}2000 V, which was of the same order
as in typical slab gel electrophoresis. As a conse-
quence, despite these pioneering efforts to perform
free solution electrophoresis with in-line monitoring,
the full potential with respect to column performance
was not yet attained. Also, complexity in instrumen-
tal design deterred follow-up by ordinary elec-
trophoresis practitioners.
A milestone for column-based electrophoresis was
set in the early 1980s, when Jorgenson et al. introduc-
ed capillary zone electrophoresis (CZE) with on-col-
umn optical detection. They found that with the inner
diameter of the capillaries scaled down to 80
m,
voltages as high as 30 kV could be applied without
incurring overheating problems. Thus the separation
time for most charged species, from small molecules
to macromolecules, was shortened to less than
30 min, which is comparable to modern chromato-
graphic methods. For the Rrst time outstanding
column efRciencies of several hundred thousand
plates was routinely obtained. The unprecedented
performance, relatively simple instrumentation, con-
current with the widespread availability of fused
silica capillary columns by the mid-1980s quickly
aroused the interests of both electrophoresis practi-
tioners and chromatographers, thus making capillary
electrophoresis (CE) one of the most exciting research
areas. Today, it has become an indispensable branch
of modern separation science. The powerful separ-
ation ability of CE was exempliRed in an early
electropherogram concerning the resolution of de-
rivated peptides originated from egg white lysozyme
(Figure 1).
This article serves as an introduction to CE. It
covers the basic principles, various aspects of instru-
mentation, separation modes and major applications.
Some future trends of CE are discussed in the Rnal
section.
Fundamentals
Electrophoretic Migration of Ions
The uniform motion of an ion under an electric
Reld can be recognized as a result of balancing
electromotive and frictional forces of the ion in
solution:
qE"6
ru [1]
where q is the effective charge of the ion concerned,
E is electric Reld, while r is the ion’s Stokes radius,  is
the dynamic viscosity of the solution, and u is the
linear velocity of the ion. The important parameter,
electrophoretic mobility (), is deRned as the ion’s
 II / ELECTROPHORESIS / Capillary Electrophoresis
Figure 1 Capillary zone electrophoresis separation of fluor-
escamine-labelled peptides obtained from a tryptic digest of re-
duced and carboxymethylated egg white lysozyme. (Adapted with
permission from Jorgenson JW and Lukacs KD (1981) Zone
electrophoresis in open-tubular glass capillaries: preliminary data
on performance. Journal of High Resolution Chromatography and
Chromatographic Communications 4: 230}231.)
linear velocity per unit of electric Reld:
u q
" " [2]
E 6
r
From eqn [2], it can be seen that the ion’s effec-
tive charge, its size and the viscosity of the solution
decide ionic mobility. Thus in a given system ionic
mobility is an intrinsic property of an ion. Usually
ionic mobility cannot be directly derived from
eqn [2], as the parameters are not easily accessible
quantities. Instead it can be measured based on rel-
evant experimental data, i.e. how long an ion takes to
travel through a certain distance under a deRnite
electric Reld, as follows:
u Leff 1 Leff 1 Leff;Ltot
" " ; " ; " [3]
E tm E tm V/Ltot V;tm
where Leff and Ltot are the effective migration length
(from inlet to detection window) and total migration
length, respectively, V is the applied voltage, and tm is
the migration time of the ion. For a CE system oper-
ated under a constant voltage, Leff, Ltot and V are all
1177
Rxed. Hence electrophoretic mobility, the inherent
attribute of an ion is directly reSected by its migration
time. This provides the theoretical basis of using
migration time as a means of identifying an ion
in CE.
Electroosmotic Flow (EOF)
Electroosmosis is a fundamental electrokinetic effect
involving movement of the bulk solution against
a charged solid surface under the inSuence of an
electric Reld. In the case of CE, the capillary inner
wall usually carries negative charges due to the de-
protonation of silanol groups. For the part of the
liquid adjacent to the capillary wall, build-up of
cations takes place to counterbalance the negative
charges on the capillary surface. According to Stern’s
double layer model the solution containing net ca-
tions can be divided into two regions, namely a rigid
layer and a diffuse double layer. The rigid layer is
immediately adjacent to the capillary wall, so the
cations within it are largely immobilized owing to the
strong electrostatic interaction with the wall. The
diffuse layer is slightly away from the wall, hence the
cations inside are mobile. Upon the application of
a voltage, these cations together with their surround-
ing hydrating water will migrate towards the cathode.
The cohesive nature of water causes the whole solu-
tion inside the capillary to be dragged forward,
generating a net Sow across the capillary. This is
named the electroosmotic Sow (EOF). The magni-
tude of the EOF can be described via the Helmultz
equation:

eo" [4]
4

where  is the dielectric constant of the buffer solu-
tion,  is the zeta-potential across the diffuse layer,
and  is the viscosity of the electrolyte. Unlike con-
ventional electrophoresis where EOF is regarded as
unfavourable and thus usually suppressed, in CE it
has several important positive implications.
First, the existence of an EOF offers a simple and
highly efRcient way of driving a separation system.
The zeta-potential is uniformly distributed within an
extremely narrow cylindrical region along the whole
capillary so the bulk electrolyte solution
is pumped out of the capillary with virtually no
pressure drop (Figure 2). A ‘plug-like’ Sow is ob-
tained, which subsequently contributes to high col-
umn performance. This is advantageous over the
conventional pumping methods such as in HPLC,
where the pressure-based Sow always introduces
a parabolic proRle thus adding to the loss of column
efRciency.
1178 II / ELECTROPHORESIS / Capillary Electrophoresis
Figure 2 The generation of electroosmotic flow (EOF) in a
silica capillary.
Second, the presence of EOF affects the apparent
mobilities of ions (Figure 3). In any electrophoretic
separation system where EOF is not fully suppressed,
the observed mobility of a charged species will be the
resultant of its effective electrophoretic mobility and
EOF:
obs"ep#eo [5]
Under normal conditions, with EOF directs towards
the cathode, obviously cations will be accelerated,
while anions will be decelerated. If the magnitude of
the EOF exceeds the mobilities of the anions, the
anions will be swept towards the detection side, thus
allowing the simultaneous analysis of cationic and
anionic species. As the magnitude and direction of
EOF will affect how long the analytes stay inside the
separation capillary, manipulation of EOF often be-
comes a core issue for effecting a satisfactory resolu-
tion. Since the formation of EOF involves two phases
(capillary wall and running buffer), any modiRcation
to their chemistries will bring about a change in EOF.
Causes of Band Broadening
As in a chromatographic process, in electrophoresis it
is necessary to contain the ionic species within nar-
row bands while creating sufRcient mobility differ-
ences. How narrow a band is depends not only on the
various dispersive factors inherent to the elec-
trophoretic process, but also on how well the whole
process is performed. The common causes of band
Figure 3 Effect of EOF on the apparent mobilities of anions and
cations.
broadening in CE include longitudinal diffusion, in-
jection-related volume overloading, thermal
effects, electrodispersion, wall adsorption, etc. These
band broadening mechanisms are deemed to be ran-
dom and independent events, so that the concept of
summation of variances can be used to evaluate the
contributions of individual factors to the overall band
broadening effect, that is:
2tot"2diff#2inj#2therm#2wall#2electr#2other [6]
A brief description of these band broadening factors
is given below.
Longitudinal diffusion In the course of electro-
phoretic transportation of an analyte band along the
capillary, the sample molecules will inevitably have
a tendency to enter the surrounding buffer solution
because of the apparent concentration difference,
leading to a wider and more dilute sample band.
According to Einstein’s diffusion equation, band dis-
persion due to longitudinal diffusion is a function of
diffusion coefRcient and time:
2diff"2Dmt [7]
Under an ideal situation, longitudinal diffusion be-
comes the only unavoidable band broadening pro-
cess. Therefore it deRnes the maximum attainable
column efRciency in CE. Based on chromatographic
theory, the maximum obtainable theoretical plates
(N) can be derived as follows:
L2 L2 L2
N" 2 " "
 2Dmt 2Dm;(L/v)
L2 V
" " [8]
2Dm;[(Leff;Ltot)/V] 2Dm
Thus the maximum column efRciency in CE is pro-
portional to the mobility and voltage, while
inversely proportional to the diffusion coefRcient.
Considering that mobilities of ions range
between 10\4 and 10\3 cm2 V\1 s\1, diffusion coefR-
cients from 10\7 to 10\5 cm2 s\1, and an applied
voltage up to 104 V, the attainable theoretical
plate number would be in the order of 105}106, which
is much higher than any conventional HPLC ap-
proach. Eqn [8] also suggests that in principle CE
should be well suited for the separation of high-mass
charged particles such as biopolymers, since their
diffusion coefRcients are extremely low. This has
been demonstrated in the most successful resolution
of DNA fragments and proteins where plate numbers
of 106 have been reached. It should be emphasized
 II / ELECTROPHORESIS / Capillary Electrophoresis
that eqn [8] is only valid under the precondition that
longitudinal diffusion plays a predominant role
among the various band broadening mechanisms. In
other words, to achieve the maximum column efR-
ciency, the electrophoretic separation should be car-
ried out in such a way that all the other potential
band dispersions are curbed well below the magni-
tude of the longitudinal diffusion effect.
Injection related volume overload During sample
injection, a Rnite volume of sample is placed onto the
capillary. The length of this starting plug will contrib-
ute directly to the Rnal band width. Treating the
original band as rectangular in shape, the variance of
this plug can be expressed by:


2
linj
2inj" [9]
12
where linj is the initial plug length. As a rule of thumb,
loss of efRciency due to any extraneous dispersion
factor should be kept within 10% of the maximum
theoretical column efRciency. Assuming a
moderate plate number in the order of 105 as deRned
by longitudinal diffusion, it can be easily estimated
that the acceptable injection length should be a few
millimetres. For the commonly employed capillaries
with inner diameter between 50 and 75
m, the above
length is equivalent to only a few nanolitres. So it is
obvious that CE’s high column efRciency will pose
very stringent restriction on the sample size. Any
attempt to increase the injection volume in an aim to
enhance detection sensitivity may result in a signiR-
cant loss of column efRciency.
Thermal gradient effect An electrophoretic pro-
cess is always accompanied by certain amount of
thermal effects due to the passage of a current
through the resistive medium (Joule heating). For
a CE system, the electrical power (P) responsible for
the generation of heat can be estimated through the
following equation:
V 2 V 2
r2c
P"V;I" " [10]
R L
where V is the applied voltage, r and L are capillary
inner radius and length, respectively, while  and
c are respectively the molar conductivity and
the concentration of the electrolyte solution. While
heat generation is uniform for the whole electrolyte
solution, heat dissipation is apparently not: the nearer
the electrolyte is to the capillary wall, the faster is
the heat transferred out to the surroundings. Con-
sequently, a temperature gradient is generally present
in the radial direction of the capillary, which is
1179
equivalent to the superimposition of a parabolic
proRle to the otherwise plug-like ion boundaries
and bulk Sow, as any temperature gradient will be
translated into viscosity and mobility gradients in the
solution. To minimize the inSuence of thermal effects
on the overall column efRciency, it is imperative to
limit the heat generation while maximizing the heat
dissipation. In this regard, the use of narrow bore
capillaries is particularly recommended because it
favours the above two aspects simultaneously. Ac-
cording to eqn [10], for a certain CE system, heat
generation may also be controlled through balancing
the buffer composition and separation potential.
Wall adsorption effect The capillary surface, like
most solid surfaces, never behaves in a completely
inert manner to foreign compounds. In HPLC, it has
been well known that peak anomalies are often the
result of some speciRc interaction (e.g. hydrogen
bonding) between the residual silanol groups and
analytes. While similar adverse effects cannot be
ruled out, in CE the problem is exacerbated by the
fact that under a typical operation condition, the
silanol groups along the wall are mostly deprotonated
to give a negatively charged capillary surface. When
an analyte with a positive charge travels along the
capillary, the electrostatic force will tend to attract
the analytes onto the wall, causing additional band
broadening. This is a feature of the analysis of pro-
teins owing to their low diffusion coefRcients and
multiple charge sites. SigniRcant efforts have been
made to tackle this problem, mostly through the
suppression of EOF or complete reversal of the
charge status of the capillary wall.
Electrodispersion Electrodispersion is the result of
Ohm’s law during the electrophoretic separation. It
may appear in two instances. First, if the conductivity
of the injection plug is larger than that of the buffer
solution, an isotachophoresis effect will occur to di-
lute the original band till the conductivity is equal to
the surrounding buffer. Second, during the separ-
ation, an analyte zone is ‘submerged’ into the buffer
solution. Any mismatch of its mobility with its co-
ions in the buffer will render the local electric Reld
different from that in the normal running buffer. If
the mobility of the sample ion is larger than that of its
co-ions, then there will be a lower electrical Reld for
the analyte zone. Thus any sample ions diffusing out
of the zone will experience a higher electric Reld, and
these ions will speed up along the migration direction.
This causes the ions at the rear to re-enter the zone,
whereas the ions at the front will drift away from the
zone. The accumulative effect of such phenomenon is
the formation of a tailing band showing a sharp
1180 II / ELECTROPHORESIS / Capillary Electrophoresis
Figure 4 Mobility mismatch-induced band broadening.
trailing boundary but a broadened leading boundary.
By a similar argument, if the analyte ion is of lower
mobility than the co-ion, a fronting band is expected
(Figure 4). The higher the sample concentration com-
pared to the buffer concentration, the more pro-
nounced are the nonuniformities with respect to con-
ductivity and Reld strength, and eventually more se-
vere is the band asymmetry.
From the above discussion it is obvious that, to
prevent possible loss of column efRciency due to elec-
trodispersion, the conductivity of the injection plug
and the actual sample concentration should be sufR-
ciently low. Theoretical study had shown that, to
conRne the electrodispersion-related band broaden-
ing at a level comparable to longitudinal diffusion,
the sample concentration should be two orders below
the buffer concentration. To some extent, it is the
electrodispersion that limits the mass loadability of
a CE system.
Instrumentation
CE can be performed with relatively simple instru-
mentation as depicted in Figure 5. A capillary con-
taining an appropriate separation medium spans two
buffer reservoirs, to which the high voltage power
Figure 5 Instrumental setup of a capillary electrophoresis
system.
supply is connected via the platinum electrodes. Fol-
lowing the introduction of sample at the capillary
inlet, a high voltage is applied, thus driving the
analytes to travel inside the capillary with different
velocities. Somewhere close to the capillary outlet, an
on-line detector is installed to monitor the separation
process. The resulting signals are fed to the data
acquisition device, and Rnally the result is presented
in the form of an electropherogram.
Apart from these fundamental components, com-
mercial CE instruments are commonly equipped with
some dedicated facilities, such as an autosampler,
pressure regulating unit, capillary thermostatting,
and comprehensive supporting software. These
added functions allow a sequential analysis of differ-
ent samples under prespeciRed conditions, thus ensur-
ing better reproducibility, accuracy and higher
throughput. Two modern commercial systems are
shown in Figure 6.
High Voltage Power Supply
A high voltage power source delivering stable DC
potential of $30 kV will satisfy the requirement of
most CE applications. Many power supplies offer
additional features such as polarity switching, con-
stant potential/current setting, and an interlocked
Figure 6 Commercial CE systems. (A) Bench top CE system.
Photograph courtesy of Bio-Rad Laboratories. (B) Portable CE
system. Photograph courtesy of CE Resources Pte. Ltd.
 II / ELECTROPHORESIS / Capillary Electrophoresis
antielectrical shock loop. In the commercial CE in-
struments, the power supply is designed with digital
communication capability, so that more information
such as a current curve can be tracked and retrieved,
and the applied voltage can be programmed.
Separation Capillary
Although capillaries made of glass or polymer
(e.g. TeSon or Nylon) have found occasional applica-
tion, fused silica capillaries are used predominantly in
CE, largely due to their strength, Sexibility and most
importantly their excellent UV transparency. Usually
the fused silica capillary is coated with a layer of
polyimide to enhance its durability. For on-column
optical detection a small segment of this coating
needs to be removed to provide the detection win-
dow. The most commonly used CE capillaries have
inner diameter between 20 and 75
m, outer dia-
meter 100}400
m and are about 30}100 cm in
length.
Sample Injection
To achieve high column efRciency and good quantit-
ative results, sample injection must be performed in
a reproducible manner. Since the injection volume in
CE is in the nanolitre range, which precludes the use
of conventional injection methods, alternative ap-
proaches have to be pursued. Hydrodynamic injec-
tion and electrokinetic injection have turned out to be
the most widely employed sampling techniques.
Hydrodynamic injection introduces a sample based
on a pressure difference in the two sides of the capil-
lary. For a laboratory-built instrument, this is realized
by simply lifting up the sample vial together with the
capillary inlet for a certain period of time (typically
a few seconds). The hydrodynamic force will siphon
a band of sample solution into the capillary. For
commercial instruments, the pressure drop is created
by either applying pressure at the inlet side, or impos-
ing a vacuum at the outlet vial. The injection volume
can be calculated based on Poiseuille’s law:
P
r4t g
r4 ht
Vinj" " [11]
8L 8L
where P and h are the pressure and height differ-
ences, respectively, while r, t,  and L represent capil-
lary inner radius, injection time, solution viscosity
and capillary total length, respectively.
Electrokinetic injection is based on the transporta-
tion of sample ions by electrophoretic movement and
EOF. Normally a lower voltage than that for separ-
ation purpose is applied for a certain amount of time
to allow analytes to migrate into the capillary. The
1181
injection volumes of individual components can
be calculated through the following equation:
(i#eo)
r2Vt
Vinj" [12]
L
where i, eo are the mobilities of the analyte and
EOF, respectively, and V is the injection voltage. For
electrokinetic injection the injected amounts of differ-
ent analytes are dictated by the mobilities of the
respective analytes. Thus it is different from hy-
drodynamic injection, where the sample plug is of
entirely the same composition as the original sample
solution. To avoid sample injection bias, hydro-
dynamic injection is preferred. However, in some
circumstances hydrodynamic injection may be im-
practicable (e.g. owing to the high viscosity or low
permeability of the separation medium). Elec-
trokinetic injection is then the only viable option,
such as in capillary gel electrophoresis. On the other
hand, sometimes electrokinetic injection may be ex-
ploited in favour of CE operation. For example, it can
be employed to diminish the interference of the
sample matrix if the components concerned are of
low mobilities. Moreover, if the sample is of low
conductivity, then sample enrichment during elec-
trokinetic injection is possible by taking advantage of
the sample-stacking effect.
Detection
In CE, sample separation is accomplished in an elec-
trolyte solution, so detection strategies analogous to
HPLC are adopted. Compared with HPLC, CE col-
umn efRciencies are at least one order of magnitude
higher, which suggests that solute bands will be nar-
row and the average concentration of the analyte
zone is several times higher than for HPLC. As far as
a concentration-sensitive detector concerned, this im-
plies a larger detection signal output. However, so far
for CE the concentration sensitivity is usually lower
than its HPLC counterparts. This apparent contradic-
tion stems mainly from the small sample size,
which poses great difRculties for detection. The fun-
damental detection schemes in CE fall into three
categories: optical (UV absorptive and Suorescent)
detection, electrochemical detection, and various hy-
phenating techniques (typically mass spectrometric
detection). The sensitivities of the different detection
methods are compared in Figure 7.
UV adsorbance This is the most commonly used
detection method for CE, mainly because of its sim-
plicity. It is frequently implemented by modifying
a HPLC-type UV detector. The normal Sow cell is
1182 II / ELECTROPHORESIS / Capillary Electrophoresis
Figure 7 Comparison of sensitivities of various CE detection methods. (Adapted with permission from Landers JP (ed.) Handbook of
Capillary Electrophoresis, 2nd edn, chap. 10, pp. 425}448. Boca Raton, FL: CRC Press.)
removed and the capillary is mounted directly be-
tween the incident lens and photocell, in conjunction
with a corresponding aperture. Such on-column de-
tection conRguration, though easy to execute, pro-
vides only a moderate detection sensitivity with the
lower detection limits around 10\5 to 10\6 mol L\1,
as it is associated with two major problems. First,
since the capillary is illuminated radially, the max-
imum optical path is equivalent to the inner diameter
of the capillary. This severely reduces the achievable
absorbance according to Lambert}Beer’s law. Sec-
ond, due to its unique cylindrical geometry, the capil-
lary tends to act like a lens to defocus the incoming
light, thus posing difRculties in light orientation and
collection. As a consequence, signal noise and nonlin-
earity are further aggravated. By bending the capil-
lary into Z-shape or creating a bubble feature on the
capillary body, the optical pathlength can be con-
siderably increased. These efforts, coupled with
improved detection optics, enhance the detection
limits to 10\7 mol L\1. The capillary can also be
connected to an HPLC-type Sow cell (pathlength
&1 mm) as in the HP system.
Fluorescence detection Fluorescence detection is so
far the most sensitive detection mode available to CE
due to the fact that the measurement is performed
against a ‘dark’ background, and Suorescence inten-
sity is less pathlength dependent but directly propor-
tional to excitation power. On-column Suorescent
detection can also be realized through adaptation of
an HPLC-type detector. Once again, due to the pres-
ence of the lens effect of the capillary, improvement
of the detection optics is necessary to ensure better
focusing of excitation light and more effective collec-
tion of emission light. An important development in
CE Suorescence detection is the introduction of laser-
based excitation sources. Owing to its outstanding
coherence, a powerful laser source can be focused
into an extremely sharp beam to illuminate the core
part of the capillary, thus producing emission light of
high intensity. As a result, detection limits of
10\9}10\11 mol L\1 can be obtained routinely,
which is at least two orders of magnitude more sensi-
tive than the conventional approach. Single molecule
detection has been demonstrated by Rne-tuning the
CE system and utilizing laser-induced Suorescence
(LIF) detection. A major drawback of Suorescence
detection is that the number of analytes with native
Suorescence is far less than that with UV absorbance.
Therefore, to make use of this highly sensitive detec-
tion scheme, derivatization of analytes by tagging
with a Suorescent group is often needed.
Indirect photometric detection For compounds
without any chrophormore, indirect detection offers
a simple, effective way to take advantage of the con-
venience of on-column optical detection. Here a UV-
absorbing or Suorescent compound with the same
sign as the sample ions is added to the running
buffer to provide a stable background. During the
electrophoretic separation, the sample ions will dis-
place a certain amount of background ions due to
electroneutrality requirements. Thus the passage of
a sample zone through the detection window will
appear as a negative peak. In principle, the quantiRca-
tion of a sample in such a way involves differentiating
a signal from two large responses, hence the detection
 II / ELECTROPHORESIS / Capillary Electrophoresis
sensitivity is lower than that for direct photometric
detection. To achieve an acceptable detection limit,
the background concentration should be kept low.
Under such circumstance, any mismatch of mobilities
between the sample ions and background co-ions will
give rise to a considerable electrodispersion, hence
leading to severely distorted peaks. Therefore,
selecting a suitable CE background is of special signif-
icance for the implementation of indirect optical
detection.
Electrochemical detection In electrochemical detec-
tion, sample bands are monitored in terms of an
electrical signal. Depending on the application, the
measured quantity can be conductivity, voltage or
current. Accordingly it is referred to as conduc-
timetry, potentiometry or amperometry, respectively.
Unlike optical detection in which the measurement
output depends strongly upon the available volume,
in electrochemical detection the signal relates only to
the part of solution that is directly contacting the
electrode surfaces. Thus, wherever volume insufR-
ciency precludes the employment of photometric de-
tection, e.g. in case where ultra-narrow bore capillary
is used to facilitate fast separation, electrochemical
detection may still be a viable choice. Moreover, as
long as the analyte is electrochemically active, the
detection can be performed directly without involv-
ing derivatization, as it may be for photometric
detection. Due to these advantages, electrochemical
detection gradually gains its popularity in CE
practice.
There are two major difRculties involved with im-
plementing electrochemical detection for CE. The
Rrst comes from the fact that, it is not easy to make an
electrochemical detector function well in the presence
of a high voltage, because the separation voltage will
produce noise that swamps the minute response of the
analytes. Thus, decoupling of the separation electric
Reld from the detection system has to be done to
facilitate the measurement. The insulation of the sep-
aration voltage can be realized by introducing a frac-
ture or a gap structure in the capillary through which
the electrical grounding is made, such that the seg-
ment behind the fracture or the gap can be used for
the accommodation of the electrochemical electrodes.
Alternatively, to shield off the high voltage, the de-
tecting elements may be mounted immediately out-
side the outlet of the separation capillary. A more
elegant way is to etch the capillary end to generate an
enlarged conical cuvette from where the detecting
electrodes are installed. In this way the interference of
high voltage can be eliminated because the electric
Reld decays very rapidly before the enlarged conical
part. The second difRculty stems from the need to
1183
fabricate microelectrodes and mount them into an
extremely small space, usually deRned by the separ-
ation capillary. While the current state-of-the-art
allows the preparation of microelectrodes with di-
mensions down to several
m, the installation of the
microsensing elements to the detection region is
a task requiring complicated microfabrication tech-
niques and great patience. All three types of electro-
chemical detection have shown their feasibility in CE
with typical detection limits in the range of
10\7}10\9 mol L\1.
Hyphenation with mass spectrometry Following
the great success in interfacing HPLC with mass spec-
trometry, it is a logical move to explore the potential
of CE-MS. For a successful implementation of
CE-MS, keeping a proper electrical contact at the MS
side is essential. This has been achieved through
a variety of means, such as contacting via coaxial
sheath Sow, or through a liquid junction. Sheathless
contact has been realized by attaching the metal-
coated separation capillary tip directly to the ion
source emitter. For sample ionization and transporta-
tion, the electrospray ionization interface (ESI) is pre-
dominantly adopted, mainly because ESI is operated
under almost atmospheric pressure, thus not conSict-
ing with CE separation. Moreover, ESI ionization
works with electrostatically induced nebulization, in
which compounds are ionized with different charge
status depending on their relative molecular masses
and shapes. In principle, this allows determination of
relative molecular mass for a large range of com-
pounds, from small molecules to large biopolymers
(such as proteins and polysaccharides). Thus it is
applicable to all types of CE analytes. As for the mass
analyser, the high efRciency of CE separation de-
mands a mass analyser with a fast scan rate. The
time-of-Sight (TOF) mass spectrometer is probably
the most promising candidate, but the quadrupole
mass spectrometer is currently the workhorse for CE-
MS owing to its commercial availability and relative-
ly low cost. Its insufRcient scan speed may be com-
pensated partially through the manipulation of the
electric Reld, i.e. when a certain analyte is migrating
out of the capillary, the applied voltage can be re-
duced temporarily, such that the analyte of interest
will remain in the ionization chamber for a longer
time. With careful optimization of parameters, CE-
MS has shown detection limits in the attomole range
(10\18 mol mass).
Separation Formats of CE
One of the advantages of CE is that many types of
separation can be performed with the instrument
1184 II / ELECTROPHORESIS / Capillary Electrophoresis

described in Figure 2. This is made possible by simply Capillary Gel Electrophoresis (CGE)
altering the separation medium inside the capillary
For this format, polymeric networks are present
and utilizing an appropriate buffer system. The com-
along the electrophoretic pathway of analytes, which
monly employed separation formats in CE can be
causes charged species to be resolved on the basis of
divided into the following.
their physical sizes rather than charge-to-mass ratios.
Accordingly this separation process is also commonly
known as ‘molecular sieving’. It is particularly suit-
Capillary Zone Electrophoresis (CZE) able for the separation of biomacromolecules consist-
ing of numerous repeat charge units, such as DNA
Among the various CE formats CZE is the simplest
fragments, SDS-denatured proteins and polysacchar-
and most popular. Here a homogenous free solution
ides (Figure 8). CGE may be considered as an adapta-
is employed to maintain a constant electric Reld along
tion of traditional gel electrophoresis into its capillary
the capillary. Ionic species are separated inside this
format. However, traditional gel electrophoresis use
supporting solution according to their different
only cross-linked polyacrylamide (so called ‘chemical
charge-mass ratios, thus forming segregate zones.
gel’), while for CGE both cross-linked polymer gels
The desired separation selectivity can be achieved
and various polymer solutions can be employed to
by simply optimizing the parameters of the carrier
create the sieving pores. The use of polymer solutions
electrolyte particularly pH, as well as ionic
instead of chemical gels facilitates renewing of the
strength and the type and concentration of various
sieving medium by simply Sushing out the original
EOF modiRers. A number of secondary equilibria
solution, and reRlling with another. It also alleviates
are available to strengthen the separation selectivity
some technical problems associated with chemical gel
further.
such as gel shrinkage, bubble formation and matrix
Although CZE is commonly performed with an
collapse, hence improving the separation repro-
aqueous buffer, it can also be implemented in a non-
ducibility.
aqueous medium by using organic solvents and
compatible conductive salts. The effect of replacing
water with an organic solvent in CE can be under-
stood from the fact that organic solvents possess
signiRcantly different viscosities and dielectric con-
stants compared with water. Thus, changes in the
magnitude of EOF and migration behaviours of
charged analytes are expected in non-aqueous CE,
which can be seen from eqns [2] and [4]. Further-
more, organic solvents differ in their capacity to
stabilize equilibria. Thus, in organic media, the
charge status of organic analytes can be dramatically
different from that in an aqueous medium, hence
leading to quite different separation selectivity. Addi-
tionally, it is evidenced that organic media are ca-
pable of promoting certain mechanisms such as inclu-
sion interaction and ion-pair formation, which en-
hance possibilities of achieving the desired separ-
ation. Moreover, with a water dominated buffer, it is
very difRcult to conduct CE separations of hydropho-
bic analytes. Under such circumstances, switching to
a non-aqueous buffer system would provide an efR-
cient solution. Finally, it is noted that, when an or-
ganic medium is utilized for CZE separations, the
electrophoretic current is reduced considerably. As
a result, even though capillaries of relatively wide
inner diameter are employed, Joule heating is at Figure 8 CGE separation of poly(uridine 5 -phosphate).
(Adapted with permission from Yin HF, Lux JA and Schomburg G
a manageable level, thus enabling the enhancement of
(1990) Production of polyacrylamide gel filled capillaries for
detection sensitivity through the usage of wide bore capillary gel electrophoresis (CGE): influence of capillary surface
capillaries. In short, non-aqueous CZE is an attract- pretreatment on performace and stability. Journal of High Resolu-
ive means for extending the applicability of CE. tion Chromatography 15: 624}627.)
 II / ELECTROPHORESIS / Capillary Electrophoresis 1185

Capillary Isotachophoresis (CITP) ampholytes with isoelectric points in a certain range

and in close proximity to each other. Such a mixture
In CITP, a leading electrolyte of higher mobility than
is commonly formed by a series of synthetic poly-
any of the sample components is Rlled into the separ-
amino polycarboxylic acids. The capillary is Rlled
ation capillary and outlet reservoir, while a termina-
with the carrier ampholyte solution and a small
ting electrolyte of lower mobility than any of the
amount of sample, and dipped into the buffer reser-
sample components is placed in the inlet reservoir.
voirs that contain acid and base, respectively. Under
The sample is applied from the capillary inlet. This
the electric Reld, different ampholytes will migrate
arrangement ensures that there will be no mixing of
along the capillary until they reach positions corre-
sample with the two electrolytes during the whole
sponding to their pI values, where they stand still.
electrophoretic process. As a consequence, upon
Collectively a stable pH gradient is formed across the
application of an electric Reld, the main change hap-
capillary. The sample components, too, will migrate
pens inside the sample band itself, i.e. the sample
until they Rnd positions equivalent to their pI values.
components tend to ‘queue up’ according to their
When such a steady state is obtained, an immobiliz-
effective mobility orders. These components are relo-
ation step (by utilizing electrophoretic movement or
cated with corresponding changes in their concentra-
pressure) is performed, and the separated analyte
tions and band lengths. Ultimately a series of con-
bands will be forced out and passed through the
secutive ‘pure zones’ containing only the individual
detector. If the carrier ampholyte mixture is well
substances are formed. This is the steady state as there
prepared and a sufRcient time is allowed for the
will be no more changes for the sample except that all
electric focusing process, a very high column efRcien-
the zones will continue to migrate out of the capillary
cy can be realized.
with an identical velocity (hence the term ‘iso-
tachophoresis’). Unlike the other types of CE separ-
ation, in CITP the concentration of a speciRc zone is Micellar Electrokinetic Chromatography (MEKC)
predetermined by the concentration of the leading
Neutral species can never be resolved with traditional
electrolyte and the relative mobility of the ion of
electrophoretic techniques since they have no elec-
interest with respect to the leading electrolyte, while
trophoretic mobility. After the birth of CE, this prob-
the sample amount is reSected by the zone length.
lem was solved by Terabe et al. through the addition
Due to the unique stepwise increase in electric Reld
of a charged surfactant to the running buffer. Typi-
starting from the leading electrolyte side, which
cally a negatively charged surfactant is added at
prevents a component from drifting off its own
a concentration above its critical micelle concentra-
band, sharp boundaries can be maintained more
tion (CMC). Under such circumstances, micelles are
readily. Therefore CITP can be implemented using
formed, which allow neutral compounds to be re-
tubes of relatively large i.d. (e.g. 0.2}0.8 mm).
tained based on their hydrophobicities. Under an
Nevertheless, CITP as an analytical tool has largely
electrical Reld, the negatively charged micelles move
lost its popularity since the advent of CZE, probably
toward the anode side, while a strong EOF moves
due to the fact that a thorough knowledge of the
toward the cathode side (Figure 9). As EOF usually
sample is required before the separation. Further-
exceeds the electrophoretic mobility of micelles, the
more, the isotachopherogram is usually step-shaped,
micelles will eventually be swept out in the same
and the steps are not time related, thus rendering
automatic identiRcation and quantiRcation difRcult.
CITP is often applied as a sample pre-concentration
step before CZE separations to enhance detection
sensitivity.

Capillary Isoelectric Focusing (CIEF)

This is another example of an adaptation of a conven-
tional electrophoresis principle to a capillary format.
CIEF exploits differences in isoelectric points (pIs),
a unique characteristic of amphoteric compounds un-
der which its acidic and basic groups dissociate to an
equal extent, so that the whole molecule exhibits no
net charge. It is only suitable for the separation of
compounds like amino acids, peptides and proteins.
The operation of CIEF relies on a mixture of carrier Figure 9 Separation mechanism of MEKC.
1186 II / ELECTROPHORESIS / Capillary Electrophoresis
Figure 10 Separation of small neutral aromatic compounds by
bile salt-based MEKC. (Unpublished results.)
direction. Thus a migration time window is deRned
within which the neutral compounds can be separ-
ated according to their afRnities for the micelle micro-
environment. If a positively charged surfactant is
used, then similar resolution of neutral analytes can
again be obtained, the only difference being that the
EOF will be reversed owing to the adsorption of
cationic surfactant molecules onto the capillary. The
polarity of the separation voltage, therefore, needs to
be reversed. The micelle-forming reagents for MEKC
are not limited to synthetic detergents } any chemical
with similar surface activity can be employed. These
include biogenic surfactants such as bile salts (Fig-
ure 10), and as some synthetic high mass-to-charge
polymers.
Af\nity Capillary Electrophoresis (ACE)
Compared with other types of CE techniques, ACE
represents a relatively recent development. Introduc-
ed in the early 1990s, ACE is performed on the basis
of speciRc or non-speciRc afRnity interactions be-
tween receptor and ligand molecules, typically bio-
molecules. Theoretically, if either receptor or ligand
is a charge species, their binding complex would
show a different electrophoretic mobility from that of
the parent molecule due to the changes in charge-
mass ratio. Thus, measuring the changes in elec-
trophoretic mobility of the receptor via CZE mode
provides an excellent way to study aspects of the
receptor}ligand binding.
Depending on the binding strength of the recep-
tor}ligand pair and the operational procedure, ACE
can be divided into three different categories, namely,
non-equilibrium mode, dynamic equilibrium mode
and immobilizing mode. Equilibrium ACE is well
suited to studying strong binding systems, in which
the sample is injected as an equilibrated mixture
of receptor and ligand, whereas the electrophoresis
medium contains only the supporting buffer. In
such cases, CE serves merely as a tool to separate
and determine the free and bound receptor molecules.
Dynamic equilibrium ACE is typically employed
for weak to moderate binding system, in which the
receptor is injected as the sample, while ligand of
varied concentration is incorporated in the running
buffer. In this case, free and bound molecules are not
separated due to the fast on-and-off kinetics, but
rather, they are detected as single peaks. The im-
mobilizing mode is self-explanatory, for which ligand
is attached to the capillary wall, or more commonly,
onto a supporting material via an appropriate bond-
ing chemistry, while sample (receptor) is driven over
the active surface by application of an electric Reld.
Again, the migration behaviour of the receptor is
a good indication of receptor}ligand interaction
strength.
ACE has become a powerful tool in diverse Relds,
including the measurement of binding constants,
the study of binding kinetics, the determination
of interaction stoichiometry, the characterization of
biomolecules and the separation of enantiomers,
etc. ACE has been applied to the investigation of
a number of biologically important systems, such as
the interaction of polypeptides with immunog-
lobulins, polypeptides, carbohydrates, nucleic acids,
drugs, etc.
ACE bears some resemblance to classical gel afRn-
ity electrophoresis and conventional afRnity
chromatography, in that they all utilize speciRc inter-
actions to effect separations. However, ACE in-
herently has an unsurpassed advantages, that is, the
buffer conditions (ionic strength and pH) can be
Rnely tuned to give a perfect mimic of the real physio-
logical environment, so that more precise character-
isation of many binding processes is possible, pro-
vided the wall adsorption problem can be addressed
properly.
Applications
The introduction of CE has resulted in a dramatic
expansion of the applicability of electrophoresis as
a separation tool. While conventional slab gel elec-
trophoresis is mainly limited to the separation of
biosubstances such as proteins and DNA fragments,
CE has been utilized to resolve a much broader
spectrum of substances, ranging from simple ions
 II / ELECTROPHORESIS / Capillary Electrophoresis
through small molecules to macromolecules. High
mass biomolecules such as proteins, DNA and poly-
saccharides can be separated by CZE as well as CGE.
Smaller charged molecules such as amino acids,
peptides, organic acids and amines can be resolved
via CZE. Simple ions, both cations and anions,
can be easily separated via CZE with conductivity
detection. CZE separations of ions with indirect UV
detection, dubbed as capillary ion analysis (CIA) has
been accepted rapidly as an alternative to ion
chromatography owing to its simplicity, speed and
low cost. The separation of various neutral com-
pounds has been made possible through MEKC or
CEC. Chiral separations, an area hardly touched by
conventional electrophoresis, is increasingly carried
out by CE. Many chiral selectors (cyclodextrin deriv-
atives, amino acids, proteins, optical micelles, etc.)
may be added to the buffer solution to induce enan-
tiodis crimination based on one of several mecha-
nisms (host}guest complexation, ligand exchange and
solubilization by micelles, etc.). The unequal stability
of dynamically formed diastereoisomers cause the
optical isomers to be moved out of the capillary
with different velocities, and enantioseparation
is achieved. Because of its far-reaching capabilities,
CE is becoming ubiquitous in almost all analytical
Relds.
Future Prospects
The introduction of CE in the early 1980s has had a
huge impact on numerous scientiRc Relds. It can be
anticipated that in the future CE will continue to
evolve into a fully Sedged analytical technique that
will beneRt many research disciplines. Based on the
characteristics of CE and its current status, several
directions deserving special attention can be en-
visaged.
CE is hailed for its high column efRciency and
outstanding mass sensitivity. The concentration sensi-
tivity of CE remains relatively low compared to
HPLC, thus limiting its applications in areas such as
trace impurity determination and environmental
analysis. Therefore, improving the detection sensitiv-
ity will continue to be a topic for development. Cre-
ative detection conRgurations, interplay from micro-
mechanic and microelectronics, together with ad-
vancement of light sources, may remarkably enhance
the sensitivity of optical detection. To facilitate the
use of electrochemical detection, efforts must be
made to provide better microelectrodes with reason-
able ruggedness. With the enhanced performance
and reduced cost, a more common use of sophisti-
cated hyphenation techniques such as CE-MS is
expected.
1187
On the other hand, the ability of CE to handle
extremely small sample quantities is attractive for the
direct probing of micro-entities. The recently
emerged single cell analysis is a good example of this.
With further improvements in sampling techniques
and detection schemes, it is believed that in the near
future, CE-based methodologies will allow us to in-
vestigate the chemistries of a wider spectrum of cells,
thus enriching our knowledge about many biological
processes essential to life.
Finally, due to its instrumental simplicity, CE is
amenable to further miniaturization. CE on a glass
chip has been successfully demonstrated by borrow-
ing microfabrication concepts from the microelec-
tronics industry. These devices feature the integration
of injection, separation and detection, as well as
sophisticated designs with intricate patterns and
multi-channel arrays, thus producing an unprece-
dented analytical platform which is fully manipulated
by applying voltages. As the plate format possesses an
excellent ability for heat dissipation, electric Relds up
to 2000 V cm\ may be applied across the separation
channels, hence shortening the analytical time scale
to minutes or even seconds. With continuous matur-
ity of relevant technologies, such an ultra-high speed
separation method may provide a solution to some
formidable tasks including DNA sequencing.
See also: II/Electrophoresis: Electrochromatography
Thin Layer.
Further Reading
Baker DR (1995) Capillary Electrophoresis. New York:
Wiley.
Engelhardt H, Beck W and Schmitt T (1996) Capillary
Electrophoresis: Methods and Potentials. Braun-
schweig/Wieshaden: Vieweg.
Guzman NA (ed.) (1993) Capillary Electrophoresis Tech-
nology. New York: Marcel Dekker.
Jorgenson JW and Phillips M (eds) (1987) New Directions
in Electrophoretic Method. Washington, DC: American
Chemical Society.
Kevin DA (ed.) (1996) Capillary Electrophoresis Guide-
book: Principles, Operation and Applications. Totowa,
NJ: Humana Press.
Kuhn R and Hoffstetter-Kuhn S (1993) Capillary Elec-
trophoresis: Principles and Practice. Berlin/New York:
Springer-Verlag.
Landers JP (ed.) (1997) Handbook of Capillary Elec-
trophoresis, 2nd edn. Boca Raton, FL: CRC Press.
Li SFY (1992) Capillary Electrophoresis: Principles, Prac-
tice and Applications. Amsterdam: Elsevier.
Vindevogel J and Sandra P (1992) Introduction to Micellar
Electrokinetic Chromatography. Heidelberg: Huthig.
Weinberger R (1993) Practical Capillary Electrophoresis.
Boston: Academic Press.
1188 II / ELECTROPHORESIS / Capillary Electrophoresis}Mass Spectrometry
Capillary Electrophoresis Detection
See II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis
Capillary Electrophoresis^Mass Spectrometry
M. Hamdan, GlaxoWellcome Medicines Research
Centre, Verona, Italy
P. G. Righetti, University of Verona, Verona, Italy
Copyright ^ 2000 Academic Press
Introduction
Capillary zone electrophoresis (CZE) is widely recog-
nized as a powerful analytical technique in its own
right, known for its high separation efRciency, short
analysis times and low-volume sample requirements.
These characteristics made CZE a popular method
for the analysis of peptide mixtures, protein digests,
drug substances and biotechnological products. The
coupling of CZE with electrospray ionization mass
spectrometry (ESI}MS), Rrst reported by Olivares et
al. in 1887, has added further capabilities, in particu-
lar for obtaining molecular mass information and
structural details when tandem mass spectrometry
(MS}MS) is used. However, it can be said that the
major advantage of such coupling is that the migra-
tion time is not the only parameter used for identify-
ing the eluted components. These times are subjected
to variations between runs, yet such variations be-
come irrelevant when, in the same run, highly diag-
nostic mass spectra are obtained.
Depending on the ionization method, CZE can be
coupled to a mass spectrometer either directly (on-
line) or indirectly (ofSine). In the latter mode of
operation, 252Cf plasma desorption and matrix assist-
ed laser desorption can be used. The online coupling
of CZE is more common and usually performed
by electrospray ionization (ESI) or fast atom (ion)
bombardment (FAB). Although online CZE/MS is
the more common form of application, ofSine
analysis has the advantage of allowing separation in
non-volatile buffers, which are highly undesired
in ESI.
It goes without saying that every analytical tech-
nique has its limitations and CZE/MS is no excep-
tion. One of the main limitations of this experimental
arrangement is its relatively poor sample concentra-
tion/ion sensitivity. Approaches to reduce such lim-
itations included online preconcentration, sample
stacking, and the increasing use of time-of-Sight
(TOF) analysers which use ESI and TOF analysers
with and without a quadrupole in between. The
innovative feature of this class of instruments is their
fast scanning, which allows the acquisition of a
number of full spectra per second. Additionally,
as all ions in each spectrum are sampled at the same
moment in time, spectra are free of mass discrimina-
tion or peak skew typical of slow scanning systems
that must scan over a narrow chromatographic/elec-
trophoretic peaks.
Capillary electrochromatography (CEC) is another
technique which is currently undergoing a rapid
phase of advancement and development. This tech-
nique was revived by Jorgenson and Lukacs in 1981;
these authors used 0.005 mol L\1 phosphate buffer,
170
m packed column and 30 kV separation voltage
to separate 9-methylanthracene and perylene. This
technique has recently become more diffuse because
of a number of advances in both CE instruments and
detection techniques including electrospray mass
spectrometry. However, on-column UV detection
and in-column laser-induced Suoroscence detection
remain the most commonly used methods. Despite its
high sensitivity, the latter method is subjected to
interferences by buffer Suoroscence. In MS detection,
the column is commonly packed right up to the point
where the sample is injected into the mass spectrom-
eter. The combination of CEC with mass spectro-
metry provides reliable molecular weights and in
many cases structural information, which makes it
highly attractive for a wide range of applications. For
more details on this topic, the reader is referred to
recent extensive reviews, covering the methodology
of CEC and its coupling to MS, by Colòn et al. (1997)
and Rentel et al. (1999). Interestingly, packed-CEC
offers the possibility of higher sample capacity and
the utilization of simpler mobile phases, which are
more compatible with MS.
Experimental Aspects
One of the main advantages of CZE is that it requires
simple instrumentation, which generally consists of
 II / ELECTROPHORESIS / Capillary Electrophoresis}Mass Spectrometry
a high-voltage power supply, two buffer reservoirs,
a capillary and a detector. Coupling of such instru-
mentation to a mass spectrometer requires the re-
placement of one of the buffer reservoirs with a suit-
able interface and some simple electronic circuit to
accommodate the presence of an additional 3}5 kV
required for the operation of the ion source.
Interfacing CZE to a mass spectrometer has been
effected in a number of ways, yet all of them can be
traced to two general conRgurations: liquid junction
interface and coaxial sheath Sow. The liquid junction
interface, Rrst described by Minard et al. in 1988, was
used to couple a CZE to a continuous-Sow FAB
source. In such a conRguration the CE capillary ter-
minates in a 20
m block that contains either the
matrix solution for FAB ionization or the sheath
solution for ES ionization. The inlet of the transfer
capillary is aligned in close proximity to the cathode
end of the CE capillary, an alignment considered
critical, since the gap must be sufRciently wide to
allow enough matrix or sheath Suid to maintain
a stable ion beam. On the other hand, the same gap
should be small enough to prevent analyte diffusion,
upon exiting the CE capillary, which would otherwise
result in peak broadening. The main disadvantage of
this conRguration is associated with the dead volume
caused by a long transfer line which together with the
mixing effects in the interface and the presence of the
FAB matrix renders the separation efRciency of this
interface poorer than that of the coaxial sheath Sow.
Liquid-junction interface has been redesigned to
allow easier mounting and alignment of the CE and
continuous-Sow FAB capillaries. The same interface
was further modiRed by Caprioli et al. in 1989, who
Figure 1 The main components of a sheath}flow probe which can be used to couple CZE to a Q-TOF or to a single/triple quadrupole
instruments. (Courtesy of Micromass, Manchester, UK.)
1189
used it for the analysis of synthetic mixtures of pep-
tides and protein digests.
The Rrst successful coupling of CE with MS was
effected by Olivares et al. in 1887, wherein the cath-
ode end of the CE capillary terminated within a stain-
less-steel capillary which completed the electrical cir-
cuit and established contact with the CE side. An
improved version of this interface was developed by
the same research group, where the metal contact at
the CE terminus was replaced by a thin sheath of
liquid Sow. In comparison with the liquid-junction
interface, the coaxial sheath}Sow interface provided
both better sensitivity and better resolution, yet this
did not mean that the interface was trouble free.
A number of difRculties can derive from the use of
sheath liquid: ionic and neutral species within this
liquid compete for protonation in the ESI process,
thus lowering the overall sensitivity. The composition
of the sheath liquid commonly includes a volatile
organic acid (1% formic or acetic acid) in a mixed
water}organic solvent, a composition which is differ-
ent from that of the electrophoretic buffer. During the
CE, sample ions and other species present in the
buffer exit the capillary at the MS end, and, simulta-
neously, counterions from the liquid sheath enter the
column and migrate toward the injection end. These
moving ion boundaries can inSuence migration order,
times and resolution. The same phenomenon has the
advantage of allowing some analysis in the presence
of difRcult-to-spray electrolytes (such as phosphate-
or borate-containing buffers).
Figure 1 gives the up-to-date version of a commer-
cial coaxial sheath}Sow interface constructed and
marketed by Micromass (Manchester, UK).
1190 II / ELECTROPHORESIS / Capillary Electrophoresis}Mass Spectrometry
Representative Examples
CZE/MS in Peptide Analysis
One of the advantages of coupling CZE to ES mass
spectrometry is the formation of multiply charged
ions which allows simple analysers such as quadru-
poles to measure the mass to charge ratios (m/z) of
relatively large biomolecules. Indeed, the majority of
CZE/MS applications have been in the Reld of biolo-
gical and biochemical research. A number of reports
have appeared on the characterization of synthetic
mixtures of peptides and proteins, where the unusu-
ally high resolution of CZE permits the separation of
sequences which may differ by a single amino acid
residue. This advantage has been exploited in the use
of CZE/ES}MS to examine a number of reaction
mixtures of peptides obtained by solid-phase syn-
thesis; two representative examples are considered
here. A reaction mixture obtained by solid-phase syn-
thesis of neuropeptide Y (NPY) analogue, [Leu31,
Pro34]-NPY. This peptide has 36 amino acid residues,
a relative molecular mass of 4222 and is known to
play a major role in the central and peripheral ner-
vous system. Online CZE/ES}MS analysis of this
mixture resulted in a fairly complex UV and total ion
current (TIC) electropherograms which, in addition
to the target peptide, contained a number of side
products. The use of mass spectrometry allowed re-
liable identiRcation of all the components of the mix-
ture, although some of the side products differed by
a single amino acid residue in sequences containing
over 30 residues. These measurements allowed unam-
biguous identiRcation of the various components of
the mixture which are summarized in Table 1.
A second and less complex reaction mixture
associated with the same synthesis was also examined
by the same technique which yielded the total ion
current electropherogram in Figure 2(A). Deconvolu-
tion of the mass spectra in Figure 2 (panels 1}4)
yielded the relative molecular masses 4222 (1), 2441
Table 1 Proposed peptide sequences, corresponding to relative molecular masses (M r) obtained from the ES mass spectra
associated with various peaks in a TIC electropherogram obtained by online CZE/MS of NPY analogue reaction mixture
Peptide chain M r calculated M r measured
Ac [Leu31,Pro34]-NPY 4222 4222$1
H-12}36-NH2 3022 3022$1
Ac[Leu31,Pro34]-NPY 4222 4222$1
Ac(5}36) 3789 3789$1
Ac(4}36) 3917 3917$1
Ac(3}36) 4004 4004$1
Ac(20}36) 2440 2441$1
Ac(7}36) 3577 3577$1
Ac(5}36) 3789 3789$1
(2), 3577 (3) and 3789 (4). The Rrst molecular mass
coincides with the desired peptide, while the other
three masses are associated with a number of incom-
plete sequences which are summarized in Table 2
which clearly shows an excellent agreement between
the calculated and measured masses, fully based on
CZE/MS measurements. Given the complex proced-
ure associated with solid-phase synthesis of peptides,
it is evident that the use of online CZE/MS is an
indispensable analytical tool for the initial character-
ization of the product and for providing a reasonable
indication on the yield of synthesis.
CZE/MS can also give reliable information on un-
expected processes in the course of solid-phase syn-
thesis of peptides. To underline this statement, two
cases are considered. A series of newly synthesized
peptides investigated, corresponding to portions of
the extracellular domain of human granulocyte}
macrophage colony-stimulating factor receptor
-
subunit. A solution containing 3 mg mL\1 of the
peptide [PRAKHSVKIRAADVRILN], Mr"2084,
was examined by full-scan CZE/MS, which yielded
UV and TIC electropherograms, each of which con-
tained three peaks of almost equal relative heights.
The associated ES mass spectra revealed the presence
of the desired peptide together with its acetylated
version in one of the three TIC peaks. It is fair to say
that without the use of online MS detection, the
identiRcation of the latter component would have
been very unlikely. A second case refers to online
CZE/MS of an NPY analogue which exhibited a TIC
electropherogram containing seven peaks, two of
which yielded Mr"4222 which implied the presence
of two different conRgurations of the same sequence
within the same crude of synthesis. This deduction
was found in accord with existing literature describ-
ing undesirable side products observed in solid-phase
synthesis of peptides and small proteins. One such
side product can be invoked by the formation of
succinimide of the Asp residue (-aspartyl peptide)
which, in the present case, has the same Mr as the
Proposed sequence
Ac-YPSKPDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
H-APAEDLARYYSALRHYINLLTRPRY-NH2
Ac-YPSKPDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
Ac-PDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
Ac-KPDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
Ac-SKPDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
Ac-YYSALRHYINLLTRPRY-NH2
Ac-NPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
Ac-PDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
 II / ELECTROPHORESIS / Capillary Electrophoresis}Mass Spectrometry 1191

Figure 2 Upper frame: total ion current (TIC) electropherogram of a crude of synthesis associated with [Leu31,Pro34]-NPY. Lower
frames: associated positive ES mass spectra of the TIC peaks: (1) tM"59.61 min; (2) tM"64.41 min; (3) tM"68.30 min; and
(4) tM"70.02 min.

NPY analogue. The mechanism responsible for such
reaction is depicted in Figure 3.
Forensic Application of CZE/MS
Most drugs of forensic interest are commonly
analysed by GC/MS or extraction followed by IR
analysis. Separation of benzodiazepines and ergot
alkaloids by GC is difRcult and differences in the IR
spectra are small. On the other hand, the analysis of
these compounds by CE/MS is characterized by short
analysis times and speciRc MS information. Forensic
drug chemists, investigating drugs of abuse, also Rnd
the analysis of drugs such as LSD challenging because
of microgram quantity dosage and the fact that
GC/MS analysis is rather demanding owing to col-
umn adsorption and thermal lability. This is also true
for psilocybin, the psychoactive agent in certain
1192 II / ELECTROPHORESIS / Capillary Electrophoresis}Mass Spectrometry

Table 2 Proposed peptide sequences, corresponding to relative molecular masses (M r) obtained from the ES mass spectra
associated with the components arising from the synthesis of [Leu31,Pro34]-NPY

Peptide chain M r calculated M r measured Proposed sequence

Ac[Leu31,Pro34]-NPY 4222 4222$1 Ac-YPSKPDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2

Ac(20}36) 2440 2441$1 Ac-YYSALRHYINLLTRPRY-NH2
Ac(7}36) 3577 3577$1 Ac-NPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2
Ac(5}36) 3789 3789$1 Ac-PDNPGEDAPAEDLARYYSALRHYINLLTRPRY-NH2

Figure 3 Representation of succinimide formation using partial sequence containing Asp6 and two adjacent amino acids, Asn7 and
Pro5 taken from the sequence of [Leu31,Pro34]-NPY. (Reproduced with permission of John Wiley & Sons Ltd.)
 II / ELECTROPHORESIS / Capillary Electrophoresis}Mass Spectrometry 1193

mushrooms, owing to the presence of the highly and barbiturates in serum. Other examples include
labile phosphate moiety. The absence of complex the chiral separation of racemethorphan and ra-
preparative extractions for CE analysis of both drugs cemorphan in urine using a cyclodextrin/SDS/pro-
allows their separation without causing undesired panol buffer and the micellar CE separation of nit-
decomposition. razepam and its metabolites in urine. An extensive
The fact that many forensic drugs have closely review is available on the use of CZE coupled to
related isomeric structures renders CE/MS a powerful tandem mass spectrometry to identify a variety of
tool for their separation and identiRcation. For drugs and metabolites. All these reviews are listed in
instance, phenethylamines are historically a difRcult the Further Reading.
group to analyse due to the large number of isomers
including both D- and L-optical isomers. Twelve op- Forensic Analysis of Inorganic Explosives
tical isomers of ephedrine, pseudoephedrine, nore-
Capillary electrophoresis is also widely used in the
phedrine, norpseudoephedrine, amphetamine and
analysis of inorganic ions in criminal cases, material
methamphetamine have been separated in under
such as black powder, Sash powder, ammonium
30 min using CE. It is worth noting that this class of
nitrate and home-made explosive mixtures. The inor-
compounds tends to give strong [M#H] signals in
ganic anions resulting from an explosive reaction of
ES ionization which makes them ideal for CZE/MS
such materials are among the most important evid-
analysis.
ence used to determine the nature of an inorganic
A separation of 18 common drugs of abuse was
explosive. For many years, the most powerful
accomplished in less than 40 min using CE buffered
tool in these investigations was ion chromatography
with phosphate/borate, with sodium dodecyl sulfate
(IC). The introduction of CE for anion analysis
(SDS) as a micellar phase and acetonitrile as an or-
provided a simpler, faster and slightly more sensitive
ganic modiRer. The main advantage of this method
technique for performing such analysis. It is interest-
over other screening techniques is that acidic, basic
ing to note that CE separations are based on differ-
and neutral drugs can be screened through a single
ences in charge-to-mass ratios of the solvated ions,
analytical method, where extraction procedures are
while IC separations are the result of complex
unnecessary.
interactions between the ions and the stationary
Clandestine manufacturers of illegal drugs pose
phase. As a result, the migration order is quite differ-
additional challenges to the forensic scientist, since
ent in the two methods and a nearly orthogonal
samples submitted for analysis tend to contain
relationship exists between the relative retention
complex mixtures of chemicals, including thermally
times.
reactive and labile components. CZE with and with-
The reliable analysis of certain alkyl-substituted
out MS detection can be a powerful tool for such type
organophosphorus acids, which are the primary hy-
of analysis. For example, the frequently encountered
drolysis products of neurotoxic agents, has become
methamphetamine can be analysed by CE allowing
very important in the last few years owing to the
the identiRcation of its isomeric composition which
likelihood of an international agreement that will
can be eventually used to construct evidence concern-
forbid the development, production and stockpiling
ing its synthetic pathway and the source of the
of chemical warfare agents and weapons. CZE
sample.
coupled to ion spray mass spectrometry was applied
An additional problem with samples from clandes-
in the negative ion mode to investigate Rve organo-
tine laboratories is the presence of unreacted precur-
phosphonic acids, which are the primary hydrolysis
sors and adulterants that may interfere with the anal-
products of neurotoxic agents. The MS spectra exhib-
ysis of the target compound(s). For instance, under
ited a very abundant [M!H]\ signal with minimal
typical GC/MS conditions, certain mixtures of chem-
fragmentation. The authors reported sensitivity in the
icals associated with methamphetamine will de-
range 10}30 pg using the single-ion recording (SIR)
rivatize the illegal drug to a compound that is not
mode.
controlled, resulting in an inconclusive analysis.
Using a number of CE methods, such interference can
be easily avoided.
CZE/MS is also highly suitable for the analysis of
Conclusions
biological samples which may contain traces of illegal The examples cited in this work have to be considered
drugs or poisons in a complex matrix. Extensive re- as only a part of the capabilities of CE with and
views cite over sixty references on the use of CE for without MS coupling. The literature cites varied areas
the analysis of drugs in biological samples including of application of such powerful methodology. As
drugs of abuse in urine, cocaine and morphine in hair new applications of CE continue to appear, the
1194 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance
advantages and importance of CE in conjunction
with mass spectrometry have also become appreci-
ated. Analytical chemists are faced with the challenge
of increasing sample complexity and decreasing
sample quantities. Because of the complexity ob-
served with most biological mixtures, there continues
to be a need for the development of a highly efRcient
separation technique in conjuction with a sensitive
and speciRc detector. The low quantities of analytes
often available require nanoseparation techniques.
The mass spectrometer is a selective and broadly
applicable detector for analytical separations. It can
provide information regarding the structure of un-
known components present in a sample mixture with
high speciRcity and sensitivity. The coupling of CE
with MS combines the extremely high-resolving
power and structural information in one system. Like
any other separation technique such as GC}MS and
LC}MS, the principal advantage of CE}MS is that
analytes are identiRed by both their differential separ-
ation and their molecular masses and/or fragmenta-
tion patterns. An analytical separation that precedes
MS analysis is often necessary to assure correct inter-
pretation of the mass spectral data.
Fast, high-efRciency separation techniques are be-
coming ever more important in the race to discover
new drugs. The potential complexity of libraries pro-
duced by automated parallel synthesis, combinatorial
and genetically manipulated natural product chemis-
tries are driving many developments in separation
sciences.
CE, CEC and nano-LC are all potential candidates
for such analyses and each has a requirement for
a fast, sensitive detection system.
Further Reading
Aumatell A and Wells RJ (1993) Journal of Chromato-
graphic Science 31: 502}508.
Capillary Electrophoresis+Nuclear Magnetic Resonance
K. Pusecker and J. Schewitz, University of Tu( bingen,
Tu( bingen, Germany
Copyright ^ 2000 Academic Press
Miniaturization is an important current trend in sep-
aration science and the development of capillary elec-
Cai J and Henion J (1995) Journal of Chromatography
A 703; 667}692.
Caprioli RM (1990) Continuous-Flow Fast-Atom Bom-
bardment. New York: John Wiley.
Caprioli RM, Moore WT, Martin M, DaGue BB, Wilson
K and Moring S (1989) Journal of Chromatography
480: 247}257.
Casazza A, Curcuruto O, Hamdan M, Bisello A and
Peggion E (1995) Journal of Chromatography A 715:
227}240.
Colòn LA, Reynolds KJ, Alicea-Maldonado R and Fermier
AM (1997) Electrophoresis 18: 2162}2174.
Foret F, Thompson TJ, Vouros P, Karger BL, Gebauer P
and Bocek P (1994) Analytical Chemistry 66:
4450}4458.
Karas M, Bahr U and Giessmann U (1991) Mass Spectro-
metry Review 10: 335}358.
Kostiainen R, Bruins AP and Hakkinen VMA (1993) Jour-
nal of Chromatography 634: 113}118.
Lurie IS (1992) Journal of Chromatography 605: 269}275.
McCord BR, Hargadon KA, Hall KE and Burmeister SG
(1994) Analytica Chimica Acta 288: 43}56.
Northrop DM, McCord BR and Butler JM (1994) Journal
of Capillary Electrophoresis 1: 58}168.
Olivares JA, Nguyen NT, Yonker CR and Smith RD (1987)
Analytical Chemistry 59: 1230}1232.
Rentel C, Gfroerer P and Bayer E (1999) Electrophoresis
20: 2329}2336.
Rovatti L, Curcuruto O, Hamdan M, Cassano E, Galoppini
C and Rovero P (1996) Rapid Communications in Mass
Spectrometry 10: 1504}1508.
Sundqvist B and MacFarlane RD (1985) Mass Spectro-
metry Review 4: 421}460.
Tagliaro F, Aiello C, Dorizzi R, Ghielmi S and
Marigo M (1993) Journal of Chromatography 638:
303}309.
Thormann W, Maier P, Marcolli C and Binder F (1991)
Journal of Chromatography 545: 445}460.
Thormann W, Molteni S, Caslavska J and Schutz A (1994)
Electrophoresis 15: 5}12.
Wernly P and Thormann W (1991) Analytical Chemistry
63: 2878}2882.
trophoresis (CE), capillary HPLC (cHPLC) and capil-
lary electrochromatography (CEC) are milestones in
this respect. The electrophoretic techniques especially
can achieve rapid and efRcient separations using only
very small volumes and they have become research
tools with widespread applications. The advantages
of this miniaturization are obvious: less sample is
 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance
required, less solvent is consumed, and the separation
times are shorter.
The second trend in separation science is toward
information-rich detection modes. Although UV-VIS
Suorescence, and electrochemical detectors provide
sensitive and simple detection, the information is gen-
erally not sufRcient for unequivocal characterization
or structural elucidation of compounds. For this pur-
pose the coupling of capillary separation methods
with electrospray mass spectrometry (ESI}MS) has
proven to be highly successful. ESI}MS is an extreme-
ly sensitive detector and in many cases information
about mass and fragmentation gives detailed struc-
tural information.
The enormous sensitivity gain in nuclear magnetic
resonance (NMR) spectroscopy in recent years also
enables the wider application of directly-coupled
HPLC}NMR. NMR spectroscopy is considered one
of the most powerful methods for determining chem-
ical, dynamic, and spatial structural properties of
organic compounds. Its capability to distinguish be-
tween most structural, conformational, and, in
special cases, optical isomers is highly complement-
ary to the information gained from MS experiments.
NMR spectroscopy is nondestructive and spectra are
recorded in solution. A combination with additional
detectors is possible, e.g. HPLC}UV}NMR, HPLC}
MS}NMR.
The miniaturization of liquid chromatography
(LC)}NMR has additional advantages. The small
volumes of eluent consumed in capillary separation
techniques make the use of fully deuterated solvents
economically feasible. Therefore, problems asso-
ciated with protonated solvents are prevented. NMR
solvent suppression techniques which can lead to dis-
tortion of parts of the spectra are no longer necessary.
Thus, the entire range of the 1H-NMR scale can be
used in one- and two-dimensional experiments. The
supplementary beneRt of conserving material is par-
ticularly interesting for valuable samples, such as
natural products or labelled proteins.
However, the NMR spectrometer is one of the least
sensitive of all possible LC detectors. The miniaturiz-
ation of the LC}NMR coupling for the application of
capillary separation techniques requires a reduction
of the detection volume by a factor of approximately
1000 in comparison to conventional LC}NMR sys-
tems. On these conditions, the techniques seemed to
be incompatible. Despite these problems, because of
the great potential of the technique, efforts have been
made in the past few years to enable coupling of
capillary techniques to NMR spectroscopy. To date,
two different experimental approaches have been
developed and evaluated. They differ mainly in the
type of radiofrequency (rf) coil that is used for the
1195
NMR detection. This variation leads to different
strengths and weaknesses of the two systems, which
allow or hinder their use for particular applications.
Set-up for CE+NMR Coupling
The design of an interface for the coupling of capil-
lary separation techniques with NMR spectroscopy is
simple in principle. Since phase transfer is not neces-
sary, detection can be carried out in the fused-silica
capillary, which is used for CE separation. The detec-
tion takes place on-column similar to a common UV
detection. The capillary is built into an NMR probe
head. Even the inlet and outlet vials of the CE can be
incorporated inside this probe, but for practical rea-
sons to allow easy sample loading, it is more useful to
retain the inlet outside the NMR spectrometer. In any
case both vials are maintained at the same height to
avoid siphon Sow. The separation equipment, e.g.
power supply or HPLC pump, have to be placed
outside the magnetic Reld. High voltage is applied to
one end of the capillary and the other vial is
grounded. With vials outside the magnet (inlet vial or
both vials) a capillary length of approximately
150}200 cm is required. This means the capillary
is three to four times longer than a common CE
capillary.
The limiting factor of the whole system is the low
sensitivity of the NMR spectroscopy. The parameters
that affect the detection limit are the type and size of
the radiofrequency coil, the detection volume, and the
so-called Rlling factor, the ratio of coil and sample
volume. A optimal detection cell Rlls the purpose-
build microcoil completely. Furthermore, the inner
diameter of the used fused-silica capillaries have to
fulRll the chromatographic requirements. Especially
for electrophoretic techniques, the inner diameter of
the capillaries is therefore limited (4100
m) and the
size of the coil has to be adapted to these require-
ments.
Two different microcoil NMR conRgurations have
been used for online coupling of capillary separation
techniques with NMR spectroscopy. One approach is
based on a solenoid rf coil wound directly on the CE
capillary (Figure 1). The detection volume for the
system is thereby determined by the inner diameter of
the capillary and the length of the coil. With 75
m
capillaries and a typical coil length of approximately
1 mm, detection volumes of 5}8 nL are obtained.
Solenoid coils are positioned in a horizontal direction
in the NMR spectrometer, perpendicular to the main
magnetic Reld. The other approach uses saddle-type
rf coils (Figure 2). The detection unit consists of a coil
Rxed onto a glass tube, into which the capillary is
inserted. Due to technical problems, the reduction of
1196 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance

Figure 1 (A) Solenoid NMR interface. (B) Instrumentation of the solenoid NMR probe system. Adapted with permission from GfroK rer
et al. (1999) Analytical Chemistry News and Features 71: 315A}321A, and Olsen et al. (1999) Analytical Chemistry, 71: 3070}3076.

the coil diameter is restricted. The smallest available
coils have diameters of 1.5}2.0 mm and a length of
5}9 mm. To optimize the Rlling factor, special NMR
capillary cells have been fabricated to a predeter-
mined size by etching a standard fused-silica capillary
with a HF solution only in the detection region. While
the inner diameter in this part was widened, the rest
of the capillary remained nearly the original diameter
of 75
m allowing the application of electrophoretic
separation techniques. By varying the duration of
etching, detection volumes between 180 and 440 nL
were obtained. In contrast to the solenoid coil, the
saddle coil is situated vertically in the NMR spec-
trometer, parallel to the main magnetic Reld.

Figure 2 (A) Saddle-type NMR interface. (B) Instrumentation of the saddle-type NMR probe system. Adapted with permission from
GfroK rer et al. (1999) Analytical Chemistry News and Features 71: 315A}321A.
 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance
Development of the Solenoid NMR
Probe System
In 1994 Wu et al. described the Rrst capillary elec-
trophoresis NMR system with a solenoid rf micro-
coil. The coils were produced by winding copper wire
around fused-silica capillaries with inner diameters of
75}530
m. The resulting NMR detection cells had
volumes of 5}200 nL. The possibility of an online
NMR detection was shown for CE separation of
amino acids. The set-up was not suitable for routine
operation, e.g. each sample had to be loaded ex-
ternally, after which the probe was inserted. The
structural information of the NMR spectra was
limited due to large line widths (7}200 Hz). The
electrophoretic current had a clear effect to the NMR
line width.
Further developments of this solenoid system fo-
cused on the NMR detection of mass limited samples.
In 1995, Olson et al. presented a new microcoil de-
sign. Several modiRcations were devised to obtain
higher resolution. The fabrication of the coil was
slightly changed and the outer diameter of the capil-
lary was increased (357
m o.d., 75
m i.d.). The
main improvement was the reduction of the effect of
magnetic susceptibility caused by the proximity of the
rf coil to the sample. A perSuorinated organic
liquid was used to match the susceptibility of the
surrounding medium to that of the coil material. This
lowered the static magnetic Reld inhomogeneities in
the sample and thus improved resolution and line
shape (line width (1 Hz). As a consequence of the
improved resolution, the sensitivity of the system
increased. Limits of detection in the range of
100 pmol were reported for arginine, sucrose and
a seven amino acid peptide using a 5-nL detection
cell.
In 1998 Subramanian et al. presented solenoid
microcoil probes for direct or inverse 13C-NMR de-
tection. Heteronuclear NMR techniques are an im-
portant component in determining full-structure in-
formation on unknown compounds. Due to the low
relative sensitivity of 13C-NMR, the detection volume
was increased in comparison to the previous design.
Capillaries with inner diameters of 700
L were used
for the fabrication of the probe. The resulting detec-
tion cells had volumes of 550}1200
L. The natural-
abundance 13C-NMR limit of detection was below
100 nmoL. One-dimensional 13C-NMR spectra and
two-dimensional 1H}13C inverse-correlation NMR
spectra were acquired using samples in the tens of
micrograms range.
In 1999, Olson et al. applied the optimized NMR
interface for CE}NMR coupling. Arginine, glycine
and triethanolamine were used as model compounds
1197
to investigate in more detail the inSuence of the elec-
trical current on the NMR signals. It was conRrmed
that for geometries, in which capillary and static Reld
are not parallel, the electrophoretic current induces
a magnetic Reld, which degrades the spectroscopic
information obtainable from the CE}NMR spectra.
To circumvent this effect, the electrophoretic voltage
was periodically interrupted to obtain high resolution
spectra. In addition, different sample-loading tech-
niques including Reld-ampliRed stacking for sample
preconcentration, were evaluated. Flow proRles were
observed by the detection of the water signal of the
loaded samples.
Development of the Saddle-type NMR
Probe System
First results of this approach were reported by Behnke
et al. in 1996. The LC capillary was mounted inside
a modiRed NMR microprobe equipped with a 2.5-
mm double-saddle Helmholtz coil. In static NMR
experiments, a line width of 3 Hz could be achieved
in 75-
m i.d. capillaries. However, this arrangement
adversely affected the Rlling factor of the system and
thus the sensitivity using this conRguration was re-
duced signiRcantly. First, coupling experiments with
this system were performed with cHPLC. In compari-
son with CE, cHPLC provides a higher sample capa-
city and offers the possibility of peak preconcentra-
tion by the application of gradient elution. Online
and stopped-Sow NMR experiments of dansyl amino
acids were carried out in 315-
m i.d. capillaries.
The sensitivity of the saddle-type system was im-
proved by Schlotterbeck et al. in 1997. Decreasing the
inner diameter of the rf coil from 2.5 to 2 mm im-
proved the Rlling factor and thus the sensitivity and
the line shape of the system. For online cHPLC-NMR
experiments, the inner diameter of the capillary was
increased to 180-
m i.d. in the detection region and
this led to a limit of detection of 150 pmol. The feasi-
bility of the interface was proved by its application to
vitamin derivatives. The structure of a so-far unknown
kitol, a retinyl acetate dimer, was determined from
one- and two-dimensional 1H-NMR spectra in both
continuous and stopped-Sow measurements.
In 1998 the conRguration was modiRed for the
application of electrophoretic techniques by Pusecker
et al. The most important alteration was a purpose-
built CE}NMR capillary. This capillary had a detec-
tion cell with an inner diameter of 190
m corre-
sponding to a detection volume of 240 nL, whilst the
rest of the capillary still had the usual CE diameter
of approximately 75
m. CE}NMR measurements
proved that for the saddle geometry, in which
capillary and static magnetic Reld are parallel, the
1198 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance
spectroscopic information is not degraded by a mag-
netic Reld gradient induced by the electrophoretic
current. The suitability of the conRguration for electro-
phoretic methods was investigated by the application
of CE and CEC}NMR spectroscopy to model sys-
tems. The favourable capabilities of the CEC}NMR
coupling in view of the increased sample capacity and
separation efRciency were demonstrated for the sep-
aration of Rve alkyl benzoates.
The optimization of the Rlling factor by a further
increase of the detection volume to 440 nL improved
sensitivity and line shape. A line width (0.8 Hz was
reported by Schewitz et al. The system was applied to
the analysis of mixtures of pharmaceuticals and drug
metabolites, nucleotides, peptides, natural products
and ingredients of soft beverages. The possibility of
CE stopped-Sow NMR experiments was shown for
an adenosine dinucleotide. The Sow was halted by
switching off the voltage at a retention time corre-
sponding to the Rrst NMR signals of the dinucleotide.
By means of this stopped-Sow mode, it is possible to
accumulate spectra for a longer period of time. This
enables the acquisition of two-dimensional NMR ex-
periments. These techniques are substantial in deter-
mining full-structure information on unknown com-
pounds. As an example, the two-dimensional TOCSY
Figure 3 Two-dimensional CE 1H}1H-TOCSY NMR spectrum of adenosine dinucleotide. Adapted with permission of Schewitz et al.
(1999) Chromatographia 50: 333}337.
NMR spectrum of the adenosine dinucleotide is
shown in Figure 3. In this spectrum cross-peaks indi-
cate those protons which are coupled to each along an
unbroken chain of couplings. Hence the coupling of
each desoxyribose is clearly observable in two separ-
ate spin-coupling connectivities.
In 1999, a Rnal improvement was made by GfroK rer
et al. who coupled a gradient-elution CEC}NMR
system to the interface. As an example, Figure 4
shows the CEC}NMR separation of paracetamol
metabolites extracted from human urine. The result is
viewed as a contour plot with the CEC separation
time on the vertical axis and the chemical shift on the
horizontal axis. The peaks that spread throughout the
Rgure arise from the residual water and acetonitrile in
the deuterated solvents of the buffer. In addition, sets
of peaks related to paracetamol and endogenous ma-
terial are observed. Those marked (1) can be assigned
to the paracetamol glucuronide. Clear identiRcation
is possible via the diagnostic shifts of the glucuronic
acid at
"3.8 and 4.2. The component marked (2) is
the paracetamol sulfate. The NMR spectrum of the
third compound is consistent with the endogenous
material hippurate (3). The appropriate individual
rows taken from the contour plot are shown on the
right-hand side of Figure 4.
 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance 1199

Figure 4 On-flow contour plot of the 1H-NMR detected CEC separation of human urine. 1H-NMR spectra on the right side are
extracted from the contour plot. (1) Paracetamol glucuronide; (2) paracetamol sulfate; (3) hippurate. Adapted with permission from
Pusecker et al. (1998) Analytical Communications 35: 213}215.

Comparison of the Systems
Both types of NMR detection system have strengths
and weaknesses. The substantial advantage of the
solenoid system is its better sensitivity. The low detec-
tion limit resulting from the perfect Rlling factor of
the system allows extremely small detection volumes.
This is important for electrophoretic separations,
where the peaks often have volumes in the order of
a few nanolitres and larger detection cells thus limit
separation efRciency. It is obvious that the arrange-
ment is well suited for mass limited samples, but
unfortunately the small detection volume is offset by
the need for high sample concentrations. Typically,
these have been '30 mM but in some examples
concentrations of '50 mM were necessary, and this
then leads to a limitation in the number of possible
applications. The approach may need special sample
injection techniques like the Reld-ampliRed stacking.
Another problem of the solenoid CE}NMR set-up
is the complex interdependence on Sow rate, electric
Reld and current. The current, which passes through
the capillary, produces a magnetic Reld gradient
that perturbs the uniformity of the main magnetic
Reld of the NMR spectrometer. Large NMR line
widths and reduced structural information are the
consequences. High-resolution NMR spectra are only
obtainable by periodic stopped Sow capillary elec-
trophoresis when the interruption of the current en-
ables the acquisition of NMR spectra. However, this
technique makes the already long separation times
even longer.
The most important advantage of the saddle system
is its easier handling capability. Already there is
a large number and a wide range of reported applica-
tions to support this assertion. There are several
reasons. Firstly, as described above, the capillary is
arranged vertically in the saddle-type probe. Because
the shim systems used for the Reld homogeneity ad-
justment in cryomagnets are optimized in the vertical
z-direction parallel to the magnetic Reld, the shim-
ming procedure is facilitated in comparison to the
solenoid system. Furthermore, the coil is not perma-
nently attached to the separation capillary, thus
allowing the capillary to be easily exchanged for
different applications without the risk of damaging
1200 II / ELECTROPHORESIS / Capillary Electrophoresis}Nuclear Magnetic Resonance
the rf coil. Finally the distance between coil and
sample is relatively large and so problems with sus-
ceptibility do not occur.
On the other hand, the size of the saddle coil is
problematic. Due to technical problems, it is not pos-
sible to reduce the coil diameter beyond a certain limit.
To obtain a suitable Rlling factor, the detection cell has
to be enlarged and the resulting detection volumes of
180}440 nL are much larger than in conventional CE
and might be thought to lead to lowered separation
efRciency of electrophoretic techniques. However, the
fabrication method of these NMR cells produces cell
proRles that avoid mixing and turbulent Sows and thus
the separation efRciency is not seriously compromised.
Assuming a sufRcient separation of the components
large detection cells can even have advantages. Longer
residence times of the components allow an improve-
ment in the signal-to-noise ratio of the NMR spectra
through the accumulation of an increased number of
scans and the higher level of solvent reduces the neces-
sary sample concentration to (10 mM.
The two systems are complementary in approach
and both have made contributions to a better under-
standing of the challenges involved in coupling CE and
NMR. A decision on which type will Rnd more wide-
spread application will be determined by future im-
provements of the systems and possibly by the type of
applications to which this technology will be applied.
Future Developments
Numerous obstacles to coupling capillary separation
techniques with NMR spectroscopy have already
been overcome. In the short term, the development is
likely to be focused on the improvement of the NMR
interface and a goal must be implementation of the
advantages of solenoid and saddle systems. An inter-
face that combines sensitivity and small detection
volumes of the solenoid arrangement with the easy
handling of the saddle-type conRguration and thus
allow routine and automated operations will Rnd
extensive use in the future.
One way to reach this goal might be the application
of the new cryo-probe technology to capillary NMR
spectroscopy. Such probes already exist for conven-
tional NMR spectroscopy and provide, in favourable
circumstances, an improvement in signal-to-noise ratio
of approximately 400%. This is achieved by cooling
the rf coil to reduce the level of thermal noise. This
might increase the sensitivity of a saddle coil by a sub-
stantial factor without the need of a size reduction.
Another improvement would be integration of UV
detection directly into the NMR probe. This would
allow the performance of real stopped-Sow measure-
ments and this would, for example, circumvent the
problem of NMR line broadening by the electrical
current for the solenoid interface.
Usually, capillary separation science techniques are
optimized to handle minimal amounts of sample.
However, because of the desirability of direct combi-
nation with NMR spectroscopy, new technologies
have to be developed that allow an effective separ-
ation of relatively large sample amounts in small
volumes without a reduction of the efRciency. There
are many possible improvements, which are necessary
in order to increase the capabilities of this new hy-
phenated technique. With forthcoming advances in
the sensitivity of NMR spectroscopy, CE}NMR and
CEC}NMR will become practical and useful
methods in situations which require separation and
structural determination of components of mixtures
in severely mass-limited situations.
See also: II/Chromatography: Liquid: Electrochromato-
graphy; Nuclear Magnetic Resonance Detectors. Elec-
trophoresis: Capillary Electrophoresis; Capillary Elec-
trophoresis-Mass Spectrometry. III/Clinical Applica-
tions: Capillary Electrophoresis.
Further Reading
Behnke B, Schlotterbeck G, Tallarek U, Strohschein S,
Tseng L-H, Keller T, Albert K and Bayer E (1996)
Analytical Chemistry 68: 1110}1115.
GfroK rer P, Schewitz J, Pusecker K, Tseng L-H, Albert K and
Bayer E (1999) Electrophoresis 20: 3}8.
GfroK rer P, Schewitz J, Pusecker K and Bayer E (1999)
Analytical Chemistry 71: 315A}321A.
Lacey ME, Subramanian R, Webb GA, Olson DL and
Sweedler JV (2000) Chemical Reviews, in press.
Olson DL, Peck TL, Webb AG, Magin RL and Sweedler JV
(1995) Science 270: 1967}1970.
Olson DL, Lacey ME and Sweedler JV (1998) Analytical
Chemistry 70: 257A}264A.
Olson DL, Lacey ME, Webb AG and Sweedler JV (1999)
Analytical Chemistry 71: 3070}3076.
Pusecker K, Schewitz J, GfroK rer P, Tseng L-H, Albert K and
Bayer E (1998) Analytical Chemistry 70: 3280}3285.
Schewitz J, GfroK rer P, Pusecker K, Tseng L-H, Albert K,
Bayer E, Wilson ID, Bailey NJ, Scarfe GB, Nicholson JK
and Lindon JC (1998) Analyst 123: 2835}2837.
Schewitz J, Pusecker K, GfroK rer P, Tseng L-H, Albert K and
Bayer E (1999) Chromatographia 50: 333}337.
Schlotterbeck G, Tseng L-H, HaK ndel H, Braumann U and
Albert K (1997) Analytical Chemistry 69: 1421}1425.
Subramanian R and Webb GA (1998) Analytical Chemistry
70: 2454}2458.
Subramanian R, Sweedler JV and Webb GA (1999) Journal
of the American Chemical Society 121: 2333}2334.
Webb AG (1997) Progress in Nuclear Magnetic Resonance
Spectroscopy 31: 1}42.
Wu N, Peck TL, Webb AG, Magin RL and Sweedler JV
(1994) Analytical Chemistry 66: 3849}3857.
 II / ELECTROPHORESIS / Capillary Gel Electrophoresis
Capillary Gel Electrophoresis
R. Freitag, Swiss Federal Institute of Technology,
Lausanne, Switzerland
Copyright ^ 2000 Academic Press
Introduction
Electrophoresis covers a number of bioseparation
techniques where charged substances are separated
due to differences in migration speed in an electrical
Reld. Since the 1950s, electrophoresis has been
routinely used by biochemists to separate, quantify
and identify large biopolymers, such as proteins,
DNA and complex carbohydrates. In the early days
electrophoresis was typically performed in slab or
rod-shaped gels, which were necessary to hold the
electrophoresis buffer and to minimize the dispersion
of the analyte bands by suppressing convection
caused by Joule heating and electroosmosis. At that
time, electrophoresis was therefore commonly gel
electrophoresis.
Electrophoretic separation is based on differences
in the charge density (the mass-to-charge ratio) of the
analytes. Small, highly charged molecules move faster
than large, low charged ones, while uncharged mol-
ecules do not move at all. This is a powerful principle,
for example for the separation of native proteins and
peptides. Other biomolecules such as DNA restriction
fragments and carbohydrates are very similar in
chemical structure and therefore almost identical in
charge densities. For these molecules gel electrophor-
esis represents a unique opportunity. Gels can be
prepared to provide a sieving effect in the relevant
mass range and thereby allow separation of charged
molecules not according to their charge density but
according to their size.
In 1981 JoK rgenson and Lukacs published a series of
landmark papers, in which they demonstrated both
theoretically and practically how electrophoresis
could be carried out in free solution as long as capil-
laries of small inner diameter were used and how the
previously unwelcome electroosmotic effect could
be used to advantage. The era of (free zone) capillary
electrophoresis (CZE) began. Gels were no longer
necessary, since the ‘wall effect’ in capillaries of less
than 100
m i.d. is sufRcient to stabilize the
Sow. Capillary electrophoresis (CE) was seen as
a complementary analytical technique to both liquid
chromatography and ‘conventional’ slab/rod gel elec-
trophoresis. In the following decade, research and
application of free zone capillary electrophoresis ex-
1201
ploded and several variants of the basic technique
were established. In 1989 the Rrst commercial CE
instrument entered the market.
Capillary gel electrophoresis (CGE) was resurrec-
ted in 1983 by Hjerten, albeit for protein separations.
It would probably have rested there, since protein
CGE subsequently encountered severe problems.
Instead it was the human genome project that helped
to establish CGE as the important analytical tech-
nique it is today. The aim of this project is to deter-
mine the sequence of the over 3 billion base pairs
that form the human genome. DNA sequencing re-
quires the separation and identiRcation of DNA frag-
ments that differ in length by a single base. The
progress of the project is largely determined by the
speed of the respective analytical techniques.
The efRciency of CGE is extremely high and up to
3 million plates per metre are routinely reached. Heat
dissipation is more effective in thin capillaries com-
pared to conventional gels. As a consequence, much
stronger Relds can be used in CGE without loss in
resolution. This results in signiRcantly reduced analy-
sis times. An example, the single base resolution of
oligoadenylates containing from 12 to 60 nucleotides,
is shown in Figure 1.
Current high throughput instruments use ultrathin
capillaries and oligonucleotide primers that are
covalently attached to different Suorescent dyes. The
restriction digests of all four primers can therefore be
analysed in the same sample. A comparison between
CGE and conventional gel electrophoresis shows that
even though more samples can be analysed in parallel
in slab gels, CGE resolves a given sequence at least
three times faster. Table 1 summarizes some addi-
tional advantages and disadvantages of CGE over
conventional gel electrophoresis.
Figure 1 Single base resolution of oligoadenylates containing
from 12 to 60 nucleotides by CGE. (Adapted with permission from
Baba Y, Matsuura T, Wakamoto K and Tsuhako M (1991) Journal
of Chromatography 558: 273}284.)
1202 II / ELECTROPHORESIS / Capillary Gel Electrophoresis

Table 1 Comparison of capillary and conventional gel electrophoresis

Advantages of CGE
Smaller sample volume
Less buffer, gel and reagent consumption. The latter is especially important, since DNA gel electrophoresis requires toxic and/or
mutagenic substances (e.g. ethidium bromide)
Shorter analysis time
Higher resolution
Higher efficiency (over several million plates per metre)
Reliable quantification of the analyte concentration by on column or off column flow through detector (UV absorbency, fluorescence,
mass spectrometer, etc.) instead of staining and semiquantitative evaluation in the gel by densitometry
Other than conventional slab gel electrophoresis, CGE can be fully automated from sample injection to separation, detection and
data processing steps
Semipreparative applications become much easier, since the molecules pass through and exit the capillary. In conventional gels, the
zones are stained within the gel, from which they can only be removed by manual cutting of blotting
Disadvantages of CGE
Restriction in sample size
CGE gels are more difficult to prepare
Capillary coating often necessary to reduce electroosmosis

Gel Filled Columns for Capillary Gel Chemical Gels

Electrophoresis
Cross-linked chemical gels and physical gels such as
agarose are just as popular in CGE as in conventional
slab gel electrophoresis. However, since the capillary
wall effect exerts its stabilizing inSuence, solutions of
entangled polymers are also used.
Naked fused silica capillaries show electroosmosis
above a pH of 3. As a consequence the electrophoresis
buffer Sows towards the cathode. In capillary gel
electrophoresis electroosmosis would lead to extru-
sion of the gel from the capillary. A possible way to
avoid this is to coat the inner silica surface with
a neutral polymer. Table 2 summarizes the more
common types of capillary coating. Ideally, the coat-
ing should be effective in suppressing electroosmosis
as well as the interaction of the analytes with the
capillary wall. The preparation should be reproduc-
ible and the coating should be stable for a long time
and over a wide pH range. Most coatings are at-
tached to the wall by covalent bonds. However, dy-
namic coatings consisting of ionic, zwitterionic or
nonionic molecules that are simply adsorbed to the
wall either as monolayer or as bilayer are also used.
A general outline for preparing a cross-linked gel
column for CGE is given in Table 3. Cross-linked
polyacrylamide (PAA) gels are prepared by radical
copolymerization of acrylamide as monomer and
} most frequently } N,N
-methylenebisacrylamide
(BIS) as cross-linker. (NH4)2S2O8 is commonly used
as initiator in combination with N,N,N
,N
-
tetramethyleneethylenediamine (TEMED). Ribo-
Savin ((5 ppm) has been used as light sensitive
initiator but requires the use of UV transparent capil-
laries. The polymerization of acrylamide can also be
started by -radiation (60Co source).
The resulting gels have well deRned, fairly small
average pore sizes of 2 to 8 nm. The %T, %C nomen-
clature is used for their characterization, with:
gacrylamide#gcrosslinker
%T"
100 mL solvent
corresponding to the monomer concentration:
gcrosslinker;100
%C"
gcrosslinker#gacrylamide
corresponding to the crosslinker concentration.

Table 2 Common types of capillary coatings

Coating Comments

Polyacrylamide with Si}O}C bond Very common, easy to prepare

Polyacrylamide with Si}C bond Improved hydrolytic stability, difficult to prepare
Polyethylene glycol Very common
Nonionic surfactants TWEEN and BRIJ series
LC type stationary phases C1, C8, C18, i.e. weakly, moderately and highly hydrophobic
GC type stationary phases e.g. DB-17 (50% phenylmethyl silicone)
 II / ELECTROPHORESIS / Capillary Gel Electrophoresis 1203

Table 3 Outline for the preparation of a cross-linked sieving matrix for CGE

Prepare stock solution of monomer, cross-linker, buffer, etc.

Cut/burn detection window if necessary
Pretreat the inner surface of the silica capillary (coating, activation, etc.)
Introduce mixture of monomer (cross-linker)/initiator/catalyst/stabilizing agent (if necessary)
Initiate polymerization
Pre-electrophoresis (to remove impurities and assure constant run conditions and a stable baseline)

Single strand oligonucleotides are typically separ-
ated in 2.5%T/3.3%C or 4%T/3.3%C gels. A higher
cross-linker concentration gives a tighter sieve and
thus higher size selectivity. An increase in the mono-
mer concentration also results in a gel matrix with
smaller pores. Lower %T/%C values result in a re-
duced selectivity but allow coverage of a wider
analyte size range. An increased column length can
compensate for the loss in resolution, albeit only at
the price of increasing the analysis time. A reduction
of the cross-linker concentration for a given mono-
mer concentration results in larger pores. A
3%T/0.5%C gel has, for example, been found to give
excellent resolution of double-stranded DNA restric-
tion fragments varying from 5 to 12 000 base pairs.
Polyacrylamide is a popular gel matrix because of
its electroneutrality. With time, however, it will be
hydrolysed to charged polyacrylate, especially at high
temperatures (high Reld strength) and extreme pH
values. When the polyacrylate concentration becomes
too high, an electroosmotic Sow can be observed,
which Rnally results in expulsion of the gel from the
capillary. In order to prevent this, the gel may be
covalently anchored to the wall coating. This calls for
polymerizable (double bonds) groups in the coating.
Alternatively, coating and gel formation can be done
simultaneously in activated capillaries.
The preparation of gel-Rlled capillaries is not
simple and only four out of Rve capillaries produced
can actually be used. Problems arise in a number of
areas, but a major cause of concern is the formation
of bubbles during production, transport and use (in-
jection!) of the gel-Rlled capillaries. Bubbles lead at
best to loss in resolution and changing separation
patterns and at worst to the total breakdown of the
electrical Reld, especially at high Reld strengths
('300 V cm\1).
If the gel and the capillary wall coating are not
physically linked, bubble formation is less of a prob-
lem, since the wall and the gel can move relative to
one another. The reduced stability of the gel under
these circumstances is another matter. Whereas capil-
laries with wall-bound gels can often be used for
more than 100 runs, the number of separation that
can be performed with the mobile gels tends to be an
order of magnitude lower.
Bubble formation during polymerization is largely
due to shrinkage of the gel during the process (higher
density of the gel than the monomer mixture).
Table 4 summarizes approaches to circumvent this
problem. In the case of the laterally aggregated
polyacrylamide gels, a hydrophilic polymer such as
polyethylene glycol (PEG) is added to the reaction
mixture. The more hydrophobic PEG coordinates
a large amount of water leaving the growing polyac-
rylamide strands to form hydrogen bridges mainly
among themselves. As a result, thick gel Rbres are
formed and subsequently stabilized by cross-linking
(Figure 2). The laterally aggregated gels have larger
pores than homogeneous polyacrylamide gels of sim-
ilar %T/%C. Their sieving behaviour depends also
on the concentration and molecular mass of the PEG
and should be determined experimentally in each
case.
Physical Gels and Entangled Polymers
In physical gels the network structure is formed by
noncovalent interactions such as van der Waals
forces and hydrogen bonds rather than chemical

Table 4 Suggestions for the polymerization of bubble-free cross-linked PAA gels

Polymerization at high pressure (400 bar) to equalize the density of the monomer mixture and the final gel
Sequential polymerization starting with the gel formation at one end of the capillary and slowly progressing to the other, for instance by:
z irradiation (1 cm min\1)
z thermal (slow immersion into a heated water bath or slow removal from a cold one)
z isotactic polymerization (the initiator is placed between the leading and the terminating electrolyte and slowly moves through the
capillary by isotachophoresis)
Laterally aggregated polyacrylamide gels (formed in the presence of a hydrophilic polymer such as polyethylene glycol)
1204 II / ELECTROPHORESIS / Capillary Gel Electrophoresis

pends on the polymer concentration. In dilute solu-

tion the polymer molecules are hydrodynamically
isolated. For a certain polymer volume fraction H
(overlap threshold) the chains begin to entangle and
interact. As a consequence a highly dynamic network
is created. Experimentally, H can be determined by
plotting the logarithm of the speciRc viscosity versus
the polymer volume fraction. For (H the curve
has a slope of approximately 1.0; for 'H the
Figure 2 Hypothetical model for ‘laterally aggregated’ gels. (a)
Homogeneous network of cross-linked PAA. (b) Heterogeneous
slope increases.
network of PAA in the presence of PEG 2000 (Reproduced with Linear polyacrylamide has to be used at concentra-
permission from Kuhn R and Hoffstetter-Kuhn S (1993) Capillary tions of at least 6%T to ensure sufRcient size discrim-
Electrophoresis: Principles and Practice, Springer Verlag.) ination. The solution is then already very viscous and
difRcult to handle. Other hydrophilic polymers have
much lower threshold values. The corresponding
cross-linking. If large analyte molecules try to trans- solutions are therefore less viscous and can be easily
verse this dynamic structure they also encounter injected and replaced. Among them are various cellu-
a size-dependent resistance (sieving effect). The type lose derivatives (overlap threshold approximately
and number of physical gels used in CGE is much 0.3%), such as methylcellulose, hydroxymethylcel-
more varied than the chemical gels. Mixed physical lulose, hydroxyethylcellulose and hydroxypropyl-
gels containing, for example, polyacrylamide and cel- methylcellulose. The sieving efRciency depends also
lulose, are also possible. Both solid and liquid sieving on the molecular mass of the cellulose. Molecules
matrices are known. Physical gels are often cheaper weighing more than 900 000 g mol\1 make good
than chemical gels. They have a more Sexible pore
structure and can be operated at higher temperatures
(50}703C) and Reld strength (up to 1000 V cm\1).
Examples of solid physical gels include agarose gels
at room temperature and linear polyacrylamide gels
with %T'8 (the %C would be zero in this case).
Such high %T polyacrylamide gels also have to be
prepared by in situ polymerization. The size selectiv-
ity is a function of the polymer concentration and can
be similar to that of the cross-linked polyacrylamide
gels (Figure 3). Agarose is a natural polysaccharide
network of 1,3-linked -D-galactopyranose and 1,4-
linked 3,6-anhydro--L-galactose. PuriRed agarose
for electrophoresis can be obtained from a number of
chemical suppliers. The matrix exhibits a wide range
of pore diameters from several hundred nanometres
to several micrometres and has high mechanical
strength. Agarose is biologically inert and stable be-
tween pH 4 and 9. The gel liqueRes around 653C and
solidiRes again at 353C. In the molten state it is easily
injected into (or removed from) the capillary. Agarose
gels can be used below and above the gelling point.
Gel Rlled capillaries (0.3}5% by weight) are stable for
a couple of days. No treatment of the capillary walls
is necessary. To improve the stability of the gel
a small amount of a polyalcohol such as sorbitol may
be added.
Entangled polymer solutions are liquid sieving ma-
trices. They are essentially low viscosity gels and can
Figure 3 Separation of lambda-DNA restriction fragments
be replaced after each analysis. Hydrolysis or con- (EcoRI). Top, 2% polyacrylamide gel; bottom, 4% polyacrylamide
tamination of the network is thus less problematic. gel. (Adapted with permission from Hebenbrock K, SchuK gerl K
Gel formation in entangled polymer solutions de- and Freitag R (1993) Electrophoresis 14: 753}758.)
 II / ELECTROPHORESIS / Capillary Gel Electrophoresis
sieving gels for DNA fragments (polymer concentra-
tion 0.5%). For protein separation dextran or poly-
ethylene rather than cellulose derivatives are used to
prepare the gels. This has the additional advantage of
allowing detection of the proteins at the more sensi-
tive wavelength of 214 nm.
Both physical and chemical gels can be further
modiRed, for example by the incorporation of cyclo-
dextrins or antibodies into the matrix, to allow more
speciRc interactions.
Separation in Capillary Gel
Electrophoresis
Two models are used in CGE to describe the sepa-
ration. The Ogston model considers the gel as a
labyrinth of interconnecting pores and channels with
an average pore (mesh) size of . The analyte mol-
ecules are pictured as rigid spheres with a radius R.
Small molecules can pass unhindered through a large
fraction of the pores, they therefore move fast but
larger molecules are slower. If the pore size of the gel
is assumed to be a function of the polymer concentra-
tion, the following equations can be used to describe
the situation:
g" exp[!Tb(R#r) ] 2 carried out on the same type of instrument as stan-
and:
g" exp[!0.25((R#r))2]
where g is the analyte mobility in the gel matrix;  is
the analyte mobility in free solution for EP0; T is
the polymer concentration; b is a constant; R is the
radius of gyration of the analyte molecule and r is the
average radius of the network pores.
The term b(R#r)2 is also called the retardation
coefRcient, k. To obtain , the mobility of the
analytes in free solution is extrapolated to a Reld
strength of zero. A plot of g versus T, the Ferguson
plot, results in a straight line with a slope of k and an
intercept of . This plot is used to determine the size
separation range of the gel. If the plot is not linear, the
assumption that the analytes are rigid spheres with
a radius R(r is no longer true.
Biopolymers are Sexible molecules rather than
rigid spheres and can therefore in reality pass through
pores with radii of less then their own radius of
gyration. The reptation model considers the analytes
as dynamic and essentially chain-like structures that
snake and wriggle headRrst through the network in
a reptation (hence the name) type of motion. In this
model the length of the Sexible macromolecule is
considered large in comparison to the distance be-
1205
tween neighbouring knots in the network. The
analyte mobility is proportional to the reciprocal of
the chain length or, for nucleic acids, to the base
number, N, of the molecule:
g+1/N
Larger molecules are again more handicapped than
smaller ones. For a Sexible macromolecule the
transition from the Ogston sieving model to the rep-
tation model takes place when R becomes approxim-
ately 1.4 times , depending on the nature (globular
proteins, rod-shaped DNA molecules) and the Sexi-
bility of their structure. For very large molecules
neither model holds true, because the electrical Reld
deforms the molecular structure. For DNA fragments
the equation is valid up to approximately 1000 base
pairs, depending on the applied Reld strength. The
mobility of larger DNA fragments is actually higher
than expected. This leads to a co-migration of larger
fragments with smaller ones, especially at high Reld
strength and in high %T gels.
Instrumentation and Methodology
CGE requires no special instrumentation and can be
dard capillary electrophoresis. All that is required in
addition to the gel Rlled column is a high voltage
power supply and a detector. Most commercial sys-
tems are automated to a high degree and also include
an autosampler/injection module and a data handling
station.
CGE uses continuous electrolyte systems. The sep-
aration of the analyte molecules is therefore a kinetic
process. In general, all buffer systems for capillary
electrophoresis can also be used in CGE. By far the
most popular are TRIS}borate buffers (50 or 100 mM
TRIS, 50, 100, 250 mM borate, pH 7.6}8.5, up to
5 mM EDTA), especially for the separation of DNA
fragments and nucleotides.
While the separation of the native molecules is
possible, denatured molecules are sometimes easier to
analyse. In this case, denaturing agents, such as high
concentrations of urea (5}8 M) are added to the separ-
ation buffer, in order to destroy the tertiary and
quaternary structure of the molecules. A special case
is polyacrylamide gel electrophoresis of proteins in
the presence of the surfactant sodium dodecyl sulfate
(SDS-PAGE). The protein structure is fully denatured
under these conditions including cleavage of the di-
sulRde bonds by a reducing agent such as mercap-
toethanol. The remaining polypeptide chains bind
SDS in a constant weight ratio (1.4 g SDS per gram of
protein) to yield detergent}protein complexes of
1206 II / ELECTROPHORESIS / Capillary Gel Electrophoresis
Figure 4 Determination of the molecular mass of recombinant
human antithrombin III. Top, calibration using pepsin (34.7 kDa);
ovalbumin (45.0 kDa); bovine serum albumin (69.0 kDa); and
phosphorylase subunit B (97.4 kDa) as markers. Bottom, recom-
binant h-AT III. (Reproduced with permission from Reif O-W and
Freitag R (1994) Journal of Chromatography 680: 383}394.)
constant charge density. SDS-PAGE in slab gels is
a standard method for purity control and mass es-
timation of proteins. SDS-PAGE in capillaries has
been used for similar purposes (Figure 4). Ethidium
bromide is a selective intercalating agent for double
stranded DNA. The positively charged molecule can
be used to optimize the separation of large DNA
restriction fragments in CGE (Figure 5). It lowers the
charge density and thereby the migration speed of the
negatively charged DNA molecules.
Sample Introduction
Two basic types of injection principle are used in
capillary electrophoresis. One is injection by pressure
(vacuum suction or gravity), i.e. hydrodynamic injec-
tion, the other is electrokinetic injection. Pressure
injection cannot be used with cross-linked gels, al-
though it may be an option for some of the low
viscosity entangled polymer solutions. In most cases it
would either not succeed in pushing the sample into
the capillary or destroy the gel.
For electrokinetic injection the capillary is placed
into the sample vial and the electric Reld is switched
on (typically between 1 and 20 s and 100 and
400 V cm\1). The analyte molecules migrate into the
capillary due to electrophoresis. Electrokinetic
injection is biased towards small, highly charged
molecules which are then over-represented in the
introduced sample. In capillary gel electrophoresis
the problem is less pronounced, since all analyte
molecules are assumed to have identical mass
to charge ratios, i.e. differ very little in migration
speed.
The surface of the capillary inlet is very important
(Figure 6). To ensure an even surface, the respective
end of the capillary should be cut of with a microtome
or snapped off cleanly after scoring with a sapphire
cleaver.
Detection
The most common detection principles in CE are
optical or column detectors (UV absorbence and
(laser induced) Suorescence, LIF). Fluorescence
and especially LIF detection allow very low limits of
detection (LOD) to be reached. Unfortunately
only a few biological molecules show native Suores-
cence. DNA molecules require pre- or postcapillary
derivatization with a Suorescing agent, e.g. a Suores-
cence-labelled hybridization probe or primer. FITC,
JOE, TAMRA, FAM and ROX have been used as
labels in DNA sequencing. Proteins have been detec-
ted using their native Suorescence at 280 to 340 nm
Figure 5 Effect of ethidium bromide addition on the resolution
of a restriction fragment digest (pBR 322 by Msp I). Top, no
ethidium bromide; bottom, 1
g mL\1 ethidium bromide. (Repro-
duced with permission from Guttmann A and Cooke N (1991)
Analytical Chemistry 63: 2038.)
 II / ELECTROPHORESIS / Capillary Gel Electrophoresis
Figure 6 Effect of the physical shape of the inlet of a gel-filled
capillary on the resolution of naproxen enantiomers. (Reproduced
with permission from Guttmann A and Schwartz HE (1995) Ana-
lytical Chemistry 67: 2279}2283.)
(tryptophan residues). However, since the quantum
yield is often low, labels such as Suorescein
isothiocyanate (FITC 494 to 525 nm) are also used
(Figure 7).
The problem with UV absorbence in CGE is the
short optical pathway (inner diameter of the capil-
lary), which leads to low sensitivity. Bubble and z-
shaped cells can be used, but the bubble cell makes
the capillary fragile, while the z-shaped cell lowers
the resolution. In gel-Rlled capillaries the sensitivity
suffers further because of the low UV transparency of
most gels. The transparency of a PAA-Rlled capillary
(6%T, 5%C) at 260 nm is 15% lower than that of
a water-Rlled one; even lower if additives like urea or
PEG are used. Instead of on-column detection, an
off-column detector connected with a sheath Sow
interface can be used. In addition, several suppliers
offer UV-transparent gels either of the cross-linked or
the entangled type. The composition of these gels is
usually proprietary.
1207
Special Techniques
Varying the magnitude and the direction of the
Reld, so-called pulsed Reld electrophoresis, can
improve the resolution especially for DNA separ-
ation, since fragments of up to 10 million base pairs
can be analysed. Since the larger molecules are as-
sumed to snake through the gel, they need to orientate
themselves in the Reld before they can advance. If the
Reld is regularly inverted, the advance of the larger
molecules is more affected than that of the smaller
ones and as a consequence their separation is im-
proved.
Voltage ramping has been shown to improve the
separations of mixtures of small and large molecules.
First high voltage is used to separate the smaller
fragments, afterwards the larger fragments are sepa-
rated with a gradually or abruptly decreased voltage.
Field programming may also aid fraction collection,
since the extremely sharp peaks of CGE are difRcult
to ‘catch’ unless the analyte migration is reduced by
lowering the voltage.
Future Developments
CGE is a rapidly maturing technique and the need for
high resolution bioanalytical techniques can be ex-
pected to increase rather than decrease in the near
future. Up to now application of capillary gel elec-
trophoresis has been dominated by DNA analysis.
The potential of the method for protein analysis has
not been realized. However, the demands of the mod-
ern biopharmaceutical industry for fast and reliable
protein characterization techniques may soon change
this.
The speed of CGE can be further accelerated by
more stable gels (higher Reld strength) or by using
several capillaries in parallel (array). If the sample
size could be reduced further, the need for template
ampliRcation, e.g. by the polymerase chain reaction
(PCR) would be reduced. Detection may be improved
by increased use of high molecular weight mass spec-
trometers (MS). They can be linked to the CE instru-
ment, for example by an electrospray ionization (ESI)
interface. As these detectors become more affordable,
their use in CGE will increase. Mass spectrometers
combine a high sensitivity with being nearly universal
detectors. When MS-MS is used more structural in-
formation of the analytes become available.
Capillary electrophoresis in many ways is already
a microtechnique. However, further miniaturization
is possible and the CE on a chip need not be far
ahead. Such microchip CE will most likely use gel
Rlled ‘capillaries’ to realize the maximum number of
theoretical plates over the short separation distance.
1208 II / ELECTROPHORESIS / Capillary Isoelectric Focusing
Figure 7 Quantification of human immunoglobulin G (h-IgG) using FITC-labelled Protein G. A, no IgG; B, 250
g mL\1; C,
1 mg mL\1. (Adapted with permission from Reif O-W, Lausch R, Scheper Th and Freitag R (1994) Analytical Chemistry 66:
4027}4033.)
Further Reading
Beale SC (1998) Capillary electrophoresis. Analytical
Chemistry 70: 279. (This journal prints every two years
a review on capillary electrophoresis. The article is writ-
ten by an expert in the Reld and covers more or less all
developments in CE including CGE of the preceding
years.)
Grossmann PD and Colburn JC (eds) (1996) Capillary
Electrophoresis: Theory and Practice. San Diego: Aca-
demic Press.
Grossmann PD and Soane DS (1991) Capillary electrophor-
esis of DNA in entangled polymer solutions. Journal of
Chromatography 559: 257. This paper treats the two
models for separation in CGE (Ogston and reptation) in
more detail.
Capillary Isoelectric Focusing
P. G. Righetti, University of Verona,
Faculty of Sciences, Verona, Italy
C. Gelfi, ITBA, CNR, Segrate, Milan, Italy
Copyright ^ 2000 Academic Press
JoK rgenson JW and Lukacs KD (1981) High resolution sep-
aration based on electrophoresis and electro-osmosis.
Analytical Chemistry 53: 1298.
Kuhn R and Hoffstetter-Kuhn S (1993) Capillary Electro-
phoresis: Principles and Practice. New York: Springer
Verlag.
Landers JP (ed.) (1994) Handbook of Capillary Elec-
trophoresis. Boca Raton: CRC.
Li SFY (1992) Capillary Electrophoresis. Amsterdam: Else-
vier.
Lunte SM and Radzik DM (1996) Pharmaceutical and
Biomedical Applications of Capillary Electrophoresis.
Oxford: Pergamon Press.
Schwartz H and Guttman A (1995) Separation of DNA by
Capillary Electrophoresis. Fullerton, CA: Beckmann
Primer 607397.
Isoelectric focusing (IEF) possibly represents the elec-
trokinetic method with the highest resolving power.
In IEF, amphoteric compounds are sorted in order of
their isoelectric points (pI ) in a steady-state pH gradi-
ent. Good resolution is favoured by both a low diffu-
sion coefRcient and a high mobility slope at the pI,
 II / ELECTROPHORESIS / Capillary Isoelectric Focusing 1209

conditions which are well satisRed by most proteins. A principal difference between IEF in a gel and in
A high Reld strength and a shallow pH gradient a capillary is that, in the latter, mobilization of the
further enhance resolution. There are two basic vari- focused proteins past the detector has to be carried
ants of IEF: (1) in soluble, amphoteric buffers, called out if an on-line imaging detection system is not used.
carrier ampholytes (CA) and (2) in insolubilized, non- Three techniques are mainly used: chemical and hy-
amphoteric buffers (the latter technique is known as drodynamic Sow mobilization (in coated capillaries)
immobilized pH gradients, IPG). In this article we and mobilization utilizing the electrosmotic Sow (in
will deal with the former, i.e. CA-driven IEF, since uncoated or partially coated capillaries). These tech-
IPGs have not as yet been implemented in capillaries. niques are discussed further below.
In fact, in IPGs, the buffers (acrylamido weak acids
and bases) have to be grafted onto a support which at
present is only a polyacrylamide gel. In addition, the Focusing in Internally Coated
gradient is created ‘artiRcially’, outside the electric Capillaries
Reld (whereas in CA-IEF the pH gradient has to be
At any pH value above pH 2 the fused silica surface
generated and maintained by the electric Reld itself),
will progressively acquire negative charges due to
and thus gel casting requires the use of a two-vessel
ionization of weakly acidic silanol groups which are
gradient mixer, with the simultaneous pouring of
fundamental constituents of any vitreous material.
a density and a pH gradient in a Sat-gel slab format.
Accordingly, close to the capillary wall (in the diffuse
In normal use IPGs additionally require that the gel
part of the double layer) there will be more positive
cassette is opened and that the polyacrylamide slab,
than negative ions (electroneutrality will thus not
with the grafted pH gradient, is exhaustively washed,
prevail in the double layer). In an electric Reld, the
so as to eliminate salts, ungrafted buffers and cata-
hydrated, positively charged surface layer will move
lysts. This preparation sequence means that preparing
toward the negative pole, thus producing an electros-
an IPG in a capillary format is not a straightforward
motic Sow (EOF), which can be observed as a bulk
process.
Suid movement. Such an EOF pump is not, in itself,
The theory of IEF and IPGs, as well as the chem-
deleterious to the analyte zone, since it has a Sat
istry of the buffers adopted, has been covered else-
proRle (except in the few nanometre thickness of the
where. Here we will focus on the techniques used in
double-layer); however, in the case of proteins, strong
capillary IEF (CIEF), namely: one- and two-step fo-
adsorption may ensue, due to multipoint attachment
cusing methods, pH gradient determination, sample
of positively charged species to the negative charges
preconcentration systems and some application
of the wall. In addition, particles migrating in direc-
examples.
tions opposite to the bulk liquid Sow might never
reach the detector. The electrosmotic mobility (
eo) is
General Considerations inversely proportional to the viscosity (in the double
layer). Thus, by coating the inner surface of the capil-
CIEF combines the high resolving power of conven-
lary with a hydrophilic, non-ionic polymer, there will
tional IEF with the advantages of automation and
be two beneRcial effects: the charges will be masked
speed. The separation of charged analytes takes place
and, in general, suppressed and, additionally, the
in a pH gradient created in a capillary by carrier
viscosity in the double layer will be so high as to
ampholytes under the inSuence of an electric Reld.
virtually eliminate EOF. A typical coating consists
The concentrating effect that occurs during the focus-
in Rrst reacting the wall with a bifunctional agent
ing step, enables components present in small quant-
(-methacryloxypropyltrimethoxysilane) and then
ities to be detected. A major difference between IEF
covalently afRxing a monolayer of linear poly-
and CIEF is the detection method: in CIEF detection
acrylamide to the dangling double bonds. Many other
is performed in most cases with an ultraviolet (UV) or
coating procedures have been described and have
photodiode array detector placed near one end of the
been reviewed by Chiari et al. (1996).
capillary. In order to visualize the stationary zones
formed in the capillary, its content must therefore be
Focusing Step
mobilized in an additional step so that they pass in
front of the detector window. To date, most of the The coated capillary is Rlled entirely with sample
reagents used in CIEF (carrier ampholytes, solution mixed into at least 1% carrier ampholytes
solubilizers and so on) have been adopted from tradi- (the sample should be desalted so as to avoid pH
tional IEF. A number of reviews have already ap- gradient drift). One end of the tube is then pressed
peared on the topic of CIEF (see the Further Reading into a 1% agarose gel, prepared either in 20 mM
section). NaOH or in 20 mM phosphoric acid (representing
1210 II / ELECTROPHORESIS / Capillary Isoelectric Focusing
the cathodic and anodic solutions, respectively). The
gel plug thus inserted into the tube end prevents
zone-deformation by hydrodynamic Sow in the tube
during the subsequent focusing step. A constant volt-
age of 4000}6000 V is applied. When the steady-state
has been attained, which occurs when the current has
dropped to about 10}25% of the starting value, the
voltage is switched off and elution started immedi-
ately as described below.
Elution and Detection Step
In such a system, due to suppression of EOF, the
focused stack of carrier ampholytes and proteins is
arrested; thus ways have to be found to transport the
stack past the detector. A number of procedures can
be adopted: (1) apply a mechanical pump to the
capillary and generate a hydrodynamic Sow (at the
end of the IEF process); (2) replace the base at the
cathode with acid or the acid at the anode with base
and (3) salt mobilization. This latter technique has
attained wide popularity. When a salt (e.g. NaCl) is
added at the anolyte, mobilization will be towards the
anode; conversely, if added to the catholyte, the train
of bands will elute at the cathode. Figure 1 shows
Figure 1 Focusing in internally coated capillaries. The focusing
(upper) and mobilization (lower) steps. In the upper drawing the
steady-state is shown as an arrested stack of proteins and CA
buffers. Only protons and hydroxyl ions move from the respective
electrodes carrying most of the current. In the lower drawing,
addition of NaCl to the cathode is shown to mobilize the stack of
proteins and CA buffers towards and past the detector port (repre-
sented as a large vertical arrow close to the cathode. (Repro-
duced by permission of Bio Rad, Hercules, CA, USA.)
a representation of this process: the upper part depicts
the steady-state, characterized by a stationary pattern
of focused proteins in an arrested pH gradient. In this
stage, current is carried mostly by protons and hy-
droxyl ions moving from the anodic and cathodic
compartments, respectively (and also by the to and
fro movement of the CA buffers about the pI posi-
tion!). On addition of the salt (typically 20}80 mM)
at the cathodic reservoir, the stack of proteins and CA
buffers is mobilized towards the cathodic side (past
the detector). The time required for mobilization is
about 15 min at 360 V cm\1. During mobilization
the current which had reached a minimum at the end
of the focusing stage (typically 1
A) rises again to as
high as 50
A. It is during mobilization that the train
of zones, titrated away from the pI by the cations or
anions (other than protons or hydroxyl ions) entering
the tube from one of the electrode reservoirs, transits
in front of the detector and is registered as a series of
bands. Ideally, proteins and peptides should best be
monitored at 210 (or even 190) nm where the absorb-
ance of the amido bond is 20}50 times higher than at
280 nm. However, at the low wavelengths the CA
buffers also produce a UV signal (rather similar for
Ampholine and Biolyte, quite different in the case of
Pharmalytes) which could be mistaken as sample
zones. Thus, in an IEF experiment, it is best to read
the sample at 280 nm.
Elution by Vacuum or Pressure Under Voltage
An alternative elution method consists in applying
a vacuum of 5 mmHg while still under high voltage.
The vacuum causes the focused proteins to Sow past
the detector, while the voltage maintains the pH
gradient and zone sharpness even in the presence of
distorting effects due to laminar Sow. First, the entire
tube is Rlled with NaOH (catholyte). Then, by hy-
drodynamic Sow, approximately two-thirds of the
capillary length is Rlled with carrier ampholytes. This
is followed by a short sample plug, which is sub-
sequently insulated from strong anolyte (which
might, by contact, denature some proteins). At this
point liquid pumping is stopped and the focusing
process takes place between the phosphoric acid as
anolyte and the NaOH as catholyte. Mobilization is
again accomplished by a vacuum-driven hydro-
dynamic Sow under voltage. The detection limit for
proteins at 280 nm is as low as 1.3 ng, while the
signal linearity is in the range of 1.3}10.7 ng (an
eight-fold concentration range). Interestingly, this
sensitivity is of the same order of magnitude as that
reported for silver staining of sodium dodecyl sulfate-
denatured protein zones and two to three orders of
magnitude higher than conventional Coomassie Bril-
liant Blue staining.
 II / ELECTROPHORESIS / Capillary Isoelectric Focusing 1211

pI Measurements
In its simplest approach unknown pI values can be
assessed by plotting the pI values of a set of markers,
co-focused with the proteins under investigation,
versus their relative mobility on elution. This plot
is linear and thus a high precision (approximately
$0.1 pH unit) is obtained (see Figure 2). pI values as
low as 2.9 and 2.75 can be determined. In another
approach monitoring the current in the mobilization
step can be adopted for pI assessments. If the peaks of
the mobilized stack of proteins are monitored simul-
taneously with the rising current due to the passage of
the salt wave in the capillary one can correlate a given
pI value (which should already be known from the
literature) with a given current associated with the Figure 3 Calibration graphs for pI determination using the cur-
transit of a peak at the detector port. The system can rent during the mobilization step as a parameter in capillary IEF.
The six experimental points represent six forms of transferrin,
thus be standardized and used for constructing a cal- containing different amounts of sialic acid and of iron. (Repro-
ibration graph to be adopted in further work, without duced from Kilar F (1991) Journal of Chromatography 545:
resorting to ‘internal standards’. One such graph cor- 403}406, by permission.)
relating current with pI values is shown in Figure 3:
this appears to be a precise method, since the error is
may lead to precipitation. However, at, or in the
only about 0.03 pH units. The use of low Mr sub-
proximity, of their pI value proteins exhibit a min-
stituted aromatic aminophenols, which fulRl all the
imum of total charge and thus a solvation minimum.
requirements for pI standards for CIEF and assure
This increases the risk of aggregation, and is further
a 0.06% pI reproducibility, have also been proposed.
enhanced by the extremely low ionic strength condi-
In another method dansylated peptides have been
tions prevailing in IEF. When protein molecules pre-
synthesized for use as pI markers for evaluating pH
cipitate they probably aggregate by hydrophobic in-
gradient formation.
teractions. It seems logical therefore to try to suppress
precipitation by supplementing the CA buffers with
On Isoelectric Precipitation
agents known to decrease hydrophobic interactions,
Proteins have nett negative and nett positive charges such as ethylene glycol (10}40% v/v) or detergents
at pH values above and below their pI values. This (1}4% w/v). The detergents should be either non-
decreases the risk of aggregation, which ultimately ionic or zwitterionic, so as to be compatible with the
focusing process; in addition, they should preferably
be transparent at 280 nm so as to minimize interfer-
ence with protein detection. There are some simple
ways to reduce protein interaction and precipitation:
one is to use dilute protein solutions (aggregation is
proportional to protein concentration); the other is to
increase the CA buffer concentration up to 4%, since
this leads to an increase in total ionic strength. Conti
et al. (1997) have investigated the use of a large
number of solubilizers and proposed mixtures of mild
agents capable of fully preserving the three-dimen-
sional structure and full activity of biomolecules.
Among the solubilizing agents are non-detergent sul-
fobetaines, taurine, Good’s buffers such as bicine and
Figure 2 Calibration graph for pI determination using a set of CAPS, polyols, such as sucrose, sorbose, sorbitol and
marker proteins. The markers (open squares) are ribonuclease mixtures thereof.
A (pI 9.45); carbonic anhydrase (pI 5.90);
-lactoglobulin (pI 5.1)
and unsulfated cholecystikinin flanking peptide (pI 2.75). The four
solid squares represent four unknown proteins whose pl s have The Detection System
been determined by linear interpolation in the calibration graph.
(Reproduced from Chen SM and Wiktorowicz JE (1992) Analyti- As discussed above, the standard absorption
cal Biochemistry 20: 84}98 by permission.) wavelength for detection in IEF is 280 nm, the typical
1212 II / ELECTROPHORESIS / Capillary Isoelectric Focusing
absorption maximum of proteins. As an alternative
method a universal concentration gradient imaging
system not requiring a mobilization step can be used.
This system is based on the Schlieren shadowgraph
methods and uses a He}Ne laser to probe the capil-
lary content and a charge-couple device (CCD) as
a detector. The capillary has to be square, not covered
by polymide and rather short (4 cm). Focusing is
completed in 2 min, resolution is of the order of
0.02 pH units and the mass detection limit appears to
be of the order of the picomole level. As an alterna-
tive, a whole column UV absorbance detection con-
sisting in transporting the entire column past the
detector (the capillary, of course, has to be UV trans-
parent) can be used. In this set-up the optical conRg-
uration consists of a 5 mW argon ion laser (with
either a 496.5 or a 514.5 nm lasing line) probing the
entire length of a 4 cm long, square glass capillary
and using as a detector a 1204-pixel CCD camera.
Focusing and imaging are usually completed in
2}4 min. The problem is that only coloured proteins
(those that absorb in this wavelength region) can be
efRciently detected (e.g. haemoglobin, myoglobin and
cytochrome-c). Finally, a mass spectrometer has been
proposed as a detector after CIEF: it should be
noted that the technique becomes two-dimensional,
since proteins are then mapped by both charge and
mass.
Focusing in Dynamically
Coated Capillaries
Three groups have reported approaches on the possi-
bility of focusing in dynamically coated capillaries.
The techniques will be reviewed below.
Dynamic Coating with Methylcellulose
Rather than completely eliminating EOF one might
try to reduce it to such an extent as to allow attain-
ment of steady-state conditions; from there on, the
bulk Sow would keep the ‘arrested’ stack moving past
the detection window. This approach would then
obviate the need for performing salt, vacuum or hy-
drodynamic mobilization; focusing and elution being
accomplished in one step. A simple way for modula-
ting EOF is to add viscous polymer solutions. The
capillary can be conditioned with 0.1% methyl cellu-
lose, mixed with sample and CA buffers and used to
Rll the entire column. It has been found that by
increasing the anolyte concentration to 25 mM phos-
phoric acid also allows the detection of acidic pro-
teins. Finally, it has been assessed how deleterious
different amounts of salt (NaCl) present in the sample
would be to the focusing process. As little as 10 mM
of NaCl in the sample sufRces to entirely destroy the
separation so salt in the analyte must be kept below
this level.
Dynamic Coating with Hydroxypropyl
Methylcellulose
In another system, the dynamic agent used for partial
coating is hydroxypropyl methylcellulose (HPMC).
In this approach, some interesting variants have been
adopted. A new capillary is Rrst rinsed for 20 min
with 1 M NaOH and then for 10 min with 0.1 M
NaOH containing 0.3% HPMC. It is during this Rnal
washing that conditioning of the capillary and partial
coating with HPMC occurs. This etching procedure
(with 1 M NaOH), followed by a short renewal of the
dynamic coating (0.3% HPMC in 0.1 M NaOH) is
shown to provide data of the highest reproducibility.
The sample proteins are dissolved in 2.5% Ampho-
line solution, without any addition of HPMC. The
anolyte is the standard 10 mM phosphoric acid solu-
tion, whereas the catholyte consists of 20 mM NaOH
in the presence of 0.1% HPMC. The sample is intro-
duced as a plug, occupying only 10}50% of the
capillary length at the anodic side, the remaining
being Rlled with catholyte. Since the entire stack of
proteins will eventually be displaced towards the
cathode by the EOF, this initial sample plug distribu-
tion allows more time to reach a good focusing pat-
tern prior to sample passage in front of the detector.
Dynamic Coating with Adsorbed Surfactants or
Polymers
In another variant, reduction of EOF via derivatiz-
ation of capillaries with a hydrophobic coating
(octadecylsilane) followed by adsorption of either
a surfactant (Brij 35, PF-108) or a hydrophilic
polymer (e.g. polyvinyl alcohol, polyvinyl pyrro-
lidone, methylcellulose) has been proposed. The pro-
cedure is as follows: the capillary is Rrst treated with
1 M NaOH for 30 min, followed by several washings
with deionized water and methanol (30 min each).
The residual methanol is evaporated in an oven at
903C for 2 h, while Sushing the capillary with a
stream of nitrogen at 400 kPa. While still in the oven
at 903C, a solution of octadecyltrichlorosilane in
50% toluene is Sushed through the capillary for 6 h.
After silylation, the capillary is rinsed for 20 min with
methanol and then with water for 30 min. Surfactant
solutions (in general 0.4%) are pumped continuously
through the capillary for an additional 6 h in order to
complete the coating process. Coating via adsorption
of detergent (or methylcellulose 4000) is shown to
reduce the EOF of the native, untreated capillary, to
approximately 3d5% of the original value but the
 II / ELECTROPHORESIS / Capillary Isoelectric Focusing 1213

residual EOF still allows adequate Sow to obviate the Ampholine is admixed with an equimolar mixture of
need for a separate mobilization step. Based on the ‘separators’, namely 0.2 M
-alanine and 0.2 M 6-
resolution of haemoglobin variants proteins that var- aminocaproic acid, which Satten the pH gradient in
ied 0.03 pH units in isoelectric point were resolvable. the focusing region of the three major components.
Figure 4 shows this separation obtained by CIEF. The
method is simple, can unambiguously detect any
Sample Preconcentration Systems
Typical preconcentration steps commonly used in
biochemical analysis (especially for macromolecules)
such as lyophilization, ultraRltration, partition be-
tween two polymer aqueous phases, osmotic removal
of water and chromatographic adsorption-desorp-
tion, will entail large losses when the sample volume
is 1}10
L or less, as is customary in capillary zone
electrophoresis (CZE). A number of electrophoretic
methods for concentrating biopolymers (especially
peptides and proteins), while partially depleting them
of strong electrolytes (often a problem in all IEF
procedures), have been described. In practice the
whole electrophoresis tube is Rlled with the sample
solution to be concentrated and then the sample is
allowed to migrate against the end of the tube where
a gradient of conductivity or viscosity exists and is
arranged in such a way as to continuously slow down
sample electrophoretic migration. The sample will
Rnally collect in a narrow zone of the tube (typically
0.2}0.5 mm in width). A 400}1000-fold concentra-
tion is obtained when a 200 mm long tube is Rlled
completely with the sample and still more if an elec-
trode vessel is also loaded with sample. As an alterna-
tive, an on-line isotachophoretic (ITP) concentration
process of very large injection volumes prior to CZE
analysis can be adopted. Sample volumes up to 25
L
can be concentrated by this system. As concentrating
large volumes would take a relatively long time, de-
pending on the migration path length, a system of
coupling a narrower bore to a larger bore capillary is
generally utilized in order to speed up the process.
Finally, after the ITP concentration step the sample
can be analysed by CZE via a T-junction connected to
another electrolyte reservoir (i.e. one has to resort to
a three-pole column).

Some Application of Examples

A vast body of applications already exists; a number
of them can be found in the reviews listed in the
Further Reading section. We will consider some se-
lected examples here. By using umbilical cord blood
which contains only three major haemoglobin (Hb)
components (Hb F, Hb A and Hb F acetylated, Fac) it
is possible to perform thalassaemia screening pro-
vided a good separation is obtained between Hb
A and Hb Fac, which have minute differences in pI
values. In order to improve the separation the pH 6}8
Figure 4 Separation of Hb F, A and Fac by capillary IEF. Back-
ground electrolyte: 5% Ampholine, pH 6}8, added with 0.5%
TEMED (panel A) and additionally with 3% short-chain polyac-
rylamide and 50 mM }-Ala (panel B). Anolyte: 20 mM H3PO4;
catholyte: 40 mM NaOH. Sample loading: by pressure, for 60 s.
Focusing run: 20 kV constant at 7
A (initial) to 1
A (final cur-
rent), 203C. Capillary: coated with poly(AAEE), 25
m internal
diameter, 23.6/19.1 total/effective length. Mobilization conditions:
with 200 mM NaCl added to anolyte, 22 kV. Detection at 415 nm.
(Reproduced from Conti M, Gelfi C and Righetti PG (1995) Elec-
trophoresis 16: 1485}1491, by permission.)
1214 II / ELECTROPHORESIS / Capillary Isoelectric Focusing
thalassaemic condition and can be easily performed
on a routine basis in a neonatal unit. Using the same
principle Conti et al. have also attempted CIEF separ-
ation of Hb A from Hb A1c (the glycated form of Hb
A), the latter component being of diagnostic value for
the long-term control of diabetic patients (glucose
Figure 5 Separation of Hb A from A1c by capillary IEF in the
absence (A) and in presence (B) of 3% short-chain polyacrylam-
ide and an equimolar mixture of ‘separators’, 0.33 M }-Ala and
0.33 M 6-amino caproic acid. Background electrolyte: 5% Am-
pholine, pH 6}8, added with 0.5% TEMED. Anolyte: 20 mM
H3PO4; catholyte: 40 mM NaOH. Sample loading: by pressure, for
60 s. Focusing run: 20 kV constant at 7
A (initial) to 1
A (final
current), 203C. Capillary: coated, 25
m internal diameter,
23.6/19.1 total/effective length. Mobilization conditions: with
200 mM NaCl added to anolyte, 22 kV. Detection at 415 nm.
(Reproduced from Conti M, Gelfi C, Bianchi-Bosisio A and
Righetti PG (1996) Electrophoresis 17: 1590}1596, by per-
mission.)
Figure 6 CIEF of a mouse monoclonal antibody using pressure
mobilization. Focusing for 2 min at 10 kV, followed by mobilization
at low pressure (0.5 psi) at (A) 10 kV and (B) 20 kV. Concentra-
tion of the marker proteins: 50 ng
L\1; concentration of desalted
antibody: 0.5
g
L\1. Ampholyte solution: 4% Pharmalyte, pH
3}10, 1% TEMED in 0.8% methyl cellulose. Anolyte: 10 mM
H3PO4; catholyte: 20 mM NaOH. (Reproduced from Schwer
C (1995) Electrophoresis 16: 2121}2126, by permission.)
binds irreversibly to Hb molecules; the percentage of
Hb A1c varies with the blood glucose concentration to
which red blood cells have been exposed during their
circulating lifetime, see Figure 5). A good separation
of monoclonal antibodies is shown in Figure 6: as
mobilization was obtained by pressure under voltage
it shows the importance of working under high volt-
age during this step.
Further Reading
Chiari M, Nesi M and Righetti PG (1995) Surface modiRca-
tion of silica walls: a review of different methodologies. In
Righetti PG (ed.) Capillary Electrophoresis in Analytical
Biotechnology. Boca Raton: CRC Press. pp. 1}36.
 II / ELECTROPHORESIS / Capillary Isotachophoresis
Conti M, Galassi M, Bossi A and Righetti PG (1997)
Capillary isoelectric focusing: the problem of protein
solubility. Journal of Chromatography A 757: 237}245.
Liu X, Sosic Z and Krull IS (1996) Capillary isoelectric
focusing as a tool in the examination of antibodies,
peptides and proteins of pharmaceutical interest. Jour-
nal of Chromatography A 735: 165}190.
Pritchett TJ (1996) Capillary isoelectric focusing of pro-
teins. Electrophoresis 17: 1195}1201.
Righetti PG (1983) Isoelectric Focusing: Theory, Methodo-
logy and Applications, Amsterdam: Elsevier.
Righetti PG (1990) Immobilized pH Gradients: Theory and
Methodology. Amsterdam: Elsevier.
Righetti PG and Bossi A (1998) Isoelectric focusing of
proteins and peptides in gel slabs and in capillaries.
Analytica Chimica Acta 372: 1}19.
Righetti PG, GelR C and Conti M (1997) Current trends in
capillary isoelectric focusing of proteins. Journal of
Chromatography B 699: 91}104.
Capillary Isotachophoresis
J. SaH deckaH and J. PolonskyH , Faculty of Chemical
Technology, Bratislava, Slovak Republic
Copyright ^ 2000 Academic Press
Introduction
Isotachophoresis (ITP) is one of the fundamental elec-
trophoretic separation techniques, where charged
constituents are separated in an electric Reld due to
their differences in their electrophoretic mobilities.
The moving boundary electrophoretic experiments
and theoretical developments were the forerunners of
isotachophoresis. Even by 1923, Kendall and Critten-
den had described separation of some metals and
acids by } as they called it } the ‘ion migration
method’, which was in fact isotachophoresis. They
concluded that ion concentrations were in accordance
with the Kohlrausch regulation function. In 1942,
Martin did his Rrst experiments on what he called
‘displacement electrophoresis’ as an analogue of dis-
placement chromatography. In 1963, Everaerts and
Martin started their work on isotachophoresis. Up to
1970 several names had been used for what Kendall
had called the ion migration method: these included
the ‘moving boundary method’, ‘displacement elec-
trophoresis’, ‘steady-state stacking’, and ‘ionophor-
esis’. In 1970, Haglund introduced a name, based on
the characteristic feature of the electrophoretic tech-
nique, namely the equal velocity of the sample zones
in the steady state: isotachophoresis (ITP). The basic
theory and early development in the Reld of iso-
1215
Righetti PG, Bossi A and GelR C (1997) Capillary iso-
electric focusing and isoelectric buffers: an evolving
scenario. Journal of Capillary Electrophoresis 4:
47}59.
Rodriguez-Diaz R, Wehr T and Zhu M (1997)
Capillary isoelectric focusing. Electrophoresis 18:
2134d2144.
Steinmann L, Mosher RA and Thormann W (1996) Char-
acterization and impact of the temporal behaviour of
the electroosmotic Sow in capillary isoelectric focusing
with electroosmotic zone displacement. Journal of
Chromatography A 756: 219d232.
Strege MA and Lagu L (1997) Capillary zone electrophor-
esis and isoelectric focusing of biotechnology-derived
proteins. Electrophoresis 18: 2343d2352.
Taverna M, Tran NT, Merry T, Horvath E and Ferrier
D (1998) Electrophoretic methods for process monitor-
ing and the quality assessment of recombinant glycopro-
teins Electrophoresis 19: 2527d2594.
tachophoresis was described in 1976 by Everaerts in
his fundamental book. An outline of the development
of isotachophoresis is given in Table 1. Some ad-
vances in isotachophoresis are described in detail
below.
ITP in Closed Systems
Up to 1990, ITP was carried out in commercial ap-
paratus in 200}500-
m i.d. narrow-bore plastic ca-
pillaries and with closed systems, i.e. no electroos-
motic Sow (EOF) occurred.
Basic Theory
Under the inSuence of an applied electric Reld, E,
ionic species will move towards the electrode with
a migration velocity, v, of:
v"m;E [1]
where m is the effective mobility of an ionic species.
The effective mobility depends on various factors, for
example, the ionic radius, shape and charge of the
ion, degree of dissociation, pH, dielectric constant
and viscosity of the solvent, and temperature.
Typically, ITP is performed with a constant current
and it is not possible to separate cations and anions in
the same run (unidirectional isotachophoresis).
It is characteristic of ITP that the sample to be
separated is injected between two different electrolyte
solutions. The Rrst solution (the leading electrolyte)
contains an ion (the leading ion) with the same charge
1216 II / ELECTROPHORESIS / Capillary Isotachophoresis

Table 1 Development of isotachophoresis

Year Development of: Attributed to:

1897 Regulation function Kohlrausch

1930 Moving boundary electrophoresis Tiselius
1942 Displacement electrophoresis Martin
1968 Capillary tube apparatus for isotachophoresis Verheggen, Everaerts
1970}1989 ITP in closed systems, 200}500
m i.d. narrow-bore plastic capillary with minimized EOF
Column-coupling ITP Everaerts
1970}1980 Thermometric, conductometric, potentiometric and UV detection Everaerts
1981 Refractometric detection Bresler
1981 Offline ITP}MS Kenndler
1983 Radiometric detection Kaniansky
1984 Fluorimetric detection Reijenga
1984 Amperometric detection Kaniansky
1985 Absorption spectra HanibalovaH
1989 ITP in open system, 100
m i.d. fused-silica capillary with EOF, online ITP}MS Udseth
1990 ITP in open system, 25}50
m i.d. fused-silica capillaries Thormann
1990 Online ITP}CZE (column coupling) Kaniansky
1991 Offline ITP}PIXE Hirokawa
1993 Bidirectional ITP Hirokawa
1995 Raman spectroscopic detection Walker

sign as that of the sample ions, but with an effective
mobility higher than that of the fastest moving
sample ion. The second solution (the terminating
electrolyte) contains an ion (the terminating ion) with
the same charge sign, but with an effective mobility
slower than that of the slowest moving sample ion.
The polarity of the electric Reld has to be such that the
leading ion migrates to the electrode that is placed on
the same side of the sample as the leading electrolyte.
After application of an electric Reld to the system,
each ionic species will have a different migration
velocity according to eqn [1] and hence the iso-
tachophoretic process starts.
The process of isotachophoresis may be divided
into two parts. In the Rrst part, the separation of the
ions proceeds and the migration velocity of the indi-
vidual ions in the mixed zones is different. In the
second part (in the steady state) the ions have already
separated from one other and all move with the same
velocity, v:
v"mLEL"miEi"mTET [2]
where L is the leading ion, i is the ith ion and T is the
terminating ion.
A schematic representation of the cationic and ani-
onic modes in ITP experiments without EOF is given
in Figure 1(A,B). As the ionic species are arranged in
order of decreasing effective mobilities, the electric
Reld strengths increase on the terminating ion side.
The increase in the electric Reld strength in the
consecutive zones induces the zone-sharpening effect.
When a zone has attained the steady state, the bound-
ary will not broaden further, which is in contrast to
zone electrophoresis, where the peaks are broad ow-
ing to adsorption and diffusion. This effect can easily
be explained. If an ion diffuses into a preceding zone,
where the electric Reld strength is lower than the
value that corresponds to its velocity, its velocity will
decrease according to eqn [2], and it will be over-
taken by its own zone. If an ion diffuses into a zone
with a higher electric Reld strength, then it will obtain
a higher migration velocity according to eqn [2], until
it reaches its own zone.
It is characteristic for the steady state that the
concentration of each component is adjusted to the
value following from the Kohlrausch regulation func-
tion in the form:
mL#mR mi zL
ci"cL ; ; [3]
mL mi#mR zi
where R is the common counterion, Zi is the ionic
charge. In the steady state, the concentration Ci of the
ith ion is always adjusted to a certain value depending
only on the concentration of the leading electrolyte
CL and on the mobility of the ions i, L and R. From
the analytical point of view this is a very important
feature of ITP. It can be concluded that for a given set
of experimental conditions, the zone length is a direct
measure of the amount of an ion present in the zone.
Another important consequence of these properties is
the concentration effect of isotachophoresis. In fact,
a species more concentrated in the original sample is
diluted during the separation and, a species originally
too dilute is concentrated during the separation.
 II / ELECTROPHORESIS / Capillary Isotachophoresis 1217

Figure 1 Schematic representation of the two modes in unidirectional ITP experiments without EOF: (A) the cationic separation of
a mixture of cations A, B, and C with mA'mB'mC; (B) the anionic separation of a mixture of anions X, Y, and Z with mX'mY'mZ.
The four modes in unidirectional ITP with EOF: (C) the cationic cations as in (A); (D) the anionic anions as in (B); (E) the reversed
cationic; (F) the reversed anionic; (G) Schematic representation of the bidirectional ITP; separation of mixture of cations A, B and
C and anions X, Y and Z. L1"leading cation; T1"terminating cation; L2"leading anion; T2"terminating anion. Only steady state is
presented. S"sample inlet; D"detector position, L"leading ion, T"terminating ion, vEOF"velocity of EOF; vITP"iso-
tachophoretic velocity; QorP"net velocity; M"semipermeable membrane. For further explanation, see text.

In ITP, the response is usually recorded against zone, and is given by the relation:
time with a detector placed at the end of capillary
(Figure 2). hi!hL
rshi" [4]
The identity of a species is characterized by the hT!hL
effective mobility (or a quantity proportional to the
effective mobility). This is usually the response of the where hi is the step height of the compound, hL is the
universal detector. It is called the height (step height) step height of the leading ion and hT is the step height
or the relative height (relative step height, rsh) of the of the reference ion (usually the terminating ion)
1218 II / ELECTROPHORESIS / Capillary Isotachophoresis
Figure 2 Graphical representation of response R from univer-
sal detector [(B) linear; (C) differential] for the different anions A,
B, and C, moving in the steady state of an isotachophoretic
analysis (A). L"leading anion; T"terminating anion;
S"sample inlet; D"detector position; M"semipermeable
membrane. For further explanation, see text.
(Figure 2B). The values obtained in this way are then
compared with those of standard species measured
under the same experimental conditions.
The quantiRcation is in general simpliRed by differ-
entiating the signals and measuring the distance be-
tween the inSection point (Figure 2C). The zone
lengths li are directly proportional to the number of
ions (ni): li"Kini. The constant Ki depends on the
equipment and the current used. A universal calib-
ration constant (the response factor RF, eqn [5],
which is independent of the diameter of the capillary,
construction of the universal detector and driving
current using during detection, has been introduced.
For each component, the RF depends only on the
concentration of the leading electrolyte:
l;I
RF"
"z";F;Q
where l is the zone length (seconds), I is the driving
current (amps), "z" is the charge of the ion
(equiv mol\1), F is Faraday’s Number (coulombs
equiv\1) and Q the amount injected (mol).
Based on the mathematical models for iso-
tachophoresis described, computer programs have
been set up for calculation of the parameters of the
different zones. Unfortunately, only a few schemes
can be used for simulating capillary isotachophoresis
at realistic current densities without causing either
severe oscillations or unexpected program termina-
tion.
Online Coupling of ITP with CZE
(Column-Coupling Instrumentation)
Column-coupling instrumentation (see below) of the
separation unit for ITP as described by Everaerts has
been shown to be suitable for online coupling of ITP
and CZE. The extensive studies of Kaniansky and
MaraH k give a good impression of the potential of
combined ITP}CZE.
The online combination of ITP and CZE is a very
effective tool for increasing the separation capability
and sensitivity of CZE. It is characterized by iso-
tachophoresis in the Rrst capillary followed by online
transfer of the sample cut into the second capillary
where zone electrophoresis proceeds.
In principle, there are three ways of performing an
ITP}CZE combination technique as far as the electro-
lyte systems are concerned. The simplest way is to use
the terminating electrolyte as the background electro-
lyte (BGE) for CZE; the second possibility is to use
the leading electrolyte as BGE and the third possibil-
ity is to use a totally different BGE.
ITP has the advantage of much higher loading
volumes, e.g. microlitres instead of nanolitres in CE.
In addition, ITP is a concentration technique. The
combination of these features makes ITP, in principle,
an ideal technique for sample pretreatment. In
ITP}CZE, a 104-fold concentration increase can be
achieved, and this even for a component present in
a 105-fold excess of the matrix.
ITP in Open Systems
Since the early 1990s, commercial instruments for
CZE have been available generally with open-tubular
fused-silica capillaries with an inner diameter be-
tween 20 and 100
m, together with an on-column
detector placed towards one end of the capillary. As
this apparatus can be used for ITP it was of interest to
study the possibilities for ITP in open systems. If ITP
[5]
experiments are performed in open-tubular fused-
silica capillaries, the negative surface charge of un-
treated fused-silica causes an EOF towards the cath-
ode. This EOF will inSuence the ITP system and four
different modes can be observed. In Figure 1(C), the
cationic ITP mode is shown. The EOF will generally
act in the direction from the anode to the cathode and
as a result the cationic ITP system will be pushed
towards the cathode with a higher velocity compared
with cationic experiments in closed systems. In
 II / ELECTROPHORESIS / Capillary Isotachophoresis
Figure 1(D), the anionic ITP mode is shown. This
mode can be applied if the velocity of the leading ion
is greater than that of the EOF during the whole
experiment. Only in this case will anions with mobili-
ties slower than that of the EOF also migrate to the
anode according to the isotachophoretic condition.
The reversed cationic mode (Figure 1E) can be ap-
plied if there is a reversed EOF (e.g. using coated
capillaries or additives to the electrolyte) with a velo-
city greater than that of the cationic system. Here the
cathode must be placed at the sample inlet end and
the anode at the detector end. Although the ITP
separation takes place in the direction of the cathode,
there will be a net velocity of the ITP system in the
direction of the detector end and components will be
detected in a reversed order compared with a normal
cationic ITP system. In Figure 1(F), the reversed
anionic mode is presented. Here the anode is
placed at the sample inlet end, the cathode at the
detector end and components will be detected in
a reversed order compared with a normal anionic ITP
system.
As the velocity of the EOF is extremely important
in the migration behaviour of ITP systems, much
effort must be put into controlling EOF. The velocity
of the EOF strongly depends on the choice of the
leading and terminating electrolyte and it also varies
with the composition of the sample. Moreover, the
velocity of the EOF continuously changes during the
analysis and is Rrst determined by the composition of
the leading electrolyte and Rnally by that of the termi-
nating electrolyte. Varying EOF velocities cause irre-
producible migration times and zone length and the
results of quantitation are erroneous. The addition of
methylhydroxyethylcellulose to the electrolytes and
sample largely suppresses the EOF in order to im-
prove quantitation. In spite of the addition of methyl-
hydroxyethylcellulose, the reproducibility of the zone
lengths with time is poor, and an internal standard is,
therefore, needed. Hence the reproducibility in ITP
quantitative analysis in open systems is a problem
similar to that in electrophoresis. Generally, closed
systems are to be preferred to open systems for quant-
itative analysis.
The presence of an EOF, however, facilitates the
development of bidirectional ITP for the simulta-
neous determination of anionic and cationic compo-
nents. In bidirectional ITP, the leading electrolyte for
cations must be simultaneously the terminating elec-
trolyte for anions, and vice versa the leading electro-
lyte for anions must be the terminating electrolyte for
cations. That is, the counterions (cations) coexisting
with the leading anions play the role of the termina-
ting cation, and the counterions (anions) coexisting
with the leading cations play the role of the termina-
1219
ting anion. In a fused silica capillary in the presence of
a cathodic EOF, cationic sample trains can be detec-
ted with a detector placed towards the cathodic end
of the capillary. However, anionic species can be
detected only at pH'6. At pH'6, the velocity of
the EOF is greater than that of the anionic ITP system
and hence the anions migrate slowest since they are
attracted to the anode, but are still carried by the EOF
towards the cathode (Figure 1G).
Instrumentation for ITP
Separation Capillary
The actual separation takes place in a PTFE (polytet-
raSuoroethylene) or a silica capillary. The separation
capacity can be increased by extending the length of
the capillary, but the analysis time and the maximum
voltage required also increases. From the instru-
mental point of view, the column-coupling system
(Figure 3) frequently used today has led to signiRcant
progress. It consists of a pre-separation unit with
a capillary of larger diameter (e.g. 0.8 mm) equipped
with the detector and bifurcation block, to which an
analytical capillary of small diameter (e.g. 0.3 mm) is
connected. At the beginning of the analysis, the driv-
ing current passes through the pre-separation capil-
lary only. The detection system in the Rrst capillary is
employed to evaluate analysis. In addition, it provides
the information necessary to control the transfer of
the analytes into the second capillary and the removal
of the sample constituents which are led out of the
separation compartment after the Rrst stage. At
a suitable moment, the driving current is switched so
that it passes through the analytical capillary and thus
introduces the required sample zones into this capil-
lary where further separation takes place. Column
coupling enables use of different leading electrolytes
in the pre-separation and analytical capillaries, there-
by inSuencing the subsequent separation, separation
of mixtures containing components in ratios up to
1 : 1000 without increasing the voltage and without
prolonging the analysis time, and application of ITP
in combination with CZE.
Electrode Chamber, Electrodes and Power Supply
The capillary is connected on each side to an elec-
trode chamber provided with a platinum electrode. In
closed systems, the chamber, Rlled with the leading
electrolyte, is connected to the capillary via a
semipermeabile membrane. The terminator chamber
is connected via a multiway switching valve, which is
open in the course of the analysis. In open systems,
the ends of the capillary are placed in electrolyte
reservoirs (electrode chamber).
1220 II / ELECTROPHORESIS / Capillary Isotachophoresis
Figure 3 Column-coupling isotachophoretic system. EC"
electrode compartment; BB"bifurcation block; D"conductivity
detector; UV"UV detector; L"leading electrolyte; T"terminating
electrolyte; M"semipermeable membrane.
A high-voltage power supply capable of delivering
500
A at up to 20}30 kV d.c. is needed. The con-
stant current regulation of the power supply must be
extremely well designed.
Injection System
In closed systems, the sample can be introduced by
a microsyringe through a septum or by a multi-port
valve system. In open systems, the sample can be
vacuum-aspirated or loaded electrokinetically.
Detection
The Rrst universal online detector was the ther-
mocouple detector. Owing to its low sensitivity, the
thermocouple detector was replaced with universal
contact detectors, which sense the electrical resistance
or potential gradient in the zones. The disadvantage
of contact detectors is polarization of the sensing
electrodes. To solve this problem, a universal contact-
less high-frequency conductivity detector was pro-
posed in the 1970s. Detection based on differences in
the refractive index of various zones was introduced
in 1981. The disadvantage of this system was the
necessity of working with high electrolyte concentra-
tions, which resulted in slow analysis.
In 1991, McDonnell and Pawliszyn developed
a new refractive index detector for ITP consisting of a
He}Ne laser or a laser diode and photodiode position
sensor. The direction of the beam is deSected when it
passes through the refractive index gradient produced
by the sample zone. By using this detector, a few
nanomoles of sample can be detected. The develop-
ment of a selective UV-absorption detector for ITP
had been an important contribution to the develop-
ment of ITP in the 1970s. The UV detector is now
a common component of commercial apparatus. In
most cases only the wavelengths 254 and 280 nm
have been utilized for detection. Arlinger had shown
in 1974 that a UV detector could be applied as
a pseudo-universal detector. UV-absorbing counter-
ions were used, for which the molar absorption was
pH-dependent. As each zone has its own deRned pH
and concentration, the pH and concentration differ-
ence gave rise to an absorbance difference sufRciently
large to be detectable.
Sometimes it can be advantageous to use a UV-
absorbing spacer in order to make the detection of
consecutive zones of nonabsorbing ionic species pos-
sible. In some instances it is possible to detect bound-
aries between two consecutive non-UV-absorbing
zones because of the trace amounts of UV-absorbing
impurities which are present in most electrolytes and
which concentrate as markers between the separated
non-UV-absorbing zones. Great attention has been
paid to development of new selective detectors for
ITP, to facilitate the identiRcation of compounds in
the detected zones. Sensing of absorption spectra in
isotachophoretic zones is one of the possibilities.
Fluorimetric detection is a highly sensitive method. In
ITP, the equipment designed initially for the dual-
wavelength UV detection has been employed for
Suorimetric zone detection. Zones of Suorescing
compounds or of compounds quenching counterion
Suorescence can be detected.
In 1991, Hirokawa introduced a new speciRc de-
tection method for metal ions. He used an ofSine
combination of ITP and particle-induced X-ray emis-
sion (PIXE), which is a multi-elemental method with
high sensitivity. As the method is based on the charac-
teristic X-rays emitted by target elements, it has
a high speciRcity for the determination of the ele-
ments even if they are not separated. Radiometric
detection of compounds labelled with a radioactive
isotope is a speciRc method. Its principle is the detec-
tion of the radiation emitted from the labelled
 II / ELECTROPHORESIS / Capillary Isotachophoresis 1221

compound zone passing the window of Geiger}MuK l-

ler tube. Electrochemical detection, owing to its high
sensitivity and speciRcity, is widely used in liquid
chromatography. Its direct use in ITP is hindered by
the presence of the driving electric Reld. To minimize
disturbances due to the driving current, post-column
amperometric detection has been employed. The sep-
arated constituents are hydrodynamically transported
from the separation compartment into the detection
Figure 4 Capillary preparative isotachophoresis with a counter-
cell. The hydrodynamic transport causes the disper-
flow of leading electrolyte. EC"electrode compartment;
sion to increase, therefore, the resolving power of SC"separation capillary; D"detector, M"semipermeable
post-column detection is lower in comparison, with membrane; L"leading electrolyte; T"terminating electrolyte;
for example, the conductivity detector. However, this FL"counter flow of leading electrolyte.
disadvantage can be outweighed by its inherent selec-
tivity and/or sensitivity.
The online combination of ITP with mass spectro- lates the separation chamber from the electrode
metry was Rrst demonstrated in 1989. The ITP/MS compartments.
interface is based on electrospray ionization. Separ- In recycling electrophoresis, in order to increase the
ations were conducted in open-tubular untreated electric charge applied to the sample, the fraction
fused-silica capillaries. The interface requirement of from each channel are continuously reinjected into
strong electroosmotic Sow did not signiRcantly de- the inlet port of the separation chamber. This instru-
grade separations and both high sensitivity (limit of mentation allows a high throughput and complete
detection 10\9 mol L\1) and high resolution can be separation of the injected sample. Typical operation
obtained. Recently, Walker has demonstrated that is batchwise, in contrast to continuous free-Sow
a Rbreoptic Raman probe can be used to obtain isotachophoresis.
real-time intracapillary Raman spectra during ITP.
Even at 2;10\5 mol L\1 initial concentration,
Raman spectra were obtained at a good signal-to-
Future Developments
noise ratio. Isotachophoresis underwent major development in
the years 1970}1990. Over the last ten years CZE has
occupied the major part of both the theory and ap-
Preparative Procedures in plications of electrophoresis. Despite this, capillary
isotachophoresis has kept its position as a special
Isotachophoresis technique with unique features. Concentrating and
Capillary isotachophoretic analysers can be used for
preparative purpose in a discontinuous arrangement
only. Once the separation has been performed, the
analysis is discontinued and the analysed compound
zone is isolated by using a microsyringe, a specially
designed fractionating valve placed at the end of the
separation capillary or a counterSow of leading elec-
trolyte (Figure 4).
Continuous free-Sow isotachophoresis (Figure 5)
was developed to fractionate large-scale samples con-
tinuously. The separation Reld of continuous free-
Sow isotachophoresis is typically a thin Rlm of Suid
Sowing between two parallel plates. An electric Reld
is applied perpendicular to the Sow direction. The
leading and terminating electrolytes and the sample
solution are continuously supplied with a multifold
peristaltic pump into one end of the electrophoretic
chamber and are collected with a multifold pump at
Figure 5 Continuous free-flow isotachophoresis. EC"elec-
the other. The leading and terminating electrolytes trode compartment; M"semipermeable membrane; L"leading
used for the electrode compartments circulate by electrolyte; T"terminating electrolyte; sample"mixture of
pumps during migration. A dialysis membrane iso- A and B.
1222 II / ELECTROPHORESIS / Cellulose Acetate
zone sharpening make it possible to obtain, in par-
ticular cases, much better results than when using
CZE. Most promising is the combination of ITP with
CZE where ITP serves as a preconcentration and
pre-separation step for analysis of samples with com-
plex matrices. Unfortunately, there is only one
manual ITP}CZE system still commercially available.
Further Reading
Boc\ ek P, Deml M, Gebauer P and DolnmH k V (1988) Analyti-
cal Isotachophoresis, pp. 5}237. Weinheim: VCH.
Boc\ ek P, Gebauer P, DolnmH k V and Foret F (1985) Recent
developments in isotachophoresis. Journal of
Chromatography 334: 157}195.
Everaerts FM, Beckers JL and Verheggen ThPEM (1976)
Isotachophoresis. Theory, Instrumentation and Applica-
Cellulose Acetate
G. Destro-Bisol, University ‘La Sapienza’,
Rome, Italy
M. Dobosz and V. Pascali, Catholic University,
Rome, Italy
Copyright ^ 2000 Academic Press
The introduction of zone electrophoresis, pioneered
by Konig in 1939, played a crucial role in the progress
of electrokinetic separations. With this technique,
molecules migrate as zones with sharp boundaries in
a supporting medium immersed in a buffer solution
under the application of an electric Reld. Zone elec-
trophoresis was quickly found to be superior in per-
formance to Tiselius’s original technique of moving
boundary electrophoresis and replaced it entirely } to
be superseded in turn by displacement electrophoresis
and isoelectric focusing (IEF). Interestingly, the term
‘zone electrophoresis’ was Rrst suggested by Tiselius
himself.
Kohn Rrst used cellulose acetate (CA) as a support-
ing medium for zone electrophoresis in 1957, as a su-
perior substitute for plain Rlter paper. Since then, CA
has been used in many electrophoretic protocols, for
both research and clinical investigations (Table 1).
Nowadays CA electrophoresis is a widespread
technique.
In this article we explain what CA is and why it is
used in electrophoresis. This is followed by a brief
overview of the uses of CA in various electrophoretic
contexts. Finally, some recent and innovative applica-
tions of CA in electrophoretic protocols are dis-
cussed.
tions, Journal of Chromatography Library, vol. 6, pp.
7}282. Amsterdam: Elsevier.
Gebauer P and Boc\ ek P (1997) Recent application and
developments of capillary isotachophoresis. Elec-
trophoresis 18: 2154}2161.
Hirokawa T, Watanabe K, Yokota Y and Kiso Y (1993)
Bidirectional isotachophoresis. Journal of Chromatog-
raphy 633: 251}259.
Hjalmarsson SG and Baldesten A (1981) A critical review
of capillary isotachophoresis. CRC Critical Reviews in
Analytical Chemistry 11: 261}352.
Kaniansky D and MaraH k J (1990) On-line coupling of
capillary isotachophoresis with capillary zone elec-
trophoresis. Journal of Chromatography 498: 191}204.
Thormann W (1990) Isotachophoresis in open-tubular
fused-silica capillaries. Impact of electroosmosis on zone
formation and displacement. Journal of Chromatogra-
phy 516: 211}217.
General Concepts
Preparation of CA
CA sheets employed in electrophoresis are made of
a molecular matrix, similar in structure to a sponge
but a thousand times smaller. This matrix is obtained
by letting acetic anhydride react with cellulose and
dissolving the product in an organic solvent, that can
evaporate quickly. After letting the solvent evaporate
in closely-controlled conditions of temperature and
humidity, a highly permeable matrix is obtained with
a uniformly distributed microporosity. The spatial
volume of the pores may account for 80% of the
total matrix size, ensuring ideal permeation by any
Table 1 Historical sequence of main applications of CA to elec-
trophoretic protocols in different areas of research and clinical
investigations
Year Application
1957 CA is used as an electrophoretic support (Kohn)
1971 Application to conventional electrophoresis of white cell
and red cell enzymes (Meera Khan)
1975 Application to isoelectric focusing of alpha-1-antitrypsin
in human serum and 6}phosphogluconate dehydro-
genase (Harada)
1984 Application to counterflow affinity isotachophoresis of
antigens in biological fluids with low protein contents
(Abelev and Karamova)
1992 Introduction of CA for protein transfer from polyacrylam-
ide gels
1993 Introduction of protocols for reusing CA

electrolytic solution. When shaped into gel sheets CA
has better resistance to the dehydration involved in
the dissipation of heat and is more easily handled.
Thus, pre-gelled CA membranes (also referred to as
Cellogel2+) are the Rrst choice of support for many
electrophoretic applications. For better handling,
some commercial versions of Cellogel TM come welded
to an inert support of polyester plastic (Mylar2+).
These commercial forms of CA pass practically un-
changed through the entire separation}staining}de-
staining cycle of a classical electrophoretic experi-
ment.
There are several major factors accounting for the
versatile electrophoretic properties of CA: (1) the
cellulose chain length, which ranges from a few to
millions of individual molecules; (2) the degree of
acetylation (from 0.1% to 44%); (3) the pore size
(between 50 A> and 10
m), the random pore distribu-
tion and the volume of the pores compared with the
solid matrix (20% to 80%). The spatial coiling of
cellulose molecules, the type and concentration of
wetting agents and the presence of residual con-
taminants may also be important factors.
CA as an Electrophoretic Medium
Migration of molecules through the CA matrix de-
pends mainly on the nett charge on the molecule, the
buffer pH and ionic strength and the intensity of the
electric Reld. The difference in surface nett charge
between the molecular species in a sample to be
separated is perhaps the most important point to
consider. Proteins are amphoteric, like their constitu-
ent amino acids, and they may be charged positively
or negatively depending on the pH of the solvent
medium (the buffer solution, in an electrophoretic
experiment).
In gel electrophoresis a sieving effect may affect the
separation, depending on the critical relationship be-
tween the spatial shape of a protein species and the
pore size of the matrix medium. Because of the ex-
tremely large cellulose matrix pores, the mobility of
proteins in CA electrophoresis is a direct function of
their surface net charge, whereas molecular weight
and shape are less important. For example, the hu-
man heavy
-2 macroglobulin (Mr: 1 000 000; pI 5.9)
moves faster than the much lighter haptoglobin (Mr
100,000; pI 6.1) in alkaline buffer solutions.
As in most electrophoretic protocols, to improve
a CA separation the ideal buffer pH and ionic
strength, strip temperature, voltage, current, elec-
troosmosis and time of separation should be selected.
The optimal ionic strength is between 0.01 and 0.1
(mequiv L\1). Although mobility is theoretically en-
hanced at high temperature, proteins are easily heat
II / ELECTROPHORESIS / Cellulose Acetate 1223
denatured so the separation temperatures should be
kept below 503C. Moreover, since CA electrophoresis
is traditionally carried out with no cooling, separ-
ation voltages should not exceed 500 V (60 V per
linear centimetre in gel strips), and the current should
be adjusted to less than 2.5 mA cm\2. CA contains
polar groups } hydroxy (OH\) and acetyl
(CH3COO\) radicals } that become charged at the
pH system and move towards the anode through the
cellulose matrix. This produces a counter-reaction,
displacing buffer toward the cathode and interfering
with the separation of the molecules of interest (en-
doosmosis). Prolonged separation times may thus
lead to the creation of artefacts due to the combined
effects of heat, buffer turbulence and the counter-
diffusion of molecules. Running times should be al-
tered accordingly.
A few fundamental properties make CA elec-
trophoresis notably superior to electrophoresis using
Rlter paper: (1) the CA matrix is homogeneous,
microporous and chemically pure, reducing molecu-
lar adsorption to a minimum; (2) instant heat dissi-
pation occurs in the matrix, which does not need to
be cooled; (3) the amount of protein needed is very
small (1 mg or less); (4) the inherent buffering}
staining}destaining system is very simple; (5) stained
CA strips have no background; (6) the standard elec-
trophoretic apparatus required is simple and inexpen-
sive (Figure 1).
For most purposes } especially for routine clinical
investigations } small-scale CA electrophoresis (with
membranes (10 cm long) is widely used (Figures 2
and 3). Larger scale membranes (usually 20 cm long)
suit a variety of research analytical purposes and
micropreparative applications.
CA in Electrophoretic Protocols
Conventional Electrophoresis
CA was originally introduced as a classical support
for analytical zone electrophoresis but found a much
broader range of applications. Essentially, it can now
be used for both analytical and preparative purposes.
Preparative applications exploit the speed of CA sep-
arations, the absence of molecular interaction, and
the easy recovery of biologically active substances
from the matrix.
CA is popular in clinical laboratories in which
some well-established routine analyses are per-
formed, e.g. for haemoglobin, serum proteins, lipo-
proteins and lactate dehydrogenase. Isoforms of
many enzymes and proteins from different tissues
come out very clear-cut on CA } a fact that is (or has
been) of particular interest for anthropogenetic and
1224 II / ELECTROPHORESIS / Cellulose Acetate

Figure 1 Description of a universal electrophoretic apparatus for CA electrophoresis (redrawn and modified from Kohn, 1957). The
CA strips (11) are supported at each end by the shoulder pieces (4) and when taut are just clear of the top edge of the centre partition (10).
The top edge of this centre partition is formed as a row of pyramids (9) which support the strip should it tend to sag. When using long
strips, strip supports (6) may be fitted to the labyrinth partitions (7) that form the connections between the buffer compartments (5)
and electrode compartments (8). Filter paper wicks (3) connect the CA strips to the buffer compartments. The internal sides of the
tank are stepped all round (2) as an aid to buffer level checking. The lid (12) fits in a recess (1) moulded all round the tank.

forensic purposes and for the biochemical character-
ization and classiRcation of various pathogenic
microorganisms such as Leishmania and Trypano-
soma species.
In addition to one-dimensional electrophoretic
methods, two-dimensional CA electrophoretic proto-
cols are also available. A summary of important ap-
plications is given in Table 2.
Detection and Quantitation
Any protein stain can be used with CA, provided that
the solution does not contain a cellulose solvent.
Aqueous staining solutions are preferred to alcoholic
ones, since with the latter strips tend to shrink and
curl unless they are passed through an aqueous bath.
Staining solutions for CA are less concentrated than
those used in Rlter paper electrophoresis, and they
can be repeatedly used with no appreciable loss of
sensitivity.
A wide range of analytical applications can be
listed with an impressive variety of fully compatible
staining methods, including Coomassie blue brilliant,
Ponceau red, Nigrosin, Schiff, gold and silver stain,
different types of immuno-staining, and many differ-
ent types of enzyme speciRc staining. A 5% (w/w)
aqueous solution of acetic acid is a universal washing
solution for reducing the background.
The simplest way of evaluating the results is by
visual inspection of stained strips, which should be
carried out against a strong light source to improve
the assessment of the separation pattern.

Figure 2 Electrophoretic separation of human haemoglobin variants A, C and S. Ponceau red staining was used to visualize
hemoglobin bands, and the anode was on top.

Figure 3 Routine clinical electrophoretic separations on CA:
(A) serum proteins; (B) lipoproteins; (C) Lactate dehydrogenase
isoenzymes. Samples were obtained from healthy patients.
Quantitative determinations can be carried out by
elution or by scanning of the stained strips. Once
stained, protein bands can be easily eluted from the
membrane by an appropriate buffer system (a classi-
cal system is Tris (2-amino-2-hydroxymethyl-
propane-1,3}diol) or Barbitone elution buffer over
Ponceau red stained bands). Alternatively, a solvent
(e.g. chloroform}ethanol 9 : 1 v/v) can be used to
dissolve the membrane and recover the protein
Table 2 Some recent applications of CA electrophoresis
Year Application
1994 Introduction of thermocooling apparatus for CA IEF
Sequential electrophoresis, with detection of 21 different
alleles in ESD-2 locus in Drosophila buzzatii
1995 Improved separation of apolipoproteins by use of surfac-
tant Tween 20
1996 Rapid screening of biochemical loci of rat
Highly sensitive detection of urinary proteins using col-
loidal silver staining
1997 Detection of superoxide dismutase isozymes to distin-
guish between tsetse blood meals of human and non-
human origin
CA electrophoresis used as the method of choice for
alpha-thalassaemia screening
IEF on CA applied to the analysis of microheterogeneity
of immunoglobulins and serum protein fraction
II / ELECTROPHORESIS / Cellulose Acetate 1225
of interest. To enhance the recovery efRciency, gelled
CA blocks (about 0.5 cm thick, instead of much thin-
ner 0.5 mm supports) can be used.
Scanning is preferred to elution for routine clinical
applications. To reduce background and increase sen-
sitivity, CA strips should be cleared prior to scanning.
As with Rlter paper it is important to use oil with the
same refractive index as the support. CA strips
cleared with oil may be returned to their original dry
state by using a solvent such as ether. By contrast,
swelling agents such as acetic acid and dioxan used in
conjunction with heat treatment, permanently clear
CA.
Isoelectric Focusing
CA has ideal features to suit IEF separations. CA is
virtually a non-sieving matrix enabling a quasi-free
fractionation of macromolecules according to their
respective isoelectric points ( pI, the pH at which
there occurs an equal number of negative and positive
surface charges). CA is easily soaked with very small
amounts of carrier ampholyte species, allowing them
to be eluted in due course with no damage to
stained}destained proteins; this in turn allows den-
sitometry measurements and storage.
Unfortunately, the combined effect of CA elec-
troosmotic Sow and the low ionic strength of com-
mercial ampholines can seriously impair the resulting
separation of proteins at their isoelectric points. To
overcome these drawbacks, CA has been variously
treated with surface active agents or with methylating
agents. Such treatments can partly } if not wholly
} reduce the osmotic Sow. Also, a high concentration
of carrier ampholytes should be used to cover broad
pH ranges (8% v/v instead of the customary 2% v/v)
and electrolyte additives at low concentration (such
as 0.2 M lysine and 0.2 M acetic acid) should help
stabilize narrow pH intervals. Untreated CA strips
give better results when 5% -mercaptoethanol and
5 M urea are used as stabilizing agents.
Alternative strategies to circumvent electroosmo-
sis, which differ in effectiveness, involve shortening
the inter-electrode distance or using ‘chemical space-
rs’ to Satten the pH gradients at the appropriate
segment of separation. These devices may help to
create high Reld strengths with low voltages. Re-
cently, thermoelectric cooling has been used to stabil-
ize CA IEF gradients.
Counter]ow Af\nity Isotachophoresis
Isotachophoresis or ‘displacement’ electrophoresis
permits simultaneous concentration and effective sep-
aration of surface-charged substances, including bio-
logical macromolecules. With this analytical method,
1226 II / ELECTROPHORESIS / Cellulose Acetate
proteins are stacked as closely spaced, narrow bands
between the ‘leading’ and the ‘trailing’ ions. Iso-
tachophoresis on CA gels takes advantage of the
absence of sieve effect in this matrix to study sets of
interacting biological macromolecules, such as anti-
gen/antibody and glycoprotein/lectin systems. How-
ever, electroosmosis once again interferes with this
application. Abelev and Karamova were able to over-
come this drawback by demonstrating that the cath-
odic counterSow, combined with the constant Sow of
liquid through the membrane, stabilizes separations.
The counterSow may be also used as a ‘conveyer belt’
to move immunoreagents through antigens or anti-
bodies immobilized onto the membrane. Abelev and
Karamova used a discontinuous buffer system, in
which the two buffers have the same cation and differ
in the anion species (chloride as the leading ion and
-alanine as the trailing ion). Under these conditions,
macromolecules are separated between the two an-
ions.
Abelev and Karamova’s method was originally
developed to analyse proteins in highly dilute bio-
logical Suids such as urine, tears, and cerebrospi-
nal and amniotic Suids, and it turned out to also
be useful for detecting low levels of urinary mon-
oclonal immunoglobulin light chains (Bence Jones
protein) and alpha-fetoprotein in various pathologi-
cal conditions.
CA as a Reusable Electrophoretic Support
CA separations are faster than those on other sup-
ports, usually with no resolution loss. However, CA
sheets cost considerably more than starch, agar,
agarose or polyacrylamide gel sheets.
Recently, a wash method has been described that
makes it possible to recycle CA strips. The procedure
has been shown to work even after using the strips for
analysis of a variety of erythrocyte isoenzymes, which
notoriously expose the support matrix not only to the
strain of the electric Reld but also to many somewhat
elaborate biochemical colorimetric treatment steps.
Surprisingly, none of these stages seem to irreversibly
affect the mechanical and physicochemical properties
of the CA. In fact, after a variety of enzyme activity
tests (adenosine deaminase, adenylate kinase, carbon-
ic anhydrase, erythrocyte acid phosphatase, esterase
D, glutathione peroxidase, glyoxalase 1, phosphog-
lucomutase and 6-phosphogluconate dehydrogenase)
Cellogel2+ returns to its original features if soaked/
washed in water and methanol for a short time. In the
course of double blind trials, no difference in band
sharpness and resolution was noticed between new
and used Cellogel2+ strips. The procedure can be
repeated two or three times if care is taken to avoid
warping strips with absolute methanol soaking or
rough handling.
Blotting Proteins from Polyacrylamide Gels to CA
Sheets
Different electrophoretic species run in the same gel
for the same time with the same electric Reld settings.
The end of a given experiment is currently set depend-
ing on the speciRc requirements of the molecules to be
separated, in zone electrophoresis as well as in IEF.
To achieve optimal resolution of different protein
constituents of the same sample, various experiments
are often carried out, only differing in voltage and
duration. To save time, a simple method involves
repeatedly blotting a polyacrylamide gel with CA
sheets at various stages of separation. The blots ob-
tained in this way can be stained and the protein
species made to show the optimal resolution.
The advantages that can be obtained from CA blots
of the same acrylamide gel are great, the most out-
standing being:
1. various stages of a single protein separation can be
tested in one experiment, to improve the protocol;
2. common and rare variants of a single elec-
trophoretic pattern can be detected, each under
optimal separation;
3. several proteins can be analysed at optimal condi-
tions in the same experiment;
4. all the allele products may be discriminated by
isotacophoretic mechanisms (in non-equilibrium
IEF) and isoelectric point (in true equilibrium IEF)
within the same run.
Conclusion
Almost uniquely among the various supports for elec-
trokinetic separations, CA electrophoresis is still in-
tensively used for both research and routine applica-
tions. The reasons for this long-lasting success are
clear: simplicity of use, low cost, versatility and cost
effectiveness. These same factors are likely to provide
the general basis for the continuing use of CA in the
future.
Acknowledgement
The drawing of Figure 1 was provided by Niccolò
Falchi of the Department of Animal and Human
Biology, University of Rome ‘La Sapienza’.
Further Reading
Abelev GI and Karamova ER (1984) CounterSow afRnity
isotacophoresis on cellulose acetate membranes. Ana-
lytical Biochemistry 142: 437d444.
 II / ELECTROPHORESIS / Deoxyribonucleic Acid, Theory of Techniques for Separation
Ambler J (1978) Isoelectric focusing of proteins on cellulose
acetate gel membranes. Clinica Chimica Acta 85:
183d191.
Destro-Bisol G and Santini SA (1995) Electrophoresis on
cellulose acetate and Cellogel: current status and per-
spectives. Journal of Chromatography A 698: 33d40.
Golias TL (1971) Helena Laboratories Electrophoresis
Manual. Beaumont, Texas: Helena Laboratories.
Grunbaum BJ, ed. (1980) Handbook for Forensic Individ-
ualization of Human Blood and Bloodstains. Gottingen:
Sartorius.
Harada H (1975) Isoelectrofocusing in cellulose acetate
membrane: the method and application. Clinica Chim-
ica Acta 63: 275d283.
Deoxyribonucleic Acid, Theory of Techniques for Separation
J. Noolandi, Xerox Research Centre of Canada,
Mississauga, Ontario, Canada
Copyright ^ 2000 Academic Press
Introduction
Separation of biochemical molecules can be carried
out in gels or polymer solutions and, in speciRc cases,
in free solution, using constant or variable electric
Relds. Gels are used primarily in deep-dish containers,
submerged in buffer, and polymer solutions are used
in glass capillaries, with inner diameters less than
100
m. Thin gels between two glass plates have been
used for separating and sequencing single-stranded
DNA molecules. We begin the theoretical discussion
by considering the separation of double-stranded
DNA molecules (dsDNA) in submarine gels under
constant electric Reld conditions.
Geometrical Sieving Model for Small
DNA Molecules in a Constant Electric
Field
Ogston was the Rrst to calculate the fractional vol-
ume available to a sphere of radius R in a gel of
a given concentration. The gel itself was modelled as
a random array made up of Rbres of radius r. Within
this description, a sphere with a radius Rr cannot
pass through the network if the sphere is not allowed
to deform. This geometrical model predicts that the
electrophoretic mobility of DNA molecules, as a re-
sult of molecular ‘sieving’, varies as:


 Rg 2
Jexp ! [1]
0 a cannot move through the gel as a random coil, rather
1227
Kohn J (1970) Electrophoresis and immunodiffusion tech-
niques on cellulose acetate membrane. Methods in
Medical Research 12: 243d260.
Meera Khan P (1971) Enzyme electrophoresis on cellu-
lose acetate gel: zymogram patterns in man}mouse
and man}Chinese hamster somatic cell hybrids.
Archives of Biochemistry and Biophysics 145:
470d483.
Righetti PG (1976) Isoelectric Focusing, Theory, Methods
and Applications. Amsterdam: Elsevier.
Schneider RG (1978) Methods for detection of hemoglobin
variants and hemoglobinopathies in the routine clinical
laboratory. CRC Critical Reviews in Clinical Laborat-
ory Sciences 9: 243d271.
where Rg is the radius of gyration of the DNA mol-
ecule,  is the free solution mobility, a is the average

pore size of the gel, and the exponential dependence
of the mobility arises from the assumption of Poisson
statistics for the distribution of spaces in a random
network of straight Rbres. This model describes the
mobility of small molecules when they Rrst encounter
the gel Rbres as obstacles to molecular motion. The
analysis of experimental data using eqn [1] is highly
model dependent, but can provide some guidance for
the development of new gel structures for more efR-
cient electrophoretic separations of small molecules.
Entropic Trapping of Small DNA
Molecules
For DNA in the entropic size regime, the deformable
molecules select the larger pores in order to maximize
locally their conformational entropy. However, in
order to accomplish this, they must squeeze through
the narrow channels connecting the larger pores. The
corresponding polyelectrolyte dynamics is dominated
by an activation process in this regime, where the
electrophoretic mobility is given by an inverse power
law ('1) over a size range that is larger than for the
Ogston regime, but smaller than for the beginning of
reptation, which is discussed in the next section.
Gel Electrophoresis of Large DNA
Molecules in a Constant Electric Field
Figure 1 shows a schematic picture of a gel matrix, in
which a DNA molecule is embedded. For a molecule
that is much longer than the average spacing between
the chemical cross-links of the gel Rbres, the molecule
1228 II / ELECTROPHORESIS / Deoxyribonucleic Acid, Theory of Techniques for Separation
Figure 1 Schematic representation of a DNA molecule in
a two-dimensional gel (A), in which the open dots represent
obstacles corresponding to the gel fibres. (B) Shows how the
obstacles hindering the motion of the molecule are approximated
by a tube and the polymer by a chain of beads; the electric field
exerts a force qE on the last bead and orients the segment leaving
the tube. The tube is defined by the molecular conformation so
that an extended conformation (with less DNA per gel pore) has
a longer tube. (C) Charge gradients (C1) along the tube axis, and
field-driven tube leakages (C2) are neglected in the biased repta-
tion model.
it must reptate (from the Latin reptare, to creep)
around the obstacles (cross-links), in a way that is
analogous to the movement of polymer molecules in
a self-entangled polymer melt. In this situation, the
natural length scale is the average pore size of the gel,
a. In terms of this length scale, we deRne a number of
different quantities that arise in a theoretical descrip-
tion of the problem. One of these is the number of gel
pore segments, N, occupied by the molecule:
M vl(hx/N)
N" [2]
Ma
were M is the relative molecular mass of the molecule
and Ma is the average relative molecular mass in a gel
pore. Under the inSuence of an electric Reld, Ma can
change as a function of time, depending on the stiff-
ness of the molecule, the pore size, the magnitude of
the electric Reld, and other factors. Another impor-
tant quantity is the scaled effective electric Reld, Eeff :
e1MaEa E
Eeff" " [3]
2kBT Ea
which deRnes the intrinsic electric Reld parameter Ea.
The net charge per nucleotide for DNA is denoted by
e1 . This quantity depends on the native charge on the
molecule, as well as on the charge screening proper-
ties of the buffer solution used for the electrophoretic
separation. A simple way of understanding the limita-
tions for separating large molecules in a gel using
constant electric Relds is as follows.
For constant a and Ma, the electrical force Fl on the
molecule in the longitudinal direction (also known as
the tube axis, if the molecule is considered to Rll
a ‘tube’ made up of occupied gel pores) is:


ri
Fl"(qE) )  [4]
i a
where q"elMa is the average net charge of the mol-
ecule in a gel pore, and rl/a is the unit vector pointing
from pore i to pore i#1 along the molecule, where
the dot in eqn [4] indicates the dot product of two
vectors. This expression reduces to:
hx
Fl"qE [5]
a
where hx is the end-to-end distance of the molecule in
the Reld direction. The force is time-dependent and
Suctuates during electrophoretic migration. Oppos-
ing the migration along the tube is a friction coefRc-
ient "0N, where 0 is the friction coefRcient per
tube segment, as deRned by eqn [2]. The instan-
taneous velocity of the chain along the longitudinal
axis is then:



Fl qE hx
vl" " [6]
  a
and the average mobility of the centre of mass (de-
Rned as the velocity per unit electric Reld in the Reld
direction) is given by:
h2x

" "
0 2 [7]
E N
where
0"q/0a, and the geometrical factor (hx/N)
in eqn [7] takes into account the fact that the vector
 II / ELECTROPHORESIS / Deoxyribonucleic Acid, Theory of Techniques for Separation
h is in general not parallel to the direction of the
electric Reld. For small electric Relds and/or small
molecules, h2x proportional to N, giving
propor-
tional to 1/N. In this regime the molecules reptate
while retaining their random-walk conformations.
For large molecules and/or high electric Relds, the
molecules become stretched in the electric Reld direc-
tion during migration, and h2x is proportional to
N2, resulting in a mobility that is independent of the
molecular mass, according to eqn [7].
Figure 2 shows a log}log plot of the reduced elec-
trophoretic mobility,
/
0, from Slater and co-
workers as a function of the scaled molecular size,
N"M/Ma, for different values of the scaled electric
Reld. From top to bottom, Eeff"1.0, 0.2, 0.1, 0.01.
For small molecular sizes, the mobility is independent
of the Reld intensity and deceases as 1/N. For large
molecular sizes, the mobility is independent of size
and increases as E2eff. The small minimum in the mo-
bility is a phenomenon known as band inversion, for
which, in a limited size range, large molecules can
move faster than smaller molecules. Noolandi and
co-workers have shown that this is a statistical effect
where linear molecules, which have two free ends,
can migrate into slowly moving ‘J’ or ‘U’ states for
different time intervals, depending on the speciRc
electrophoresis conditions. Since the bands of DNA
molecules are not in order of increasing size in this
Figure 2 Log}log plot of the reduced electrophoretic mobility

/
0 as a function of the scaled molecular size N"M/Ma for
different values of the scaled constant electric field Eeff (see text).
For small molecular sizes the mobility is independent of the
constant field intensity and decreases with size as 1/N. For large
molecular sizes, the mobility is independent of size and increases
as E 2eff. The electrophoretic mobility has a shallow minimum for
intermediate molecular sizes.
1229
region, one has to be careful not to mislabel the
molecular fragments.
In summary, the basic reason for the constant pla-
teau mobility in Figure 2 is that for stretched mol-
ecules the electrical driving force and the opposing
friction both scale linearly with length, and the result-
ing ratio is independent of length. For freely draining
molecular coils in free solution, the same effect oc-
curs, with the result that it is not possible to separate
freely draining polyelectrolytes according to size in
free solution electrophoresis. However, there are
ways to overcome this limitation at least partly for
free solution electrophoresis, as we discuss later. Next
we turn to overcoming the limitations of constant
Reld gel electrophoresis.
Pulsed Field Gel Electrophoresis of
Large DNA Molecules
In constant Reld gel electrophoresis the maximum size
of dsDNA that can be separated is about 50 kilo-
bases. In pulsed-Reld gel electrophoresis separations
up to a few megabases can be achieved. In this section
we explain why such an increase in separation lati-
tude takes place.
The plateau mobility is reached because of molecu-
lar stretching in a constant electric Reld. Schwartz and
Cantor showed that the way to avoid this is to change
the electric Reld constantly, either in magnitude
and/or direction. The disadvantage of doing this is
that it can take a long time to separate large mol-
ecules (separating dsDNA molecules a few megabases
in size can take a few days). The advantage, of course,
is that it is possible to separate large molecules at all,
provided that the variations in the electric Reld are
chosen properly. Changes in the magnitude and/or
direction of the electric Reld force the molecules to
adapt to the new electrophoresis conditions. How-
ever the adaptation time (also known as the ‘relax-
ation’ time) is very speciRc to the molecular size,
average gel pore diameter, and electric Reld changes.
It follows that there are a large number of ways to
implement this process. Some pulsed-Reld gel separ-
ations use electric Relds of the same magnitude and
only change the Reld direction in two dimensions.
Others keep the electric Reld in one dimension (for-
wards and backwards) for different time periods and
with different amplitudes). We use the one-dimen-
sional case developed by Turmel and co-workers as
an example of how separations can be carried out by
tuning the pulse times to the relaxation times of the
molecules in the gel.
Figure 3 is a schematic illustration of the displace-
ment of two types of molecules, molecular masses
1230 II / ELECTROPHORESIS / Deoxyribonucleic Acid, Theory of Techniques for Separation
Figure 3 Pulse shape for gel separations of large DNA mol-
ecules, using zero-integrated-field electrophoresis (ZIFE). As ex-
plained in the text, this pulse shape allows the proper size-migra-
tion distance to be maintained at all times, and gives rise to
a sharp drop in the electrophoresis mobility as a function of size
for fixed pulse parameters. For given electric field amplitudes, the
reverse pulse time, t , defines the relaxation time of the largest
\ veloped over the past decade in a number of different
molecule that can adapt to the lower electric field and have a net
forward displacement during one complete cycle.
M1 and M2 (M1(M2), when subjected to a high Reld
forward pulse, followed by a low Reld backward
pulse of longer duration than the forward pulse. Dur-
ing the pulse duration t# , with Reld intensity E# , the
displacements of molecules M1 and M2 are

(E#, M1)E#t# and
(E#, M2)E#t# respectively,
where the dependence of the mobilities on the Reld
strength and molecular size has been explicitly in-
dicated. During the reverse pulse of duration t , with
\ 
absolute Reld intensity E , the reverse displacement
\
of the shorter molecule is !
(E , M1)E t , if t is
\ \\ \
long enough to allow the shorter molecule to relax to
the new electric Reld in the backward direction. The
net displacement of the shorter molecules over the
time period t##t is then
(E#, M1)E#t#!
\

(E , M1)E t . However, for the longer molecule
\ \\
M2, the reverse displacement is !
(E#, M2)E t , if
\\
the time interval t is too short to allow the longer
\
molecule relax to the lower electric Reld intensity
E in the reverse direction. The net displacement of
\
the longer molecule is then
(E#, M2)E#t#!

(E#, M2)E t , and vanishes when E#t#"E t .
\\ \\
This is known as the zero-integrated Reld condition.
With this condition the displacement of the shorter
molecule becomes [
(E#, M1)!
(E , M1)]E#t#,
\
which is in the forward direction since

(E , M1)(
(E#, M1) if E (E#. The zero-inte-
\ \
grated pulse form allows the proper molecular size
versus displacement in the gel to be maintained at all
times, avoiding the problem of band inversion. Al-
though the tight DNA bands achieved by this process
is an advantage, the running time for experiments is
longer than for some other pulse schemes. Its useful-
ness depends on the speciRc requirements of the prac-
titioner. The pulse times are determined experi-
mentally using known molecular size markers, which
show a sharp drop in mobility when the relaxation
times are reached for a given size range. Using this
information, a general electric Reld pulse scheme can
be programmed for separating unknown DNA size
distributions.
Capillary Zone Electrophoresis (CZE)
Capillary electrophoresis of DNA is a speciRc
example of a technique used to separate molecules
known as capillary zone electrophoresis (CZE), in
which fused silica capillaries have their ends inserted
in electrolyte reservoirs which also contain elec-
trodes. This powerful analytical technique, as dis-
cussed by Gordon and co-workers, has been de-
academic and industrial laboratories. Here the walls
of the fused silica capillaries have a negative charge
resulting from the ionization of the surface silanol
groups in aqueous solution. When a potential differ-
ence is established between the electrodes, a bulk Sow
of Suid towards the cathode takes place. This is called
electroosmotic Sow, and results from the electrical
double layer formed at the wall}electrolyte interface.
The electroosmotic velocity is given by:
0E
Veo"
eoE" [8]
where " /0, and
eo is the electroosmotic mobil-
ity, E is the electrical Reld, 0 the permittivity of the
vacuum,  the relative permittivity,  the zeta poten-
tial,  the viscosity,  the double layer thickness and
the surface excess charge density. The double layer
thickness is inversely proportional to the square root
of the molar concentration of the buffer solution. The
electroosmotic Sow enables the separation of charged
molecules according to their different electrophoretic
mobilities. In this method, positively charged molecu-
les are eluted Rrst because their electrophoretic
motion and the electroosmotic motion are in the same
direction. Negatively charged molecules try to move
in the opposite direction, but if their electrophoretic
mobility is less than the electroosmotic mobility, the
net result is that they are eluted later than the positive
molecules. Thus electroosmotic Sow can be con-
sidered as a built-in pump which is useful for carrying
out electrokinetic separations.
 II / ELECTROPHORESIS / Deoxyribonucleic Acid, Theory of Techniques for Separation
Clearly this process cannot separate neutral mol-
ecules if only an electrolyte is used. Sometimes
micelles are dissolved in the separation electrolyte,
resulting in some separation of uncharged species
because of their partitioning between the micelles and
the electrolyte. More commonly, one packs the capil-
lary with a stationary phase, which can be bonded to
the walls of the capillary, or formed by close packed
small particles. The Sow proRle in the capillary has
a nearly Sat (plug-like) proRle instead of the usual
parabolic proRle for Poiseuille Sow when a viscous,
incompressible Suid is pushed through a cylindrical
tube by a pressure difference. The nearly Sat Sow
proRle makes the separation and detection of small
amounts of analyte easier. This technique is known as
capillary electrochromatography (CEC). The packing
of the capillaries can be carried out electrokinetically,
using commercially available small (a few
m) porous
silica particles.
A number of different theoretical strategies have
been used to increase the sensitivity of analyte detec-
tion. Programming of the electric Reld to achieve
uniform sensitivity for on-line detection has been
used, and theories for maximizing the signal-to-noise
ratio for the case of laser-induced Suorescence and
UV absorbance detection in capillary electrophoresis
have been developed. In particular, for rather general
conditions, the ratio of the number of Suorescent
molecules in two analyte bands can be shown to be
proportional to the ratio of the heights of the band
peaks. This is useful for Suorescent analytes, and for
cases where nonSuorescent analytes are labelled with
Suorescent probes prior to detection.
Capillary Electrophoresis of DNA
Molecules in Free Solution
In free solution the electrophoretic mobility of a free
draining polyelectrolyte coil is equal to the ratio of its
electric charge to its friction coefRcient. Since both
quantities scale linearly with molecular mass, the
mobility is independent of molecular size. Hence the
only way to separate molecules of different size in this
case is to break the scaling symmetry of the charge
and/or friction with size. As shown by Voelkel and
Noolandi, adding a molecular unit with a different
charge and friction to one of the ends of a linear
polyelectrolyte, such as DNA, will change the free
solution mobility to:
M#qeff

(M)"
0 [9]
M#eff
where
0 is the free solution mobility of the unlabelled
free draining polyelectrolyte, qeff is the ratio of the
1231
effective charge carried by the label expressed in
terms of the charge per monomer unit of the polyelec-
trolyte, and eff is the friction of the label expressed in
terms of the friction per monomer unit of the poly-
electrolyte. In this approximation the hydrodynamic
interaction of the label and the polyelectrolyte is
neglected. The free solution mobility is now a func-
tion of the molecular mass. However the effectiveness
of this strategy depends on the size of the label,
compared with the rest of the molecule, and the
amount of the band broadening caused by Brownian
motion and other effects. In practice this approach
seems viable for separating single-stranded DNA
molecules up to several hundred bases long for DNA
sequencing applications. While gel separations of
single-stranded DNA molecules can routinely be car-
ried out for sizes well over 500 bases, free solution
separations, which are only effective for shorter frag-
ments, can be carried out only an order of magnitude
faster.
Capillary Electrophoresis of DNA
Molecules Using Dilute Polymer
Solutions
A separation mechanism that combines aspects of
both free solution and gel electrophoresis is separ-
ation in capillaries using ultrathin polymer solutions.
Here the DNA molecules drag along the polymer
molecules that they encounter during migration. The
capture and release of the unentangled polymer mol-
ecules results in the separation of DNA molecules.
This new type of separation mechanism, developed
by Barron and co-workers, is based on hydrodynam-
ics, drag forces and molecular collisions, and is best
suited for high throughput applications. The theoret-
ical basis for the separation is still under develop-
ment.
Capillary Electrophoresis of DNA
Molecules Using Entangled Polymer
Solutions
As pointed out by Slater, entangled solutions, as op-
posed to a gel, involve physical, instead of chemical,
cross-links. For polymer solutions, we consider con-
centrations c'cH, where cH is called the critical con-
centration for entanglements:
M
cH" [10]


4 3
Rg
3
1232 II / ELECTROPHORESIS / Deoxyribonucleic Acid, Theory of Techniques for Separation
M is the relative molecular mass, and Rg is the radius
of gyration of the polymer molecule. For c'cH the
polymer coils overlap and a loosely associated poly-
mer network is formed. From the theory of polymer
networks in solution we can characterize the mean
pore size by a ‘blob’ dimension:


3/4
cH
"1.4Rg [11]
c data that are available with the use of laser-induced
However we must bear in mind that we are dealing
with loosely associated networks, and that the phys-
ical cross-links are temporary, as opposed to chemic-
ally cross-linked networks. As discussed by Viovy
and Heller, for low electrophoretic mobilities
the dissociation and reassociation times of the
network become important, because they occur on
the same timescale as the transit time of the DNA
molecules through the network. As a consequence,
it is not possible to achieve better separations of
DNA molecules through a concentrated polymer
solution than through a gel, however polymer solu-
tions are more easily processed (injected or removed)
through capillary channels than cross-linked gels
because of their low viscosities. Some polymer solu-
tions have strongly temperature-dependent proper-
ties, which allows more latitude in the processing
conditions.
Future Developments
Separations of biomolecules are being carried out on
smaller and smaller devices, using miniaturized Suid-
handling devices and detection systems. Making use
of recent rapid developments in the area of microelec-
tromechanical systems (MEMS), which exploits ad-
vances in microlithography for the semiconductor
industry, new biochips and biosensors have been de-
signed that enable faster analytical and diagnostic
techniques to be carried out than was possible with
macroscopic devices. As shown by Chee and co-
workers, hundreds of thousands of DNA oligo-
nucleotide probes can be assembled on a glass micro-
chip, and combined with micromachined capillary
electrophoresis injectors and separators. Complete
hybridization patterns can be revealed in a matter
of minutes, using laser-induced Suorescence. The im-
plications for the entire biotechnology industry are
revolutionary. Coupled with the knowledge obtained
from the Human Genome Program, in which the
estimated 100 000 human genes are in the process of
being discovered and sequenced, thousands of genetic
variations can be analysed in a single experiment,
making possible the rapid localization of disease-
causing genes.
From the theoretical point of view, there are several
areas where advances are necessary to keep up with
the rapid developments in instrumentation. First, new
software codes are required to enable the rapid de-
ciphering and processing of the massive amounts of
bioinformation that are being generated. Second, de-
tection systems with higher resolution than is current-
ly available must be designed to interpret the spectral
Suorescence. Finally, theoretical modelling of the be-
haviour of electrolytes and biomolecules in micro-
channels in the presence of electric Relds will be
useful in understanding the ultimate capability of
microdevices for the applications that we have
mentioned.
See Colour Plate 40.
Further Reading
Barron AE, Blanch HW and Soane DS (1994) A transient
entanglement coupling mechanism for DNA separation
by capillary electrophoresis in ultradilute polymer solu-
tions. Electrophoresis 15: 597}615.
Chee M, Yang R, Hubbell E et al. (1996) Accessing genetic
information with high-density DNA arrays. Science 274:
610}614.
Gordon MJ, Huang X, Pentoney SL Jr and Zare RN (1988)
Capillary electrophoresis. Science 242: 224}228.
Noolandi J, Rousseau J, Slater GW, Turmel C and Lalande
M (1987) Self-trapping and anomalous dispersion of
DNA in electrophoresis. Physics Review Letters 58:
2428}2431.
Ogston AG (1958) The spaces in a uniform random suspen-
sion of Rbres. Transactions of the Faraday Society 54:
1754}1757.
Schwartz DC and Cantor CR (1984) Separation of yeast
chromosome-sized DNAs by pulsed Reld gradient gel
electrophoresis. Cell 37: 67}75.
Slater GW, Mayer P, Hubert SJ and Drouin G (1994) The
biased reptation model of DNA gel electrophoresis:
a user guide for constant Reld mobilities. Applied and
Theoretical Electrophoresis 4: 71}79.
Slater GW (1996) Electrophoresis theories. In: Heller C
(ed.) Analysis of Nucleic Acids by Capillary Elec-
trophoresis, ch. 2. Berlin: Vieweg Press.
Turmel C, Brassard E, Forsyth R, et al. (1990) High resolu-
tion zero integrated Reld electrophoresis (ZIFE) of DNA.
In: Birren B and Lai E (eds). Current Communication in
Molecular Biology. Electrophoresis of Large DNA Mol-
ecules, p. 101. Cold Spring Harbor, NY: Cold Spring
Harbor Laboratory.
Viovy JL and Heller C (1996) Principles of size-based separ-
ations in polymer solutions. In: Righetti PG (ed) Capil-
lary Electrophoresis in Analytical Biotechnology, ch. 11.
Boca Raton, FL: CRC Press.
Voelkel A and Noolandi J (1995) Mobilities of labeled and
unlabeled single stranded DNA in free solution elec-
trophoresis. Macromolecules 28: 8182}8189.
 II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
Detection of Proteins in Electrophoresis
See II / ELECTROPHORESIS / Proteins, Detection of
Detection Techniques: Staining, Autoradiography and Blotting
P. J. Wirth, National Cancer Institute, Bethesda, MD,
USA
Copyright ^ 2000 Academic Press
Introduction
Polyacrylamide gel electrophoresis (PAGE) is a highly
reliable and widely used technique for the separation,
identiRcation and characterization of proteins and
protein mixtures. Although two-dimensional (2D)-
PAGE, which combines protein isoelectric focusing
(IEF) in the Rrst dimension with sodium dodecyl
sulfate (SDS)-PAGE molecular sieving in the second
dimension, provides the highest resolution allowing
one to separate 1000}2000 individual polypeptide
spots on a single gel, 2D-PAGE is technically very
demanding. However, in the vast majority of applica-
tions, one-dimensional (1D)-PAGE, speciRcally SDS-
PAGE, provides sufRcient resolution and is especially
well suited for the simultaneous analysis of multiple
protein samples on a single gel. Since its introduction
in 1951, very few modiRcations to the basic protocols
for preparing and running 1D-PAGE gels have been
made, although considerable advances have been in-
troduced for the detection and analysis of PAGE
separated proteins.
Post-electrophoretic gel staining is the most fre-
quently used method for the detection of individual
protein bands or spots on 1D- and 2D-PAGE gels,
respectively, although procedures for prestaining pro-
teins prior to PAGE have been described. Detection is
usually performed either in situ within the polyac-
rylamide gel matrix itself or following Western elec-
troblot transfer of proteins from PAGE gels on to
polymeric membrane support matrixes. Detection
systems include organic dye and metal salt-based
staining protocols, Suorescent group tagging, speciRc
protein}ligand/receptor interactions, enzymic activ-
ity detection, as well as group-speciRc (e.g. glyco-,
phospho-, lipoproteins, etc.) staining and immuno-
logical detection of antibody}antigen complexes.
Alternatively, proteins which have been labelled with
radioactive molecules, either prior to or post-elec-
trophoretically, can be visualized using autoradio-
1233
graphic and Suorographic detection on X-ray Rlm.
Although a plethora of protein staining and visual-
ization protocols have been described, none is totally
ideal and often the use of multiple protein staining
and/or protein labelling procedures is necessary.
Protein Staining
Organic Dyes
Many of the organic dyes and stains that have been
adopted for the detection of proteins in polyacrylam-
ide gels and on membranes have been derived from
dyes originally developed for the textile industry.
Currently the most commonly used dyes include
Amido Black, Procion Blue RS, Ponceau S, Alcian
Blue, Fast Green FCF, Coomassie1,2 Brilliant Blue R-
250 (R"reddish hue) (CBB-R) and Xylene Cyanine
Brilliant G (confusingly referred to as Coomassie Bril-
liant Blue G-250) (G"greenish hue) (CBB-G). Re-
cently, inorganic metal ion-based staining procedures
have been developed that provide highly sensitive
methods of protein visualization.
CBB-R and CBB-G are the most sensitive, conve-
nient and economical to use of the commonly avail-
able dyes, and have become the reagents of choice for
protein staining. CBB-R is a nonpolar, sulfated aro-
matic dye that is generally used in methanolic acetic
acid solutions where excess CBB is removed from the
gel matrix by destaining. An acidic environment is
required for optimal CBB staining to enhance ionic
interactions between the dye molecules and basic
amino groups of the protein as well as to augment
dye}protein interactions due to hydrogen bonding,
van der Waals attraction and hydrophobic bonding.
 Any reference to a trademark or proprietary product does not
constitute endorsement of that product by the US government and
does not imply its approval to the exclusion of other products.
 CoomassieTM is a registered trademark held by the Imperial
Chemical Industries (ICI). Equivalent CBB dyes under their own
trademarks include Serva Blau (Blue) R or G, PAGE blue 83 (R) or
90 (G), Kenacid blue R, Supranocyanin 6B or G, Brilliant Blue
R and Microme no. 1137.
1234 II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
Typically, CBB staining is performed using a 0.1}
0.2% CBB-R (w/v) in aqueous (v/v) 45% methanol
and 10% acetic acid. The duration of staining is
dependent on gel thickness and polyacrylamide com-
position and destaining is performed using either
passive diffusion or electrophoretic destaining. After
destaining, gels can be stored in 7.5% acetic acid in
which the dye}protein complex is Rxed and the col-
our is relatively stable.
An alternative CBB-based procedure exhibiting
very low background staining has been described
using CBB-G. Incubation of gels in a colloidal suspen-
sion of CBB-G in aqueous trichloroacetic acid (TCA)
results in the formation of dye}protein complexes, in
which CBB-G interacts with proteins only at the sur-
face of the gels and does not penetrate into the gel
matrix, thus minimizing background staining. Major
protein bands are visible within 5}10 min and for
optimal staining of less abundant proteins gels should
be left in the staining solutions for several hours to
overnight followed by destaining in 5% TCA.
Amido Black (Buffalo black NBR, naphthalene
black 12B, aniline blue black, napththol blue black,
acid black 1 and amido schwartz) was probably the
Rrst dye used to stain proteins in polyacrylamide gels;
however, its use today is less frequent because of the
availability of more sensitive CBB-based protocols.
None the less, Amido Black still enjoys numerous
applications because of its rapid staining and destain-
ing properties.
A dye especially well suited for quantitation ap-
plications is Fast Green FCF (food green 3). Fast
Green exhibits a greater linearity of staining as com-
pared to CBB-R and also has the capacity to form
stable coloured complexes with histones, thereby
making it a useful group-speciRc stain. In contrast to
CBB, Fast Green does not bind to carrier ampholytes
and can be used for staining of proteins in IEF gels.
Recently, protein staining procedures utilizing
calconcarboxylic acid N,N
1-(2-hydroxy-4-sulfo-1-
naphthylazo)-2-hydroxy-3-naphthoic acid, Erioch-
rome Black T/rhodamine B, Evans blue/rhodamine B,
and CBB/Bismark Brown R have been introduced
with reported sensitivity comparable or better than
CBB in SDS-PAGE gels. A very useful general protein
stain that also displays group-speciRc staining is
Stains-allTM, a cationic carbocyanine dye, which
stains sialoglycoproteins and phosphoproteins blue
and almost all other proteins red. Table 1 summar-
izes some of the more commonly used organic dyes
for protein staining.
Silver Staining
For most applications, visualization of proteins with
CBB is sufRciently sensitive. However, if one is inter-
ested in determining the absolute purity or trace
amounts of a protein then more sensitive techniques
must be utilized. To accomplish this, Merril de-
veloped an ultrasensitive silver staining procedure
based on photographic principles. Silver staining
is upward to 100-fold more sensitive than CBB-R
with sensitivity comparable to, or greater than,
autoradiography, for selected polypeptides. It should
be noted, however, that many proteins respond dif-
ferently to silver staining. Some proteins may not
stain at all, so sensitivity values for silver staining may
vary from protein to protein.
Silver staining or silver shadowing procedures can
be divided into three basic catagories: diamine or
ammoniacal silver stains; chemically developed non-
diamine type; and photoreduction silver stains. The
diamine or ammoniacal silver stains utilize am-
monium hydroxide to form soluble silver}diamine
complexes and proteins are visualized by acidiRca-
tion, usually with citric acid in the presence of formal-
dehyde. Diamine stains are rather time-consuming
(overnight Rxation and 6 h staining) but are parti-
cularly good for the staining of gels thicker than
1 mm. The nondiamine chemical development stains
are generally more rapid than the diamine stains but
display higher backgrounds and are best suited
for gels 1 mm or thinner. Image development of
nondiamine stains occurs as a result of selective
reduction of silver ions to elemental metallic silver
by formaldehyde under alkaline pH. The photo-
reduction silver stains are the most rapid, allowing
the visualization of protein patterns within 10 min
after electrophoretic separation; however, they are
the least sensitive of the silver-based staining
methods.
While most proteins stain monochromatically with
silver, yielding brown or black spots and bands, cer-
tain silver stains can produce varying shades of black,
blue brown, red and yellow and the staining of indi-
vidual proteins appears to be group-speciRc. Lipo-
proteins tend to stain blue while glycoproteins appear
yellowish-brown, or red. Colour formation has
been shown to be highly dependent on the size and
distribution of the silver grains within the gel as well
as the refractive index of the gel and standardized
colour-based silver staining kits are commerically
available.
Reverse Staining
In contrast to the positive-staining procedure de-
scribed above, alternative but generally less sensitive
staining procedures based on the formation of insol-
uble metal (zinc, copper and potassium) salts have
also been described. These methods, commonly
 II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting 1235

Table 1 Reagents useful for protein visualization on gels and membrane supports

Stain /dye Sensitivity Gel Membranes Comments

NC PVDF Nylon

Organic dyes
Coomassie Brilliant Blue (CBB-R) 100}1000 ng yes yes yes no Permanent, convenient to use, sensitive,
and (CBB-G) anionic, relatively high background
(CBB-R), ('1 h)a
Ponceau S 1}2
g yes yes yes no Reversible, low background, rapid staining/
destaining ((2 min)
Amido Black 1}2
g yes yes yes no Permanent, low background, anionic
stain, stains histones and oroso-
mucoids ('30 min)
India Ink (colloidal carbon) 80}100 ng no yes yes yes Permanent, very sensitive, low back-
ground, sensitivity dependent upon
source/lot of ink ('2 h), useful for
charged nylon membranes
Fast Green FCF 1}2
g yes yes yes no Permanent, linearity of staining'CBB,
does not bind IEF ampholytes, stains
histones (30 min)
Stains-AllTM 1}2
g yes yes yes no General protein and nucleic acid stain,
stains RNA (bluish purple), DNA (blue),
most proteins (red) sialoglycoproteins
and phosphoproteins (blue)
Periodic acid}Schiff (PAS) 25}100 ng yes yes yes no General protein and glycoprotein stain.
Silver enhancement increases sensitiv-
ity (&0.4 ng)
Procion Blue RS 1}2
g yes yes yes no Anionic, hydrophobic dye, occasionally
used
Alcian Blue 500}700 ng yes yes yes no Useful for staining glycosaminoglycan
Eosin Y 10 ng yes yes yes no General protein stain, stains sialoglycop-
roteins
Congo Red 500 ng yes yes yes no Low background, rapid (5 min), anionic
stain
Fluorescent stains
Dansyl chloride 100}200 ng yes yes yes no Stains all types of proteins, including pro-
teoglycans
SYPRO Red/Orange 100 ng yes no no no Exhibits a greater linearity of staining than
CBB, SDS-PAGE only, does not require
protein fixation, will not stain nucleic
acids
Fluorescamine 6}10 ng yes yes yes no Neither fluorescamine nor its degradation
products are fluoresecent, low back-
grounds
Fluorescein isothiocyanate 50 ng yes yes yes no Rapid, can be used as pre-electrophoretic
tag
Metal salt complexes
Silver 1}3 ng yes yes yes no Permanent, very sensitive, stains nucleic
acids, hundreds of modifications, time-
consuming
Colloidal gold (AurodyeTM) 1}3 ng yes yes yes no Permanent, very sensitive, silver en-
hancement increases sensitivity
Potassium chloride 10}100 ng yes no no no Reversible, SDS-PAGE only, negative
stain, whiter bands on opaque back-
ground
Iron (FerridyeTM) 1}3 ng yes yes yes yes Permanent, useful for nylon membranes
Copper (iodide/chloride) 10}100 ng yes no no no Reversible, rapid (5 min), SDS-PAGE
only, negative stain, clear bands on
opaque background
Copper phthalocyanine 10}100 ng yes yes yes no Reversible, rapid ((2 min), sensitive
3,4,4,4 tetrasulfonic acid
Zinc / imidazole SDS 10}100 ng yes no no no Reversible, metal chelate stain, SDS-
PAGE only, negative stain, clear bands
on opaque background, does not re-
quire protein fixation, protein eluted
from gels with high efficiency ('90%)

a
Staining / destaining time. NC, Nitrocellulose; PVDF, polyvinylidene fluoride.
1236 II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
referred to as negative or reversible staining, are lim-
ited to SDS-containing gels and produce a semi-
opaque background on the gel surface where proteins
are detected as whiter or transparent bands or spots
when viewed against a black background or when
back-lit. Staining procedures are rapid (within
15 min), display intermediate sensitivity between that
of CBB and silver staining, and since minimal protein
Rxation is required, proteins are readily eluted from
gels ('90%) for biochemical characterization in-
cluding Western immunoblotting, amino acid com-
position analysis and Edman N-terminal amino acid
microsequencing.
Fluorescent Protein Labelling
Fluorescent methods for protein visualization are ex-
tremely sensitive but are less frequently used than the
CBB/silver staining protocols due to their relative
complexity (e.g. they require ultraviolet illumination
for protein visualization, and Suorescence signal in-
tensities diminish with time) and increased cost. Pro-
teins can be tagged either pre- or post-electropheti-
cally with Suorescent sensitive dye(s) via covalent
interaction of the dye with terminal -NH2 groups of
the proteins. Fluorophores most commonly utilized
to label proteins prior to electrophoresis include dan-
syl chloride (1-dimethylaminonaphthalene-5-sulfonyl
chloride), Suorescamine (4-phenylspiro(furan-2[3H]-
1-phthalan-3,3 dione), MDPF (2-methoxy-2,4-
diphenyl-3-(2H)-furanone), DACM (N-(7-dimethyl-
amino-4-methylcoumarinyl)maleimide) and OPA (o-
phthaldialdehyde). Although dansyl chloride was the
Rrst Suorescent dye used for pre-electrophoretic
labelling of proteins, Suorescamine has found in-
creasing use since neither Suorescamine nor its hy-
drolysis products are Suorescent. Fluorescamine is
capable of detecting as little as 6 ng of myoglobin
while MDPF, which is 2.5 times more sensitive than
Suorescamine, is very useful for quantitative applica-
tions and displays a linear staining response from 1 to
500 ng.
Proteins can also be detected post-electrophor-
etically using Suorescent reagents such as ANS
(1-anilinonaphthalene-8-sulfonate), Bis-ANS, Suor-
escamine, p-hydrazinoacridine and OPA. Since
labelling is usually performed under nondenaturing
conditions, these reagents can be used quite advant-
ageously for the rapid detection of proteins during
preparative electrophoresis. Generalized procedures
for both pre- and post-labelling with Suorescent dyes
have appeared in reviews by Hames and Rickwood
and Merril. Table 1 brieSy summarizes the major
metal-based and Suorescent dyes used to stain pro-
teins on polyacrylamide gels.
Autoradiographic Detection
Labelling of proteins either prior to or post-elec-
trophoretically using radioactive isotopes remains the
most sensitive method for protein detection. Indi-
vidual radiolabelled protein bands or spots are usu-
ally detected in one of three ways: liquid scintillation
counting, autoradiography and Suorography. Modi-
Rcations to facilitate the detection of proteins ex-
pressed at very low concentrations (e.g. transcription
factors, cytokines, single copy gene products) or pro-
teins labelled with low energy -particle-emitting
radioisotopes, such as [3H], include indirect auto-
radiography which utilizes intensifying screens for
signal enhancement and Suorography.
Autoradiography
In autoradiography dried polyacrylamide gels con-
taining radiolabelled proteins are placed in direct
contact with the appropriate X-ray Rlm (e.g. Kodak
X-Omat AR, Kodak SB-5, Kodak BMS2, Kodal
BMR2, Fuji RX) where radioactive emissions (-par-
ticles and/or /-radiation) react with the silver hali-
des in the Rlm emulsion, resulting in the formation of
elemental silver atoms which are visualized following
photographic development of the Rlms. [14C]-, [35S]-,
[32P]-, [125I]- and [131I]-labelled proteins are readily
detected using direct autoradiography while [3H]-
labelled proteins are very weakly detected due to
severe quenching of their low energy -emissions by
the polyacrylamide gel matrix.
Fluorography and Indirect Autoradiography
To enhance the detection of low abundance proteins
and proteins labelled with low energy -type emitters,
Suorographic and indirect autoradiographic method-
ologies have been developed. Both procedures pro-
vide enhanced autoradiographic imaging of low to
medium energy -particle emitters (3H, 14C or 35S)
and involve the conversion of the emitted energy from
the respective isotopes to photons of visible light,
which become the predominant exposing radiation.
This is accomplished either by the incorporation of an
organic scintillator, PPO (2,5-diphenoxyoxazole),
directly into the polyacrylamide gel matrix prior to
Rxation, drying and exposure to Rlm (Suorography)
or by the use of calcium tungstate X-ray intensifying
screens (indirect autoradiography), as originally de-
veloped by Bonner and Laskey and Mills, respective-
ly. For optimal sensitivity, Rlm exposure utilizing
X-ray intensifying screens such as Kodak X-
OMATIC and Dupont Cronex Lightning Plus or
Cronex Quanta II should be performed at low tem-
peratures (!703C) to stabilize latent image forma-
tion. This results in up to a 30}40-fold increase in the
 II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
detection of [125I] and 8}10-fold increase in sensitivity
to [32P]. Additional rare earths (europium-activated
barium Suorochloride or terbium-activated mixtures
of lanthanium oxysulRde and gadolinium oxysulRde)
are available and these appear to be more efRcient
than calcium tungstate for [32P] detection but result in
higher background Rlm darkening. The sensitivity of
Suorography can be further increased by pre-Sashing
((1 ms) or hypersensitizing the Rlm before main
exposure. This step has the added beneRt of correct-
ing the so-called toeing effect or nonlinear relation-
ship between the radioactivity in a sample to the
absorbance of the Rlm image, thus permitting quant-
itative measurements. If an autoradiogram or Suoro-
gram is too faint then it is possible to intensify the
images up to 10-fold by incubation of the X-ray Rlm
in [35S]-thiourea which complexes with the silver ions
in the Rlm to form silver [35S] sulRde.
Storage Phosphor Imaging
Two of the most serious limitations to the use of
X-ray Rlm for the visualization of isotopically label-
led proteins are relative insensitivity to low energy
-radiation and a nonlinear, limited dynamic range of
Rlm darkening to radiation exposure. An alternative
to X-ray Rlm for the detection and quantiRcation of
autoradiography is photostimulable storage phos-
phor imaging. Storage phosphor imaging exhibits
a dynamic exposure range of more than Rve orders
of magnitude (100 000 : 1 versus 300 : 1 for X-ray
Rlm) and a 10}250-fold greater sensitivity than
autoradiography to -emissions. Dried gels contain-
ing radiolabelled proteins are exposed to imaging
screens composed of a thin layer of BaFBr : Eu#2
crystals in an organic binder in the same manner in
which X-ray Rlm is exposed. Incident radiation (-
particles, -rays, X-rays) from labelled proteins indu-
ces excitation of the Eu#2 ions in the phosphor com-
plex which stores this energy as a latent image. The
latent images are scanned with a helium}neon laser
which releases the stored energy as blue photons and
the intensity of luminescence is quantitated. [14C],
[35S], [32P], [33P], [125I] and [131I] are readily detected
and quantitated.
Labelling of Proteins with Radioactive Isotopes
The most commonly used isotopes include [14C],
[35S], [32P], [3H] and [125I], although metal isotopes
such as [59Fe], [45Ca], [63Ni] and [75Se] have been used
to identify iron, calcium and nickel binding proteins
and Se-cysteine-containing proteins, respectively.
Whereas [32P]-orthophosphate has been used for the
introduction of radioactive phosphate groups into
proteins, the substitution of [33P] for [32P] has gained
in popularity because of signiRcantly increased res-
1237
olution band sharpness as well as increased safety
factors afforded by the lower energy [33P] emitter as
compared to [32P].
In vitro metabolic labelling of cells or tissue sec-
tions in short term culture by incorporation of iso-
topically labelled amino acid(s) precursor molecules
during the cellular growth phase is usually performed
using either [3H]-leucine or [35S]-methionine/cysteine.
The use of [35S]- is favoured because of its higher
energy -emitter potential (0.167 vs. 0.018 MeV),
higher speciRc activity ('1000 vs. 50 Ci mmol\1)
and lower cost than [3H]-labelled molecules. The ex-
tent of incorporation of [35S]-methionine is, however,
dependent upon methionine content of the individual
proteins and proteins lacking methionine would be
undetected using [35S]-methionine labelling. This
problem has been cricumvented by labelling with
[14C] amino acid mixtures, although this method is
less favoured due to signiRcantly higher cost and
lower speciRc activity (50 mCi mmol\1) of [14C] ver-
sus [35S]. Proteins which are post-translationally
modiRed via glycosylation can be labelled with
[3H]/[14C]-galactose, mannose, N-acetyl-glucose and
galactoseamine (carbohydrates), respectively, while
lipoproteins and certain membrane-associated poly-
peptides can be labelled with [3H]/[14C] palmitate and
myrisoylate.
Radioactive Stains
The use of radioactive stains for the in situ detection
of proteins has found limited applications because of
the availability of relatively few radiolabelled re-
agents. [59Fe]-ferrous bathophenanthroline has been
used to label radioactively a series of protein markers
in polyacrylamide gels post-electrophoretically using
simple staining and destaining procedures.
Protein Blotting
One the major advances in the analysis of proteins
has been the development of post-electrophoretic
techniques for the transfer and immobilization of
proteins from the polyacrylamide gel matrix to thin
support membranes. Originally based on DNA
Southern and RNA Northern blotting principles, pro-
tein-blotting protocols were similarly developed by
Towbin and co-workers utilizing the electrophoretic
elution of proteins separated by PAGE to nitrocel-
lulose (NC) sheets. The major advantage of protein
electroblotting is that separated proteins are transfer-
red from the gel matrix, where their access to detec-
tion reagents is severely hindered, to the surface of
a membrane where the protein molecules are readily
accessible. Although protein blotting has tradition-
ally been associated with the immunodetection of
1238 II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
antigen}antibody complexes (Western immunoblot-
ting), blotted proteins are amenable to analysis and
characterization via a multitude of visualization and
overlay techniques. These include general protein
staining and autoradiography, group-speciRc ligand
binding, receptor}ligand interaction, enzymic activity
determination as well as amino acid composition and
primary amino acid sequence analysis of individual
spots or bands. A single protein blot offers numerous
advantages not afforded by polyacrylamide gels. It is
easily handled and manipulated, can be stored for up
to 1 year and the blot can be used for multiple
successive analyses. Once a signal has been obtained
and recorded, the blot can be erased by removing the
probe or stain while retaining the original protein
pattern on the membrane and the blot reprobed.
Protein blotting has been said to add a second and
third dimension to 1D and 2D-PAGE, respectively.
Various types of membranes have been used for
protein blotting and immobilization. Nylon and NC
sheets (thin Rlm on cellulose esteriRed with nitric
acid) were Rrst used but recently different types of
polyvinylidene Suoride (PVDF) membranes have
been introduced. A detailed description of protein-
blotting methodology, including the advantages and
disadvantages of the various blotting membranes, is
beyond the scope of this article (see Further Reading).
Immunological Detection
Immunological or group-speciRc detection of pro-
tein(s) on blotted membranes is far and away the
most utilized method for protein detection but is
limited by the available of appropriate antibod-
ies/ligands. Following protein transfer, membranes
are incubated with dilute protein solutions (e.g. bo-
vine serum albumin, gelatin or instant nonfat dry
milk) to block nonspeciRc binding sites on the mem-
branes. The membrane-bound proteins are incubated
with either monoclonal or polyclonal antibodies di-
rected against speciRc target antigens or group-speci-
Rc ligands (e.g. Concanavilin A for the detection of
glycoproteins, 59Fe and 45Ca to detect iron- and cal-
cium-binding proteins, and [32P]-labelled DNA to
detect DNA binding proteins). NC and PVDF
membranes are most frequently used since nylon
membranes with their intrinsically higher binding
afRnity for proteins are more difRcult to block. If
antibodies are used, then blots are incubated with
a second antibody that has been directed toward the
primary antibody and has been tagged with a reporter
label. Typically, the second anti-species antibody may
be radiolabelled (125I) or conjugated with an enzyme
(e.g. horseradish peroxidase (HRP), alkaline phos-
phatase (AP) or -galactosidase). The resulting ((anti-
gen}13 antibody)}23 antibody) complex is detected
either autoradiographically or colorimetrically using
an appropriate chromogenic substrate. The advant-
ages of enzyme-conjugated antibodies are ease of
handling and storage and rapid development of col-
our (min vs. day). Sensitivity is usually in the range of
0.1}10 ng of antigen/band. While this sensitivity is
approximately 10}100-fold less sensitive than
autoradiographic or Suorographic detection, it is pos-
sible to achieve a similar level of enhanced sensitivity
using peroxidase}antiperoxidase (PAP) sandwiching.
Recently, modiRcations have been developed for
the use of light emission (luminescence) as an end
point for protein}antibody detection on membranes.
Although substrates which produce colour complexes
exist for both AP and HRP, the HRP luminescent
system, in which blue light is generated by the HRP-
mediated oxidation of luminol, is the most sensitive
system. This system, commonly known as the ECL
(enhanced chemiluminescence) Western blotting sys-
tem, provides excellent signal-to-noise ratio and is
extremely rapid and sensitive.
In the basic (nonenhanced) chemiluminescent reac-
tion, HRP is used to oxidize a peracid, resulting in
a raised oxidation state of the haem Fe in HRP.
Relaxation of this excited state to initial (ground)
state occurs in a two-step process. At each relaxation
a luminol radical is formed and, as each radical
decays, light is emitted. However, in the ECL reac-
tion, an enhancer molecule is added which reacts with
the haem Fe in place of the luminol molecule, result-
ing in the formation of enhancer radicals which them-
selves react to produce luminol radicals and light is
emitted. The enhancer molecules increase light emis-
sion greater than 1000-fold over luminol alone. Light
emission on membranes rises rapidly over the Rrst
5 min, remains at maximum for 15 min, and then
declines with a t1/2 of 60 min. Typical exposures for
ECL are of the order of a few seconds to minutes and
are capable of detecting 1 pg or less of protein.
Total Protein Staining
Blotted membranes can be stained with many of the
general protein stains used for polyacrylamide gels
including Amido Black, CBB, Ponceau S, Fast Green
and India ink (Table 1). Amido Black and Ponceau
S are preferred to CBB-R because stained membranes
can be destained quickly to leave very low back-
grounds, whereas CBB-R gives higher backgrounds.
India ink (colloidal carbon) is the most sensitive of
the above dyes and can detect as little as 80 ng of
protein but staining sensitivity is highly dependent
upon dye source and lot. Silver staining is also pos-
sible as well as the use of colloidal gold and iron sol
stains. Both silver and gold stains can detect as little
as 1}5 ng protein on NC and PVDF membranes and
 II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis
the sensitivity of gold stain can be further enhanced
by incubation with a silver lactate solution such that
as little as 400 pg of protein per band can be detected.
Although nylon or charged nylon membranes possess
the greatest protein-binding capacitites (450 vs.
80
g cm\2 (NC/PVDF)), staining of nylon mem-
branes is very problematic. Anionic organic dyes as
well as colloidal gold and silver are not useful for
staining nylon membranes due to extremely high
backgrounds. However, colloidal sols are especially
useful for the detection of proteins on nylon mem-
branes. On nylon membranes the positively charged
colloidal iron particles bind to negatively charged
SDS-denatured proteins and protein staining can be
intensiRed using potassium ferricyanide, which gives
deep blue-stained bands with low backgrounds. India
ink and a modiRed silver stain have been reported to
have been used to stain charged nylon membranes.
A less frequent, but none the less useful method for
the visualization of protein bands on NC and charged
nylon membranes involves protein iodination in situ
with chloramine T/potassium iodide, followed by
formation of a purple complex between the bound
iodine and starch.
Autoradiographic Detection
Electroblotting of proteins radiolabelled with 14C or
35
S permits more efRcient autoradiography since the
gel matrix is no longer present to quench the -
emissions. The minimum level of 14C or 35S that can
be detected in 24 h is about 400 dpm cm\2. While
Suorography is necessary to detect 3H on polyacryl-
amide gels, 3H exposure can be detected directly
on electroblots using autoradiography, although
2;104 dpm cm\2 is required for detection in 24 h.
The efRciency of detection for all isotopes is enhanced
if Suorography is employed (100 dpm and
500 dpm cm\2 for 14C/35S and 3H, respectively).
Future Developments
PAGE, in particular 2D-PAGE, remains the method
of choice for the separation of complex protein mix-
tures. This has necessitated the development of highly
sensitive protein visualization protocols incorporat-
ing both nonradioactive and radioisotopic imaging
Detectors for Capillary Electrophoresis
Thomas Kappes and Peter C. Hauser, University of
Basel, Switzerland
Copyright ^ 2000 Academic Press
1239
methodologies. The development of methods for the
transfer of polypeptides from gels to membranes
where they are readily accessible to react with stains,
speciRc antibodies, group-speciRc ligands and de-
tailed structural characterization, including amino
acid microsequencing and mass spectral analysis, has
permitted the identiRcation of previously unidentiRed
proteins. Further developments are likely to take
place in low background staining polyacrylamide for-
mulations and modiRed membrane support matrices
in which proteins may be bound either covalently or
which form reversible covalent bonds. Such proteins
can be easily and selectively eluted for more detailed
biochemical studies. Future advances are likely to
take place in the development of more sensitive and
group-speciRc dyes/stains and increased speed and
sensitivity of detection systems such as the enhanced
bioluminescent and chemiluminescent systems, as
well as the development of faster and more sensitive
photographic detection Rlm.
Further Reading
Bonner WM (1983) Use of Suorography for sensitive iso-
tope detection in polyacrylamide gel electrophoresis and
related techniques. Methods in Enzymology 96:
215}222.
Dunbar BS (ed.) (1994) Protein Blotting, p. 242. New
York: Academic Press.
Gershini JM (1988) Protein blotting: a manual. Methods in
Biochemistry Analysis 33: 1}58.
Hames BD and Rickwood D (eds) (1990) Gel Electrophor-
esis of Proteins, p. 383. New York: IRL Press.
Laskey RA and Mills AD (1977) Enhanced autoradio-
graphic detection of 32P and 125I using intensifying
screens and hypersensitized Rlm. FEBS Letters 82:
314}316.
Merril CR (1990) Gel staining techniques. Methods in
Enzymology 182: 477}488.
Towbin H, Staehelin T and Gordon J (1979) Elec-
trophoretic transfer of proteins from polyacrylamide
gels to nitrocellulose sheets. Procedure and some ap-
plications. Procedures of the National Academy of
Science (USA) 76: 4350}4354.
Wirth PJ and Romano A (1995) Staining methods in gel
electrophoresis, including the use of multiple detection
methods. Journal of Chromatography (A) 698:
123}143.
Detection is a particularly critical issue in capillary
electrophoresis (CE) because of the extremely small
cell volumes available. Considerable effort has gone
into overcoming this limitation and a bewildering
1240 II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis
variety of methods has been described, ranging from
the straightforward adaptation of existing
chromatography detectors to less obvious and highly
experimental techniques. The path has not been
smooth, but many obstacles have turned out to be
less serious than anticipated. Optical methods have
proved very useful despite the short pathlengths
involved. Electrochemical detection methods, intuit-
ively considered incompatible with the applied high
voltage and long neglected in favour of optical means,
are now readily implemented. However, the develop-
ment of detection methods is still in Sux and it may
take several more years before maturity is reached
and different methods have found their established
roles for particular applications.
The Detection Challenge
The internal diameters of the capillaries employed in
CE range from 100
m down to about 5
m and
a single analyte zone is approximately 1 mm long.
Because the detection volume has to be smaller than
the peak volume available, detection volumes range
from about 1 pL to 1 nL. In high performance liquid
chromatography (HPLC), in contrast, detection vol-
umes of at least 1
L are available. One would there-
fore expect sensitivities for CE to be several orders
of magnitude lower than those in HPLC and, as
a consequence, the detection limits to be much
inferior. However, in CE the sample does not experi-
ence signiRcant dilution before it reaches the detector,
as is the case in HPLC, because of the Sat Sow proRle
in CE. Therefore, the sensitivities are not in fact as
signiRcantly degraded in comparison with HPLC as
might be expected. Nevertheless the issue of detection
limits is still critical in CE and detector sensitivity is
not always adequate. Preconcentration by electros-
tacking is sometimes advocated, but this method is
only possible for samples with low ionic strength and
generally leads to poor precision unless an internal
standard is employed.
Because of the small detection volumes, on-column
detection schemes are required to avoid band broad-
ening, rather than detector cells attached in an off-
column arrangement as is the case in chromatogra-
phy. A unique property of detection in electrophor-
esis, which is not shared with chromatography, is the
fact that there is a dependence of the peak area
(expressed on a time basis) on migration velocity.
However in practice this is usually of no concern.
Detection methods may be grouped according to
whether a bulk property of the solution (such as
conductivity, refractive index) or a speciRc attribute
of the analytes (such as optical absorption or Suores-
cence, redox activity or membrane permeability) is
monitored. Detectors used in the Rrst case tend to be
more universal, but generally suffer from the presence
of a large background signal against which small
changes have to be distinguished. This often leads to
poor signal-to-noise (S/N) ratios and hence relatively
high detection limits. The exploitation of speciRc in-
teractions is generally better in this regard, but each
method is usually only applicable to a certain class of
analytes. Some of the speciRc detection methods also
allow additional information on the analyte to be
gathered, which may be desirable as migration times
can never be taken as absolute proof of identity.
These detectors may be termed ‘information rich’,
and include for example mass spectrometers, photo-
diode arrays and voltammetric detectors.
Important general characteristics of detectors are
their sensitivity, dynamic range, and linearity. The
term sensitivity generally denotes the gradient of the
calibration curve but the precision of the measure-
ment (S/N ratio) has to be considered as well for
a complete evaluation. Often, the term sensitivity is
used to indicate the lowest concentration that may be
detected (limit of detection, LOD) and these para-
meters are of course interrelated. In CE detection
limits are sometimes quoted in terms of the detectable
mass or number of moles, as impressive Rgures in the
pico- or atto-gram or -mole range can be given be-
cause of the small sample volumes used. However,
the standard concentration limits are much more use-
ful and meaningful. The dynamic range is en-
compassed by the detection limit and by a maximum
where a loss of sensitivity occurs. Wide dynamic
ranges are desirable as they simplify sample prepara-
tion, but they often go hand-in-hand with relatively
poor precision. The upper concentration limit in cap-
illary electrophoresis is generally determined by the
ionic strength of the background buffer (typically
1}10 mmol).
The choice of detector is guided by the requirement
of the application in terms of detection limit, selectiv-
ity and information requirements but to a large de-
gree also by commercial availability, cost, robustness
and ease of use. Some features of the major detection
methods are summarized in Table 1.
Optical Methods
Optical detection methods are more widely employed
than any other detection means. Commercial CE in-
struments with optical absorption detectors were in-
troduced in 1989 and are available from a variety of
instrument manufacturers. The detectors employed
have often been adapted from devices used in HPLC
and this may be part of the reason for the prevalence
of the ultraviolet (UV) absorption detection method.
 II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis 1241

Table 1 Main detection methods for capillary electrophoresis

Method Features Detection limits a (mol L\1)

UV/Vis absorption Readily available commercially 10\7

Indirect UV/Vis absorption Compromise with poorer detection limits for nonabsorbing species 10\5
such as most inorganic ions
Fluorescence Good detection limits but most species require derivatization; 10\9
available commercially
Laser fluorescence Elaborate; excellent detection limits; available commercially 10\11
Conductometry Good for small ions; available commercially 10\6
Amperometry Simple, but only possible for electroactive ions; not available 10\8
commercially
Mass spectrometry Provides information on peak identity; expensive; interfaces 10\8
available commercially

a
The values given should be considered as rough guides only, as these are often very much dependent on species and instrumental
set-up. UV/Vis, ultraviolet}visible.

Fluorescence-based detectors are not as widely used
but are also on the market.
To carry out on-column detection the usual poly-
imide protection coating has to be removed from the
column by burning, by dissolution with hot sulfuric
acid, or by mechanical scraping, to form a window
into the capillary. The material is fairly brittle, so
that care has to be taken to avoid breakage once the
protective cladding has been removed. Fused silica
capillaries are transparent even below 200 nm, so
that the near-UV range is readily accessible.
The basic cell arrangement for absorbance
measurement through a capillary is illustrated in
Figure 1. Generally, besides the light source, there is
a monochromator or optical Rlter to deRne the
wavelength employed, a lens and aperture, and
a photodetector. Variations of this arrangement are
possible. Most commonly wavelengths in the UV
range from about 250 nm down to 185 nm are em-
ployed, using different types of sources such as
deuterium lamps, but instruments that include the
visible range are also available. Variable wavelength
as well as Rxed wavelength arrangements are in use. It
is important to get a high light intensity transmitted
onto the detector for best S/N ratio. The usual UV
light sources, such as deuterium lamps, are larger
than the optical cell and it is only possible to focus
Figure 1 Schematic representation of absorbance detector.
1, Light source; 2, monochromator or optical filter; 3, lens; 4,
aperture; 5, capillary; 6, photodiode or photomultiplier tube.
a small fraction of the radiation emitted through the
cell even with the best available lenses. Ball lenses,
mounted directly adjacent to the capillary, are often
employed. Apertures are required to minimize the
amount of stray light reaching the detector. Optical
Rbres can be used for transmission of the radiation
as this allows efRcient electrical shielding of the
photodetector and at the same time the distal ends
form the optical apertures. Absorbance detectors
based on light-emitting diodes (LEDs) and laser
diodes have also been demonstrated. These devices
give high baseline stability because of the absence of
Sicker noise present in discharge lamps and allow the
construction of battery operated instruments because
of their low power consumption. However, these de-
vices are not available for the UV wavelength range.
The circular cross-section of the capillary is far
from ideal for absorbance measurements because it is
not possible to pass collimated light through the
interior of the tube without refraction. This means
that changes in the refractive index of the solution are
a potential source of interference. In practice, how-
ever, the only serious limitation appears to be the
short optical pathlength, which leads to low sensi-
tivity according to the Lambert}Beer law. For this
reason the largest capillary diameters that allow
efRcient cooling are usually employed in absorbance
detection, typically with an internal diameter of
50}75
m. Different methods of increasing the sensi-
tivity in absorbance detection have been described.
These include the use of rectangular capillaries, capil-
laries bent in a Z-shape to obtain a longitudinal light
path, multipass cells by multiple reSections in silver-
coated capillaries, and so-called bubble cells formed
on the capillary itself. Only the last approach is reas-
onably easily implemented, and it appears to be the
only one that is commercially available (albeit at
a cost much higher than that of ordinary capillaries).
1242 II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis
The internal diameter of the capillary is widened in
the detector region by a factor of about three, thereby
increasing the sensitivity by the same magnitude.
A different approach to increase the sensitivity of
absorption measurements is the use of thermooptic
methods. Here the heat evolved following the absorp-
tion of light is sensed indirectly. In the thermal lens
method the refraction of a laser light beam is mea-
sured, using a relatively simple arrangement. Two
light beams perpendicular to each other are em-
ployed. One of the beams is of a wavelength that is
absorbed by the analytes. The heat evolved through
absorption of light leads to a refractive index gradient
in the capillary which is monitored by the second
beam. A variation on this technique has been re-
ported that uses intensity-modulated light. This leads
to a vibration of the capillary that may again be
detected with a second probe beam. Ordinary refrac-
tive index detection has also been described using the
deSection of a laser beam but neither of these two
techniques has gained much acceptance.
Most organic analytes possess chromophoric
groups that show intrinsic absorbance in the near}UV
range, so most methods are based on this wavelength
region. It has been demonstrated that for these species
it is preferable to use wavelengths that are as short
as possible (below 200 nm) for the best sensitivity.
Photodiode-array detection is also possible. This
technique yields additional qualitative information
on the identity of the detected species and allows peak
inhomogeneity to be detected. However, the S/N
ratio and therefore the detection limit, which is al-
ways critical in CE, are degraded because of the
reduced integration time available, and the method
requires considerable computing power, because of
the large amount of data acquired. Ions that do not
show absorbance in the UV/Vis range, such as inor-
ganic species or completely saturated organics, may
be determined by indirect methods. These methods
rely on the displacement of dye molecules of equal
charge as the analyte species (to maintain electroneu-
trality) so that a decrease in absorbance is detected.
This is more demanding on the stability of the system
than the direct absorbance method and the detection
limits are generally higher. However it is the only
method employing optical absorbance detectors to be
available for most inorganic anions. Chromate is of-
ten used as the background ion but other species,
some for the visible wavelength range, have also been
reported. Inorganic cations can also be detected by
indirect means, but many of them are best determined
via the formation of coloured complexes using non-
discriminating ligands.
Fluorescence detection is also possible and is com-
mercially available. However, few species display in-
trinsic Suorescence, so derivatization reactions have
to be employed. Derivatization may be classiRed as
pre-column, on-column or post-column according to
the scheme employed. Fluorescence has the great ad-
vantage of much higher sensitivity than absorbance
measurements. Detection limits approaching single
molecule detection have been achieved. Lasers appear
to be ideal light sources for Suorescence measure-
ments, as the light is produced in a tightly focused
beam well matched to capillaries, but inexpensive
sources are not available for the UV range and
available lasers are often plagued by insufRciently
stable output intensities. This limits their use, espe-
cially for the more universal indirect detection
scheme. Nevertheless, impressive results have been
obtained for microbiological applications (e.g. in
neuroscience) that include the analysis of single cells.
Chemiluminescence detection is usually based on
the inSuence of analytes on the efRciency of one of
several available chemiluminescence reactions. The
achievable sensitivities are very high, a feature this
method has in common with Suorescence. Its imple-
mentation is similar to post-column Suorescence
detection in that a pumped reagent stream has to be
merged with the column efSuent in a suitable small-
scale mixing device prior to detection in a light-tight
enclosure with a photomultiplier tube.
Electrochemical Methods
Electrochemical detection techniques in general are
developing rather more slowly than optical tech-
niques, even though some of the earliest examples of
open tubular electrophoresis were based on elec-
trochemical detection. It was considered that the
electrical Reld applied for separation was a serious
hindrance. Also, the exposure of the detector to the
buffer solution (which is not the case for the optical
methods) is a potential source of problems as the
electrodes may corrode or be affected in other ways.
While the common optical detection methods have
reached maturity, the same cannot be said for the
electrochemical methods. Of the three reported
methods, namely conductimetry, amperometry and
potentiometry, the former is the only one that is
commercially available at this time. Nevertheless,
these methods have attracted considerable attention
and in general are much simpler than other methods.
In an early approach to conductivity detection, the
cell was formed by drilling a hole perpendicularly
through the capillary with a laser and then inserting
two small wires that faced each other. In this way the
two detector electrodes were not exposed to a voltage
gradient. Another approach, which is still used by
some workers for amperometric detection, is to
 II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis
Figure 2 Common arrangements for electrochemical detec-
tion. (A) Decoupled configuration with on-column detection using
a micro-electrode. (B) Wall-jet arrangement possible with capillar-
ies with internal diameters of 50
m or less, showing a relatively
large electrode at a suitable distance from the capillary end.
decouple the detector from the electrical Reld. This is
achieved by creating a small gap in the capillary
and using a sleeve typically made of an ion exchange
membrane to provide a contact to the electrical earth.
This arrangement is illustrated in Figure 2A. The
analytes are pushed forward to the detector elec-
trode(s) by the pressure created by the electroosmotic
Sow. This arrangement is ideal in an electrical sense
but cumbersome to implement. However, it was later
realized that if electrodes are positioned immediately
outside the end of capillaries that have internal
diameters of 50
m or less, then the electrical bias on
the detector is minimal. This arises because the
cross-section of the liquid volume outside the capil-
lary is considerably larger than that inside, so that the
remaining voltage drop between the end of the capil-
lary and the electrophoretic earth electrode located
a few millimetres away is a few hundred millivolts
only. Inside the capillary voltage drops of 30 V mm\1
are typically encountered. Also, the electrical current
through the capillaries is considerably lower for
smaller internal diameters. This so-called wall-jet ar-
rangement (Figure 2B) is the one used in commercial
instruments for conductivity detection, and this same
conRguration is also frequently employed for the
other two electrochemical detection methods. A fur-
ther feature of the wall-jet conRguration is the use of
electrodes with diameters larger than the internal
diameter of the capillary, which was found to be
possible without signiRcant loss of peak resolution.
1243
This allows the construction of relatively simple cells
for the alignment of capillary ends and electrodes.
In conductimetric detection it is essentially the
same property which is responsible for the separ-
ation, namely the mobility in the electrical Reld giving
rise to a detector signal. This means that in principle
any species that can be separated by CE may be
detected by conductimetry. However, the need for an
electrolyte in the running buffer leads to the presence
of a background signal against which the analyte
signal has to be measured. As the analyte displaces
ions of the same charge (the same feature exploited in
indirect absorbance detection), it is the difference in
conductivity (caused by a difference in mobility be-
tween background and analyte ions) that leads to
a detectable signal. To optimize the sensitivity the
conductivity of the background buffer should be low,
a requirement that conSicts with the need for match-
ing the mobility of the buffer to that of the analytes to
prevent peak tailing or fronting. A compromise there-
fore has to be made. For analytes with low conductiv-
ity indirect detection may be employed using a back-
ground electrolyte with high conductivity.
Conductivity detection can also be carried out in
a contactless conRguration with two tubular elec-
trodes placed over the capillary. These then form
capacitors (albeit with small capacitance values) with
the liquid, whose conductivity can be probed with an
applied high frequency alternating current. Electrode
degradation is prevented in this mode. The sensitivity
of conductivity detection can be improved by the
so-called suppressed detection technique, known
from ion chromatography, in which the background
conductivity is largely removed by using a weak acid
or base that is rendered neutral by ion exchange
before the detection cell. However, for CE an ar-
rangement similar to that used for electrical decoup-
ling is required for suppression. This is difRcult to
implement and the method has not found wide use.
Amperometric detection may be employed for ions
that are electroactive, i.e. that can be reduced or oxi-
dized at electrodes. Different classes of species show
this property, including heavy metal ions, certain
inorganic anions, and many different organic molecu-
les that incorporate reactive groups such as phenols,
aldehydes, amines, etc. Many applications of am-
perometric detection have been reported but these
have certainly not been fully explored yet. As the
detection limits of amperometry tend to be good this
approach is useful when low concentrations are to be
determined. Please note that the terms ‘electrochemi-
cal detection’ and ‘EC detection’ are often employed
with the sole connotation of amperometric detection,
a usage that has evolved in the context of HPLC
detection methods. This may lead to confusion as
1244 II / ELECTROPHORESIS / Detectors for Capillary Electrophoresis
conductimetry and potentiometry clearly are electro-
chemical methods as well.
At this stage detector cells for amperometry have to
be constructed in-house as (at least to our knowledge)
no commercial units are available. However, all the
other parts required to set up a CE instrument, in-
cluding potentiostats to operate the detector, are
available commercially in modular form. In the wall-
jet arrangement, the voltage applied to the working
electrode by the potentiostat circuitry is superim-
posed by a voltage bias that is not only dependent on
the applied separation voltage but also on parameters
such as buffer composition, capillary diameter and
the exact position of the electrode. For this reason
some workers continue to use the decoupled detector.
Amperometric detection in CE in principle requires
a total of four electrodes at the detector end of the
column. Besides the detector electrode (the working
electrode of the potentiostat circuitry) and the elec-
trophoretic earth, a reference electrode and a counter
(or auxiliary) electrode are required. It is possible to
simplify this conRguration by employing the elec-
trophoretic earth as a pseudo-reference and as a
counter electrode as well.
Different electrode materials may be used in am-
perometric detection including gold, platinum and
glassy carbon to suit different applications. The use of
copper wire electrodes has proved to be useful as
several oxidation reactions are catalysed on this mater-
ial. Pulsed amperometric detection (PAD) may be em-
ployed when reaction products lead to a fouling of the
electrode. Voltammetric detection in which the applied
electrode potential is swept rapidly and repeatedly
over the range of interest to gain additional informa-
tion on the peak identity via the redox potential is also
possible. Somewhat higher detection limits may have
to be accepted, however, for these pulsed methods.
Potentiometric detection with ion selective elec-
trodes is the least reported of the three electrochemi-
cal detection methods. The matching of a separation
method with a sensor (rather than a detector which
by deRnition is not selective) may appear to be a con-
tradiction, but ion selective electrodes are in fact
rarely highly selective and may be tailored to be
responsive to a range of ions. So-called Hofmeister
electrodes discriminate solely on the basis of the lipo-
philicity of the anions or cations, and are therefore at
least in principle well suited for the determination of
singly charged organic species. Early reports on this
technique were based on micropipette ion selective
electrodes known from physiological studies on single
cells. These electrodes consisted of glass capillaries,
with tip diameters of a few micrometres, that were
Rlled with a viscous organic solvent incorporating an
ionophore and acted as ion selective membranes.
However, these electrodes were not very robust and
have now been superseded by more reliable miniature
coated-wire ion selective electrodes. These detectors
have been used to detect a variety of inorganic and
organic species that otherwise could not be detected
with CE, or could only be detected with difRculty. It
is possible to use a copper wire electrode as a simple
potentiometric detector for amino acids in CE.
In summary, the three electrochemical methods
may be considered to be complementary. Conductiv-
ity detection is a versatile general method that works
best for small ions of high mobility. Amperometric
detection is useful for electroactive ions and good
detection limits can be expected. Potentiometric de-
tection has currently been relatively poorly explored,
but may prove to be a useful alternative for large,
singly charged ions that cannot be detected am-
perometrically or by direct optical absorption
measurements.
Other Methods
Detection methods other than optical or electro-
chemical methods have also been reported. One
method that is fairly widely used is mass spectromet-
ric detection, including detection by inductively-
coupled plasma mass spectrometry, and commercial
interfaces are available. Detection by nuclear mag-
netic resonance is also an established technique.
These methods are covered elsewhere in this encyclo-
pedia. Radioisotope detection has been reported and
good detection limits have been achieved.
Future Developments
The acceptance of CE depends to a large extent on the
availability of suitable robust detection methods with
good detection limits. Shortcomings appear to exist
for trace analysis of organic species that do not Suor-
esce, and for inorganic species. Thermooptic and
chemiluminescence methods are promising in this re-
gard as are the electrochemical methods of am-
perometry and conductivity. The last two methods
appear to have reached some degree of maturity and
it is hoped that these will become more readily avail-
able commercially.
See also: II/Chromatography: Liquid: Detectors: Ultra-
violet and Visible Detection; Detectors: Mass Spectro-
metry; Detectors: Fluorescence Detection.
Further Reading
Baker D (1995) Capillary Electrophoresis. New York: John
Wiley.
 II / ELECTROPHORESIS / Discontinuous Electrophoresis
Doble P and Haddad PR (1999) Indirect photometric detec-
tion of anions in capillary electrophoresis. Journal of
Chromatography A 834: 189.
GarcmH a Campan a AM, Baeyens WRG and Zhao Y (1997)
Chemiluminescence detection in capillary electrophor-
esis. Analytical Chemistry 69: 83A.
Jandik P and Bonn G (1993) Capillary Electrophoresis
of Small Molecules and Ions. Weinheim: VCH Pub-
lishers.
Landers JP, ed. (1997) Handbook of Capillary Electrophor-
esis. Baton Rouge: CRC Press.
Li SFY (1993) Capillary Electrophoresis, Principles, Prac-
tice and Applications. Amsterdam: Elsevier.
Liu BF, Liu LB and Cheng JK (1999) Analysis of inorganic
cations as their complexes by capillary electrophoresis.
Journal of Chromatography A 834: 277.
Discontinuous Electrophoresis
M. J. Doktycz, Oak Ridge National Laboratory,
Oak Ridge, TN, USA
Copyright ^ 2000 Academic Press
Introduction
Electrophoresis is one of the most powerful tools in
the arsenal of separation scientists. It is commonly
employed in the Reld of biochemistry, where separ-
ation of complex mixtures of proteins or nucleic acids
is a continuing challenge. Numerous variants of elec-
trophoresis have been described with the goal of
optimizing the speed and effectiveness of the separ-
ations. One important electrophoretic variable is the
separation matrix. It provides the retarding forces, or
sieving qualities, that counter the electrophoretic
transport. These forces can ultimately effect the sep-
aration and can be altered by the matrix type or
concentration. Cross-linked or linear forms of poly-
mers such as agarose or acrylamide are common
choices. Different formats for the electrophoresis me-
dium can also have dramatic effects on the resolution
and separation time. This is exempliRed by recent
uses of microcapillary formats which greatly speed up
separations. Another component that dictates the
speed and resolution of electrophoretic separations is
the charge-carrying buffer ion. The buffer is a univer-
sal component of electrophoresis, independent of gel
constitution or format. This component is often over-
looked, though attention to this aspect can be beneR-
cial in developing electrophoresis-based separation
techniques.
Proper buffer selection offers several practical ad-
vantages, including optimum separation times, in-
1245
Lucy CA and Wu Q (1998) Characteristics and calibration
of conductivity detection in capillary electrophoresis.
Journal of Chromatographic Science 36: 33.
Nouadje G, SimeH on N, Nertz M and Couderc F (1996)
ED lectrophorèse capillaire et deH tection par Suorescence
induite par laser. Analusis 24: 360.
Polesello S and Valsecci SM (1999) Electrochemical detec-
tion in the capillary electrophoresis analysis of inorganic
compounds. Journal of Chromatography A 834: 103.
Saz JM and DmH ez-Masa JC (1994) Thermo-optical spectro-
scopy: new and sensitive schemes for detection in capillary
techniques. Journal of Liquid Chromatography 17: 499.
Voegel PD and Baldwin RP (1997) Electrochemical detection
in capillary electrophoresis. Electrophoresis 18: 2267.
Weinberger R (1993) Practical Capillary Electrophoresis.
Boston: Academic Press.
creased band concentration and reduced effects of
diffusion. The resolution of closely migrating species
results from the proper choice of the pH, concentra-
tion, and type of buffer ion. These physical character-
istics deRne the conductivity of the electrophoretic
medium and affect the transport of the molecules to
be separated. Inorganic ions, such as chloride anion,
have high conductivities in comparison to the ionized
form of weak acids and bases. Such high mobility ions
offer little advantage when used for the elec-
trophoretic separation of large, less mobile bi-
omolecules but slower-moving ions, such as those of
weak acids or weak bases, are more useful choices.
These not only buffer the pH but, due to the slower
mobility of these ions, lead to better separation of
charged macromolecules.
Zonal electrophoresis utilizes a single buffer in the
gel and reservoirs. An alternative to the continuous
buffer, zonal separations is a discontinuous system
where multiple ionic components are used. The pres-
ence of multiple ionic components in electrophoresis
leads to discontinuities in the voltage gradient, pH
and ionic strength due to the different physical mobil-
ities of the ions involved. These different mobilities
lead to the formation of discrete zones of ions that,
under equilibrium conditions, travel at a constant
rate in an applied electric Reld. Adjustment of these
mobilities involves alteration of the ion concentration
and potential gradient of the zone. Sharp boundaries
can exist between these zones, with the ionic concen-
tration being dictated by the Kohlrausch regulating
function. The technique is similar or identical to
a number of electrophoresis techniques that are
known as discontinuous multiphasic, multizonal,
 II / ELECTROPHORESIS / Discontinuous Electrophoresis
Doble P and Haddad PR (1999) Indirect photometric detec-
tion of anions in capillary electrophoresis. Journal of
Chromatography A 834: 189.
GarcmH a Campan
a AM, Baeyens WRG and Zhao Y (1997)
Chemiluminescence detection in capillary electrophor-
esis. Analytical Chemistry 69: 83A.
Jandik P and Bonn G (1993) Capillary Electrophoresis
of Small Molecules and Ions. Weinheim: VCH Pub-
lishers.
Landers JP, ed. (1997) Handbook of Capillary Electrophor-
esis. Baton Rouge: CRC Press.
Li SFY (1993) Capillary Electrophoresis, Principles, Prac-
tice and Applications. Amsterdam: Elsevier.
Liu BF, Liu LB and Cheng JK (1999) Analysis of inorganic
cations as their complexes by capillary electrophoresis.
Journal of Chromatography A 834: 277.
Discontinuous Electrophoresis
M. J. Doktycz, Oak Ridge National Laboratory,
Oak Ridge, TN, USA
Copyright ^ 2000 Academic Press
Introduction
Electrophoresis is one of the most powerful tools in
the arsenal of separation scientists. It is commonly
employed in the Reld of biochemistry, where separ-
ation of complex mixtures of proteins or nucleic acids
is a continuing challenge. Numerous variants of elec-
trophoresis have been described with the goal of
optimizing the speed and effectiveness of the separ-
ations. One important electrophoretic variable is the
separation matrix. It provides the retarding forces, or
sieving qualities, that counter the electrophoretic
transport. These forces can ultimately effect the sep-
aration and can be altered by the matrix type or
concentration. Cross-linked or linear forms of poly-
mers such as agarose or acrylamide are common
choices. Different formats for the electrophoresis me-
dium can also have dramatic effects on the resolution
and separation time. This is exempliRed by recent
uses of microcapillary formats which greatly speed up
separations. Another component that dictates the
speed and resolution of electrophoretic separations is
the charge-carrying buffer ion. The buffer is a univer-
sal component of electrophoresis, independent of gel
constitution or format. This component is often over-
looked, though attention to this aspect can be beneR-
cial in developing electrophoresis-based separation
techniques.
Proper buffer selection offers several practical ad-
vantages, including optimum separation times, in-
1245
Lucy CA and Wu Q (1998) Characteristics and calibration
of conductivity detection in capillary electrophoresis.
Journal of Chromatographic Science 36: 33.
Nouadje G, SimeH on N, Nertz M and Couderc F (1996)
ED lectrophorèse capillaire et deH tection par Suorescence
induite par laser. Analusis 24: 360.
Polesello S and Valsecci SM (1999) Electrochemical detec-
tion in the capillary electrophoresis analysis of inorganic
compounds. Journal of Chromatography A 834: 103.
Saz JM and DmH ez-Masa JC (1994) Thermo-optical spectro-
scopy: new and sensitive schemes for detection in capillary
techniques. Journal of Liquid Chromatography 17: 499.
Voegel PD and Baldwin RP (1997) Electrochemical detection
in capillary electrophoresis. Electrophoresis 18: 2267.
Weinberger R (1993) Practical Capillary Electrophoresis.
Boston: Academic Press.
creased band concentration and reduced effects of
diffusion. The resolution of closely migrating species
results from the proper choice of the pH, concentra-
tion, and type of buffer ion. These physical character-
istics deRne the conductivity of the electrophoretic
medium and affect the transport of the molecules to
be separated. Inorganic ions, such as chloride anion,
have high conductivities in comparison to the ionized
form of weak acids and bases. Such high mobility ions
offer little advantage when used for the elec-
trophoretic separation of large, less mobile bi-
omolecules but slower-moving ions, such as those of
weak acids or weak bases, are more useful choices.
These not only buffer the pH but, due to the slower
mobility of these ions, lead to better separation of
charged macromolecules.
Zonal electrophoresis utilizes a single buffer in the
gel and reservoirs. An alternative to the continuous
buffer, zonal separations is a discontinuous system
where multiple ionic components are used. The pres-
ence of multiple ionic components in electrophoresis
leads to discontinuities in the voltage gradient, pH
and ionic strength due to the different physical mobil-
ities of the ions involved. These different mobilities
lead to the formation of discrete zones of ions that,
under equilibrium conditions, travel at a constant
rate in an applied electric Reld. Adjustment of these
mobilities involves alteration of the ion concentration
and potential gradient of the zone. Sharp boundaries
can exist between these zones, with the ionic concen-
tration being dictated by the Kohlrausch regulating
function. The technique is similar or identical to
a number of electrophoresis techniques that are
known as discontinuous multiphasic, multizonal,
1246 II / ELECTROPHORESIS / Discontinuous Electrophoresis

displacement, isotachophoresis and moving bound-
ary electrophoresis. The primary advantages of dis-
continuous buffer systems over continuous buffer,
zonal separation are their ability to concentrate dilute
samples, enhance resolution between closely migra-
ting species and provide deRned reference fronts.
Implementation
Most of the electrophoretic techniques that exploit
the differential migration of ions in an electric Reld
trace their origins to the work of Kohlrausch. In
1897, Kohlrausch presented the equations that de-
scribe ionic migration in an electric Reld. In later
decades, Tiselius furthered electrophoretic techniques
and accomplished the separation of serum proteins
using moving boundary electrophoresis. In 1958,
Poulik published the use of the discontinuous buffer
technique for the separation of proteins in a starch gel
and, a few years later, Ornstein and Davis described
the theory of discontinuous buffer systems for the
separation of serum proteins in polyacrylamide gels.
The Ornstein and Davis ‘disc’ electrophoresis system
concentrated the sample by placing it between buffers
of different mobility and then performed the separ-
ation using zone electrophoresis. In 1965, Richards
applied the technique for the separation of nucleic
acids. Since this period, discontinuous buffer systems,
or similar techniques, have seen consistent usage in
protein separations. Theoretical treatments that aid
in the design of appropriate buffer systems have been
extended. This has led to the description of various
buffer systems for the separation of both acidic and
basic proteins and the development of ‘spacer’ ions
for resolving closely migrating proteins. The applica-
tion to nucleic acid separations has been sporadic.
This is perhaps due to the lower reliance on charge
differences for the separation of nucleic acids. In
contrast to proteins, nucleic acids have a near con-
stant mass-to-charge ratio giving them a common free
electrophoretic mobility. Typically, zonal gel systems
that fractionate solely on size are used. However,
discontinuous buffer systems do offer advantages for
certain nucleic acid separations.
In practice, a leading ion is chosen such that the
mobility of this ion is greater than all others in the
system. This ion is incorporated into the separation
matrix (or in some applications, an anti-convective
matrix). The trailing ion(s), with a mobility slower
than the leading ion, is placed in the top buffer reser-
voir while a counterion, common to all zones, is
placed in the bottom buffer reservoir. For open-faced
gel systems, the samples are loaded directly on to the
separation matrix, while for vertical slab systems the
sample can be loaded on top of the gel. As the moving
boundary is established, the sample will be concen-
trated in a very thin moving boundary. The analytes
to be separated will stack themselves in the order of
their ionic mobility and, in the case of discontinuous
electrophoresis, may be further fractionated by the
use of a sieving gel. The resultant bands are easily
detected with conventional techniques, using stains or
Suorescent or radioactive labels. A description of
a typical experimental set-up, and the identity of the
various ionic species for an anionic separation, is
shown in Figure 1.
Only the concentration of the leading ion can be
chosen freely. During electrophoresis, the ionic con-
centration and potential gradient of the trailing zones
will be regulated to compensate for the lower mobil-
ity. This regulating function is known as the Kohl-
rausch regulating function. Under equilibrium condi-
tions, the leading and trailing ions migrate at the
same rate (isotachophoresis). This causes the ion
concentration in a trailing zone to be lower, and the
voltage gradient to be higher, than that in a leading

Figure 1 Pictorial description of the progress of an electrophoretic separation using a discontinuous buffer system. The initial set-up
is shown on the left. A trailing anion, B\, is placed in the cathodol buffer reservoir while the leading ion, A\, is placed in the gel. This
leading anion has a mobility that is greater than all other anions in the system. A common counterion, C#, is used throughout the
reservoirs and gel. The centre panel shows the sample being concentrated by the moving boundary created by the dissimilar anions. As
this boundary traverses through the gel, the sample components, with mobilities that are slower than the trailing anion, will be
fractionated in the trailing zone. Those ions with an intermediate mobility will be retained in the moving boundary. This is shown on the
right.
 II / ELECTROPHORESIS / Discontinuous Electrophoresis
zone. These differences compensate for the lower free
mobility. The migration rate of the ions can be fol-
lowed by observing the position of the boundary
between the zones. This boundary is self-sharpening,
as ‘trailing ions’, if present in the leading zone, will be
in a low Reld region and slow down, while ‘leading
ions’ in the trailing zone will be in a high Reld region
and speed up. This boundary can be conveniently
demarcated by incorporation of a low concentration
of a dye with a mobility intermediate between the
leading and trailing zone. Alternatively, conductivity
or thermal changes can be monitored to detect the
passing of the boundary. Multiple boundaries can
exist with each one deRning a distinct zone. Since the
concentration of ions in the zone is regulated, the size
of the zone will depend on the quantity of material
and the concentration of the leading ion. Thus, dilute
samples can be concentrated and different ions can be
sorted based on their mobilities.
Selection of appropriate buffer systems usually re-
quires some knowledge of the charge and mobility of
the analytes to be separated. Many buffer systems,
which span the entire range of pH values, have been
described. These systems are especially useful for the
separation of proteins that are charged under speciRc
pH values. The mobilities of the charged analytes to
be separated are also critical. For molecules that are
subsequently to be fractionated within the trailing
zone, the trailing ion should have a mobility that is
similar to or intermediate between the ions to be
separated. Tuning the mobility of the trailing ion to
that of the analyte can minimize the conductance
changes across the sample. This can lead to sharp
band proRles and enhance the resolution of closely
migrating species. Reference to previously tabulated
buffer mobilities will aid in the selection of an appro-
priate buffer system. The development of new buffer
systems can be accomplished empirically or through
simple application of the equations describing discon-
tinuous electrophoresis.
Advantages
The primary advantage of discontinuous buffer sys-
tems is the ability to concentrate the sample zone. The
passage of the moving boundary has the effect of
sweeping the sample into an extremely thin starting
band. For analytical applications this can lead to
reduced band widths and higher resolution separ-
ations. This feature has obvious advantages for the
characterization of closely migrating species. For
preparative applications, the result is the concentra-
tion of dilute samples. Another advantage of discon-
tinuous buffer systems is their use as an analytical
tool for deRning relative mobilities. The mobility of
1247
the ion front is easily deRned and can serve as a refer-
ence for deRning relative mobilities. Furthermore, the
mobility of the front is reproducible and independent
of the gel matrix. These features can be convenient
for analytical and forensic applications. Additionally,
the mobility of sample ions can be altered by chang-
ing the mobility of ions in the trailing zone. This can
lead to the tailoring of separations by deRning the size
range that can be fractionated. Sample ions that are
not of interest can remain trapped in an ion front,
allowing examination of particular ions. Finally, dis-
continuous buffer systems are compatible with vir-
tually any gel format, as its use is independent of the
gel matrix or the physical format of the gel. Since it
alters only the buffer system, the only complication is
selecting an effective buffer. Fortunately, the charac-
teristics of many buffer systems have been described
and designing new systems is a straightforward task,
as described below.
Figure 2 exempliRes some of the features of discon-
tinuous buffer systems for nucleic acid separations.
Shown here is a portion of a DNA sequencing gel
using a denaturing formate}glycine discontinuous
buffer system. An intermediate zone, with a mobility
between the formate leading ion and the glycine trail-
ing ion, was inadvertently created from contamina-
ting ions in the gel or sample loading buffer. For
analytical application, care should be taken to ensure
that extraneous ions are absent. However, the pres-
ence of this ion species demonstrates the effects of
stacking limits on the size selection and band concen-
tration of DNA sequencing products. DNA bands
smaller than 106 bases are trapped in the Rrst ion
front while DNA sizes 115}166 are trapped in the
second front. The DNA sizes that are trapped in the
ion fronts deRne the stacking limits and cannot be
separated because their mobility is between that of
the migrating ion zones. Since the DNA is not freely
electrophoresing, these stacking limits could be alter-
ed by changes to the gel matrix. The intermediate ion
zone demonstrates other features. Its size is directly
proportional to the amount of contaminating ion that
is present. Larger amounts of contaminating ion
would expand this zone. Additionally, the eight DNA
bands in this zone are much sharper than the bands
migrating behind the second front, even after the
30 cm migration distance. This is due to the concen-
trating effects of the leading ion front and the closely
matched mobility between these DNA bands and the
ions in the zone.
Theoretical Description
Calculation of the ion concentrations and basic char-
acteristics of the leading and trailing ion zones are
1248 II / ELECTROPHORESIS / Discontinuous Electrophoresis

electroneutrality:
# #
1 #C1 "0"C2 #C2
CA\ C B\ C
[3]

allows deRnition of the basic relationship between the

concentration of ions in the different zones and their
mobilities:

CA\
1 mA\(mB\!mC#)
B\ " [4]
C2 mB\(mA\!mC#)

This relationship can be deRned for every boundary

in a discontinuous electrophoresis system. The rela-
tionship is valid for strong electrolytes and weak
electrolytes. For weak electrolytes, the concentration
value represents the total of the ionized and un-
ionized forms. The fraction of the trailing species that
is ionized, X(n), can be determined after calculation
of the amount of counterion that crosses the bound-
ary into the trailing zone. When using a weak electro-
lyte as the counterion, a Rnal expression for the
counterion concentration in the trailing zone is:


# # mC# (mA\#mC#)
CC2 "CC1 #CA\
1 1! [5]
mA\ (mC##mB\)
Figure 2 The differential effects of stacking limits on the migra-
tion of DNA sequencing products on a 6% polyacrylamide gel where the counterion concentrations are the total
(containing 5% bis-acrylamide cross-linker) with a 50 mmoL L\1, concentration (both ionized and un-ionized forms).
pH 9.0 formate leading ion, glycine trailing ion. The sequence
between the two fronts, AAATTGTT, corresponds to fragment
The fraction ionized, X(n), can then be calculated
lengths of 107}114 bases; the sequence behind the front, CTGG, from the ion equilibrium constants of the trailing ion
corresponds to fragment lengths of 167}170 bases. Fragment and counterion. The net mobility accounts for the
lengths between 115 and 166 bases are concentrated in the actual transport of the trailing acid species and is
second front while fragment lengths(107 bases are concen- deRned as:
trated in the first front.

net mobility(n)"mnX[n] [6]

easily accomplished through use of a few simple
equations. The starting point is the Kohlrausch regu- The net mobility should be tuned closely to the
lating function (R) that describes the moving bound- sample’s mobility. Sample mobilities faster than the
ary condition: trailing ion net mobility will be retained in the ion
C front. Details regarding these equations and more
"Rn [1] rigorous deRnitions can be found in numerous refer-
m
ences.
where C is the concentration of the ion in zone n and A few characteristics of discontinuous buffer sys-
m is the mobility of the ion. These values are signed tems are apparent from examining the above equa-
according to the charge of the ion. For the set-up tions. First, the equilibrium concentration of the trail-
described in Figure 1, the regulating functions char- ing ion is independent of its initial concentration. It is
acterizing the two zones would be equal under mov- determined by the free mobilities of the ions in the
ing boundary conditions and: system and the concentration of the leading ion.
Therefore, only the leading ion conditions can be
# #
CA\
1 CC1 CB\
2 CC2 chosen freely. Second, since the trailing ion has
# " # [2] a slower mobility than that of the leading ion, the
mA\ mC# mB\ mC#
concentration of the trailing ion must be lower than
where the concentration subscripts denote the that of the leading ion. This bears directly on the
zone. This relationship, along with the condition for conductance and potential gradient of the trailing
 II / ELECTROPHORESIS / Discontinuous Electrophoresis 1249

zone. The zone conductance, , can be calculated
from the ion concentrations, free mobilities and Fara-
day’s constant (F). It can also be related to the Reld
strength (volts (V) per unit length (l)) through Ohm’s
law, as shown (where i is the current and A the
cross-sectional area of the gel):
"F [(net mobility(B\))(CB\
2 )
# i
#(net mobility(C#))(CC2 )]"
A(V/l)
Since the ion concentrations and ion mobilities in
the trailing zone are lower than that in the preceding
zone, the potential gradient in a trailing zone will be
higher than the leading zone. This is what allows the
trailing ions to migrate at the same rate as the leading
ions, despite their lower physical mobility. Tuning the
ionic strength is critical for optimizing separations. It
affects the size of the migrating zone, the speed of the
separation and the joule heating. The joule heating
further inSuences resolution. Therefore, a high poten-
tial gradient for a given current, or heat output, will
be preferable.
The above equations can be implemented on simple
spread sheet software to determine the physical char-
acteristics of discontinuous buffer systems. A few
example buffer systems are shown in Table 1. These
buffer systems have been designed to vary the trailing
ion type and the trailing ion net mobility while keep-
ing the ion speed constant. A Tris-formate buffer, at
a formate concentration of 50 mmoL L\1, has been
selected as a common leading ion. Tris is used as
a common counterion. The ion speed and voltage
gradient would be determined by the applied current.
These buffer systems exemplify some of the charac-
teristics of discontinuous buffer systems. As can be
seen, the trailing ion concentration is lower than the
leading ion concentration and a wide range of net
mobilities can be achieved when using a common
leading ion. Also notice that the net mobility, or
actual transport of the trailing ion, is lower than the
free mobility when using weak electrolytes. Such buf-
fer systems can be used to assess the stacking limits,
or mobility, of the analytes to be separated. Sample
ions that migrate in the ion front will have a mobility
intermediate between the leading and trailing ions.
Those sample ions that have a mobility slower than
the net mobility of the trailing ion will migrate more
slowly than the ion front and will electrophorese
[7]
within the trailing zone.
Other examples of calculated buffer systems are
shown in Table 2. These buffer systems vary the
trailing ion type while keeping the ionic strength of
the trailing phase constant. This is accomplished by
varying the leading ion concentration to allow for
a predetermined ion concentration in the trailing
zone. All the buffer systems in Table 2 use a formate
leading ion of varying concentration and have a com-
mon trailing ion concentration of 30 mmol L\1. Even
with a common trailing ion concentration, a 50%
change in the conductance can be obtained. This
allows for increased voltages to be applied without
increasing the current. As with the buffer systems
described in Table 1, a range of trailing ion net mo-
bilities is obtained. Other buffer systems, that deRne
different ranges of net mobilities, different pH values
or different conductivities, could be similarly cal-
culated to address particular separation problems or
analytical characterizations.
Outlook
Electrophoretic techniques that exploit the differen-
tial migration of ions are over 100 years old. Though
the applications and physical formats have evolved,
the underlying technique continues to endure. The
iteration of discontinuous buffer systems has had
intermittent use since its introduction over 30 years
ago. This is presumably due to the perceived added

Table 1 Examples of discontinuous buffer system using constant leading ion conditionsa

Trailing ion m Mobility Leading ion pH Concentration of Ion concentration Net mobility  Trailing zone
species (;104 cm2 V\1 s\1) trailing ion species (mmol L\1) (;104 cm2 V\1 s\1) conductance
(mmol L\1 ) (cm2 \1 mol\1)

Hepes 1.45 8.0 26.4 25.1 1.38 0.98

Tricine 2.18 8.0 33.6 27.1 1.76 1.25
Asparagine 2.80 8.5 38.2 23.2 1.70 1.20
Glycylglycine 2.85 8.0 38.5 27.8 2.06 1.47
Taurine 3.27 9.0 41.0 27.2 2.17 1.54
Glycine 3.74 8.0 43.4 9.4 0.81 0.58

a
The leading ion in all cases is 50 mmoL L\1 formate (ion mobility of 5.50;10\4 cm2 V\1 s\1) buffered with Tris (ion mobility of
2.60;10\4 cm2 V\1 s\1) to the indicated pH. The counterion in all cases is Tris.
1250 II / ELECTROPHORESIS / Electrochromatography

Table 2 Examples of discontinuous buffer systems with constant trailing ion concentrationa

Trailing ion species Concentration of Ion concentration Net mobility Trailing zone Leading ion
trailing ion species (mmol L\1) (;104 cm2 V\1 s\1) conductance concentration
(mmol L\1) (cm2
\1 mol\1) (mmol L\1)

Hepes 31.6 30 1.38 1.17 60

Tricine 37.2 30 1.76 1.38 55
Asparagine 49.4 30 1.70 1.56 65
Glycylglycine 41.5 30 2.06 1.58 65
Taurine 45.2 30 2.17 1.70 55
Glycine 138.5 30 0.81 1.83 160

a
The leading ion in all cases is formate (ion mobility of 5.50;10\4 cm2 V\1 s\1) at the indicated ion concentration. The counterion in all
cases is Tris (ion mobility of 2.60;10\4 cm2 V\1 s\1).

complexity of the technique or ignorance of the tech-
nique’s advantages. However, the beneRts of sample
stacking, mobility tailoring and an ionic reference
front are unchanged and unique when compared to
zonal buffer systems. With new applications and
challenges for electrophoretic separations, renewed
attention to the technique is certain.
Further Reading
Allen RC and Budowle B (1994) Gel Electrophoresis of
Proteins and Nucleic Acids. Berlin: Walter de Gruyter.
Allen RC and Doktycz MJ (1996) Discontinuous elec-
trophoresis revisited: a review of the process. Applied
and Theoretical Electrophoresis 6: 1}9.
Davis BJ (1964) Disc electrophoresis II. Method and ap-
plication to human serum proteins. Annals of the New
York Academy of Science 121: 404}427.
Doktycz MJ (1993) Discontinuous electrophoresis of DNA:
adjusting DNA mobility by trailing ion net mobility.
Analytical Biochemistry 213: 400}406.
Jovin TM (1973) Multiphasic zone electrophoresis. I.
Steady-state moving-boundary systems formed by
different electrolyte combinations. Biochemistry 12(5):
871}878.
Jovin TM (1973) Multiphasic zone electrophoresis. IV.
Design and analysis of discontinuous buffer systems
with a digital computer. Annals of the New York Acad-
emy of Science 209: 477}495.
Martin AJP, and Everaerts FM (1970) Displacement elec-
trophoresis. Proceedings of the Royal Society of London
A 316: 493}514.
Morris CJOR and Morris P (1976) Theoretical aspects of
electrophoresis. In: Separation Methods in Biochemis-
try, pp. 705}760. New York: Halsted Press.
Ornstein L (1964) Disc electrophoresis } I. Background and
theory. Annals of the New York Academy of Science
121: 321}349.
Richards EG and Lecanidou R (1971) Quantitative aspects
of the electrophoresis of RNA in polyacrylamide gels.
Analytical Biochemistry 40: 43}71.
Vesterberg O (1989) History of electrophoretic methods.
Journal of Chromatography 480: 3}19.
Yarmola E and Chrambach A (1995) Band width measure-
ment in automated gel electrophoresis apparatus: DNA
dispersion in a discontinuous system and in a single
buffer. Electrophoresis 16: 345}349.

Electrochromatography

T. Shafik and A. G. Howard, University of Southamp-
ton, Southampton SO17 1BJ, UK
Copyright ^ 2000 Academic Press
Introduction
The application of electric Relds to planar chromato-
graphic media in order to drive and/or enhance separ-
ations is as old as planar chromatography itself.
There is evidence to suggest that paper electrophor-
esis was Rrst performed several years prior to the Rrst
reports of paper chromatography. Research in the
Reld has been intermittent, with periods of consider-
able activity separated by long periods of inactivity.
This is at least partly due to attention being diverted
away from planar methods to modern high-efRciency
column techniques. It is quite conceivable that mod-
ern chromatography would be very different, with
a much stronger focus on planar techniques, had
more development work been carried out on thin
layer electrochromatography (TLE).
The subject is enjoying something of a revival with
signiRcant advances having been made during the
 II / ELECTROPHORESIS / Electrochromatography
1990s. If this trend continues, planar electro-
chromatography may take its place as a powerful tool
in the modern analytical laboratory offering new
modes of separation at speeds that are currently un-
available in conventional thin-layer chromatography
(TLC).
De\nitions
The term TLE will be used to refer to all techniques
carried out within chromatographically active thin
layers and with electric Relds being employed to inSu-
ence the separation. This will encompass a number of
different modes of separation, including those that
are dominated by electrophoretic analyte migration
but which, due to the nature of the layer, include
some element of chromatographic retention. Conven-
tional gel-based electrophoresis is therefore distin-
guished from TLE by the electrophoresis in TLE
being carried out in a chromatographically active
layer of material such as silica, alumina or cellulose.
TLE therefore excludes gel electrophoresis carried
out in cast gel slabs of materials such as polyacrylam-
ide that are commonly used for the separation of
proteins and DNA. Paper electrophoresis is now
largely obsolete and is not therefore covered.
Advantages of TLE
Thin-layer chromatography has proved to be one of
the most successful chromatographic techniques ever
devised. This success owes largely to its simplicity and
versatility compared with column techniques. The
main drawback of TLC is the low solvent velocity
achieved through capillary action. This leads to long
separation times and means that the optimum Sow
velocity (for chromatographic efRciency) is seldom
reached. Electric Relds can be used to cause the migra-
tion of the analyte, either by directly exerting a force
on the molecules, or by causing the eluting solvent to
Sow, carrying with it the analyte. This can result in
migration velocities of one to three orders of magni-
tude greater than those achieved through capillary
action and can introduce an electrophoretic compon-
ent to a separation, facilitating the separation of difR-
cult mixtures.
Electrophoresis
This is the migration of ionized species under the
inSuence of an electric Reld. Ions in an electric Reld
experience a force proportional to their charge, caus-
ing them to accelerate in the direction of the Reld.
Ions in solution, not interacting with a solid support,
reach a terminal velocity dependent on the magnitude
of the force and their interaction with the solvent.
Therefore, different ions will migrate through the
1251
solvent at different rates. This can be used to separate
different ions and is known as free solution elec-
trophoresis.
If the solvent is within a bed of chromatographi-
cally active media such as silica, the migrating ions
can interact with the solid support, adding a
chromatographic component to the separation. At
this point, it is simpler to refer to the separation as
electrochromatography, which encompasses both
electrophoretic and chromatographic effects.
The ionic mobility u can be expressed in terms of
the ion’s velocity v in a given electric Reld strength
E by u"v/E. This only applies to free solutions, since
migration through the channels of a solid support
follows a tortuous path that deviates from a straight
line by a quantity dependent upon the properties of
the solid support. The important point is that the
migration velocity, both in free solution and through
a bed of chromatographic media, is proportional to
the potential applied. In order to achieve greater
migration velocities, it is necessary to apply higher
potentials.
Electroosmosis
Electroosmotic Sow arises from the formation of an
electrical double layer at a solid}liquid interface. This
is due to the presence of charged species on the solid
surface; either in the form of surface ionized groups
(e.g. SiO\ in the case of silica) or because of the
preferential adsorption of ions from the solution. In
most cases it is a combination of both. The surface
charges are counterbalanced by ions in solution,
which form an immobile, strongly bound layer near
the surface and a mobile, solvated layer extending
into the liquid. Under the inSuence of an applied
potential, the solvated layer of counterions moves,
causing bulk solvent Sow (Figure 1).
This means of inducing solvent Sow has been suc-
cessfully applied to capillary column chromatogra-
phy, producing capillary electrochromatography
(CEC). Using electroosmotic Sow (EOF) to pump
solvent through a column generates a ‘plug’ Sow
proRle, which is distinct from the parabolic proRle
generated by hydraulic pumping. This results in re-
duced band broadening in CEC. It also allows the use
of very Rne chromatographic supports, which would
be impossible to use in pressure-driven systems owing
to back-pressure constraints. These factors combine
to give high linear Sow rates, of the order of
1 mm s\1, allowing very fast, efRcient separations.
As with electrophoretic migration, the EOF velo-
city increases linearly with applied potential. This
makes it generally desirable to apply higher potentials
in order to achieve faster migration rates.
1252 II / ELECTROPHORESIS / Electrochromatography
Figure 1 A schematic representation of double layer formation
at a silica surface.
Modes of Migration
In standard TLC, the migration of a sample molecule
is controlled by its interaction with the bed of
chromatographic media and the partition of the sol-
ute into the eluting solvent. In electrically driven
TLC, the solutes may be made to move in a number of
different ways.
If the sample molecules are ionic in the solvent
used, the application of an electric Reld will exert
a force on them and they will migrate electrophoreti-
cally. As they migrate they are also subject to
chromatographic partitioning between the solvent
and the stationary phase. The individual components
separate from each other by migrating at different
velocities, each compound having a characteristic mi-
gration velocity that reSects the conSict between elec-
trophoretic migration and chromatographic reten-
tion.
If, however, the sample molecules are uncharged in
the solvent, they will only migrate if the solvent is
made to Sow. Whereas in standard TLC this is
achieved by capillary action, in electrically driven
TLC, with the right solvent and adsorbent, it can be
achieved through EOF. The separation that results
from an electroosmotically driven TLE experiment is,
in the absence of electrophoretic effects, similar to
that obtained by conventional TLC, but it is obtained
much more rapidly.
There is a third mode by which solvent Sow can be
induced through a thin layer chromatographic media
during an electrochromatography experiment. When
a current Sows through a wetted layer of chromato-
graphic material the layer heats up. The power which
has to be dissipated from the plate depends on the
current Sow and hence the resistance of the wetted
plate. When the solvent is unevenly distributed
through the plate this leads to an uneven evaporation
of solvent from the plate causing capillary solvent
migration to occur as a direct result of Joule heating.
This effect is more extreme with vertically mounted
plates, which under gravity drain solvent to the base
of the plate. In some experiments, evaporative Sow
can be considerably larger than that which is gener-
ated by electroosmosis. This can be falsely identiRed
as EOF and is particularly evident in experiments
carried out with vertically mounted plates. In hori-
zontally mounted plates, with a solvent reservoir at
each end of the plate, solvent is replenished least
quickly at the center of the plate. If the rate of evapor-
ation is initially assumed to be uniform across the
plate, then the middle of the plate will dry out more
quickly. This leads to solvent Sow from both ends of
the plate towards the middle being superimposed
upon any EOF.
The current Sow through the solvent-wetted
chromatographic layer is dependent on the overall
electrical resistance of the plate, which is a function of
the ionic density in the solvent. These ions may orig-
inate from soluble ionic species in the chromato-
graphic material, dissolved ions in the solvent or
dissociated solvent molecules. The smaller the ionic
density, the higher the overall plate resistance and the
smaller the current. Unless adequate cooling is pro-
vided, the input of power will cause a temperature
rise in the thin layer. This will lead to evaporation of
the solvent, the rate of which will depend upon the
rate of power inSux, the volatility of the solvent and
degree of external cooling. This is the main limitation
controlling the magnitude of the potential that can be
employed in TLE to achieve faster migration rates.
Various methods of cooling the plates have been
used in order to reduce solvent evaporation. Immer-
sion of the plate in a solvent that is immiscible with
the eluting solvent has been employed in various
separations, with CCl4 being the most popular cool-
ant for aqueous eluent systems. The purpose of the
solvent bath is to provide direct cooling to the plate
surface. This approach was experimentally clumsy
and limited the range of analytes, since analyte solu-
bility in the ‘coolant’ must be considered. It was later
dropped in favour of the use of cooling pads in con-
tact with the TLE plate, achieving cooling rates in
excess of 0.1 W cm\2. With high-conductivity aque-
ous systems this arrangement allowed the applied
potential to be raised to 160 V cm\1, generating mi-
gration velocities of up to 0.1 mm s\1.
Solvents with limited volatility, such as higher alco-
hols, propylene carbonate and formamides have been
used to reduce evaporation. This approach did not
however gain popularity owing to several experi-
mental limitations, the most important of which is the
difRculty in removing the eluting solvent from the
 II / ELECTROPHORESIS / Electrochromatography
chromatographic material following an experimental
run. This is usually necessary in order to visualize the
separated compounds.
Historical Development
The development of TLE occurred in tandem with
that of paper electrophoresis (Table 1). This is not
surprising, since both techniques require similar ap-
paratus and reagents, and are generally used to
achieve the same types of separation. TLE has always
had an advantage over the paper technique in terms
of chromatographic performance. The Rner and more
uniform surface structure achievable on thin layers
results in considerably reduced band broadening
when compared with Rbrous media.
The earliest experiments were carried out in the
1940s using paper and layers of silica gel. A wide
range of analytes was separated, largely employing
aqueous systems. The high conductivity of the aque-
ous systems limited the applied potential to
10}50 V cm\1 but the electrophoretic separations
achieved at these potentials were still a considerable
improvement on those obtained by the equivalent
paper chromatography/TLC separations. The run
times were typically of the order of 1}3 h, but some
experiments, particularly protein separations, were
run for as long as 24 h.
The development of gas chromatography (GC)
and high-pressure liquid chromatography (HPLC)
Table 1 Developments in thin-layer electrochromatography
1937 Earliest recorded use of paper electrophoresis, sep-
aration of snake venom proteins followed by UV
detection (Konig)
1946 Earliest recorded use of ‘thin-layer’ electrophoresis,
in a slab of silica jelly. Method used for the separ-
ation of amino acids and peptides (Consden et al.)
1954 First use of electroosmotic flow as driving force to
effect a separation. Polysaccharides separated on
collodion membranes (Mould and Synge)
1961 Separation of amines and amino acids by thin-layer
electrophoresis (Honnegar)
1963 Investigation of the characteristics of solution flow in
thin-layer electrophoresis on a range of thin-layer
chromatography (TLC) media. (Kowalczyk)
1974 High-speed separation of organic compounds on
silica TLC plates and in columns in electric fields
(Pretorius et al.)
1994 Planar electrochromatography on non-wetted thin
layers (Pukl et al.)
1998 TLE separation of non-polar dyes on commercial
reversed-phase TLC plates using electroosmotic
flow (Nurok et al.)
1999 Quantification of electroosmotic and separation of
basic pyrimidines by thin-layer electrochromatogra-
phy (Howard and Shafik)
1253
enabled high-efRciency separations to be achieved
and diverted attention away from planar techniques.
This led to thin-layer electrophoresis/electro-
chromatography being largely abandoned in the late
1960s in favour of the column techniques. Very few
publications between 1970 and 1998 cite the use of
TLE.
Following a period of active research into planar
techniques between 1940 and 1960, interest in TLE
has been sporadic at best. Long periods of inactivity
have been punctuated by occasional reports of tech-
nical advances and/or applications of the technique.
Possibly one of the most important of these is a paper
by Pretorius et al., which described very high-speed
separations both on TLC plates and in columns utiliz-
ing EOF to mobilize solvent. This paper set a preced-
ent, by using nonaqueous and low aqueous solvents,
potential gradients of around 1000 V cm\1 could be
used } potentials at least Rve times greater than had
been previously employed. The experiments carried
out in columns were quickly followed up by several
other research groups and are considered the direct
predecessor of modern capillary electrophoresis (CE)
and CEC.
Experimental Techniques
Development Chambers
Historically, the majority of TLE experiments
have been carried out in the horizontal mode,
with solvent reservoirs at both ends of the plate.
This set-up continues to be used today by several
groups, and is shown in Figure 2. The plates are
supported with the chromatographic surface either
up or down.
Solvent is transported to and from the plates by
wicks, which also serve as electrical contacts. A range
of materials has been used as wicks, including Rlter
paper, sintered glass and felt. The electrodes are usu-
ally submerged in the solvent and made of an inert
conductive material, such as silver, stainless steel,
carbon or platinum.
A less popular approach involves the use of vertical
plates, with solvent at the base of the plate (Figure 3).
Electrical contact is made via the solvent at the base
Figure 2 A horizontal tank design for thin-layer electro-
chromatography.
1254 II / ELECTROPHORESIS / Electrochromatography
Figure 3 A vertical tank design for thin-layer electrochromato-
graphy with bottom solvent feed.
and an electrode Rxed at the top of the plate. While in
such systems the solvent is sometimes described as
migrating up the plate because of electroosmosis, this
may not always be strictly true. The main limitation
of this arrangement stems from the lack of solvent
reservoir at the top of the plate, and uneven solvent
evaporation from the plate can impart a capillary-
driven component to the solvent migration.
This makes it difRcult to differentiate between
electroosmotic solvent Sow and capillary solvent
Sow. Some workers have placed a second solvent
reservoir at the top of the plate, and generated elec-
troosmosis in a downward direction.
Chambers are usually sealed from the atmosphere
in order to provide a solvent-saturated atmosphere,
thus reducing evaporation. Some arrangements em-
ploy a cover plate, usually glass, in direct contact with
the chromatographic surface in order to minimize
evaporative effects.
Plates
A wide range of stationary phases, mobile phases and
operating conditions have been employed in TLE.
The layer is frequently 50}200
m thick on a backing
material of glass or organic polymer. Aluminium-
backed plates are not suitable for use in TLE because
of their electrical conductivity. TLE has been carried
out on all thin layer adsorbants used for TLC, with
silica, microcrystalline cellulose and alumina attract-
ing the greatest interest.
Plates are generally 10}20 cm long and 5}20 cm
wide. Longer plates tend to suffer from evaporative
Sow more than short ones and are generally avoided.
The width of the plate is limited only by the current
that the power supply is able to deliver at the required
potential.
Solvent Systems
The majority of TLE separations have so far been
carried out in aqueous buffer systems similar to those
used in TLC. More recently, the use of nonaqueous
systems has been shown to be useful in achieving
faster and more efRcient separations.
Potentials, Currents and Power Supplies
Potentials used in aqueous TLE are in the range of
10}100 V cm\1, with currents of around 10}
100 mA. This generates power levels of around
1}100 W. At the higher power levels, plate cooling is
essential in order to prevent drying out. When
nonaqueous systems are used, potentials between 200
and 2000 V cm\1 are used, with currents of
0.01}2 mA, generating between 0.02 and 20 W.
Cooling of plates run at higher potential is seldom
employed. Cooling is rarely necessary and very difR-
cult to achieve due to the inherent incompatibility of
high thermal conductivity and good electrical insula-
tion characteristics in materials.
Starting Conditions
The layers are usually pre-wetted with the eluting
solvent following sample application. This is
achieved by spraying or dipping. Some experiments
have been carried out using dry plates, but with
limited success.
Sample Application and Visualization
The same methods of applying and viewing sample
spots and bands used in TLC are employed in TLE.
Applications
Thin layer electrochromatography can be divided into
three main forms depending on the major factor gov-
erning the separation. While not mutually exclusive,
since most separations include some element of the
other modes, these broadly arise from electrophoretic
solute migration, electroosmotic solvent Sow and the
natural spin-off from the heating effects arising from
the applied potential, electrothermal elution.
The most commonly encountered examples of TLE
are based around electrophoretic separations in aque-
ous solvent. Not surprisingly, given the historical
success of paper electrophoresis, several workers have
used thin layers of microcrystalline cellulose. In addi-
tion, cellulose acetate and silica have been used for
the separation of proteins. Other applications have
included the separation of starches, amino acids (and
various derivatives) (Figure 4), organometallic com-
pounds (Figure 5) and transition metal ions.
Electrophoretic separations are not limited to aque-
ous solvent systems and the higher resistance of
nonaqueous solvents gives the advantage of lower
currents and reduced heating effects. The separation
 II / ELECTROPHORESIS / Electrochromatography 1255

Figure 4 The separation of amino acids by two-dimensional

thin layer electrophoresis}thin-layer chromatography with an
aqueous electrolyte. The chromatographic media was plastic-
backed cellulose layer. Electrophoresis in the first dimension
using a 4%(v/v) aqueous formic acid electrolyte was followed by
chromatographic elution in the second dimension with butanol-
0.4%pyridineacetic acid (22 : 10 : 10, v/v/v). Adapted from E.
McEvoy-Bowe (1985) Journal of Chromatography 347: 199}208,
with permission.

of a number of dyes using ethanol as the solvent is

shown in Figure 6. In this separation electro- Figure 6 Nonaqueous thin-layer electrochromatography
(TLE) and conventional thin-layer chromatography (TLC) of a
osmotic Sow effects were suppressed to reveal the dye mixture: (a) Oil Blue, (b) Rhodamine B, (c) Neutral Red,
electrophoretic migration of the charged dyes, result- (d) Diazine Green, (e) Brilliant Green. The chromatography media
ing in completely different elution orders. was silica (electroosmotic flow suppressed) and the solvent
In TLE, solvent migration from EOF is easily con- was ethanol.
fused with capillary-induced Sow resulting from local-
ized solvent evaporation. Broadly speaking, EOF is to
be expected from wet polar solvents, protic solvents or
from those that are capable of autoprotolysis.
With vertical tank systems, and particularly those
employing nonpolar solvents, there must remain
some uncertainty over whether thermal effects have
been responsible for any solvent migration observed.
This is the case in the pioneering planar systems
studied by Pretorious (Figure 7), in which nonpolar
solvents such as benzene were allegedly used. Our
attempts to reproduce this work with a vertical tank
system resulted in an electrically driven solvent

Figure 5 A thin-layer electropherogram of platinum chloro-
amine complexes. The chromatographic media was microcrystal-
line cellulose thin layers and electrolyte was 0.1 M NaClO4, at
500 V for 5 min. Adapted from M Lederer and E Leipzig-Pagani
(1998) Analytica Chimica Acta 358: 61}68, with permission.
Figure 7 An early thin-layer electrochromatography (TLE) sep-
aration of nonionic compounds. The chromatography media was
dichlorodimethylsilane-treated silica and the solvent was unspeci-
fied. Adapted from V. Pretorius et al. (1974) Journal of Chromato-
graphy 99: 23}30, with permission.
1256 II / ELECTROPHORESIS / Electrochromatography
Figure 8 Conventional thin-layer (left) and electroosmotic
(right) separation of pyrimidines employing identical silica layer
chromatographic media and eluting solvent (ethanol) (TLC ca.
15 min; TLE 7 kV, 90 s). Adapted from AG Howard and T Shafik
(1999) Journal of Chromatography 844A: 333}340, with per-
mission.
Electrochromatography in Thin-Layer
Electrophoresis
See II / ELECTROPHORESIS / Electrochromatography
Electrophoresis Using Cellulose Acetate
See II / ELECTROPHORESIS / Cellulose Acetate
Electrophoresis: Discontinuous
See II / ELECTROPHORESIS / Discontinuous Electrophoresis
migration and chromatographic separation resulting
largely from thermal effects and not EOF. By chang-
ing to more polar solvents, in a horizontal tank, we
have shown that true electroosmotic Sow could be
achieved. The separation of a number of pyrimidines
on silica eluted with ethanol showed elution charac-
teristics similar to those obtained by conventional
TLC, but with higher separation efRciency and in one
tenth of the time (Figure 8).
More recently, high-voltage nonaqueous TLE em-
ploying electroosmosis as the main driving force has
been applied to the separation of a wide range of
acidic, basic and neutral organic compounds, with
considerable success.
Further Reading
Howard AG, ShaRk T, Moffatt F and Wilson ID (1999)
Journal of Chromatography 844: 333}340.
Kowalczyk JS (1996) Chemical Analysis (Warsaw) 41:
157}171.
Poole CF and Wilson ID (1997) Journal of Planar
Chromatography 10: 332}335.
Pretorius V, Hopkins BJ and Schieke JD (1974) Journal of
Chromatography 99: 23}30.
Sargent JR (1975) Methods in Zone Electrophoresis, 3rd
edn. Poole: BDH Chemicals Ltd.
Smith IW (1960) Chromatographic and Electrophoretic
Techniques, vol. 1. London: Heinemann Medical Books
Ltd.
Tsuda T (1995) Electric Field Applications in Chromatogra-
phy, Industrial and Chemical Processes. Weinheim: VCH.
 II / ELECTROPHORESIS / Immunoelectrophoresis
Gel Electrophoresis in Capillary
Electrophoresis
See II / ELECTROPHORESIS / Capillary Gel Electrophoresis
Immunoelectrophoresis
F. Lampreave, M. Pin
eiro, S. Carmona and M. A. Alava,
Facultad de Ciencias, Universidad de
Zaragoza, Zaragoza, Spain
Copyright ^ 2000 Academic Press
Development
The discovery that antigen}antibody interaction
could be produced not only in liquids, but also on gel
media, such as agar or agarose gels, with the for-
mation of insoluble immunoprecipitates, opened the
door to the development of the gel diffusion tech-
niques for immunoprecipitation analysis. The Rrst of
these techniques, known as double diffusion, was
introduced by Ouchterlony in 1948. In this tech-
nique, the antigens (proteins) and the corresponding
antibodies (immunoglobulins, Ig) are located on
a thin agar gel, in small and separated wells. The
simple diffusion of the antigen and the antibody pro-
duce precipitation lines between the two wells where
the interaction of these molecules occurs.
Advances in gel immunoprecipitation techniques
occurred in 1953, when Grabar and Williams de-
scribed immunoelectrophoresis (IE), in which the
high resolution of electrophoresis and the speciRcity
and sensitivity of the immunological reactants are
combined. The immunoelectrophoretic techniques
had a very rapid development during the 1960s and
early 1970s, because they are adequately suited to
the analysis of complex mixtures of proteins. The
Rrst attempts to improve the immunoelectrophoretic
technique pursued two main objectives: to increase
its speed (IE was rather slow, mainly due to the
time-consuming immunodiffusion step) and
to achieve quantitative methods (IE is basically
qualitative).
In a short period of time new techniques based
on crossed immunoelectrophoresis (CIE), rocket
immunoelectrophoresis (RIE), counter IE, crossed-
afRnity IE (CAIE) and charge shift IE, were
introduced.
1257
This article describes all the techniques in which
electrophoresis and immunoprecipitation steps per-
formed in agar or agarose gels are combined. Diffu-
sion-type techniques, such as double immunodiffusion
and the quantitative radial immunodiffusion, will not
be described. Other immunochemical techniques, such
as enzyme-linked immunosorbent assay (ELISA) and
Western blotting, in which immunoprecipitation is not
produced, will be considered elsewhere.
Conventional Immunoelectrophoresis
The scheme in Figure 1A shows the principles of the
techniques, such as it was originally introduced and
generally used by many workers. The Rrst step con-
sists of electrophoresis in 1% agar}agar or agar-
ose gels, prepared in different buffers at pH ranging
from 8.2 to 8.6 (Veronal 0.025 mol L\1, pH 8.2 is
one of the buffers frequently used). The sample to be
analysed, usually a complex mixture of proteins, is
applied in small wells located in the middle of
agar}agar gels, or partially displaced towards the
extreme nearest to the cathode when agarose gels are
used. Most of the proteins, for example in blood sera
or plasma, have a negative net charge in buffers with
pH higher than 7.0. Thus, when a continuous electri-
cal Reld is applied to the gel, the tendency of these
proteins is to move towards the anode with a migra-
tion rate which mainly depends on charge-to-size
ratio. However, the migration rate is in part reduced
by the electroendosmotic effect due to the negative
charges in polymeric molecules of the gel. This elec-
troendosmotic effect is greater in agar}agar than in
agarose gels. The extension of the electrophoretic run
can be precisely Rxed by controlling the migration of
a marker. One of the markers routinely used is the
protein stain Amido black. The second step is
a double immunodiffusion performed in the same gel
plate (Figure 1A). For this, longitudinal channels are
cut parallel to the direction of the electrical Sow and
separated 4 mm from the original wells where the
samples were applied. The channels are Rlled with the
1258 II / ELECTROPHORESIS / Immunoelectrophoresis
Figure 1 (A) Schematic representation of the principle of con-
ventional immunoelectrophoresis. 1, Electrophoresis of the
sample in agar or agarose gels; 2, double immunodiffusion of the
separated proteins and the corresponding antiserum. (B) Applica-
tion of the technique to the study of specificity of various antisera.
In all wells, fetal pig serum was applied. In the channels, the
following antisera were assayed: 1, anti-fetal pig serum, 2, anti-
transferrin; 3, anti-albumin; 4, anti--fetoprotein; 5, anti-1-anti-
trypsin; 6, anti-1-acid glycoprotein; 7, anti-fetuin. (With per-
mission from Lampreave, F, GonzaH lez-RamoH n N, MartOH nez-
Ayensa S, HernaH ndez MA, Lorenzo HK, GarcOH a-Gil A and Pin
eiro
A (1994) Electrophoresis 15: 672}676.)
corresponding antiserum and the plate is maintained
in a humid, sealed box, for a period of between 24
and 48 h. Under these conditions, the combination of
the radial diffusion of the proteins from the circular
or ovoid spots obtained after the electrophoretic run,
with the uniform diffusion of the antibodies, occur-
ring in a perpendicular direction to the channel, pro-
duces arcs of precipitation in the different elec-
trophoretic zones.
The IE patterns can be directly visualized or photo-
graphed, in the wet gel, by dark-Reld illumination. In
this method the oblique light from a circular source
placed below the plate is directed through the gel and
transmitted at an angle of about 253. The immuno-
precipitates are visible as white lines on the dark
background. To preserve the plates and before stain-
ing the immunoprecipitates, it is necessary to wash
the gel plates extensively so that all the unprecipi-
tated materials are removed. The wash is accomp-
lished by immersing the gel plates in a buffered
saline solution (0.01 mol L\1 phosphate, NaCl
0.15 mol L\1 buffer, pH 7.4) for 1}3 days, and with
several changes of liquid. Alternatively, a quick pro-
cedure can be carried out by placing several paper
towels on the gel plate and applying a moderate
pressure for some minutes. In this way, the liquid and
all the soluble material are removed from the gel and
absorbed by the paper towels. Then, the plate is
submerged in the saline solution until the gel almost
recovers its original thickness. Repeating this dry-
ing/soaking cycle, it is possible to achieve effective
and rapid washing of the plates. Finally, the plates are
dried and the immunoprecipitates stained, commonly
with Coomassie blue or Amido black. The sensitivity
of the method, using these conditions, allows the
detection of proteins with concentrations ranging
from 3 to 20
g mL\1. SpeciRc staining methods have
been introduced to facilitate the identiRcation of single
proteins. For example, Sudan black can be used to
detect lipoproteins and periodic acid (Schiff reagent)
for polysaccharides and glycoproteins. There are also
speciRc staining procedures for some proteins, such as
ceruloplasmin, hemopexin, etc.
IE has been used to analyse complex mixtures of
proteins from tissue extracts; to detect impurities
during the monitoring of protein puriRcations; to
detect differential expression of protein during
growth and differentiation; to study the expression of
single proteins during pathological situations and to
detect protein polymorphisms. Figure 1B shows an
example of the application of the IE technique. In the
plate, the speciRcity of different antisera against six
proteins isolated from fetal pig sera (albumin, -
fetoprotein, 1-acid glycoprotein, 1-antitrypsin, fe-
tuin and transferrin) is analysed by attaching them to
 II / ELECTROPHORESIS / Immunoelectrophoresis
a fetal pig serum by this technique. As reference, the
same fetal pig serum is analysed against a polivalent
anti-fetal pig antiserum.
Counter Immunoelectrophoresis
This technique is a modiRcation of conventional IE,
that is performed in agar}agar gels at pH 8.0. Under
these conditions, the antibodies are positively charged
whereas the antigens present a negative net charge.
Antigen and antibodies are applied in wells of 3 mm
diameter, as indicated in Figure 2. Antibodies are
placed on the well nearest to the anode and the
sample in that close to the cathode. By applying
a voltage across the gel the antigen and the antibodies
move towards each other, forming lines of precipita-
tion between the two wells. For this technique, the
utilization of agar}agar gels is convenient since the
medium, at the pH commonly used, generates a sig-
niRcant electroendosmotic Sow that increases the
cathodic movement of the antibodies. This technique
can be used to detect both antigen and antibodies;
rapidity is the main advantage.
Crossed Immunoelectrophoresis
Crossed IE is another approach in IE which is well
suited for qualitative and quantitative analysis. The
method, which was originally named two-dimen-
sional electrophoresis or antigen}antibody crossed
IE, offers not only higher resolution and simpler in-
terpretation of the results than in conventional IE but
it can also be standardized for quantitative analysis.
The principles of this technique are presented in Fig-
ure 3A and consist of two electrophoretic runs, both in
1% agarose gel plates. Veronal 0.05 mol L\1 pH 8.6,
1 mmol L\1 calcium lactate is a buffer that is frequently
used. Samples are applied in wells of around 3 mm
Figure 2 Principle of counter IE. This technique is developed in
agar}agar gels, at pH 8.0. Antigens move towards the anode and
antibodies to the cathode, due to its charge and to the electro-
endosmotic effect. (With permission from Lampreave F and
Pin
eiro A (1992) Concentration of major plasma proteins in serum
and whole-tissue extracts of porcine fetuses during development,
Journal of Reproduction and Fertility 95: 441}449.)
1259
diameter and a current applied for about 3}4 h at
a potential gradient of 10 V cm\1 (to achieve a total
electrophoretic run of around 6 cm). Afterwards, a lon-
gitudinal strip of 1 cm width, which includes the separ-
ated protein fractions, is cut off and transferred to
a plate with the second agarose gel (prepared in the
buffered solution as above) that contains the antiserum
corresponding to the material to be tested.
The second run is carried out in a perpendicular
direction to the Rrst separation, at a potential gradi-
ent of around 5 V cm\1. The time of this electro-
phoretic step varies depending on the mobility of the
antigens. For complex mixtures, and depending on
the relative proportions between antigens and anti-
bodies, periods of around 6}10 h may be needed.
Under these conditions, precipitation peaks are pro-
duced in which the height prevails over the width.
Furthermore, the area of each peak is now related to
the amount of protein contained in the sample.
IdentiRcation of proteins in the crossed im-
munoelectrophoretic patterns, though simpler than in
the patterns obtained by conventional IE, presents
difRculties when complex mixtures of proteins are
analysed. For that reason, some modiRcations to
crossed IE, such as fused rocket IE, line IE, tandem
crossed IE, crossed line IE and crossed IE with inter-
mediate gel, have been introduced.
Crossed IE techniques enable the easy comparison
of the content of a determined protein in different
samples. Figure 3B shows an example of the potential
of this technique, applied to the study of the acute-
phase proteins in pigs. The comparison of the crossed
immunoelectrophoretic patterns of blood serum from
the same pig before (left) and 48 h after the induction
of acute inSammation by turpentine injection (right,
acute-phase serum), permits detection of important
differences in the concentration of some plasma pro-
teins. Especially notable is the increase of peak 9, that
corresponds to the major acute-phase protein in pigs,
called Pig-MAP, which could be detected through this
technique.
Crossed IE can also be applied to the study of
membrane proteins. For this it is convenient to
solubilize them with nonionic detergents, such as
Triton X-100, that better preserve the structural and
functional properties of these proteins. The detergent
does not cancel out the antigenic properties of the
membrane proteins, nor does it impede the antigen}
antibody reaction. In the characterization of mem-
brane proteins it is important to know whether the
proteins studied possess hydrophobic domains that
anchor them to the hydrocarbon interior of the bi-
layer, or whether they are externally bound to the
membrane. Charge-shift immunoelectrophoresis per-
mit the user to easily obtain that information even in
1260 II / ELECTROPHORESIS / Immunoelectrophoresis

Figure 3 (See Colour Plate 41). (A) Schematic illustration of crossed IE. 1, Agarose electrophoresis of the sample (O"origin,
corresponding to the well where the sample is applied); 2, a longitudinal strip of the first-dimension gel is transferred into a second-
dimension gel containing a polyvalent antiserum. The second-dimension IE is performed perpendicularly to the first-dimension run. (B)
Example of use of the crossed IE technique. The blood serum from the same pig is analysed (left) before and (right) 48 h after
turpentine injection. The changes in the protein concentrations, induced by inflammation, can easily be studied by analysing the area
(or height) of the different-numbered peaks. (With permission from Lampreave, F, Alava MA and Pin
eiro A (1996) Trends in Analytical
Chemistry 15: 122}129.)

complex mixtures. Membrane proteins are solubilized
either with Triton X-100 alone or mixed with other
detergents, for example, Triton X-100/sodium deoxy-
cholate (an anionic detergent) and Triton X-100/cetyl-
trimethylammonium bromide (a cationic detergent).
Afterwards, these three membrane extracts are ana-
lysed comparatively by crossed IE. The detergent-in-
duced shift in mobility provides a method to distin-
guish between hydrophilic and amphiphilic proteins.
Rocket Immunoelectrophoresis
This immunochemical method, introduced by Laurell
in 1966, is suitable for the quantitative estimation of
proteins. In this case (Figure 4A) there is only one
electrophoretic run carried out in 1% agarose gel
(plates of uniform thickness of about 1.5 mm) in
Veronal buffer, pH 8.6 as described before, but in-
cluding an appropriate quantity of an adequate anti-
serum. The samples are applied in separated wells of
3 mm diameter, located near the extreme of the plate
in contact with the cathode. To obtain quantitative
results, different dilutions from a primary standard (a
solution of the puriRed protein of known concentra-
tion) or a secondary standard (a serum previously
evaluated using a primary standard) are applied on
the gel. The electrophoresis is accomplished over peri-
ods ranging from 4 to 6 h depending on the charge of
 II / ELECTROPHORESIS / Immunoelectrophoresis
Figure 4 (A) Schematic representation of rocket IE. The elec-
trophoresis is carried out in agarose gel containing specific anti-
serum against the protein to be determined. The area of the
rockets (or its height) is proportional to the content of protein in the
sample. (B) Application of this technique to the quantification of
human -fetoprotein. To improve sensitivity, the first immuno-
precipitates (developed with rabbit antiserum to human -feto-
protein) were treated with glucose oxidase-labelled sheep
antibodies to rabbit immunoglobulins. From left to right, solutions
containing -fetoprotein concentrations of 720, 360, 180 and
120 ng mL\1 were assayed.
the antigen and on the relationship between the
amount of antigen and antibodies included in the gel
plate. The immunoprecipitates formed in this case
have the shape of a rocket, the area (or height) of
which is proportional to the concentration of the
antigen applied in the well. The concentration of the
protein in unknown samples can be determined by
reference to the calibration line obtained by repre-
senting the height versus the concentration of each
standard.
This technique can be applied to any protein with
a net charge that differs from that of the anti-
bodies. Some proteins, such as, for example IgG,
hardly move during electrophoresis using a pH 8.6
buffer. In these cases, the mobility of the antigen to be
determined can be increased by carbamylation of the
sample with KCNO.
1261
Rocket IE is the most sensitive of the conventional
immunoelectrophoretic methods and allows one to
accomplish the quantiRcation of proteins (using
stains to visualize the immunoprecipitates) up to con-
centrations close to 1
g mL\1. This sensitivity can be
increased by combining immunoenzymatic methods.
Figure 4B shows an example of one of these methods
applied to determine human -fetoprotein. During
the electroimmunodiffusion, the Rrst antigen}
antibody reaction is produced using a rabbit anti-
serum containing the speciRc antibodies (Ig). In a sec-
ond step, the plate is incubated with glucose oxidase-
labelled sheep antibodies to rabbit Ig. Finally, the
glucose oxidase bound to immunoprecipitates is re-
vealed by incubating the plate with a solution that
contains glucose, MTT-tetrazolium and phenazide
methasulfate. The immunoprecipitates stain blue-
violet over a fainter background of similar colour.
This method provides reproducible results and allows
the quantiRcation of -fetoprotein at concentrations
as low as 50 ng mL\1.
Crossed-af\nity
Immunoelectrophoresis (CAIE)
AfRnity electrophoresis refers to any technique in
which two or more components speciRcally interact
during an electrophoretic run. AfRnity electrophor-
esis in agarose gels, combined with subsequent im-
munochemical detection (CAIE), was introduced by
Bog-Hansen in 1975 as a useful tool for the character-
ization of biospeciRc macromolecular interactions.
This technique permits, among other things, the dem-
onstration of ligand-binding proteins, enumeration of
binding sites and estimation of binding constants,
with the added advantage of being adequate for
determination of very small quantities in impure
materials.
An important Reld of application of CAIE has been
the study of the interaction of glycoproteins and lec-
tins. Lectins are animal and mostly plant-derived pro-
teins that speciRcally interact with the carbohydrate
components of glycoconjugates. There is a consider-
able amount of information in the literature about the
biochemical properties of numerous lectins and on
the type of glycan structures that they can recognize.
Using different lectins in the Rrst electrophoresis run
(afRnity electrophoresis step), CAIE permits the anal-
ysis of the heterogeneity of glycoproteins in complex
mixtures. The serum protein glycoforms, after frac-
tionation by lectin afRnity electrophoresis, can be re-
vealed with speciRc antibodies and Rnally quantiRed.
Figure 5A shows a schematic illustration of the
CAIE technique. First, the serum proteins are
subjected to electrophoresis (generally at pH 8.2}8.6)
1262 II / ELECTROPHORESIS / Immunoelectrophoresis
Figure 5 (A) Schematic representation of the CAIE technique. O, origin, corresponding to the well where the sample is applied in the
first-dimension electrophoresis. (B) CAIE patterns of human 1-acid glycoprotein from (right) healthy individuals and (left) patients with
inflammatory processes. The numbered peaks correspond to microforms of 1-acid glycoprotein in decreasing order of mobility.
in lectin-containing agarose gels (Rrst-dimension gel).
The lectin is added to the melted agarose (at around
553C) before pouring the gel. Under these conditions
most of the lectins used in CAIE do not have sufRcient
mobility. The proteins are fractionated in their differ-
ent glycoforms, whose mobility depends on their cor-
responding afRnity for the lectin. The Rrst-dimension
gel is then transferred into a second-dimension
agarose gel containing speciRc antibodies against the
protein to be analysed. The second-dimension elec-
trophoresis produces for many serum glycoproteins
fused precipitating peaks, which corresponds to the
different microforms present in the sample. The CAIE
patterns can be visualized by the methods described
in the previous immunochemical techniques. The
amount of each glycoform is related to the corre-
sponding immunoprecipitating area that can be cal-
culated by planimetry or by triangulation.
Figure 5B shows, as an example, the CAIE patterns
of 1-acid glycoprotein from human serum. In the Rrst
dimension gel, 1 mg mL\1 of the lectin Concanavalin
A was included. The second-dimension gel contained
speciRc rabbit antiserum against human 1-acid gly-
coprotein. The different peaks have been labelled in
decreasing order of mobility. Peak 1 corresponds to
glycoforms of the human 1-acid glycoprotein that
did not react with the lectin and contains tri- and
tetra-antenary glycans; peak 2 corresponds to weakly
reactive glycoforms of the protein containing one
bi-antennary glycan; Rnally, peak 3 corresponds to
strong reactive glycoforms containing at least two
bi-antennary glycans.
Equipment and General Methods
Agarose gels (1% w/v) are prepared in the described
buffers on glass plates of different size according to
need. For example 9;12 cm plates are suited for
conventional IE (they permit the analysis of eight
samples or seven antisera) and also for RIE (in this
case, about 20 samples can be applied). For the CIE
and CAIE techniques, 4;9 cm and 7.5;10 cm glass
plates can be used for the Rrst and the second elec-
trophoretic run, respectively. In all cases, one of the
surfaces of the glass plates is coated with an agarose
solution (1% w/v, in distilled water) that, after being
dried, serves as an anchor for the agarose gel. To
ensure regular thickness throughout the gel plate,
which is essential for the reproducibility of quantitat-
ive methods, the melted agarose is poured between
two glass plates separated by rigid spacers of around
1.5 mm thickness.
When antibodies, puriRed proteins or lectins are to
be added to the gel, the agarose solution must be
previously equilibrated around 553C in a thermo-
statized water bath to avoid protein denaturation. The

growth of microorganisms is avoided by the addition
of preservatives to the agarose solutions, for example
0.1% sodium azide or 0.01% sodium merthiolate.
Simple equipment is commercially available,
though much can be made in-house. The electrophor-
esis system only requires two electrophoretic tanks
(provided with platinum wire and connected to the
electrodes), and an adjustable power supply deliver-
ing voltage up to 400 V at 400 mA. The connection
between the agarose plates and the buffer in the tanks
can be accomplished through Rlter-paper wicks pre-
viously wetted in electrophoretic buffer (the same as
that used for preparing the gel). To avoid excessive
warming of the gel during electrophoresis, it is conve-
nient to cool the agarose plates using a system con-
nected to tap water. This allows the electrophoresis to
be run at room temperature in the laboratory, alter-
natively, it can be carried out in a cold room at 53C.
A great number of polyvalent and speciRc antisera
prepared in goat, sheep or rabbits can be obtained
from different suppliers or obtained in-house.
Present and Future Developments
Though most immunoelectrophoretic techniques de-
scribed here were developed at least 25 years ago,
they still enjoy great popularity and continue to be
excellent tools for biochemists and immunologists. IE
and CIE are very useful techniques for the character-
ization of complex mixtures of proteins and for the
study of certain pathological situations that evolve
with changes in plasma protein patterns. CAIE is a
powerful technique for detecting glycoprotein micro-
forms using different lectin speciRcities. Advances in
the characterization of new lectins with restricted
speciRcity represent a future development in this Reld.
CAIE can also be applied to many afRnity systems,
including the important contribution of monoclonal
antibodies in the afRnity electrophoresis step.
Immunoelectrophoretic techniques are time- and
antisera-consuming techniques. These limitations
Isoelectric Focusing
P. G. Righetti and A. Bossi, University of Verona,
Verona, Italy
C. Gelfi, ITBA, CNR, Segrate, Milan, Italy
Copyright ^ 2000 Academic Press
Isoelectric focusing represents a unique electrokinetic
method in that it is based on steady-state patterns
II / ELECTROPHORESIS / Isoelectric Focusing 1263
could be improved by including the capillary methods
commonly used in capillary electrophoresis systems.
Further Reading
B+g-Hansen TC, Bjerrum OJ and Ramlau J (1975) Detec-
tion of biospeciRc interaction during the Rrst dimension
electrophoresis in crossed immunoelectrophoresis.
Scandinavian Journal of Immunology 4 (Suppl. 2): 141.
Breborowicz J and Mackiewicz A (eds) (1992) AfTnity
Electrophoresis: Principles and Application. Boca
Raton: CRC Press.
DolnmH k V (1997) Capillary zone electrophoresis of proteins.
Electrophoresis 18: 2353.
Grabar P and Williams CA (1953) MeH thode permettent
l’eH tude conjugeH e des propieH teH s eH lectrophoretiques et im-
munochimiques d’un meH lange de proteH ines. Application
au seH rum sagnuin. Biochimica Biophysica Acta 10: 193.
Helenius A and Simons K (1977) Charge shift electrophor-
esis: simple method for distinguishing between am-
phiphilic and hydrophilic proteins in detergent solution.
Proceedings of the National Academy of Science of the
USA 74: 529.
Laurell CB (1966) Quantitative estimation of proteins by
electrophoresis in agarose gel containing antibodies.
Analytical Biochemistry 15: 45.
Ouchterlony OG (1948) In vitro method for testing the toxin-
producing capacity of diphtheria bacteria. Acta Patho-
logica Microbiologica Scandinavica 25: 186.
Ouchterlony OG and Nilsson LA> (1986) Immunodiffusion
and immunoelectrophoresis. In: Weir M, Herzenerg LA,
Blackwell C and Herzenberg LA (eds) Handbook of Ex-
perimental Immunology 4th edn, vol. 1. Immunochemisty,
Ch. 32. Oxford: Blackwell ScientiRc Publications.
Ressler N (1960) Two-dimensional electrophoresis of pro-
teins antigens with an antibody containing buffer. Clini-
ca Chimica Acta 5: 795.
Uriel J (1971) Color reactions for identiRcation of anti-
gen}antibody precipitates in gels. In: Williams CA and
Chase MW (eds) Methods in Immunology and Immuno-
chemistry, vol. III, p. 294. New York: Academic Press.
Williams CA (1971) Immunoelectrophoretic analysis. In:
Williams CA and Chase MW (eds) Methods in Immuno-
logy and Immunochemistry, vol. III, p. 235. New York:
Academic Press.
attained by amphoteric species (mostly proteins and
peptides) along a pH gradient under the inSuence of
an electric Reld. Due to a continuous balancing of
diffusion away from the pI (isoelectric point) and
pI-driven electric forces, extremely sharp zones are
obtained, characterized by a very high resolving
power. This article will consider conventional isoelec-
tric focusing (IEF) in soluble, amphoteric buffers; and
immobilized pH gradients (IPG) in insolubilized,
1264 II / ELECTROPHORESIS / Isoelectric Focusing
non-amphoteric buffers. In the latter case, guidelines
will be given on how to optimize linear and nonlinear
pH gradients and examples will be shown on the
unique resolving power of the technique.
Conventional Isoelectric Focusing
In principle, pH gradients could be obtained by diffu-
sion of non-amphoteric buffers but such ‘artiRcial’
gradients would be altered by changes in electric
migration and diffusion of the buffer ions. Thus,
Svensson in 1961 introduced the concept of ‘natural’
pH gradients, created and stabilized by the electric
current itself. The buffers used in this system required
two fundamental properties: Rrst amphoterism, so
that they could reach an equilibrium position along
the separation column and secondly ‘carrier’ ability.
This last concept is more subtle, but just as funda-
mental. Any ampholyte cannot simply be used for
IEF; only a carrier ampholyte, that is a compound
capable of transporting the current (a good conduc-
tor) and capable of carrying the pH (a good buffer).
With this notion, and with Vesterberg’s elegant syn-
thesis of such ampholytes in 1969, present-day con-
ventional IEF was born.
Some Basic Theoretical Concepts
Here some basic equations governing the IEF process
will be considered. The most important is the one
governing the distribution proRle of an ampholyte
about its isoelectric point. Under steady-state condi-
tions (obtained by balancing the simultaneous
electrophoretic and diffusional mass transports),
Svensson derived the following differential equation
describing the concentration proRle of a focused
zone:


dC
C
i/qk"D [1]
dx
where C is the concentration of a component in
arbitrary mass units per arbitrary volume unit;
is
the electric mobility in cm2 V\1 s\1 of ion constituent
except H# and OH\, with positive sign for cationic
and negative sign for anionic migration; i is the elec-
tric current in A; q is the cross-sectional area in cm2 of
electrolytic medium, measured perpendicularly to the
direction of current; k is the conductance of the
medium, in \1 cm\1; D is the diffusion coefRcient in
cm2 s\1 of a given ionic component with mobility
;
and x is the coordinate along the direction of current
increasing from 0 to the anode towards the cathode.
Each term in eqn [1] expresses the mass Sow per
second and square centimetre of the cross-section: to
the left being the electric and to the right the diffu-
sional mass Sows. If eqn [1] is re-written in the form:




i
dx dC
"D [2]
q k C
it is seen that it is possible to integrate it if
is known
as a function of pH and D as a function of C. SpeciR-
cally, if the conductance, the diffusion coefRcient,
and the derivative:
  
d
d
d(pH)
p"! "! [3]
dx d(pH) dx
(where p is the ratio between the protein titration
curve and the slope of the pH gradient over the
separation axis) can be regarded as constant within
the focused zone, then
"!px and one obtains the
following analytical solution:
 
(pix2)
C"C0 exp ! [4]
(2qkD)
where x is now deRned as being zero at the concentra-
tion maximum C0. This is a Gaussian concentration
distribution with inSection points at:


qkD
xi"$ [5]
pi
where xi represents the width of the Gaussian distri-
bution of the focused zone measured from the top of
the distribution of the focused ampholyte to the in-
Section point (one standard deviation). The course of
the pH gradient is d(pH)/dx and d
/d(pH) represents
the titration curve of the ampholyte. It should be kept
in mind that this Gaussian proRle holds only if and as
long as the conductivity of the bulk solution within
the zone is constant. Constant conductivity along
a pH gradient is quite difRcult to maintain, especially
as one approaches pH extremes (below pH 4 and
above pH 10), if for no other reason, because the
non-negligible concentration of H# and OH\ pres-
ent in the bulk liquid begins to contribute strongly.
Another important equation regards the resolving
power in IEF, expressed in (pI) units, i.e. in the
minimum difference of surface charge between two
adjacent proteins that the IEF technique is able to
resolve. If two adjacent zones of equal mass have
a peak-to-peak distance three times greater than the
distance from the peak to inSection point there will
be a concentration minimum approximating the two
outer inSection points. Taking this criterion for re-
solved adjacent proteins, Rilbe derived the following
equation for minimally but deRnitely resolved zones:


D[d(pH)/dx]
(pI)"3 [6]
E[!d
/d(pH)]

This equation shows that good resolution should be
obtained with substances with a low diffusion coefRc-
ient (D) and a high mobility slope [d
/d(pH)] at the
isoelectric point } conditions that are satisRed by all
proteins. Good resolution is also favoured by a high
Reld strength (E) and a shallow pH gradient
[d(pH)/dx]. It will be seen that, whereas in conven-
tional IEF the limit to the resolving power is approx-
imately 0.01, in IPGs it is 0.001 pH units.
The Carrier Ampholyte Buffers
Recall that the buffer capacity of an ampholyte in the
isoprotic state decreases with increasing pK across
the isoprotic point, linearly at Rrst, then exponenti-
ally. Let us take as an example a hypothetical am-
pholyte, with pK1"4.6 (a carboxyl group) and
pK2"6.2 (an amino group), having thus a pI"5.4
and pK"1.6. If we titrate this ampholyte in the pH
4}7 range, encompassing the two pKs, and if we plot
the accompanying buffering power (), degree of dis-
sociation () and slope of the pH gradient, we will
have the plot shown in Figure 1(A). It can be seen that
there is still a substantial buffering power at the pI of
the ampholyte, with a corresponding degree of ioniz-
ation less than unity, and that the titration curve is
smooth throughout the pH gradient explored, with
Figure 1 Degree of ionization () and buffering power () of
a good (A) and a poor (B) carrier ampholyte. (A) computer simula-
tions obtained assuming a pK1"4.6 and a pK2"6.2 (pI"5.4).
The ampholyte was titrated in the pH 4}7 interval. (B) Computer
simulation obtained by assuming a pK1"4.6 and a pK2"9.3
(pI"6.95). The ampholyte was titrated in the pH 4}10 interval.
Note the sharp sigmoidal transition in the pI region in (B), sugges-
ting total lack of buffering power (Wenger P and Righetti PG,
unpublished observations).
II / ELECTROPHORESIS / Isoelectric Focusing 1265
only a small deviation about the pI of the ampholyte,
indicating that this species is indeed a ‘good’ carrier
ampholyte. Now take an ampholyte with pK1"4.6
but with pK2"9.3, thus with a pI"6.95 and
pK"4.7. If we now titrate it in the pH 4}10 range,
again encompassing the two pK values, we will have
the graph shown in Figure 1(B). It can be seen now
that at the theoretical pI value the ampholyte does
not have any appreciable buffering power and that it
is fully ionized. In addition, it is not only isoelectric
at pH"6.95, but indeed almost at any pH in the
interval 5}9, as seen by the abrupt sigmoidal shape
in the pI environment. This species will be a ‘bad’
carrier ampholyte, useless for a well-behaved IEF
fractionation.
An important prerequisite for a good carrier am-
pholyte is that it has a high conductivity at its pI.
Regions of low conductivity will absorb much of the
applied voltage, thus reducing the Reld strength and
hence the potential resolution in other parts of the
gradient. It has been demonstrated that good con-
ductivity is associated with small values of pI}pK.
This is also true for the buffering capacity of an
ampholyte. Thus, the parameter pI}pK (equivalent to
2 pK) becomes the most important factor in selecting
1
carrier ampholytes exhibiting both good conductivity
and buffering capacity ().
Methodology
The structure of carrier ampholytes (CA) and their
general properties are illustrated in Figure 2. CAs are
oligoprotic amino carboxylic acids, each containing
at least four weak protolytic groups, at least one
being a carboxyl group and at least one a basic
nitrogen atom, but no peptide bonds. In a typical
synthesis, a mixture of oligoamines (four to six nitro-
gens in length, linear and branched) reacts with an
}-unsaturated acid (typically acrylic or itaconic
acids), at a nitrogen}carboxyl ratio of 2 : 1.
The mechanism of developing a pH gradient in IEF
is illustrated in Figure 3. Before passage of the cur-
rent, the column is at constant pH (Figure 3A) and
the multitude of amphoteric buffers is randomly dis-
tributed, resulting in a reciprocal neutralization.
However, each individual CA species will have its
own titration curve (see Figure 2, lower left side)
deRning different mobilities in the electric circuit.
After starting the experiment the different CAs will
migrate at different velocities in the column, the most
acidic and most basic compounds being the fastest
moving ions. As a result of this sorting process, a pH
gradient will form, sigmoidal at Rrst (Figure 3B), with
an uneven voltage gradient. After 1 h, the various CA
buffers will have separated further, and at this point
an almost linear pH gradient has been established
1266 II / ELECTROPHORESIS / Isoelectric Focusing

Figure 2 Composition of Ampholine. On the upper left side a representative chemical formula is shown (aliphatic oligoamino
oligocarboxylic acids). On the lower left side, portions of hypothetical titration curves of carrier ampholytes are depicted. Right: different
pH cuts for wide and narrow range ampholytes (by permission of LKB Produkter AB).

which spans the pH range deRned by the pIs of the Gaussian proRles. Now the system has achieved
ampholytes (Figure 3C). After 1.5 h the CAs have a steady-state, i.e. a balance between electrophoretic
separated into symmetrical zones with overlapping transport and diffusion away from the pI, and no

Figure 3 Calculated time development of a focusing process involving 10 ampholytes in a closed vessel. The pI s of the ampholytes
are evenly distributed in the pH 8.0}8.9 range. The initial distribution of the amphoteric buffers is indicated in (A). The calculation was
performed assuming a constant voltage (100 V cm\1) across the system. The anode is positioned to the right in the diagrams. Each
x-axis represents the distance from the cathode on the same scale as in (D). (Reproduced with permission from Schaefer-Nielsen,
1986.)

further mass transport is expected (Figure 3D). As
long as the local concentration of the different CA
species does not change the slope of the pH gradient
will be kept constant with time. Proteins will keep
migrating against this CA distribution proRle event-
ually reaching their pI position.
By and large, most analytical IEF runs are per-
formed in horizontal chambers: the polyacrylamide
gel slab rests on a cooling block [generally made of
glass or coated aluminium or even beryllium oxide
(used as the heat shield of the space shuttle)]. This
horizontal conRguration allows one to dispose of
electrode reservoirs and of all the hydraulic problems
connected with vertical chambers (tight seals, etc.): in
fact, anolyte and catholyte are soaked in Rlter paper
strips resting directly on the open gel surface. In
addition, most modern chambers contain a cover lid
with movable electrodes which can be adjusted to any
gel length (generally from 10 to 25 cm electrode dis-
tance). Since thick gels (e.g. 2 mm thick) generate
thermal gradients through the gel thickness, resulting
in skewed zones (essentially all horizontal chambers
have cooling only on one gel face) ultrathin gels
(0.2}0.5 mm) supported on a reactive polyester foil
(Gel Bond PAG) are preferred today.
Immobilized pH Gradients (IPG)
IPGs are based on the principle that the pH gradient,
which exists prior to the IEF run itself, is
copolymerized, and thus immobilized within the
polyacrylamide matrix. This is achieved by using as
buffers a set of up to 10 non-amphoteric, weak acids
and bases, called Immobilines, having the following
general chemical composition: CH" 2 CH}CO}NH}R,
where R denotes either three different weak car-
boxyls, with pKs of 3.1, 3.6, and 4.6 (Table 1), or Rve
tertiary amino groups, with pKs of 6.2, 7.0, 8.5, 9.3
Table 1 Acidic acrylamido buffers
pK Formula
1.0
3.1
3.6
4.6
II / ELECTROPHORESIS / Isoelectric Focusing 1267
and 10.3 (Table 2). This set of eight weak buffers is
complemented by a strong acid (pK of approximately
1, 2-acrylamido-2-methyl propanesulfonic acid) and
a strong base (pK'12, quaternary aminoethyl-
acrylamide), used only as titrants. During gel polym-
erization, these buffering species are incorporated
into the gel (84}86% conversion efRciency at 503C
for 1 h), by the same free radical reaction used to
activate the acrylamide double bond. Figure 4 shows
a segment of a hypothetical structure of an Immobi-
line matrix and the process of focusing two proteins
in it. It is seen that only the proteins migrate to their
steady-state position, whereas the Immobilines re-
main Rxed at their original grafting position in the
gel, where a Rxed ratio of buffering/titrant ions de-
Rnes the pH locally. This means that the pH gradient
is stable indeRnitely (but it has to pre-exist before the
onset of polymerization) and can only be destroyed if
and when the polyacrylamide gel is hydrolyzed.
Given the sparse distribution of Immobilines in the
gel they behave as isolated charges, able to effectively
contribute to the ionic strength of the medium. In
conventional IEF, on the contrary, at steady-state
the ionic strength is exceedingly low (less than
1 mequiv L!1) since the focused carrier ampholytes
form an inner salt, and this often results in protein
precipitation and smears both at the pI and in its
proximity. In IPGs the high ionic strength existing in
the matrix (typically 10 mequiv L!1) induces protein
solubilization at the pI value (thus CA-IEF is similar
to a ‘salting-out’ milieu and IPGs to a ‘salting-in’
environment).
Immobiline-based pH gradients can be cast in the
same way as conventional polyacrylamide gradient
gels by using a density gradient to stabilize the Im-
mobiline concentration gradient, with the aid of
a standard, two-vessel gradient mixer. As shown
earlier, these buffers are no longer amphoteric as in
Name Mr
2-Acrylamido}2-methylpropanesulfonic acid 207
2-Acrylamidoglycolic acid 145
N-Acryloylglycine 129
4-Acrylamidobutyric acid 157
1268 II / ELECTROPHORESIS / Isoelectric Focusing
Table 2 Basic acrylamido buffers
pK Formula
6.2
7.0
8.5
9.3
10.3
'12
conventional IEF, but are bifunctional: the buffering
group is located at one end of the molecule and at the
other end there is the acrylic double bond which will
disappear during the grafting process. The three car-
boxyl immobilines have rather small temperature co-
efRcients of ionization (dpK/dT) in the 10}253C
range due to their small standard heats of ionization
(approximately 1 kcal mol\1) and thus exhibit negli-
gible pK variations over this temperature range. On
the other hand, the four basic immobilines exhibit
rather large pKs in the same temperature range (as
much as pK"0.44 for the pK 8.5 species) due to
their larger heats of ionization (6}12 kcal mol\1).
Therefore, for reproducible runs and pH gradient
calculations, all the experimental parameters have
been Rxed at 103C. Temperature is not the only
variable that will affect immobiline pKs (and there-
fore the actual pH gradient generated): additives in
Figure 4 Hypothetical structure of an Immobiline gel and mechanism of the focusing process. The acrylamido acid and basic groups
are shown grafted on the polyacrylamide matrix. Two proteins are shown migrating in the gel at the times t"0, at t"1 and finally at the
steady-state, where they reach they respective pI values (pI1 and pI2) as points of zero net charge (by permission of LKB Produkter AB).
Name Mr
2-Morpholinoethylacrylamide 184
3-Morpholinopropylacrylamide 198
N,N-Dimethylaminoethylacrylamide 142
N,N-Dimethylaminopropylacrylamide 156
N,N-Diethylaminopropylacrylamide 184
N,N-N-Triethylaminoethylacrylamide 198
the gel that will change the water structure (chao-
tropic agents such as urea) or lower its dielectric
constant, and the ionic strength itself of the solution,
will alter pK values.
Narrow and Ultranarrow pH Gradients
We deRne the gradients (in the gel slab) from 0.1 to
1 pH unit as ultranarrow and narrow gradients, re-
spectively. Within these limits one can generally work
on a ‘tandem’ principle, i.e. choosing a ‘buffering’
Immobiline (e.g. a base or an acid), having its pK
within the desired pH interval, and a ‘non-buffering’
Immobiline (e.g. an acid or a base), having its pK at
least 2 pH units removed from either pHmin or pHmax
of the pH range. The latter will therefore provide
equivalents of acid or base, respectively, to titrate
the buffering group, but will not itself buffer in
the desired pH interval. For these calculations one
 II / ELECTROPHORESIS / Isoelectric Focusing 1269

can resort to modiRed Henderson}Hasselbalch equa-

tions and to rather complex nomograms or simply
adopt tabulated recipes, 1 pH unit wide, which start
with the pH 3.8}4.8 interval and end with the
pH 9.5}10.5 span, separated by 0.1 pH unit in-
crements (58 such recipes have been tabulated). If
a narrower pH gradient is needed this can be derived
from any of the 58 pH intervals tabulated by a simple
linear interpolation of intermediate Immobiline
molarities.
Extended pH Gradients
For wider pH intervals, several buffering species have
to be mixed and the situation becomes considerably
more complex. This has been solved with the aid of
computer programs designed speciRcally for this
purpose. The basic Rndings are: Rrst for generating
a linear pH gradient the buffering power has to be

Figure 6 Effect of changes in the pK of the acidic titrant. A ref-

erence Immobiline mixture was titrated to the same pH value with
fictitious acids whose pK was 0.5, 1.0, 1.5, 2.0 and 2.5 pH unit
lower than the gradient’s limit (in this case, pHmin"3.5) and the
pH course was calculated for the five cases. The insert is a plot of
the percentage variation of deviation from linearity as the titrant’s
pK increases (from Gianazza et al., 1983, with permission of
Elsevier Science Publishers).

constant throughout the desired pH interval (this is

Figure 5 Effect of changes in the number of (evenly spaced)
buffering components. The optimal concentrations of fictitious
buffers (bases) with pKs differing by 1, 1.25, 1.66 and 2.5 pH
units, were calculated so as to cover the pH 4.5}8.5 range. The
resulting courses of  power are shown as a function of pK. The
insert is a plot of percentage variation, in comparison with the
case pK"1, of the ranges of deviation of pH (left scale) and of
 (right scale). Note that the smoothest  power is obtained with
pK"1 (from Gianazza et al., 1983, with permission of Elsevier
Science Publishers).
best achieved when the pK values are spaced at 1 pH
unit intervals, see Figure 5). Secondly, to avoid devi-
ations from linearity, the titrants should have pKs
well outside pHmin and pHmax of the wanted pH range
(in general, at least 2 pH units removed from the
limits of the pH interval) (see Figure 6). As a conse-
quence of this, for pH ranges wider that 3 pH units,
two additional Immobilines are needed as titrants:
one strongly acidic (pK (1) and one strongly basic
(pK '12). There are two ways of generating ex-
tended pH intervals. In one approach the concentra-
tion of each buffer is kept constant throughout the
span of the pH gradient and ‘holes’ of buffering
power are Rlled by increasing the amounts of the
buffering species bordering the largest pKs; in the
1270 II / ELECTROPHORESIS / Isoelectric Focusing

a trouble-free operation, as all the complex comput-

Figure 7 Non-linear pH 4}10 gradient. Ideal (dotted line) and
actual (solid line) formulation courses. The shape for the ideal
profile was computed from data on the statistical distribution of
proteins pI s. The relevant histogram is redrawn in the figure inset
(from Gianazza et al., 1985; by permission of VCH).
other approach (varying buffer concentration) the
variation in concentration of the various buffers
along the width of the desired pH gradient results in
a shift of their apparent pKs with a concomitant
evening-out of the pK values. With the available
recipes, preparation of any Immobiline gel is now
ing routines have already been performed and no
further calculations of any type are required.
Non-linear, Extended pH Gradients
IPG formulations have been given only in terms of
rigorously linear pH gradients. While this has been
the only solution adopted so far, it might not be the
optimal one in some cases. Altering the pH slope in
some portions of the gradient might be required in
those pH regions overcrowded with proteins. The
reasons for resorting to non-linear pH gradients are
given in the histogram of Figure 7. With the relative
abundance of different species it is clear that an opti-
mally resolving pH gradient should have a gentler
slope in the acidic portion, and a steeper course in the
alkaline region. Such a general course has been cal-
culated by assigning to each 0.5 pH unit interval in
the pH 3.5}10 region a slope inversely proportional
to the relative abundance of proteins in that interval.
The ideal (dotted) curve in Figure 7 was obtained by
such a procedure. What is also important here is the
establishment of a new principle in IPG technology,
namely that the pH and density gradients stabilizing
it need not be co-linear. The possibility exists of
modulating the former by locally Sattening of pH

Figure 8 IEF of conalbumin in an IPG pH 4.5}6.5 gradient. Gel: 5%T, 3%C polyacrylamide, equilibrated in 10% glycerol. All samples
were applied in round basins punched through the gel thickness at the cathodic side as 20
L droplet (20}500
g protein). Staining with
Coomassie Blue R-250 in ethanol/acetic acid in presence of copper sulfate. Notice that, although the gel thickness is only 0.5 mm, there
is no overloading effect in such a wide interval of protein concentration (from Righetti PG and Ek K, unpublished observations).

Figure 9 Focusing of histones in an IPG pH 10}12 nonlinear
interval. Gel: 6% T, 4% C polyacrylamide matrix, containing an
IPG 10}12 gradient, reswollen in 7 M urea, 1.5% Nonidet P-40
and 0.5% Ampholine pH 9}11. The gel was run at 103C under
a layer of light paraffin oil at 500 V for the first hour, followed by
increasing voltage gradients, after sample penetration, up to
1300 V for a total of 4 h. The samples (2 mg mL\1, 50
L seeded)
were loaded in plastic well at the anodic gel surface. Staining with
Coomassie Brilliant Blue R-250 in Cu2#. Histone samples (from
left). (1) VII-S (Lys-rich); (2) VI-S; (3) II-AS and (4) VIII-S (Arg-
rich, subgroup F), from calf thymus. The pI 10.6 marker (cyto-
chrome C) is in track 5 on the right side (from Bossi et al., 1994, by
permission of Elsevier Science Publishers).
gradients for increased resolution, while leaving unal-
tered the latter.
Although only one example of a nonlinear ex-
tended pH gradient is given here, clearly the possibil-
ity exists of modulating in the same fashion any
narrower pH interval.
Examples on the Resolving Power
What can IPGs achieve in practice? Figure 8 gives an
example of a separation carried to the limit of
a small-scale preparative load. Even when conal-
bumin is greatly overloaded, up to 500
g in a single
Isoelectric Focusing in Capillary Electrophoresis
See II / ELECTROPHORESIS / Capillary Isoelectric Focusing
II / ELECTROPHORESIS / Isoelectric Focusing 1271
track, no smears or precipitations occur, while faint
bands become visible. Another interesting example,
at the very limit of any focusing technique, is given in
Figure 9. Here histones are seen focused at the
steady-state in a very alkaline pH 10}12 gradient. It
can be appreciated that all histones have a pI in the
pH range 11}12, as they should, given their amino
acid composition. Previous data obtained by conven-
tional IEF had attributed to them pIs in the pH 9}10
interval, clearly grossly underestimated.
Further Reading
Bossi A, GelR C, Orsi A and Righetti PG (1994) Isoelectric
focusing of histones in extremely alkaline immobilized
pH gradients: comparison with capillary electrophor-
esis. Journal of Chromatography A 686: 121}128.
Gianazza E, Dossi G, Celentano F and Righetti PG (1983)
Isoelectric focusing in immobilized pH gradients: gen-
eration and optimization of wide pH intervals with
two-chamber mixers. Journal of Biochemical Biophysi-
cal Methods 8: 109}133.
Gianazza E, Giacon P, Sahlin B and Righetti PG (1985)
Non-linear pH courses with immobilized pH gradients.
Electrophoresis 6: 53}56.
Righetti PG (1983) Isoelectric Focusing: Theory, Methodo-
logy and Applications. Amsterdam: Elsevier.
Righetti PG (1990) Immobilized pH Gradients: Theory and
Methodology. Amsterdam: Elsevier.
Righetti PG, van Oss CJ and Vanderhoff JW (eds) (1979)
Electrokinetic Separation Methods. Amsterdam Elsevier.
Rilbe H (1976) Theory of isoelectric focusing. In: Catsim-
poolas N (ed.) Isoelectric Focusing, pp. 14}52. New
York: Academic Press.
Rilbe H (1996) pH and Buffer Theory: a New Approach.
Chichester: John Wiley.
Schaefer-Nielsen C (1986) Computer simulation of pH
gradient formation in isoelectric focusing. In: Dunn MJ
(ed.) Gel Electrophoresis of Proteins, pp. 1}36. Bristol:
Wright.
Svensson H (1961) Isoelectric fractionation, analysis and
characterization of ampholytes in natural pH gradients.
Scandinavica I. The differential equation of solute con-
centration at steady state and its solution for simple
cases. Acta Chemica Scandinavica 15: 325}341.
Vesterberg O (1969) Synthesis of carrier ampholytes for
isoelectric focusing. Acta Chemica Scandinavica 23:
2653}2666.
1272 II / ELECTROPHORESIS / Isotachophoresis
Isotachophoresis
T. Hirokawa, Hiroshima University, Higashi-hiroshima
Japan
Copyright ^ 2000 Academic Press
Introduction
Isotachophoresis (ITP) is one of the electrophoretic
techniques useful for the analysis and isolation of
ionic substances. The Rrst successful analysis
using this method was reported for alkali earth
metals and amino acids by Longsworth in 1953.
The use of glass capillaries was Rrst reported
by Martin and Everaerts in 1967 after the successful
separation of metal cations using a glass capillary
by Konstantinov and Oshurkova in 1963 and of
carboxylic acids using a paper strip by Schumacher
and Studer in 1964. Although various names
had been used for this method, the name ITP
was proposed by Haglund in 1970 and has been
widely accepted. It is interesting to note that the root
of the present capillary electrophoretic methods is
ITP.
An ITP separation is due to the different elec-
trophoretic mobilities of the sample components in
a similar manner to the other electrophoretic tech-
niques. However, ITP has some characteristic fea-
tures which distinguish it from the other elec-
trophoretic techniques. This article, summarizes the
theoretical background of ITP separations and then
describes the analytical and preparative equipment
used. Finally, separation strategies are given together
with typical examples of ITP.
Electrophoretic Mobility and Velocity
Effective mobility
of an ionic substance in a solu-
tion can be expressed as a function of many factors as
follows:

"f (m0, K , pH, T, , , I, Ks, Ccomp) [1]
where m0 is the absolute mobility of a solvated ion,
Ka the acid dissociation constant, pH the pH of the
solution, T the temperature,  the viscosity of the
solvent used,  the dielectric constant of the solvent,
I the ionic strength, Ks the stability constants of the
complexes or ion pairs formed, and Ccomp the concen-
tration of the complex-forming agent or the ion pair-
forming agent.
Eqn [1] shows that the effective mobility is a com-
plex function of the properties of the sample ion, the
solvent used and the coexisting ions. A basic idea of
the electrophoretic separation is to vary the mobilities
of the ions being separated by varying some of the
factors in the above function.
The migration velocity of the ion (v)i can be ex-
pressed as:
vi"
E [2]
where E is the potential gradient of the electrical Reld.
Consequently the difference of the electrophoretic
velocities among separands is the driving force of the
electrophoretic separation.
The name of ‘isotachophoresis’ comes from the
Greek for equal (iso, iso) velocity (tacho, tachoz)
sample dragging (phoresis, foreesqai). Although this
name characterizes its principle as described later, if
the sample velocities were the same throughout the
migration process, there would be no separation.
Principle of ITP
Operational Electrolyte System and Separation
Principle
Two different electrolytes (a leading and a termina-
ting electrolyte) are used in ITP and this is the impor-
tant point which distinguishes it from other elec-
trophoresis method. A sample solution is injected at
the boundary of the two electrolyte solutions as
shown in Figure 1(A). The leading electrolyte is usu-
ally a pH-buffered electrolyte containing leading ions
(L) with the same sign as the sample ions and appro-
priate counterions with pH-buffering ability. Usually,
Cl\ and K# or NH# 4 are used as the leading ions
because of their large mobilities. The terminating
electrolyte contains terminating ions (T) together
with appropriate counterions.
For successful ITP separations, the effective mobili-
ties of A and B in Figure 1 should fulRl the following
relationship:

L'
A'
B'
T [3]
At the initial stage of migration, a homogeneous
mixed zone (A#B) is formed as shown in
Figure 1(B), where A and B migrate with different
velocities under the same potential gradient. In this
case, vA is greater than vB in the mixed zone. A forms
a pure zone in the leading side of the mixed zone, and
 II / ELECTROPHORESIS / Isotachophoresis 1273

remains, the following relations are valid for a

constant current:

vL"vA"vB"vT [5]
EL
L"EA
A"EB
B"ET
T [6]

From eqns [3] and [6] the following relationships

between the potential gradient of zones (Figure 1)
apply:

EL(EA(EB(ET [7]

Therefore the separated zones can be detected by the

use of a potential gradient detector and a conductivity
detector besides spectroscopic detectors.
It should be noted that the pH of the ITP sys-
tem is different in the different zones. The pH buffer-
ing counterions are continuously supplied from the
leading electrolyte, and they regulate the pH of the
sample zones. Usually the following relationship is
valid for an anion analysis:

Figure 1 Separation process in isotachophoresis. (A) Before
migration. (B) Separation process forming a mixed zone AB.
(C) Complete separation (a steady state). L, leading zone; A, B,
sample zone; T, terminating zone; E, potential gradient.
B forms a pure zone at the terminating side of the
mixed zone. After a while, the mixed zone diminished
and the sample components A and B are separated to
form independent zones (Figure 1C). It should be
noted that the migrating zones differ from those in
zone electrophoretic migration. Once they are separ-
ated, they never mix again and the zone lengths are
kept constant according to the sample amount as long
as the migration current is applied. This is the iso-
tachophoretic steady state.
When the migration order is as shown in
Figure 1(C), the following relations are valid:
pHL(pHA(pHB(pHT [8]
The reversed relationship holds in a cationic analysis.
The difference pHL!pHT is usually less than 1
when the pH buffering counterion is selected proper-
ly. Since the pH of the zones are thus not constant in
contrast to zone electrophoresis, the effective mobil-
ity dependence of the samples on the pH of the
leading electrolyte cannot be estimated straightfor-
wardly. And this makes the separation optimization
of isotachophoresis difRcult in comparison with zone
electrophoresis.
The pH buffers conventionally used are sum-
marized in Table 1 for anion analysis. To keep good
buffering capacity of the leading electrolyte, the pHL
Table 1 pH buffers used for anionic analyses
Buffer pK a pH L range

A,A'
A,B Glycylglycine 3.140 2.6}3.6

-Alanine 3.552 3.0}4.0

B,A'
B,B [4] -Aminocaproic acid 4.373 3.8}4.8
Creatinine 4.828 4.2}5.4
Histidine 6.040 5.4}6.4
where
A,A and
B,B denote the effective mobility of Imidazole 7.150 6.4}7.4
A and B ions in the steady state zone and
A,B and TRISa 8.076 7.4}8.4

B,A denote the effective mobility of A ions in the Ammediolb 8.780 8.2}9.2
B zone and that of B ions in the A zone. This relation Ethanolamine 9.498 9.0}10.0
keeps the boundary between A and B zones very a
TRIS, tris(hydroxymethyl)aminomethane.
sharp (self-sharpening effect). b
Ammediol, 2-amino-2-methyl-1,3-propanediol. These buffers
At the isotachophoretic steady state where the zone are used to adjust the pH of an electrolyte containing a leading ion
lengths of all samples are constant and no mixed zone such as hydrochloric acid.
1274 II / ELECTROPHORESIS / Isotachophoresis
should satisfy the following relationship:
pKQ!0.5(pHL(pKQ#0.5
where pKQ is the pKa of the buffers used. The max-
imum buffering capacity is obtained at pHL"pKQ.
The isotachophoretic steady state is achieved when
the following four conditions are fulRlled:
E The leading and the terminating electrolyte are
chosen properly to satisfy eqn [3].
E A constant migration current is applied.
E Mass balance of the counterion is kept: the molar
amount of the pH buffer Sowing into the sample
zone in a unit time should be equal to the amount
Sowing out the sample zone.
E Electroneutrality is kept in each zone as in normal
electrolyte solutions.
ITP is bidirectional in principle, since elec-
trophoretic phenomena are bidirectional. In fact,
isotachophoretic stacking zones can be formed simul-
taneously for anionic and cationic components in
a sample when a suitable electrolyte system is chosen.
In bidirectional ITP, the anolyte and the catholyte are
the leading electrolyte and the terminating electrolyte
for anions, and vice versa for cations.
Although the selection of the operational electro-
lyte system for bidirectional isotachophoresis is not
too difRcult, the difference between the pH of an
anolyte and that of a catholyte is restricted.
Figure 2 Isotachopherograms and experimental definition of RE values. L, leading zone; A, B, C, sample zone(s); T, terminating
zone; h, step height; h, step height corresponding to an asymmetric potential of a potential gradient detector, which can be determined
using theoretical RE value of internal standard; RE, corrected RE value of sample. The subscript S denotes component A, B or C.
Qualitative Analysis
The effective mobility of a sample component is
[9]
uniquely determined under a given set of experi-
mental conditions and this allows qualitative analy-
sis. In practice, some qualitative indices have been
proposed using some different deRnitions on the basis
of the ratio of the potential gradients or the conduc-
tivities of the separated zones. We have proposed RE,
which is deRned for the component A as follows:
RE,A"EA/EL"
L/
A [10]
Figure 2 shows the experimental deRnition of
RE values using step heights, where the asymmetric
potential of the potential gradient detector (h) is
corrected by the use of an internal standard. By
comparing such qualitative values of samples with
those of reference standards, tentative qualitative
analysis can be done. Additional information by
a UV/VIS online detector may be useful. For exact
identiRcation, ITP zones are fractionated and the
fractions analysed by independent analytical
methods.
Quantitative Analysis
In ITP, the concentration of ions in the steady-state
zone is determined by their effective mobilities and by
the concentration of the leading ion. Therefore, dilute
components in a sample are concentrated according

to Kohlrausch’s regulating function and conversely
the concentrated components are diluted during mi-
gration. The following relation is valid among the
total equivalent concentrations of the samples (C t):
C tL'C tA'C tB'C tT [11]
The quantitative index is the time-based zone
length as shown in Figure 2. The absolute zone length
of the sample A (lA in cm) can be deRned as:
lA"1000n/(C tA
r2) [12]
where n is the amount of applied sample (moles) and
r the radius (cm) of the separation tube at a detector.
The zone passing time, t (s) is equal to the actual
zone length divided by the ITP velocity, v (cm s\1).
When both the sample ion and the buffer ion are
monovalent, tA can be expressed as follows:
tA"Fn(1#
Q/
A)/i [13]
where F is the Faraday constant,
Q the mobility of
the buffer ion,
A that of sample A, and i the migra-
tion current.
Instrumentation
Separation System for ITP Analyser
A typical diagram of ITP analyser is shown in
Figure 3 for unidirectional ITP. A separation tube
connecting two electrodes is made of polytetra-
Suoroethylene (PTFE) or fused silica, whose inner
diameter is 0.2}0.5 mm. A PTFE tube as large as
1 mm inner diameter is frequently used before detec-
tion as a pre-column to increase column hold up or
electric charge for better resolution. An additional
leading electrode chamber can be used to apply high
current during pre-separation to reduce analysis time.
The chamber is connected appropriately to the separ-
ation tube before the detector. The leading and the
terminating electrode chamber are typically 20 mL in
volume.
The separation tube is also Rlled with a leading
electrolyte and a terminating electrolyte. The bound-
ary is formed at a sample injection port and a sample
solution is introduced there typically by using
a microsyringe. Sample components are separated
during migration and their zones are detected using
appropriate detectors.
Detectors
The quality of the detector employed determines and
limits the qualitative and quantitative analysis of ITP.
II / ELECTROPHORESIS / Isotachophoresis 1275
Figure 3 ITP apparatus with potential gradient detection
(PGD), high-frequency contactless conductivity detection
(HFCCD), and ultraviolet detection (UVD). HV, high-voltage
power supply (constant current); V/F, voltage-frequency conver-
ter; F/V, frequency-voltage converter.
There are two demands on the detector; Rrstly to
accurately reSect the separation occurring in ITP and
secondly to obtain the isotachopherogram from the
analysis with high reproducibility.
Detectors for ITP can be divided into universal and
speciRc types. The signal from universal detectors is
directly proportional to the effective mobilities of the
ionic species, and these detectors detect zones of all
components separated in the narrow-bore tube. Ther-
mometric, potential gradient and conductivity de-
tectors belong to this class. The detection limit of
potential gradient and conductivity detectors is sub-
nanomole but that of thermometric detector is rather
high. On the other hand, speciRc detectors such as UV
spectrophotometers allow the identiRcation of some
components directly, or at least can give additional
information about zones.
In Figure 3, three detection systems using a
high-frequency contactless conductivity detector
(HFCCD), a potential gradient detector (PGD), and
an ultraviolet detector (UVD) are shown. Since the
sensing electrodes of PGD directly contact the solu-
tion in the capillary, this system needs a device to
isolate high voltage. In Figure 3, a photocoupler is
used for this purpose (IP-1B, IP-2A, IP-3A, Shimadzu,
Kyoto, Japan, production discontinued). A trans-
former is used in the usual (contact type) conductivity
detection (a.c. method). Although the sensitivity of
a contactless detector is lower than the direct contact
1276 II / ELECTROPHORESIS / Isotachophoresis
detector, the merit of HFCCD is obviously that the
detection system needs no such isolation device.
Preparative ITP Apparatus
Capillary type ITP is useful not only for analytical
purposes but also for preparative purposes. Capillary
type (CITP) is useful for batch processing of a small
amount of sample.
In addition to direct cutting of the capillary section
containing the target of interest, preparative methods
in CITP can be classiRed into three types, as shown in
Figure 4. Figure 4(A) shows a preparative ITP system
reported by Arlinger for the fractionation of the en-
tire sample zones. This system was used in the LKB
Tachofrac (Bromma, Sweden, 1983, production dis-
continued). The zones were swept gradually by a
counterSow of a leading electrolyte (ex. 3 L min\1)
on applying migration current, and the fractions were
Rxed on a cellulose acetate strip. The separated zones
were successively pushed out through a T-branch by
applying a counterSow and the zones were continu-
ously Rxed on the strip by an electric spray. The linear
Figure 4 Preparative methods in capillary-type isotachophor-
esis. (A) Arlinger’s counterflow method. (B) Modified Arlinger’s
method. (C) Microsyringe method. (D) Fractionation valve
method. A and B, samples; L, leading electrolyte; T, terminating
electrolyte; Inj., sample injection port; SP, a counterflow pump;
Det., a detector.
velocity of the counterSow was set only a few percent
higher than the isotachophoretic migration velocity
so as not to dilute the sample by the leading electro-
lyte. The fractions on the strip can be analysed by
immunological and radioactive methods. The zymo-
gram technique can be used directly on the strip. The
fractions have to be eluted, for analysis by other
methods.
A dropwise fractionating method was developed
utilizing a counterSow technique. The schematic dia-
gram of the apparatus is shown in Figure 4(B). When
the sample zone is pushed out from a T-branch,
a spray effect is usually observed due to electro-
static forces. This can be a convenient interfacing
technique but it disturbs dropwise fractionation. The
electric spray and Suctuation of the drop rate due to
electrostatic forces are suppressed by a very simple
electrostatic device: As shown in Figure 4(B), the
exiting fraction is surrounded by a copper coil, which
is connected to a nozzle. The fractions are collected
directly into small test tubes on the fraction collector
through the coil. By using this technique, complete
recovery of the mobile components in the injected
samples is possible with minimum risk of loss and
contamination. It should be noted, however, that
mixing of adjacent sample zones cannot be avoided.
The average volume of one drop was ca. 5 L and the
deviation was estimated as #10%. A few nanomoles
of the sample components are contained in a drop.
The concentration of samples in the fractionated
drops or the amount of the target in a fraction was
adjustable by changing the Sow rate of the leading
solution. A typical counterSow of a leading electro-
lyte was ca. 12  L min\1, which is much higher than
the Arlinger-type apparatus.
Figure 4(C) shows another method reported by
Kobayashi et al., where the separated sample zones
are discontinuously isolated by using a microsyringe.
Kobayashi et al. used a potential gradient detector
(PGD) with a sample-removal port to fractionate the
target zone immediately after the tail of the zone was
detected by the PGD. Although the method was not
intended for the successive fractionation of the entire
sample zones, the ease of operation is notable.
This technique was employed for IP-1B and IP-2A
instruments (Shimadzu, Kyoto, Japan, production
discontinued).
Figure 4(D) shows other discontinuous fractiona-
tion technique using a specially designed fractionat-
ing valve placed at the end of the separation capillary.
After trapping the target zone in the valve, the zone is
Sushed out.
In addition, the separation tube used was a series of
four separation tubes (inner diameter of the tubes,
5}0.5 mm) in order to increase the amount of

sample separated. The tube of 5 mm inner diameter
was made of acrylic resin and the maximum inject-
able sample volume was 2.5 mL.
Free-Wow apparatus Since no solid media are used
in free-Sow electrophoresis (FFE), the most important
point in instrumentation is the stabilization of the
separated zones for any electrophoresis mode. Un-
stable zones may be caused by unstable operational
electrolyte, sample Sow, heat convection, density-
driven Sow, electroosmosis, etc. Bier et al. and
Thormann et al. summarized several different
methods for stabilizing the zones. A Sat-type FFE is
treated here, although there are several different ap-
proaches using different geometries, such as a thin
Rlm between parallel plates, a cylindrical laminar
Sow between two coaxial cylinders, etc.
Continuous FFE apparatus utilizing a thin Sowing
Suid was originally designed by Hanig for zone elec-
trophoresis. Prusik and Wagner et al. designed and
constructed similar apparatus, suggesting that several
modes of electrophoresis can be used. At present, the
only FFE apparatus available is the Octopus continu-
ous electrophoresis apparatus from Dr. Weber
GmbH (Kirchheim-Heimstetten, Germany). By using
this apparatus, up to several grams of pure substances
can be prepared daily, although the amount depends
on the properties of the sample. Figure 5 illustrates
the electrolyte circuits of an FFE system (Octopus)
Figure 5 Electrolyte circuits of a free-flow electrophoresis ap-
paratus (Octopus, Dr. Weber GmbH). Outer dimensions of separ-
ation chamber are 640;180;80 mm, Separation area is
500;100 mm (variable thickness 0.4}2.0 mm). P, pumps.
II / ELECTROPHORESIS / Isotachophoresis 1277
when operated in continuous free-Sow isotachophor-
esis (CFFITP) mode. The effective size of a typical
separation chamber is 10-cm wide, 50-cm high and
0.4-mm thick. The sample solution is supplied with
a multifold peristaltic pump together with an anolyte
and a catholyte. OverSow of the separation chamber
is collected as 96 fractions. The Sow rate is variable in
the range 0.3}100 mL h\1. The sample residence
time is variable in the range 1}40 min. High Sow rate
and small residence time allow stable Sow and conse-
quently stable position of zones.
The anolyte and catholyte are circulated by pumps
during migration. A dialysis membrane isolates the
separation chamber from the electrode compart-
ments. The electrolyte solutions may be denatured
(the pH will change) after a few hours operation. The
separation chamber can be thermostatted and separ-
ation can be monitored with a VIS CCD detection
system which can be positioned near the end of the
chamber. To obtain pure fractions, the positions of
the sample zones at the end of the sample chamber
should be stable. The positions are dependent on
several factors such as the electric Reld strength, tem-
perature of the electrolytes, Sow rate, and sample and
buffer composition. Since these factors are closely
correlated with each other, careful control is needed.
SufRcient residence time and separation distance are
necessary especially when mobility differences are
small. For this purpose, a larger separation chamber
or a counterSow technique should be used as reported
by Prusik (a 50;50 cm square chamber with a thick-
ness of 0.5 mm).
Separation Strategy and Typical
Applications
Operational electrolyte conditions should be opti-
mized to obtain the best quality of separation by
changing electrolyte parameters so that the difference
of the effective mobilities of the target components
should be as large as possible. Strictly speaking, the
separability depends on the mobility difference in the
mixed zone (see Figure 1). However, to a Rrst ap-
proximation, the mobility differences at the steady
state or the difference of the RE values may be used
for the criterion for optimization of separations.
For separation optimization in electrophoresis,
a theoretical approach is sometimes very useful. In
isotachophoresis especially, the pH and the ionic
strength of the separated zones are different to one
other. The optimum electrolyte system can be deter-
mined by iterative computer calculations where
the effective mobilities of the species being separated
at the steady state are calculated using their
1278 II / ELECTROPHORESIS / Isotachophoresis

physicochemical constants, and the differences of the
effective mobilities compared. This method may be
called computer-aided separation optimization,
which enables the determination of the electrolyte
conditions such as the pH without time-consuming
‘trial-and-error’ experiments. Some examples of
simulation are also included in this section.
pH Effect
When the sample contains weak electrolytes, the pH
of the leading electrolyte (pHL) should be carefully
chosen to obtain high separability among the compo-
nents, since the effective mobilities of weak electro-
lytes change drastically according to the pH of
the solution. This pH effect on the effective mobility
is the most important and should be examined Rrst.
Figure 6 shows the simulated and the observed
isotachopherograms for a test mixture of pyrophos-
phoric (P2O7), triphosphoric (P3O10), tetraphosphoric
(P4O12) acids, and 15 nucleotides, AMP, ADP, ATP,
CMP, CDP, CTP, GMP, GDP, GTP, IMP, IDP, ITP,
UMP, UDP and UTP. The pHL was adjusted to 4.7
with creatinine. The optimum pHL range was rather
limited, because CTP may form a mixed zone with
UDP at low pHL and with ATP at the high pHL.
Obviously from this example, the pH of the leading
electrolyte should be carefully chosen in order to
obtain high separability for weak electrolytes. For
this purpose, computer simulation of the iso-
tachophoretic steady state is useful, when the mobil-
ity and dissociation constants are available. In fact,
the optimum pHL of the above separation has been
determined by simulation.
Solvent Effect
Since the mobility is strongly affected by the viscosity
of the solvent and the dissociation constants depend
on the dielectric constant of the solvent, the use of
a nonaqueous solvent or a mixed solvent may im-
prove electrophoretic separations (the solvent effect).
The other advantage of the use of a nonaqueous
systems is that it enables the analysis of substances
with low water solubility. Many solvents have been
successfully applied for isotachophoresis, such as
methanol, ethanol, dioxane, acetone, propanols, di-
methylformamide, etc. The migration behaviour in
a nonaqueous solvent is very different from that
in an aqueous solvent. The use of a mixed solvent

Figure 6 Simulated and observed isotachopherograms of 19 anions, AMP, ADP, ATP, CMP, CDP, CTP, GMP, GDP, GTP, IMP,
IDP, ITP, UMP, UDP, UTP, pyrophosphoric acid (P2O7), triphosphoric acid (P3O10), tetraphosphoric acid (P4O12) and propionic acid
(Prop) at pHL"4.7 (creatinine buffer). The terminator was pelargonic acid (Pel). Current, 75
A. Sample amounts, 1.2}3 nmol. The
leading electrolyte contained 0.2% hydroxypropylmethylcellulose (HPMC). Buffer, creatinine.C tL , concentration of leading ion.
 II / ELECTROPHORESIS / Isotachophoresis 1279

(aqueous}nonaqueous) provides further possibilities
for mobility control to improve separations.
Figure 7(A) and (B) show the simulated and ob-
served isotachopherograms at pH 8 using triethanol-
amine buffer when methanol is used as solvent. The
leading ion is perchlorate. Obviously, the separation
is complete. On the other hand, insufRcient separ-
ation was predicted for the aqueous system where the
leading ion was chloride when pH "3.6 (-alanine
*
buffer), as shown in Figure 7(C). Apparently the step
heights of 2-naphthalenesulfonate, picrate and o-
chlorobenzoate were similar and they may form
a mixed zone in the actual analysis. No satisfactory
separation was estimated when pH effects in the
aqueous system were used.
Thus it is evident that the separation behaviour in
isotachophoresis is strongly affected by the solvent
used. As demonstrated above, high separability can
be expected by the use of methanol solvent for par-
ticular samples but it should be noted that carboxylic
acids are esteriRed gradually on standing in methanol
solution, although the production of esteriRed com-
pounds during analysis is negligible.
Complex-forming Effect
The use of a complex-forming agent is a traditional
technique to improve separability especially in the
case of metal ions. To achieve complex-forming equi-
libria, a constant amount of the complex-forming
agent should be supplied continuously to the sample
zone. The complexing agent can be supplied as the
counterion in the leading electrolyte, as neutral ligand
(crown ether and cyclodextrin), or as the terminating
ions.
Complex formation is very useful for the separation
of metal ions. A typical example of the iso-
tachophoretic separation of metal ions utilizing the
complex-forming effect was reported for lanthanides,
where -hydroxyisobutyric acid (HIBA) was used as
the complexing agent. The concentration of the com-
plexing agent and the pH of the leading electrolyte
should be optimized to obtain a good separation. By
adding malonic acid to the main agent HIBA, 15
rare-earth ions (lanthanide and yttrium) were success-
fully separated as shown in Figure 8.
Conclusion
Isotachophoresis is a useful analytical technique with
high reproducibility. However, ITP is not a very fam-
iliar technique to many chromatographers. Possible
reasons are absolute sensitivity of ITP is in the sub-
nanomole range, which is low in comparison with
recent capillary electrophoresis (CE) techniques;
auto-samples are not available; and stepwise record-
ing is not readily accepted by chromatographers.
However, it should be noted that the relative sensitiv-
ity of ITP is comparable with CE especially when the
detection method is the same (e.g. UV detection),
since a relatively large volume of a sample (e.g.

Figure 7 The observed (A) and the simulated isotachopherogram (B) of (1) bromide, (2) chloride, (3) picrate, (4) formate, (5)
2-naphthalenesulfonate, (6) o-chlorobenzoate, (7) p-chlorobenzoate, (8) 3-hydroxybutyrate, (9) crotonate, (10) propionate and (11)
4 and the buffer used was triethanolamine (pHL"8.06).
pelargonate ions in methanol as solvent. The leading ion was 10 mmol L\1 ClO\
(C) The simulated isotachopherogram of the above samples except for bromide. The leading ion was 10 mmol L\1 chloride and the
buffer used was -alanine (pHL"3.6). The terminator was pelargonate. Imp., impurity of the used electrolyte system.
1280 II / ELECTROPHORESIS / Micellar Electrokinetic Chromatography
Figure 8 The observed isotachopherogram of 15 rare-earth
ions (lanthanide ions and yttrium ion). HIBA, the complex-forming
agent -hydroxybutyric acid. The leading ion, 20 mmol L\1 NH# 4 ;
pH buffer"2-ethyl-n-butyric acid (pHL"4.8). The sample
amount was 0.33 mmol L\1;5 L. Migration current"40
A.
The terminator is carnitine hydrochloride. (Carn.) imp., impurity of
the used electrolyte system.
Isotachophoresis in Capillary Electrophoresis
See II / ELECTROPHORESIS / Capillary Isotachophoresis
Mass Spectrometry Detection in
Capillary Electrophoresis
See II / ELECTROPHORESIS / Capillary Electrophoresis-Mass Spectrometry
Micellar Electrokinetic Chromatography
M.-L. Riekkola, Laboratory of Analytical Chemistry,
University of Helsinki, Finland
Copyright ^ 2000 Academic Press
100 L) can be injected. For preparative purpose, ITP
is sometimes better than CE especially when the
sample size is relatively large. In order to utilize the
favourable features of ITP, an automated apparatus is
needed or a method should be found to use commer-
cial CE apparatus for ITP.
Further Reading
Babskii VG, Zhukov MYu and Yudovich VI (1989)
Mathematical Theory of Electrophoresis. New York:
Plenum.
Bocek P, Deml M, Gebauer P et al. (1988) In: Radola BJ
(ed.) Analytical Isotachophoresis. Basel: VCH.
Everaerts FM, Beckers JL, Verheggen ThPEM (1976) Iso-
tachophoresis. Theory, Instrumentation and Applica-
tion. Amsterdam: Elsevier.
Gebauer P, Caslavska J and Thormann W (1991) Journal of
Biochemical and Biophysical Methods 23: 97.
Hirokawa T, Nishino M, Aoki N et al. (1983) Table of
isotachophoretic indicies. I. Simulated qualitative and
quantitative indicies of 287 anionic substances in the
range pH 3}10. Journal of Chromatography 271:
D1}D106.
Li SFY (1992) Capillary Electrophoresis, Principles,
Practice and Applications. Amsterdam: Elsevier.
Moscher RA, Saville AD and Thormann W (1992) The
Dynamics of Electrophoresis. Weinheim: VCH.
Pospichal J, Gebauer P and Bocek P (1989) Measurement
of mobilities and dissociation constants by capillary
isotachophoresis. Chemical Reviews 89: 419}430.
Introduction
Micellar electrokinetic capillary chromatography
(MEKC), Rrst introduced by Shigeru Terabe and co-
workers in 1984, has extended the potential of
 II / ELECTROPHORESIS / Micellar Electrokinetic Chromatography
capillary electromigration techniques to the separ-
ation of uncharged analytes. With its impressive sep-
aration efRciency and Sexibility, MEKC has become
a popular technique especially in the pharmaceutical
and biomedical Relds.
Above their critical micelle concentration (CMC),
surfactant monomers added to an electrolyte solution
form aggregates called micelles. Individual micelles
are not signiRcantly larger than the solutes being
separated. On account of their small size and large
number, they have a high surface area-to-volume
ratio. Their structures are dynamic, with the average
residence time of a surfactant monomer in the micelle
being in the order of 1 ms or less. Separation in
MEKC is based on the partitioning of analytes be-
tween the micelles and the aqueous phase, in the
presence of electroosmotic Sow. The micelles act as
a pseudo-stationary phase. The mechanism of the
analyte}micelle interaction is mainly determined by
hydrophobic and electrostatic interactions. MEKC
was originally developed to exploit the advantages
of capillary electrophoretic techniques (high efRcien-
cies, the requirement of only minute amounts of
sample and reagent, fast analysis time) in the separ-
ation of neutral solutes of closely similar structure,
but it is also applicable to the separation of charged
Figure 1 Schematic depiction of separation in micellar electrokinetic capillary chromatography.
1281
compounds. A basic capillary electrophoresis (CE)
instrument is used, and the separations are carried out
usually in uncoated fused silica capillaries after hy-
drodynamic injection.
Separation in MEKC
When one or more micelle-forming surfactants are
added to the electrolyte solution at concentrations
above their CMC, partition of the analytes into the
micellar pseudo-stationary phase increases the selec-
tivity of the separation system. The overall separation
of compounds is based on their differential solubiliz-
ation into the micelles and on the migration velocities
of the micelles under the electric Reld, in the presence
of electroosmotic Sow (EOF). The separation prin-
ciple for an anionic surfactant is illustrated in
Figure 1.
The separation of neutral analytes is based on their
partitioning between the aqueous phase and the
micellar stationary phase. When solutes interact
strongly with the micelles their migration time is
comparable to that of the micelles, tmc, allowing the
solutes to serve as micelle markers. Neutral analytes
migrate with times t1 and t2, which lie inside a win-
dow formed by the migration times of the neutral
1282 II / ELECTROPHORESIS / Micellar Electrokinetic Chromatography
Figure 2 Migration window for neutral solutes in micellar elec-
trokinetic capillary chromatography. EOF, electroosmotic flow.
electoosmotic Sow marker, teo, and the micelle
marker, tmc (Figure 2). A relatively polar molecule
(e.g., acetone, acetonitrile, formamide, methanol,
1-propanol or tetrahydrofuran) can be used as
electroosmotic Sow marker, and usually a highly
hydrophobic, neutral compound such as Sudan III,
Sudan IV, dodecanophenone, Orange OT, or Yellow
OB as micelle marker. The migration window is Rnite
because the micelles themselves migrate out of the
capillary. Even though the peak capacity is restricted
by the migration window, high separation efRciencies
can be achieved. A wide migration time window is
favourable for high resolution, but then a long analy-
sis time may be required.
The micellar phase is not a true stationary phase
because it is moving along the capillary towards the
detector. When the analyte is permanently retained,
its migration time (tm) is identical with the migration
time of the micelle (tmc). Therefore, the term ‘reten-
tion factor’ used in chromatography should be re-
placed by the term ‘partition factor’ in MEKC. The
partition factor kmekc is described as:
nmc
kmekc"
naq
where nmc and naq are the numbers of the analytes in
micellar and aqueous phases, respectively. In the case
of a neutral analyte, kmekc can also be calculated
directly from the migration times:
tm!teo
kmekc"
teo(1!tm/tmc)
However, there may be variations in kmekc depending
on EOF and the micelle marker; in particular, the
choice of the micelle marker may have a signiRcant
effect on the value.
The selectivity
can then easily be determined by
the ratio of the partition factors of two compounds:
kmekc2

"
kmekc1
The most effective way to alter the selectivity of
nonpolar analytes in MEKC is to change the micellar
phase by changing the type of surfactant. When com-
pounds are neutral, factors such as concentration of
electrolyte and micellar solutions, pH, voltage and
temperature have a relatively minor effect on the
selectivity of the system. When the compounds are
charged, on the other hand, variations in pH may
induce changes in the dissociation of the compounds,
affecting their charge, and thereby the solute}micelle
ionic interactions and electrophoretic mobilities.
The resolution in MEKC is determined by the
equation given by Terabe et al.:




(N
!1 kmekc2 1!teo/tmc
Rs"
4
1#kmekc2 1#(teo/tmc)kmekc1
where N is the plate number. The resolution of the
system depends on the efRciency, the selectivity, the
partition factor and the migration time window.
Surfactants
Unique selectivities are achieved in MEKC through
appropriate choice of anionic, cationic, nonionic and
zwitterionic surfactants (Table 1). Surfactants are
molecules with distinct hydrophobic and hydrophilic
parts. The CMC increases dramatically with the alkyl
chain length of the surfactant. At Kraft temperature,
TKr, the solubility of the surfactant increases rapidly.
TKr is the point at which surfactant solubility equals
the CMC. The Kraft point varies with the surfactant,
increasing with the length of the alkyl chain. Surfac-
tant concentrations above the CMC and temperature
above the Kraft point are required for the formation
of micelles. Changes in temperature, concentration of
surfactant, pH, ionic strength, additives in the aque-
ous phase and structural groups in the surfactant may
cause changes in the size, shape and aggregation num-
ber of the micelles. In aqueous media, surfactants
with bulky or loosely packed hydrophilic groups and
long, thin hydrophobic groups tend to form spherical
micelles, while those with short, bulky hydrophobic
groups and small, close-packed hydrophilic groups
tend to form lamellar cylindrical micelles. Factors
that decrease the electrostatic repulsion between the
head groups of ionic surfactants favour micelle
 II / ELECTROPHORESIS / Micellar Electrokinetic Chromatography 1283

Table 1 Typical surfactants used in MEKC, with their critical micelle concentration (CMC)

Surfactant CMC (mM) Temperature (3C)

Anionic
Sodium dodecyl sulfate (SDS) 8.2 25
Sodium tetradecyl sulfate (STS) 2.1 25
Sodium decyl sulfate 33 40
Sodium dodecyl sulfonate 11.4 40
Sodium N-lauroylmethyl-N-taurate 8.7 25
Lithium perfluorooctane sulfonate (LiPFOS) 6.3 25
Cationic
Cetyltrimethylammonium bromide (CTAB) 0.92 25
Cetyltrimethylammonium chloride (CTAC) 1.3 30
Tetradecyltrimethylammonium bromide (TTAB) 3.6 25
Dodecyltrimethylammonium bromide (DTAB) 16 25
Dodecyltrimethylammonium chloride (DTAC) 20 25
Cationic fluorosurfactant (Fluorad FC 134) na
Nonionic and zwitterionic
Octyl glucoside (OGLU) 25 25
Polyoxyethylene (23) dodecanol (Brij-35) 0.1 na
Polyoxyethylene (20) sorbitane monooleate (Tween 80) 0.01 na
Polyoxyethylene (20) sorbitane monolaurate (Tween 20) 0.059 na
3-[3-(Chloroamidopropyl) dimethylammonio]-1-propane- 4.2}6.3 na
sulfonate (CHAPS)
Chiral surfactants
Sodium N-dodecanoyl-L-valinate (SDVal) 2 na
Sodium N-dodecanoyl-L-glutamate (SDGlu) na
Digitonin (DIG) na
Bile salt surfactants
Sodium cholate (SC) 12.5 25
Sodium deoxycholate (SDC) 10 25
Sodium taurocholate (STC) 4 25
Sodium taurodeoxycholate (STDC) 6 na
Sodium glycodeoxycholate na

na, not available.

formation leading to lower CMC in electrolyte solu-
tions than in pure water.
Many physical properties change dramatically at
the CMC. These changes can be exploited by deter-
mining the CMC of surfactants in CE electrolyte
solutions, for example by measuring surface tension,
light scattering, refractive index, electrical conductiv-
ity or electrophoretic mobility. The data are plotted
against surfactant concentration, and a change in the
slope corresponds to the CMC. However, the CMC
obtained may differ according to the method used
because micellization is a gradual aggregate growth
which occurs over a Rnite concentration range. CMC
values for the most commonly used surfactant, so-
dium dodecyl sulfate, in selected electrolyte solutions
are listed in Table 2.
Anionic Surfactants
Anionic surfactant systems are preferred in MEKC
because the electrophoretic migration of the micelles
is in the opposite direction to the electroosmotic Sow,
and the micelles do not interact with the negatively
charged walls of the fused silica capillaries. Anionic
surfactants with alkyl chain and polar group, such as
sodium decyl sulfate, sodium N-lauroyl-N-methyl-
taurate, sodium tetradecyl sulfate, and especially so-
dium dodecyl sulfate (SDS) are the most widely used.
Simultaneous separation of neutral and positively
charged compounds is not possible at low pH because
the EOF is too slow to carry the micelles to the cathode.
Most studies with anionic surfactants have been
carried out under neutral or basic conditions. The
most frequently used anionic surfactant, SDS, forms
relatively spherical micelles with hydrophobic tail
groups oriented towards the centre and charged head
groups along the outer surface. The surfaces of SDS
micelles possess a large net negative charge, giving them
a large electrophoretic mobility toward the anode.
Another group of anionic surfactants, which has
been widely used in separations of both neutral and
ionic analytes, is bile salts. Bile salts have a hydroxyl-
substituted steroidal backbone with hydrophilic and
1284 II / ELECTROPHORESIS / Micellar Electrokinetic Chromatography

Table 2 CMC values of SDS in selected electrolyte solutions at 253C

Electrolyte solution CMC (mM) Method of determination

50 mM AMPSOa (pH 9.0) 3.6 Conductometric titration

50 mM AMPSOa (pH 9.0) 3.9 CE
50 mM AMPSOa (pH 8.7) 2.7 Surface tension
20 mM PIPESb, 20 mM NaOH (pH 7.0) 3.8 Conductometric titration
100 mM BESc, 100 mM NaOH (pH 7.0) 3.1 Conductometric titration
100 mM borate, 50 mM phosphate (pH 7.0) 2.9 Conductometric titration
5 M urea, 100 mM borate, 50 mM phosphate (pH 7.0) 4.4 Conductometric titration
20% DMSO (v/v), 25 mM sodium tetraborate, 50 mM sodium dihydrogen 6 Conductometric titration
phosphate (pH 7.0)
20% acetone (v/v), 25 mM sodium tetraborate, 50 mM sodium dihydrogen 6.3 Conductometric titration
phosphate (pH 7.0)
20 mM sodium tetraborate (pH 9.2) 3.1 CE
20 mM sodium tetraborate (pH 8.0) 5.5}9.6 CE
5 mM sodium tetraborate}acetonitrile (85 : 15, v/v) 7.3 CE
5 mM sodium tetraborate (pH 9.2) 5.3 CE
100 mM sodium tetraborate, 100 mM sodium dihydrogen phosphate (pH 6.0) 2 CE
100 mM sodium tetraborate, 100 mM sodium dihydrogen phosphate (pH 6.5) 2.4 CE
100 mM sodium tetraborate, 100 mM sodium dihydrogen phosphate (pH 7.0) 3.1 CE
100 mM sodium tetraborate, 100 mM sodium dihydrogen phosphate (pH 7.7) 4 CE
50 mM CHESd (pH 10.0) 2.9}5.2 CE
50 mM CHESd (pH 10.0) 2.7}5.4 CE
80 mM CHESd (pH 10.0) 1.6}2.2 CE
100 mM CHESd (pH 10.0) 1.2}2.4 CE
50 mM ammonium acetate (pH 9.0) 1.7}2.7 CE

a
AMPSO"3-[(1,2-dimethyl-2-hydroxyethyl)amino]-2-hydroxypropanesulfonic acid; bPIPES"piperazine-N,N -bis(2-ethanesulfonic
acid) monosodium salt; cBES"N,N -bis(2-hydroxyethyl)-2-aminoethanesulfonic; dCHES"2-(N-cyclohexylamino)ethanesulfonic
acid.

hydrophobic faces and they form helical micelles. Bile
salts have a lower solubilizing effect on hydrophobic
compounds than does SDS.
Cationic Surfactants
Unlike anionic surfactants, positively charged surfac-
tants, monomers and micelles are strongly attracted to
the negatively charged surface of the fused-silica capil-
lary wall and thus have a signiRcant effect on EOF.
Cationic surfactants such as long-chain alkylam-
monium salts may even cause a reversal of EOF
through electrostatic interactions with the capillary
surface, and this may occur at surfactant concentra-
tions below the CMC. The capability for reversed EOF
has been successfully exploited in MEKC separations.
Neutral and Zwitterionic Surfactants
Although neutral surfactants with zero elec-
trophoretic mobilities cannot be exploited in the
MEKC separation of nonionic solutes, they can be
applied to the separation of ionic solutes. Since prob-
lems with Joule heat do not arise when nonionic
surfactants are used at high concentration, large volt-
ages can be used even when surfactants are added to
the buffer in high concentration. Like the neutral
surfactants, the zwitterionic surfactants do not
contribute to the net conductivity of the electrolyte
solution.
Mixed Micelles
Selectivity in MEKC can often be improved by using
mixed surfactants. Clearly different selectivities from
those obtained with the corresponding single micelles
can be achieved, Some mixed micellar systems are
presented in Table 3.
High Molecular Mass Surfactants
The high molecular mass surfactants used in MEKC
are either oligomers of monomeric surfactants or
block copolymers with surface-active properties. It
has been proposed that the micelle is formed of
a single molecule, and accordingly it has been termed
a ‘molecular micelle’. Because their CMC values are
close to zero, molecular micelles are considered to be
highly stable irrespective of the experimental condi-
tions.
Surfactants and Cyclodextrins
Cyclodextrins (CD) are the most popular chiral selec-
tors for chiral separations by MEKC. The separation
mechanism is based on differential partitioning of
solutes between the micellar and CD aqueous phase.
 II / ELECTROPHORESIS / Micellar Electrokinetic Chromatography 1285

Table 3 Selected mixed micellar systems used in MEKC The sensitivity of mass spectrometry (MS), and the
possibility of obtaining molecular information on
Mixed micellar system Surfactants in the mixturea compounds, make the on-line coupling of MEKC
Anionic}nonionic surfactants SDS and Brij-35 with MS highly attractive. Electrospray ionization
SDS and Tween 60 (ESI) has been one of the most popular ionization
SDBS abd Brij-35 techniques in coupled CE}MS. Although MEKC is
SDS and Tween 20 a convenient separation technique for neutral
Bile salts and polyoxyethylene-
analytes, problems are encountered in the on-line
4-dodecyl ether
Anionic}anionic surfactants SDS and sodium cholate MEKC}ESI}MS interface connection because the
SDS and sodium octyl sulfate micelles in the electrolyte solution are nonvolatile and
SDS and bile salts tend to contaminate the MS. A number of approaches
Two different bile salts have been developed to overcome the problems of
LiPFOS (fluorocarbon) and
separating neutral compounds, while at the same time
LiDS (hydrocarbon)
Anionic}cationic surfactants Fluorosurfactants FC 128 and preventing micelles from entering the mass spectrom-
FC 134 eter. These include use of the heart-cut technique,
Anionic}zwitterionic surfactants SDS and SB-12 high molecular mass surfactants, a semipermeable
Nonionic}nonionic surfactants Tween 20 and Tween 80 membrane interface, anodically migrating micelles,
Triton X-100 and Brij-35
and the partial Rlling technique. An electro-
Cationic}cationic surfactants TTAC and OTAC
TTAB and DTAB spray}chemical ionization interface is a possibility
for certain types of online MEKC}MS applications.
a
SDS"sodium dodecylsulfate; Brij-35"polyoxyethylene (23)
dodecanol; Tween 20"polyoxyethylene (20) sorbitane monolau- Applications
rate; Tween 60"polyoxyethylene (20) sorbitane monostearate;
SDBS"sodium dodecyl benzenesulfonate; LiPFOS : lithium MEKC has been applied to a wide variety of com-
perfluorooctane sulfonate; LiDS"lithium dodecyl sulfate; SB- pounds, including phenols and chlorinated phenols,
12"N-dodecyl-N,N -dimethyl-3-ammonio-1-propanesulfonate;
amino acids, several pharmaceuticals and their
TTAC"tetradecyltrimethylammonium chloride; OTAC"octyl-
trimethylammonium chloride; DTAB"dodecyltrimethylam- metabolites, porphyrins, peptides, nucleic acids, nuc-
monium bromide. leosides and oligonucleotides. The capability for di-
rect injection of biological Suids (plasma, serum,
urine) is a special feature of electrokinetic capillary
Most of the surfactants used in separations have been
analysis. Effective solubilization of the biological
anionic.
matrix components by surfactants, and increased sel-
ectivities due to hydrophobic interactions with the
Optimization of Separation micellar pseudo-stationary phase are evidently ad-
vantageous in bioanalysis. The use of MEKC for
Resolution in MEKC is a highly complex and non-
therapeutic and diagnostic drug monitoring has also
linear function of experimental variables and is very
proven to be of considerable value.
difRcult to optimize systematically. In a search for the
optimal conditions for separation, several mathemat-
ical models have accordingly been developed. Often
Future Directions
just a few test runs are needed to predict the best The great advantage of MEKC is the feasibility to
overall running conditions, though this naturally de- manipulate the selectivity simply by changing the
pends on the number of parameters included in the composition of the micellar phase. Even though sev-
optimization strategy. When more than one surfac- eral surfactants have shown their potential to act as
tant is added to the electrolyte solution, the situation micellar pseudo-stationary phase, the versatility of
is complicated by the possible micelle}micelle interac- the technique and the range of applications can
tions. Examples of the statistical optimization be further extended by developing new synthetic
schemes used in MEKC are listed in Table 4. micelle-forming surfactants like polyelectrolytes or
exploiting mixed micelles or biomembranes as
pseudo-stationary phases. Understanding the mecha-
Detection nisms involved will greatly facilitate the systematic
Of the various detection systems employed in MEKC optimization of the large number of experimental
separations, optical systems are the most extensively parameters leading to better, faster, easier, and more
used, and ultraviolet detectors (UV) used in conjunc- reliable separations. In addition, studies are still
tion with commercial CE instruments are a typical needed to clarify new possibilities to couple MEKC
solution. with mass spectrometry.
1286 II / ELECTROPHORESIS / Microtechnology

Table 4 Statistical optimization schemes used in MEKC

Optimized parameter Parameters varied Modelling

Selectivity and resolution pH, [SDS], [borate] CCD c, desirability functions

Selectivity and resolution pH, [SDS], [sodium cholate], [AMPSO]a CCD, desirability functions
Resolution [acetonitrile], [urea] Iterative regression strategy
Resolution 9 for a stepwise screening, followed by Fractional factorial design, full factorial
3: pH, [SDS], [acetonitrile] design, RSM d
Yield for the derivatization of some Reaction time, T, ionic strength, pH, Fractional factorial design, CCD, RSM
dipeptides [isopropanol]
Selectivity pH, [SDS] Iterative regression strategy
Resolution [SDS], [acetonitrile] CABRO II e
Precision and efficiency [SDS], V, T FUMI f
Resolution T, V, ionic strength, [SDS], [HPMC]b, PLS g
[-cyclodextrin]
Resolution [SDS], [urea] CABRO II
Resolution pH, [SDS] CAMOS h
Resolution pH, [buffer], [SDS], [SDS#sodium Plackett}Burman statistical design
heptyl sulfate], [acetonitrile]
Resolution [SDS], [N,N-dimethylformamide], ORM i
ionic strength
Resolution pH, [SDS], [tetrabutylammonium salt] ORM
Resolution pH, [SDS] ORM
Resolution [SDS], [isopropanol], [-cyclodextrin] Full factorial design
Resolution pH, [SDS] Full factorial design

a
AMPSO"3-[(1,2-dimethyl-2-hydroxyethyl)amino]-2-hydroxypropanesulfonicacid; b HPMC"hydroxypropyl methylcellulose; ccentral
composite design; d response surface modelling; ecomputer-assisted bivariate resolution optimization II; f function of mutual information;
g
partial least squares; hcomputer-assisted multivate optimization strategies; ioverlapping resolution mapping.
V, voltage; T, temperature.

Further Reading micelles, chiral selectors and non-aqueous media.

Camilleri P (1998) Capillary Electrophoresis, Theory and
Practice, 2nd edn, pp. 135d182. New York: CRC Press.
Guzman NA (1993) Capillary Electrophoresis Technology,
pp. 65d87, 693d704. New York: Marcel Dekker.
Khaledi MG (1998) High-Performance Capillary Elec-
trophoresis, pp. 77d140. New York: John Wiley.
Poole CF and Poole SK (1997) Interphase model for
retention and selectivity in micellar electrokinetic
chromatography. Journal of Chromatography A 792:
89d104.
Riekkola M-L, Wiedmer SK, ValkoH IE and SireH n H (1997)
Selectivity in capillary electrophoresis in the presence of
Journal of Chromatography A 792: 13d35.
Rosen MJ (1989) Surfactants and Interfacial Phenomena,
2nd edn, pp. 108d206. New York: John Wiley.
Terabe S, Otsuka K, Ichikawa K, Tsuchiya A and Ando
T (1984) Electrokinetic separations with micellar solu-
tions and open-tubular capillaries. Analytical Chemistry
56: 111d113.
Terabe S (1989) Electrokinetic chromatography: an inter-
face between electrophoresis and chromatography.
Trends in Analytical Chemistry 8: 129d134.
Vindevogel J and Sandra P (1992) Introduction to Micellar
Electrokinetic Chromatography, pp. 1d231. Heidelberg:
HuK thig Buch Verlag.

Microtechnology
T. McCreedy, University of Hull, Hull, UK
Copyright ^ 2000 Academic Press
Introduction
Electrophoresis is an established separation tech-
nique, frequently used for mixtures ranging from pro-
teins and DNA to small anions and cations. However,
perhaps its greatest strength lies in its remarkable
ability to separate charged macromolecules. Reports
describing electrophoretic separations started to ap-
pear in the 1930s, but the most signiRcant develop-
ments really took place in the 1940s and 50s when
separations with a paper or gel support matrix were
used for the separation of macromolecules. The early

methods used relatively large scale apparatus, but
during the later 1960s, and early 1970s, reports ap-
peared describing separations being performed in
small bore tubes Rlled with buffer solution. This work
was extended in the early 1980s, with capillaries
being a key feature of the basic methodology. This
was the start of capillary electrophoresis (CE); how-
ever, it was not until the mid 1980s that great interest
was shown towards a new approach to separation
science. From that moment, development and com-
mercialization came very quickly and soon there were
a number of commercial instruments available for
routine laboratory use.
It is not possible to cover all aspects of electrophor-
esis in an article such as this; indeed there are several
topics that have been omitted. Fluid logic devices and
freeze}melt switching are two such examples; an-
other important area not included is the use of paral-
lel bundles of microcapillaries that permit multiple
analyses to be performed at a high throughput.
The basic element of any CE system is the separ-
ation capillary, typically 10}100
m internal dia-
meter and 30}100 cm long. Each end of the capillary
is located in a small reservoir, which contains buffer
solution and a platinum anode or cathode; typically
potentials of up to 30 kV can be applied between
them. Detection is achieved by a range of in-line
detection methods, such as ultraviolet absorbance
and other detection methods, such as mass spectro-
metry, can be interfaced to the capillary.
Separation is achieved due to the differing elec-
trophoretic mobilities of the analytes in the sample,
but in addition electroosmotic Sow (EOF) takes
place. This phenomenon gives rise to bulk Sow of the
solution in the capillary without the need for an
external pump. For a unmodiRed silica capillary, the
direction of Sow would be from the anode to the
cathode, which enables all uncharged species to be
carried to the detector. This technique offers very
high separation efRciencies and rapid analysis. This
feature, coupled with the simplicity of the instrumen-
tation, makes the technique ideally suited to minia-
turization.
Interest in miniaturizing analytical systems in not
new; indeed, the idea of a micro total analysis system
(often referred to as
TAS) has been mooted for some
time within the scientiRc community (see, for
example, the paper by Martin cited in Further Read-
ing). The ideal approach is to include sample manip-
ulation and detection on a chip-sized device; this has
given rise to the term ‘lab on a chip’. Such systems
employ microstructures fabricated on glass or other
substrates to form integrated devices rather than at-
tempting to construct miniaturized systems from
discrete components. However, there is also consider-
II / ELECTROPHORESIS / Microtechnology 1287
able interest in the development of discrete
components, such as micropumps. The conference
proceedings from the recent Micro Total Analysis
Systems ’98 give some indication as to the diversity of
the developments. While on-chip injection is feasible,
some prior degree of preparation may still be neces-
sary. For example, particulate matter would quickly
block the channels, so pre-Rltering would be required
in such situations. Before examining in more detail
electrophoresis on chips, it is important to consider
the fabrication of such microchannel devices.
Fabrication of Electrophoresis
Devices
There are numerous fabrication methods available,
and the complexity of possible designs is virtually
limitless. A popular fabrication technique is the use of
photolithographic masking in conjunction with wet
or chemical etching. The simplest case would be the
fabrication of a single channel in a piece of glass.
First, the glass would be coated with a layer of depos-
ited metal and subsequently photoresist, e.g. by spin
coating, then the pattern mask is placed on top of the
photoresist. This masked surface is subsequently ex-
posed to ultraviolet light, which transfers the pattern
on to the photoresist. The unprotected area can then
be removed, along with the underlying metal surface.
An etching solution, such as hydroSuoric acid/nitric
acid, is used to etch away the glass, forming the chan-
nel in the chip. This surface of the chip protected by the
metal and photoresist layer does not etch. This pro-
cess can be seen in Figure 1. There are a few problems
with this approach; the Rrst is that only certain mater-
ials can be etched. The second is that, as the channel is
etched deeper, the width also increases. This becomes
more of a problem as the depth increases, resulting in
channel with nonvertical sides. This problem can also
create difRculties at channel intersections, which do
not have true intersecting corners due to the acceler-
ated etching of the exposed corners.
There are many alternatives to the wet etch ap-
proach. Dry etch processes include reactive ion and
laser etching; these offer a way to cut precise channels
of small dimensions. Silicon is gaining in popularity
as an alternative substrate to glass for chip fabrica-
tion, and by employing more than one etching tech-
nique complex devices can be produced.
In order to use polymeric materials, such as silicone
rubber, or Suoropolymers for chip fabrication, new
approaches are required. This may take the form of
stamping, imprinting or injection moulding of the
polymeric material. The approach offers a signiRcant
alternative to wet etching of channels directly,
1288 II / ELECTROPHORESIS / Microtechnology
Figure 1 The fabrication process for a separation chip fab-
ricated from silica. The first step is to place the mask on top of the
silica base plate covered in deposited metal and photoresist (step
1). After this has been exposed to UV light, the chip is developed
to remove the exposed photoresist and metal. It is then etched,
e.g. with hydrofluoric/nitric acids (step 2); etching does not occur
where the metal and photoresist remains. The final stage (step 3)
is the bonding of the cover plate on to the base plate. The two
etched channels can clearly be seen.
primarily since it allows the use of a wide range of
new materials, and the prospect of mass production.
It requires a template to be constructed, often by wet
etching or mechanical milling. This template can be
considered as the negative image of the channels, and
is often Rnally produced in a more durable material,
such as nickel. From this robust template, it is pos-
sible to mass-produce thousands of channel systems
with considerable speed.
All of these methods create half the chip; the next
step is to attach the cover plate, i.e. the other half of
the chip. It is common to locate the holes for the
necessary reservoirs in this plate; the reservoirs them-
selves are frequently constructed by attaching cylin-
ders, e.g. truncated pipette tips, to the top plate. For
glass and silica-based systems, it is a simple step to
bond the top plate on to the channels by a heating and
cooling cycle (the cooling cycle is required to avoid
thermally stressing the glass). The Rxing of the top
plate to polymeric materials can be more complex;
however, perhaps the simplest method is to use a ther-
mally activated adhesive to laminate the top plate on
to the chip. Typical channel dimensions are 200
m
wide by 60
m deep, and vary in length from 5 mm to
several centimetres. Of course, many other channel
dimensions can be created. Some typical patterns can
be seen in Figure 2.
Theoretical Considerations
There are two important effects that need to be con-
sidered when discussing electrophoresis in micro-
channels; these are similar to the more conventional
capillary electrophoresis. The Rrst is electrophoretic
mobility, and the second is electroosmotic Sow
(EOF). EOF is otherwise referred to as electroendos-
motic Sow.
Electrophoretic Mobility
This process forms the basis for the separation in the
channel, and dictates the migration velocity of a given
ion in the channel. The electrophoretic mobility (e) is
related to the migration velocity (v) by eqn [1], where
E is the electric Reld strength:
v"eE [1]
The units of e, v and E are cm2 V\1 s\1, cm s\1,
and V cm\1 respectively. The electrophoretic mobil-
ity is proportional to the ionic charge and frictional
forces. Thus, if two mobile species differ in either
their charge or the frictional forces, then separation
will occur. Since uncharged molecules have an electro-
phoretic mobility of zero, movement will not occur;
this is why electrophoresis cannot separate neutral
molecules. For ions of the same size, e will be greater
for ions with greater charge while for ions of the same
charge, e will be greater for smaller ions.
Electroosmotic Flow
This is a process which gives rise to the Sow of buffer
through the channel. It can be quite signiRcant,

Figure 2 Some typical channel arrangements. Reservoirs A and B start and terminate the separation channel. Reservoirs C and
D permit a known amount of sample to be injected into the separation channel. The reservoirs E and F permit the addition of other
reagents to the separation channel.
reaching linear velocities of around 5 cm min\1 or
greater. The rate of movement due to EOF is normal-
ly greater than the electrophoretic mobility, thus en-
suring that all ionic (and uncharged) species pass the
detector. However, unlike electrophoretic mobility,
EOF will only occur in the presence of an electrical
double layer at the surface of the channel. In Figure 3,
an axial view of a channel etched in glass can be seen;
the surface is covered in silanol groups.
When the pH of the buffer is above &pH 9, all the
silanol groups are ionized. Cations from the buffer
migrate towards the negative wall of the channel, and
a double layer is formed. When a voltage is applied
across the channel, these cations migrate towards the
cathode, thereby inducing bulk Sow. Electro-driven
Sow has a characteristically Sat proRle compared to
the parabolic proRle observed for pressure-driven sys-
tems. This signiRcantly reduces the dispersion due to
Sow, and is considered to be a reason for the high
efRciency separations possible. Another reason for
the low dispersion observed is that the Reynolds num-
bers for liquids in such a system are very low, which
results in limited dispersion. The electroosmotic mo-
II / ELECTROPHORESIS / Microtechnology 1289
bility (
EOF) is given in eqn [2] where  is the viscosity
of the buffer,  is the dielectric constant of the buffer
and  is the zeta potential (charge on the capillary
wall):
EOF"(/) [2]
The EOF velocity can be calculated from eqn [3]
which has striking similarities to eqn [1]. Here, the
EOF velocity (v) is related to the electroosmotic mo-
bility (EOF), and the electric Reld gradient (E):
v"EOFE [3]
From this, it is apparent that the overall velocity of
the ionic species is the algebraic sum of the migration
velocity, and the EOF velocity. By summing the two
velocity terms and subsequent rearrangement of the
equation, the actual velocity (va) of an ionic species is
given by eqn [4]:
va"(E#EOF)E [4]
Situations do arise, such as during the analysis of
anions with high electrophoretic mobility, when the
1290 II / ELECTROPHORESIS / Microtechnology
Figure 3 The double layer formed in silica channel. The layers of cations which collect along the walls of the channel will migrate
towards the cathode when a voltage is applied. This gives rise to the electroosmotic flow (EOF) with the characteristic flat flow profile.
direction of EOF needs to be reversed. This can be
achieved by coating the walls of the channels with
a cationic surfactant. This gives an apparently posit-
ive charge to the walls, so that anions (not cations)
will form the double layer. Then, when the potential
is applied, EOF will be in the opposite direction. Since
the inSuence of the double layer is generally con-
sidered to extend less than 1
m into the solution,
overlap of the double layer should not be an issue for
channels of greater than 5
m minimum dimension.
However, for channels of smaller dimension, the Sat
Sow proRle model may no longer be valid, and great
care should be exercised in describing the Sow.
To prevent EOF completely, the walls of the chan-
nel need to be rendered neutral. In silica channels, this
ought to be achievable by coating the walls with a
compound such as trimethylchlorosilane, to end-cap
all terminal silanol groups. However, in practice, it is
impossible to eliminate all EOF since residual surface
charge remains. Since many microsystems are now
being constructed from polymeric substrates, EOF
normally does not occur to any appreciable extent.
This is due to the absence of ionizable or charged
surface groups. In this situation, EOF could be in-
duced by coating the walls of the channel with
a charged compound, such as cetyltrimethylam-
monium bromide.
Practical Considerations
Perhaps the key practical consideration is whether
integrated on-chip detection will be employed, or
whether the separated compounds will be transferred
to another device, such as a mass spectrometer. In
a similar context to conventional capillary electro-
phoresis separations, on-chip detection is the ideal
option, since it minimizes dispersion and the dead
volume associated with the transfer of analytes from
the chip to a detector. The dead volume will normally
be far in excess of the separation volume, thus band
broadening will be a serious problem.
The other key issue is sample introduction. The
simplest system relies on the EOF to introduce the
sample into the separation capillary. Consider the
channel arrangement in Figure 4. The channels are
etched into silica, and no deactivating treatment is
applied. Under normal conditions (I), the applied
voltage between reservoirs A and B induces EOF. In
addition, the potential Reld gradient will give rise to
electrophoretic separations.
Since only buffer is Sowing, this does not give rise
to any apparent separation effect. When the voltage is
manipulated such that it is now between reservoirs
C and D (II), EOF is induced between the reservoirs,
thus the sample is introduced, and occupies a small
section of the main channel. Once the voltage is
restored between A and B, the separation step begins
(III). Here, the sample is moved by the EOF towards
reservior D, and separation occurs due to elec-
trophoretic mobility.
In situations where EOF is insigniRcant due to the
absence of surface charge, the injection step relies
either on the electrophoretic movement of the
analytes or an applied pressure. There is, of course,
a potential problem with electrophoretic mobility,
and that is the discriminatory effects observed
between analytes of high and low electrophoretic
mobility. Pressure, on the other hand, offers a simple

Figure 4 Sample introduction into the separation channel.
(I) When the voltage is applied between reservoirs A and B, the
separation channel is filled with running buffer. (II) To inject
a sample, the voltage is applied between C and D: the sample
moves into a short section of the separation channel. (III) With
the voltage restored across A and B, the sample moves along the
separation capillary, and separation occurs.
and nondiscriminatory route for sample introduction.
This can be achieved by either applying pressure to
one reservoir in order to force the analyte through the
system, or by deformation of the chip (in situations
where the polymer is Sexible). In either case, a valve-
less injection method is used; this greatly simpliRes
the operational aspect of these systems.
Applications
In this section, several types of application will be
considered. While much of the discussion will be
related to the separation of compounds on chips using
electrophoresis, it is impossible to neglect the poten-
tial of EOF alone for Suid mobility, which is unaffec-
ted by back-pressure.
Electrophoretic Separations
Much of the literature available on chip-based elec-
trophoretic separations features capillary zone elec-
II / ELECTROPHORESIS / Microtechnology 1291
trophoretic (CZE) separations; however, there are
many other types of separation possible, such as iso-
tachophoresis and electrokinetic focusing.
Perhaps the simplest applications are based on
CZE within silica microchannels. Here EOF and elec-
trophoretic mobility can be utilized, the EOF for
injection and bulk Sow of solutions through the capil-
lary, and electrophoretic mobility for the actual sep-
aration process. A typical separation capillary would
be 50 mm long, 45
m wide and 8
m in depth, with
an applied potential in the range 600}1200 V along
the 50 mm length. The types of samples that can be
separated by this technique are extensive (not surpris-
ing, given the diversity of the applications for conven-
tional CZE) but include small anions and cations,
monoclonal antibodies, theophylline and DNA frag-
ments. There are a number of potential detectors, but
those based on optical or electrochemical methods
are the most frequently used.
Electrochemical detection can easily be incorpor-
ated on to a microchip, but requires the detector to be
located after the high voltage section of the channel.
This is necessary to prevent the high voltage causing
interference with the detection. This can be achieved
in such a system as described above, by locating the
electrochemical detector in the channel just after the
ground electrode. The EOF occurring in the channel
would pump the Suid along the channel from the
ground electrode to the detector electrodes. Over this
short region, band broadening should not pose a sig-
niRcant problem. It is similar in principle to the por-
ous junction technique widely used in conventional
CE. It is possible to achieve limits of detection of
micromolar levels or better with electrochemical
detection.
Spectroscopic methods fall into two main classes
} absorbance and Suorescence. Absorbance measure-
ments are simple to effect, but commonly suffer from
relatively low sensitivity. This is primarily due to the
channel dimensions resulting in a very small path
length. Measurements across a 50
m channel would
give rise to a very small absorbance, since absorbance
is proportional to path length. It is possible to in-
crease the path length (Figure 5), but absorbance
measurements do not have the sensitivity of Suores-
cence measurements, although they are generally ap-
plicable to a wider range of analytes. In addition, for
practical reasons, dual-channel systems are not easily
set up and this can lead to instability in the detector
signal.
Fluorescence measurements can provide limits of
detection in the picomolar range (varying from
around 2 pmol L\1 upwards), and have even been
reported for counting single chromophore molecules.
Generally, the excitation source is directed along the
1292 II / ELECTROPHORESIS / Microtechnology
Figure 5 UV detection can be made more sensitive by increas-
ing the path length of the measurement. When the absorbance of
the analytes is measured across the channel at point x, the path
length is equal to the channel width (typically 50
m). By making
the measurement at point y, the path length is equal to the
channel length (typically 3}5 mm).
channel, in order to minimize the scatter from the
walls of the channels. By careful alignment, it is
possible to minimize the background, and obtain very
sensitive measurements. There are many other detec-
tion methods, including optical waveguide sensors
and chemiluminescence, but Suorescence detection
currently offers the most sensitive analysis.
Similar results can be obtained with channels pro-
duced from polymeric support materials. There is one
issue that must be addressed with certain materials,
e.g. plastics; that is, the background Suorescence that
is frequently observed. This can be due to the actual
substrate, or the adhesive used to seal the chip. Care-
ful selection of materials helps to reduce the problem.
However, it is the prospect of the mass production of
thousands of chips with hundreds of channels per
chip from just one master template that is particularly
attractive. Once mass production is achieved, the
devices will become truly disposable.
Electrokinetic Induction of Flow
The factor often overlooked with microsystems is the
value of EOF for Suidic manipulations. EOF will
occur at potentials of 100}200 V cm\1, and can be
used to drive Suids through channels, and indeed
physical objects such as cells, e.g. Escherichia coli. It
is possible to transport whole cells around the chan-
nels on a microchip. In addition to the EOF, there will
also be electrophoretic separations occurring, but in
practice, these are small compared to the EOF on
uncoated silica surfaces. To be of practical use, it is
necessary to have the ability to make meaningful
measurements on the contents of the cells. This can be
most easily achieved by lysis of the cells with deter-
gent. It would then be possible to measure com-
pounds, which would otherwise have been trapped
within the cell wall. Since the volume of the channels
is small, the released compounds will not be
extensively diluted, and the time to analysis
will be very short; this is particularly important if the
aim is to study rates of reaction or unstable com-
pounds.
EOF serves to deliver the sample beyond the
high voltage area if it is intended to use off-chip
detection. For example, to transfer the separated
compounds from a separation chip to a mass spec-
trometer, EOF can be used to deliver the compounds
to an electrospray interface. Indeed, it is possible to
generate the electrospray between the terminal end of
the capillary and a suitably located conductor, with-
out the need to apply a conductive coating to the end
of the chip.
The Future
Electrically driven separations on microchips have
a number of advantages over conventional CE. The
Rrst is the further reduction in reagent consumption,
in terms of both sample and buffer solutions. This
will reduce the running costs of the system, and also
the associated waste disposal costs. Second, as
methods to mass-produce the devices become more
widespread, the cost will decrease. This will allow
totally disposable systems to be used. Finally, perhaps
the most important advantage will be the portability
of analyser systems, which will be able to be used in
remote on-site mode. Without doubt, microtechnol-
ogy revolutionized the electronics industry, and it will
do the same for much chemical analysis.
There remain two other areas where chips or micro
reactors will Rnd important uses, and that is in chem-
ical discovery and manufacturing. Combinatorial
chemistry can produce thousands of compounds per
day, which places a signiRcant demand on the analy-
sis. The fabrication of multiple analysis channels with
high spatial resolution may offer an analytical solu-
tion for this problem; however, appropriate detection
must also be available. Mass spectrometry interfaced
 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis
to an electrophoretic separation chip is one possible
answer. However, there is no reason why reactions
cannot be carried out on such devices. Since it does
not require the investment of a large chemical plant,
the reactions can be performed where required, thus
reducing the need to transport hazardous chemicals
across countries. Since many reactors can be con-
structed on a single chip, and many chips located in
the same area, it is evident that this technology will
provide hazardous or chemically unstable chemicals
where they are required.
Further Reading
Altria KD (ed.) (1996) Capillary Electrophoresis Guide-
book, Principles, Operation, and Applications. New Jer-
sey: Humana Press.
Harrison DJ and Van den Berg A (eds) (1998) Micro Total
Analysis Systems ’98. Dordrecht: Kluwer Academic
Publishers.
Nonaqueous Capillary Electrophoresis
S. H. Hansen, I. Bj[rnsdottir and J. Tj[rnelund,
Royal Danish School of Pharmacy, Copenhagen,
Denmark
Copyright ^ 2000 Academic Press
Electrophoresis is a separation technique that is nor-
mally performed in an aqueous environment. This is
due to the fact that the separation mechanism is based
on the difference in migration rate of charged species
in an electric Reld. Species (ions/molecules or par-
ticles) with a difference in their charge over size ratio
will exhibit a difference in migration rate. Most
charged species are fairly soluble in aqueous media
and thus water is the most obvious solvent for elec-
trophoresis. However, in a number of nonaqueous
solvent systems, it is possible to obtain sufRcient
conductivity to perform electrophoresis. If such sys-
tems are utilized with the technique of capillary elec-
trophoresis, a number of advantages compared to
aqueous systems are obtained in the separation of
small molecules. Nonaqueous electrophoresis of bi-
opolymers like polysaccharides, nucleic acids and
proteins is not of practical use due to lack of solubility
of such molecules in organic solvents.
Nonaqueous Capillary Electrophoresis
Only a few attempts to perform nonaqueous paper
electrophoresis have been described and these articles
1293
Haswell SJ (1997) Developments and operating character-
istics of microSow injection analysis systems based on
electroosmotic Sow. Analyst 122: 1Rd1OR.
Manz A and Becker H (1997) Microsystem Technology in
Chemical and Life Sciences. Berlin: Springer.
Madou M (1996) Fundamentals of Microfabrication. Boca
Raton: CRC.
Martin AJP (1962) Opening lecture. In: Van Swaay M (ed.)
Fourth International Symposium on Gas Chromatogra-
phy. London: Butterworths.
Oefner PJ, Bonn GK and Chiesa C (1995) Encyclopaedia of
Analytical Chemistry, pp. 1041}1152. London: Aca-
demic Press.
Pethig R and Markx GH (1997) Applications of dielec-
trophoresis in biotechnology, Trends in Biochemistry
15: 426}432.
Regnier F (1999) The evolution of analysis in life science
research and molecular medicine: the potential role for
separations. Chromatographia 49: S56dS64.
Tsuda T (ed.) (1995) Electric Field Applications, pp. 47}73.
Weinheim: VCH.
were reviewed in 1978. In 1984 nonaqueous capillary
electrophoresis (NACE) was brieSy mentioned in
a single publication, but not utilized further. How-
ever, since 1993 the use of nonaqueous media for
capillary electrophoresis has seen renewed interest in
the separation of drug substances due to the high
separation selectivity obtained in these systems.
The electrophoretic migration of the solutes is in-
Suenced by the nature of the solvent or solvent mix-
ture used for the electrophoresis medium in three
main ways:
1. The mobility may change due to changes in the
size of the solvated ion.
2. The dielectric constant of the organic solvent may
inSuence the equilibrium of the protolytic dis-
sociation. The higher the value of the dielectric
constant, the higher the degree of ionization of
acids and bases.
3. The acid}base property of the solute, expressed by
its pKa value, may change due to the differenti-
ating effect of many organic solvents.
The latter effect of the three is the most signiRcant,
as the dissociation constant, Ka, may change many
orders of magnitude for different solvents.
The increased selectivity of separation in organic
solvents compared to aqueous systems is due to the
fact that the levelling effect of water is eliminated. If
1294 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis

Table 1 Classification of organic solvents according to their Br+nsted acid}base behaviour

Solvent designation Relative acidity Relative basicity Examples

Neutral # # MeOH, glycerol, phenol, tert, butyl alcohol

Amphiprotic Protogenic # ! Sulfonic acid, formic acid, acetic acid
Protophilic ! # Liquid ammonia, FA, NMF
Dipolar protophilic ! # DMSO, DMF, tetrahydrofurane, 1,4-dioxan,
pyridine
Aprotic Dipolar protophobic ! ! MeCN, acetone, nitrobenzene, sulfolane, PC
Inert ! ! Aliphatic hydrocarbons, benzene,
1,2-dichlorethane, tetrachloromethane

! indicates weaker and # indicates stronger acid or base than water. DMF, N,N-dimethylformamide; DMSO, dimethyl sulfoxide: FA,
formamide; MeCN, acetonitrile; MeOH, methanol; NMF, N-methylformamide; PC, propylene carbonate. Solvents in italic are the ones
that are preferred for NACE. Reproduced with permission from TjCrnelund J and Hansen SH (1999) Journal of Biochemistry and
Biophysical Methods 38: 139}153.

strong acids or bases are dissolved in water, they all
show up with about the same acid or base strength. If
the same acids or bases are dissolved in organic sol-
vents they will exhibit very different protolytic behav-
iour depending on the degree of dissociation, which
again depends on the solvent in question.
Important factors inSuencing the choice of organic
solvent or solvent mixture for a given separation are
volatility, the dissolving power towards suitable elec-
trolytes, viscosity and dielectric constant, UV trans-
parency and, last but not least, the effect on the
separation selectivity of the system. Information on
the viscosity and volatility, the auto protolysis con-
stant, the dielectric constant at standard conditions
and the UV transparency of the neat solvents may be
found in the literature. In contrast, data on solvent
mixtures and systematic studies of how to choose
solvents and electrolytes in order to control the
selectivity of the electrophoretic system are limited
and thus the choice of separation media is still a mat-
ter of trial and error. Solvents may be classiRed ac-
cording to their Br+nsted acid}base behaviour; a sim-
pliRed version of this classiRcation is shown in
Table 1.
Practical Considerations
Choice of Organic Solvent
The physical chemical properties of the organic sol-
vents preferred for NACE are given in Table 2 and, as
mentioned above, the physical constants have a major
impact on the choice of solvent or solvent mixture for
a given electrophoretic separation. Some of the more
practical considerations are the chemical resistance of
parts in the CE instrument towards the solvent, the
volatility of the solvent, the solvating power of the
solvent towards electrolytes, the UV transparency
and the viscosity of the solvent.

Table 2 Physicochemical parameters of selected solvents

Solvent Viscosity,
(cP) Dielectric constant,  /
pKauto Tboil (3C) UV cutoff
(nm)
(1 cm cuvette)

Water 0.89 78.4 89.9 14 100 (200

FA 3.3 111 33.6 16.8 210 275
NMF 1.65 182 110.3 10.7 182 275
DMF 0.8 36.7 45.9 29.4 153 260
DMSO 1.99 46.7 23.4 33.3 189 260
MeOH 0.544 32.7 60.6 17.2 65 205
PC 2.5 64.4 25.7 Not detected 242 200}230
Protolysis
MeCN 0.34 37.5 110.3 Not detected 82 200}230
Protolysis
Glycerol 945 42.5 0.045 * 290 205
Acetic acid 1.0430 6.152 5.91 14.45 118 *

All values are at 253C unless otherwise stated in subscript. For abbreviations, see Table 1. Reproduced with permission from TjCrnelund
J and Hansen SH (1999) Journal of Biochemistry and Biophysical Methods 38: 139}153.
 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis 1295

Figure 1 Electropherograms of imipramine and four derivatives. (A) 50 mmol L\1 6-aminocaproic acid pH 4.0; (B) 50 mmol L\1
6-amino caproic acid pH 4.0 with 25 mmol L\1 of 3-(N,N-dimethylmyristylammonium) propanesulfonate and 15 mmol L\1 of Tween威 20
added. Apparatus: Quanta 4000. Conditions: 64 cm (56 cm to the detector);75
m i.d. capillary, hydrostatic (10 cm) injection for 15 s,
ambient (27}303C), 20 kV (62
A) and UV detection at 214 nm. (C) 25 mmol L\1 ammonium acetate and 1 mol L\1 acetic acid in
acetonitrile. Apparatus: HP3DCE instrument. Conditions: 64 cm (55.5 cm to the detector);50
m i.d. capillary, injection of 3 s at 5 kPa
(50 mbar), 253C, 25 kV (7
A) and UV detection at 214 nm. Adapted with permission from BjCrnsdottir I, TjCrnelund J and Hansen
SH (1996) Selectivity enhancement in capillary electrophoresis using nonaqueous media. Journal of Capillary Electrophoresis 3:
83}87.

Solvents with a high vapour pressure and thus
a high volatility (e.g. methanol (MeOH) and aceto-
nitrile (MeCN)) may be inconvenient for automated
analysis in some instruments due to problems with
evaporation of the electrophoresis medium from the
run buffer vials as well from the sample vials. In CE
the detection is often performed by measuring the UV
absorbance of the analyte at a relatively short
wavelength (e.g. at 214 nm or below) in order to
increase the sensitivity. However, many organic sol-
1296 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis

Figure 2 Electropherograms of five basic drugs with equal or very similar mass over charge ratio. (A) 50 mmol L\1 6-aminocaproic
acid pH 4.0; (B) 50 mmol L\1 6-amino caproic acid pH 4.0 with 25 mmol L\1 of Tween威 20 added. Apparatus and conditions as in
Figure 1A. (C) 25 mmol L\1 ammonium acetate and 100 mmol L\1 sodium acetate in methanol#acetonitrile (1 : 1 v/v) and 25 kV
(23
A). Apparatus and other conditions as in Figure 1C. Adapted with permission from BjCrnsdottir I, TjCrnelund J and Hansen SH
(1996) Selectivity enhancement in capillary electrophoresis using nonaqueous media. Journal of Capillary Electrophoresis 3: 83I87.

vents have a UV cutoff at 214 nm or above (Table 2).
Nevertheless, solvents like MeCN and MeOH may be
used for measurements performed at a wavelength as
low as 200 nm as the light path through the capillary
is very short compared to the 1 cm cuvette used for
the determination of the UV cutoff wavelength. The
amides and dimethylsulfoxide can only be used when
detection at wavelengths above c. 245 nm is sufRcient
of the application.
On the positive side, organic solvents often inten-
sify the Suorescence relative to what is observed for
given solutes in aqueous media. This has been used to
decrease detection limits in NACE for analysis of
tetracyclines in biological matrices.
Choice of Electrolyte
The choice of electrolyte is important and will inSu-
ence the separation. However, due to the low solubil-
ity of many electrolytes in organic solvents, it can be
difRcult to Rnd a suitable electrolyte. The more polar
solvents, like MeOH, DMSO, formamide, N-methyl-
formamide and N,N-dimethylformamide, possess
 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis 1297

a good solvating power towards the electrolytes com-

monly used in NACE. So far, ammonium acetate has
been the most frequently used electrolyte in NACE
systems and acetic acid or sodium acetate have often
been used in combination with ammonium acetate
in order to adjust the acid}base properties of the
electrophoresis medium. Quaternary ammonium
salts have also been used a number of times with
success, e.g. in the separation of phenols and car-
boxylic acids. More rarely, Tris, magnesium acetate,
citric acid, formic acid, triSuoroacetic acid and
methanesulfonic acid have been used.
When coupling CE to mass spectrometry (MS), it is
an advantage to choose a volatile electrolyte, e.g.
ammonium acetate, in order to limit background
noise or cluster ion formation.
Other Additives
A number of polyalcohols and surfactants such as
the Tweens威 have been used as additives. Their
primary function is to decrease the electroosmotic
Sow (EOF) and thus prolong the time for elec-
trophoretic separation.
Also chiral separations are possible in NACE
using either cyclodextrines or chiral counter ions as
additives.
Reversal of EOF
The separation of anionic solutes in CE may lead to
extended time of analysis due to their migration in the
direction opposite to EOF. One method of decreasing
the analysis time is to reverse the EOF, thus making
the anions migrate in the same direction as the EOF.
In aqueous CE, the addition of long alkyl chain
trimethylammonium ions is used for this purpose, e.g.
in the analysis of inorganic anions and phenols. This
principle may also be used in NACE. However, the
long alkyl chain trimethylammonium ions are not
able to form hemimicelles at the inner capillary sur-
face when using nonaqueous solvents and thus the
EOF is not reversed. Addition of the polycation hexa-
dimethrine bromide to the nonaqueous electrophor-
esis medium may result in suitable and stable systems
with reversed EOF, even when used at fairly low Figure 3 Electropherograms of cis-trans- and diastereo-isomers. (A)
concentrations (0.001}0.05%). Separation of cis- and trans-flupenthixol decanoate using 50 mmol L\1
ammonium acetate and 1 mol L\1 acetic acid in methanol#acetonitrile
(1 : 1, v/v), above: cis-flupenthixol decanoate with 0.5% trans-isomer
added; below: trans-flupenthixol decanoate. Conditions: 64 cm (55.5 cm
Applicability of NACE to the detector);50
m i.d. capillary, injection for 3 s at 5 kPa (50 mbar),
253C, 30 kV (9
A) and UV detection at 230 nm. Test solution:
In CE the separation of solutes is due to differences
5.0 mg mL\1 of the sample in methanol#acetonitrile (1 : 1 v/v). Peak
in the charge over size ratios and thus very similar identity: 1, cis-flupenthixol decanoate; 2, trans-flupenthixol decanoate;
substances may be difRcult to separate in aqueous U, unknown. (B) Separation of dipeptides (diastereomers); (C) separ-
CE unless special mechanisms like micellar elec- ation of quinine and quinidine (diastereomers). Conditions as in (A) with
a detection wavelength of 214 nm. Adapted with permission from Han-
trokinetic chromatography (MEKC) are involved. sen SH, BjCrnsdottir I and TjCrnelund J (1997) Separation of cationic
Of course this involves addition of one or more sur- cis-trans (Z-E) isomers and diastereomers using nonaqueous capillary
factants. electrophoresis. Journal of Chromatography A 792: 49}55.
1298 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis

Table 3 Applications of NACE in analysis of food, pharmaceuticals and biological fluids

Solvents Electrolytes Analytes

Applications within food

NMF-dioxane (1 : 1 v/v) 40 mmol L\1 Tris, 2.5 mmol L\1
anthraquinone-2-carboxylic acid
NMF 500 mmol L\1 magnesium acetate
tetrahydrate
Propylene carbonate Tetraalkylammonium ions, long chain
trimethylammonium ions
20 mmol L\1 tetradecylammonium
bromide (vitamin K1 and preservatives)
Free saturated long chain fatty acids
(n-C14-n-C26). Separation of dimeric and
trimeric acids and hydrogenated fish oil
Tetracycline (TC), oxytetracycline (OTC),
chlorotetracycline (CTC), demeclocycline,
4-epitetracycline, anhydrotetracycline,
4-epianhydrotetracycline, and
desmethyltetracycline
TC, OTC and CTC in milk and plasma
Phenanthrene, -naphthol, preservatives:
methylparaben, ethylparaben and
propylparaben, thiourea (EOF marker)
and vitamin K1

Applications within pharmaceuticals

10}100% MeOH
Mixture of MeOH and H2O
MeOH
MeOH and mixture of MeOH and MeCN
MeOH, MeCN, mixture of MeOH and
MeCN, formamide, NMF, DMF, DMA,
DMSO
MeOH : MeCN : DMF (45 : 49 : 6 v/v/v)
MeOH : MeCN (1 : 1 v/v)
Formamide, NMF or DMF
NMF, formamide and mixtures of both
NMF
MeOH
MeCN
Formamide
Ammonium acetate, acetic acid Haloperidol and synthetic putative
metabolites, pyrazoloacridine and
mifentidine
Ammonium acetate, acetic acid Haloperidol, cimetropium and mifentidine
5 mmol L\1 ammonium acetate, Haloperidol and its synthetic putative
100 mmol L\1 acetic acid metabolites, pyrazoloacridine and its
synthetic putative metabolites, mifentidine
and its synthetic putative metabolites
Ammonium acetate, tetrabutylammonium Tamoxifen and four phase I metabolites
bromide, tetrabutylammonium hydrogen
sulfate and tetrapentylammonium bromide
25 mmol L\1 ammonium acetate, Imipramine, di-desmethylimipramine,
0}1 mol L\1 acetic acid or 100 mmol L\1 desmethylimipramine, methylimipramine
sodium acetate and imipramine-N-oxide. Maprotiline,
Application: 25 mmol L\1 ammonium amitriptyline, litracene, protriptyline and
acetate, 1 mol L\1 acetic acid in MeCN nortriptyline. Application: imipramine
N-oxide and impurities
25 mmol L\1 ammonium acetate, Tetracycline and three degradation products.
10 mmol L\1 citric acid and Tetracycline, oxytetracycline, doxycycline,
118 mmol L\1 methanesulfonic acid desmethyltetracycline and
chlortetracycline
20 mmol L\1 ammonium acetate, Morphine analogues, antihistamines,
1 mol L\1 acetic acid antipsychotics and stimulants
Citric acid or acetic acid mixed with Tris. Racemic mixtures of chlorphedianol,
Chiral selectors: -CD, -CD and chlorcyclizine, ethopropazine, mianserin,
derivatized -CD. Addition of long chain nefopam, primaquine, propiomezine,
alkyl ammonium salts investigated trihexyphenidyl, trimeprazine,
trimipramine and thioridazine
25}200 mmol L\1 -CD, 10 mmol L\1 NaCl Dansylated amino acids
5}100 mmol L\1 -CD and 10 mmol L\1 Dansylated amino acids
NaCl
Ammonium acetate, acetic acid, quinine N-3,5-dinitrobenzylated amino acids,
($)-1,1-binaphthyl-2,2-diyl hydrogen
phosphate and
N-[1-(1-naphthyl)ethyl]phthalomic acid
($)-Camphorsulfonic acid potassium or Atenolol, bisoprolol, bunitrolol, metroprolol,
sodium salt, 1 mol L\1 acetic acid pindolol, propranolol, salbutamol,
0.2 mol L\1 Tween 20 ephedrine, epinephrine, cisapride and
synthetic impurities
Tetra-n-butylammonium perchlorate. Chiral 1-Naphthylethylamine, 1-phenylethylamine,
selector: (#)-18-crown-6-tetracarboxylic phenylalanine, DOPA, tryptophan,
acid norephedrine, noradrenaline and
2-amino-1,2-diphenylethanol
 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis 1299

Table 3 Continued

Solvents Electrolytes Analytes

Formamide, NMF, DMF, DMA, DMSO,
MeOH, MeCN and mixtures of MeOH
and MeCN
MeOH : MeCN (75 : 25)
NMF
Mixtures of MeOH and MeCN
Mixtures of MeOH and MeCN
MeOH
25 mmol L\1 ammonium acetate,
1 mol L\1 acetic acid
25 mmol L\1 ammonium acetate,
1 mol L\1 acetic acid
500 mmol L\1 magnesium acetate
tetrahydrate
Ammonium acetate, ammonium chloride,
acetic acid, trifluoroacetic acid, formic
acid, methane sulfonic acid
Sodium acetate
20 mmol L\1 CAPS and 0}40 mmol L\1
Brij 35
Morphine, codeine, normorphine, thebaine,
noscapine and papaverine. Application:
morphine in opium tincture
Morphine
Oxytetracycline in an ointment
Cis-trans (Z-E) isomers of chlorprothixene,
thiothixene, clopenthixol, flupenthixol,
flupenthixol decanoate, clomiphene and
diastereomers: L-Ala-L-Phe, L-Ala-D-Phe;
quinine, quinidine, cinchonine and
cinchonidine
A range of penicillins, cephalosporins and
nonsteroidal anti-inflammatory drugs
Mesoporphyrin, coporphyrin,
pentaporphyrin, hexacarboxylporphyrin,
heptacarboxylporphyrin and uroporphyrin

Applications within biological fluids

10}100% MeOH in H2O 20 mmol L\1 ammonium acetate,
1% acetic acid
NMF 500 mmol L\1 magnesium acetate
tetrahydrate
MeOH 5 mmol L\1 ammonium acetate,
100 mmol L\1 acetic acid
MeOH : MeCN (1 : 1 v/v) 50 mmol L\1 ammonium acetate,
159 mmol L\1 sodium acetate and
0.002% (w/v) hexadimethrine bromide
Pyrazoloacridine, two metabolites and
a synthetic degradation product in urine
Tetracycline (TC), oxytetracycline (OTC),
chlortetracycline (CTC), demeclocycline,
4-epitetracycline, anhydrotetracycline,
4-epianhydrotetracycline and
desmethyltetracycline. TC, OTC and CTC
in cow milk and human plasma
Mifentidine and three metabolites in rat liver
homogenate
Acetylsalicylic acid and three metabolites:
salicylic acid, salicyluric acid and gentisic
acid in plasma and urine

Reproduced with permission from BjCrnsdottir et al. (1998) Electrophoresis 19: 2179.

NACE provides high separation power of very sim-
ilar substances without using additives like surfac-
tants or cyclodextrines. In Figures 1 and 2 the separ-
ation of very similar substances using NACE are
compared to separation in an aqueous CE and
a MEKC system. As seen in Figure 2, even substances
expected to have identical mass over charge may be
separated in a short time compared to the aqueous
systems. Figure 3 shows the separation of cis-trans
isomers and diastereoisomers. These isomers are also
expected to have the same mass over charge ratio.
The use of NACE in the analysis of food, pharma-
ceuticals and biological Suids has been reviewed by
Bj+rnsdottir and co-workers and in Table 3 an over-
view of applications is given. An important practical
consequence of using a NACE separation medium is
that the organic phases resulting from either simple
extractions or from eluents from solid-phase extrac-
tions can be injected directly into the system, thereby
saving time.
Furthermore, some NA solvents seem promising
for CE-MS experiments due to the volatility of the
solvents and the relatively low current generated in
the organic solvents. The low current is comparable
to the current generated in electrospray MS interfaces
and therefore the stability of online CE-MS is
optimized.
Two questions are often raised in connection
with practical work with NACE. How important
is it that the electrophoresis medium is really
nonaqueous? This is not crucial. A content of
water up to 1% will not inSuence the separation
efRciency and selectivity signiRcantly. Is it possible
to perform quantitative analysis using NACE? Yes,
if steps against evaporation are taken when
volatile solvent are used, the reliability of the
methods is comparable to that of aqueous
systems (Table 4). A number of applications
including validated quantitative methods are given in
Table 4.
1300 II / ELECTROPHORESIS / Nonaqueous Capillary Electrophoresis

Table 4 Validation data from quantitative NACE analysis of food, pharmaceuticals and biological fluids

Analytes Linearity (r 2) Repeatability of inj. (%RSD) LOD Accuracy

Free saturated long chain 0.994 (n16-n20) Inter-day: 2.1}39% (n"6) 0.025 mmol L\1 nd
fatty acids (n-C14-n-C26) 0.985 (n22-n26) at three concentration
levels
Tetracyclines in plasma and 0.999 Inter-day repeatability of the 25 ng mL\1 of TC, 97.2% Oxytetracycline (n"6,
milk method: 3.6}10.2% (n"6) OTC or CTC %RSD"4.2%) and 63.3%
at three concentration levels (n"6, %RSD"3.6%) at
two conc. in milk
Vitamin K1, propylparaben 0.993 &3% (n"6) at three nd 98% Vitamin K1 (n"6,
and methylparaben concentration levels for all %RSD"4.95%) 96%
three analytes Propylparaben (n"6,
%RSD"4.45%) 82%
Methylparaben (n"6,
%RSD"2.26%)
Tetracycline and three 0.993}0.998 Inter-day: 3.4}13% (five nd nd
degradation products concentration levels)
Oxytetracycline in an '0.999 Inter-day: 2.8}4.4% for peak nd 96.1}97.3% Oxytetracycline at
ointment area (n"6) at three conc. three concentration levels
levels. For migration time: (n"6)
'0.8% within day (n"8)
and (3.3% in-between
days (n"6)
Morphine in pharmaceutical '0.999 Inter-day: 2.0% (n"6) 0.2
g mL\1 100.7}101.2% (three
products concentrations)
Acetylsalicylic acid and three Good in the range: Inter-day: plasma: 0.8}5.0% LOQ : 5
g mL\1 Plasma: 65}99%, urine:
metabolites: salicylic acid, 5}500
g mL\1 (n"6), urine: 1.0}5.4% in plasma and 75}97%
salicyluric acid and gentisic (n"6) at three conc. levels 25
g mL\1 in
acid in plasma and urine urine

nd, not determined; LOQ, limit of quantitation. Reproduced with permission from Bj+rnsdottir et al. (1998) Electrophoresis 19: 2179.

Concluding Remarks Further Reading

The primary advantages of using nonaqueous media
for CE may be outlined in four statements:
1. The separation selectivity is improved by using
neat organic solvents and the selectivity can
easily be altered by changing the nature of the
organic solvent or using mixtures of organic sol-
vents.
2. Analysis of hydrophobic compounds is facilitated
as their solubility is higher in organic solvents than
in aqueous media.
3. Sample preparation is facilitated as extracts ob-
tained with organic solvents may be injected dir-
ectly into the nonaqueous system (e.g. the eluate
from a solid-phase extraction cartridge, when us-
ing MeOH or MeCN as the eluent, may be used
for CE without further treatment).
4. The relatively low current generated in organic
solvents combined with the volatility of the sol-
vents seems to be promising for CE-MS experi-
ments.
Altria KD (1998) Analysis of Pharmaceuticals by Capillary
Electrophoresis, p. 223. Braunschweig/Wiesbaden: F.
Vieweg.
Bj+rnsdottir I, Tj+rnelund J and Hansen SH (1998)
Nonaqueous capillary electrophoresis } its applicability
in the analysis of food, pharmaceuticals and biological
Suids. Electrophoresis 19: 2179.
Hansen SH, Tj+rnelund J and Bj+rnsdottir I (1996) Selectiv-
ity enhancement in capillary electrophoresis using
nonaqueous media. Trends in Analytical Chemistry 4:
175.
Korchemnaya EK, Ermakov AN and Bochkova LP (1978)
Electrophoresis in nonaqueous and mixed solvents. Jour-
nal of Analytical Chemistry USSR (Engl. transl.) 33: 635.
Kenndler E (1993) Organic solvents in capillary electrophor-
esis. In: Gusman NA (ed.) Capillary Electrophoresis
Technology, pp. 161}183. New York: Marcel Dekker.
Sarmini K and Kenndler E (1997) Review, InSuence of
organic solvents on the separation selectivity in capillary
electrophoresis. Journal of Chromatography A 792: 3.
Valko IE, SireH n H and Riekkola M-L (1997) Capillary
electrophoresis in nonaqueous media: an overview. LG-
GC International 10: 190.
 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis 1301

Nuclear Magnetic Resonance Detection in

Capillary Electrophoresis
See II / ELECTROPHORESIS / Capillary Electrophoresis-Nuclear Magnetic Resonance

One-dimensional Polyacrylamide Gel Electrophoresis

P. G. Righetti, University of Verona, Verona, ZE became a reality when hydrophilic gels (acting
Italy as an anticonvective support) were discovered.
Copyright ^ 2000 Academic Press Grabar and Williams in 1953 Rrst proposed the use of
an agar matrix (currently abandoned in favour of
a highly puriRed agar fraction, agarose). They also
Electrophoresis is based on the differential migration combined, for the Rrst time, electrophoresis on a hy-
of electrically charged particles in an electric Reld. As drophilic support with biospeciRc detection (im-
such, the method is applicable only to ionic or munoelectrophoresis). Barely two years after that,
ionogenic materials, i.e. substances convertible to Smithies (1955) applied another gel, potato starch.
ionic species (a classic example being neutral sugars, The starch blocks were highly concentrated matrices
which form negatively charged complexes with bor- (12}14% solids) and subsequently introduced a new
ate). In fact, with the advent of capillary zone elec- parameter in electrophoretic separations: molecular
trophoresis (CZE) it has been found that a host of sieving. Human sera, which in cellulose acetate or
neutral substances can be induced to migrate in an paper electrophoresis, were resolved in barely Rve
electric Reld by inclusion in charged micelles, e.g. of bands, now produced a spectrum of 15 zones. The
anionic (sodium dodecyl sulfate, SDS) or cationic most important discovery, however, came with the
(cetyltrimethylammonium bromide, CTAB) surfac- introduction of polyacrylamide gels and disc elec-
tants. Even compounds that are not ionic, ionogenic, trophoresis; this discovery was thoroughly debated in
or complexable can often be analysed by CZE as they
are transported past the detector by the strong elec-
troosmotic Sow on the capillary walls.
Basically, if one plots the velocity of a zone against
the pH in the same zone, electrophoretic techniques
can be divided into four main types: zone elec-
trophoresis (ZE) together with moving-boundary
electrophoresis (MBE), discontinuous (disc) elec-
trophoresis, isotachophoresis (ITP) and isoelectric fo-
cusing (IEF). Figure 1 represents this classiRcation. It
can be seen that IEF and ITP are based on principles
that are ‘perpendicular’ to ZE and MBE. In particu-
lar, in IEF, once steady-state conditions have been
attained, all proteins reach a zero-velocity (v) thus
they remain immobile (v"0, pH-axis). It is then
clear that ITP closes the ring of possibilities: all zones
move with the same velocity, but at different pH.
Alternatively, electrophoretic techniques may be enu-
merated in chronological order, as follows: moving
boundary electrophoresis (MBE), zone electrophor-
esis (ZE), disc electrophoresis, isoelectric focusing
(IEF), sodium dodecyl sulfate/polyacrylamide gel Figure 1 Classification of the four modes of electrokinetic tech-
niques. The velocity of a zone is plotted against the pH in the
electrophoresis (SDS-PAGE), two-dimensional (2-D)
same zone. (A) zone and moving boundary electrophoresis; (B)
maps, isotachophoresis (ITP), staining techniques, discontinuous electrophoresis; (C) isoelectric focusing; (D) iso-
immobilized pH gradients (IPG), and capillary zone tachophoresis. (Reproduced with permission from Routs RJ
electrophoresis. (1971) PhD thesis, University of Eindhoven.
1302 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis

a classic electrophoretic volume, which appeared in Discontinuous Electrophoresis

December, 1964 in the Annals of the New York
Academy of Sciences (a collectors item!). This was
like the explosion of a supernova in the Rrmament of
electrokinetic methodologies. Although most of the
above-mentioned techniques belong to the category
of one-dimensional PAGE, we will only mention
three of them here in detail: disc electrophoresis, SDS
electrophoresis and pore-gradient-gel electrophoresis.
The other techniques such as IEF and ITP, being
steady-state methods, are best performed in non-
sieving media. The technique of moving-boundary
electrophoresis died out long ago.
In 1959 Raymond and Weintraub described the use
of polyacrylamide gels (PAG) in ZE, which offered
UV and visible transparency (starch gels are opal-
escent) and the ability to sieve macromolecules over
a wide range of sizes. Figure 2 gives a scheme of
reaction for producing polyacrylamide gels from the
standard mixture of monomers, acrylamide and the
cross-linker Bis. It should be noted that although this
matrix should be neutral (except where accidental
hydrolysis of acrylamide to acrylic acid occurs), in
reality it is not completely devoid of charges; at the

Figure 2 The polymerization reaction of acrylamide. The chemical formula of acrylamide, N,N
-methylenebisacrylamide (Bis) and of
the initiators (peroxysulfate and N,N,N
,N
-tetramethylethylendiamine, TEMED) are shown. On the right-hand side, growing poly-
acrylamide chains, in equilibrium with free monomers, are illustrated. In this particular case, it is assumed that the chain termini are
TEMED molecules, although peroxysulfate could be just as well incorporated.
 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis 1303

chain termini either initiator, N,N,N
,N
-tetra-
methylethylenediamine (TEMED), or sulfate, could
be incorporated which would impart positive or
negative charges, respectively. The fact that polyac-
rylamides always exhibit a residual electroosmotic
Sow towards the cathode suggest that an excess of
negative charges is incorporated over positive ones
(TEMED).
In 1964, Ornstein and Davis created discontinuous
(disc) electrophoresis by applying to PAG a series of
discontinuities (of leading and terminating ions, pH,
conductivity, and porosity), thus further increasing
the resolving power of the technique. In discontinu-
ous disc electrophoresis (the principle of which is
outlined in Figure 3), the proteins are separated on
the basis of two parameters: surface charge and mo-
lecular mass. The matrix is divided into three sections
(from bottom up): a ‘separation’, or ‘running’ gel,
a ‘spacer’ or ‘stacking’ gel, and a sample gel. A sharp
discontinuity exists at the running/stacking interface:
the bottom gel is a tightly knit sieve (with small
pores), while the second and third layers are minimal-
ly sieving, open-pore structures. At the same inter-
face, a second discontinuity exists in pH. In fact, the
running gel is titrated at pH 8.9, whereas spacer and
sample gels are buffered at pH 6.7. This gel region at
pH 6.7 is also a low conductivity region (third discon-
tinuity), which means that a voltage gradient will be
generated in this zone when an electric current is
passed through it. Below and above it (in the cathodic
chamber) high-conductivity regions are found.
A fourth discontinuity exists at the interface between
the upper gel end and the liquid in the cathodic
compartment: below it only Cl\ (leading, L) ions are
present, while above it only glycinate (trailing or
terminating, T) ions are found.
Why is there the need for such a complicated sys-
tem? This intricate set up must satisfy the Kohlrausch
regulating function, which is at the heart of ITP (in
fact, movement of ions in the Rrst two gel segments
will be according to ITP rules). If all the ions in the
system are arranged in such a way that L'P'T
(where  is the mobility of leading, protein and termi-
nating ions, respectively), then, upon playing a volt-
age gradient, they will migrate down the gel cylinder
with equal velocities and the boundary between each
adjacent species will be maintained. As soon as the
electric circuit is closed, Cl\ (fastest moving) ions are
swept down the column towards the anode. Just be-
hind this boundary, all protein ions will start arrang-
ing themselves in order of their mobilities, with the
lowest pI component next to the Cl\ boundary and

Figure 3 Principle of discontinuous disc electrophoresis. (A) Sample in sample gel; (B) sample concentration in stacking gel;
(C) sample separation in running gel. From top to bottom, the following phases are encountered: glycine buffer at pH 8.3 in the cathodic
reservoir; sample gel and spacer gel, both titrated to pH 6.7; small-pore running gel, titrated to pH 8.9; and glycinate buffer again in the
anodic reservoir at the bottom. In part (C) it is seen that, as the glycinate boundary sweeps down the gel past the protein zones, the pH
increases from 8.9 (in A and B) up to 9.5. (Reproduced with permission from Ornstein, 1964.)
1304 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis
the highest pI species closing the procession. The last
‘wagon of the train’ is the glycinate (terminating) ion;
and explains why the sample and spacer gels are
titrated at pH 6.7. Gly has a theoretical pI of 6.1 but,
as shown by its titration curve, it is almost isoelectric,
even at pH 6.7; its anionic mobility, therefore, is
extremely small, in any event smaller than the slowest
protein ion.
Thus, in the sample and spacer gels, two basic phe-
nomena occur: (1) all protein ions are sorted out and
physically separated according to their pIs, and (2)
each protein ion is strongly concentrated in extremely
thin starting zones (the disc barely a few micrometres
thick and a concentration process of up to 1000- to
10 000-fold). This isotachophoretic ‘train’, however,
does not have a long life; as it enters the running gel,
the train ‘runs off the tracks’. Only the ‘locomotive’
(Cl\) of the train is unaffected; the various protein
wagons now overrun each other, since they experience
a strong frictional force, due to the highly sieving
matrix, so that now their velocity is a function of their
charge/size ratio. In addition, as the almost isoelectric
Gly enters the pH 8.9 zone, its negative charge density
strongly increases so that it jumps ahead and closely
follows the Cl\ ion. As Gly sweeps down the running
gel, the pH increases from pH 8.9 to 9.5 (approaching
the pK value of the Gly amino group) so that now the
net charge on Gly is !0.5. As a consequence of this
further jump in pH, all proteins experience an addi-
tional mobility increment. One might wonder why,
after taking on such an experimental burden in form-
ing the ITP train, one should then destroy it and
continue the run in the plain zone electrophoretic
mode. There are reasons for this. First, the ITP train,
while maintaining high resolution due to lack of degra-
dation of zone boundaries, has the main defect that the
zones are contiguous and continuous, i.e. they are not
separated by blank zones of plain buffer. As a result,
when staining the gel, one would only see a single,
continuous zone of protein ions, with no visible separ-
ation between zones. Second, whereas the sharp pro-
tein discs formed during the stacking (ITP) process are
separated solely by surface charge, during migration in
the running gel, separations continue on the basis of an
additional parameter i.e. the mass. The small loss of
resolution due to diffusion of the protein discs in the
running gel is more than compensated for by the
resolution increments due to size (coupled to
charge) fractionation in this gel zone. Although
disc electrophoresis is no longer in vogue, it was an
extremely useful analysis technique for at least 20
years after its inception. Moreover, the general prin-
ciple has not been abandoned and it is used today as
a stacking technique in both SDS and capillary elec-
trophoresis.
Disc electrophoresis could also be used for deriving
physico-chemical parameters of the proteins under
analysis. In 1964, Ferguson showed that one can
derive parameters which are proportional to both the
surface charge and the mass of the macromolecule.
This can be accomplished by plotting the results of
a series of experiments with polyacrylamide gels of
varying porosity. For each protein under analysis, the
slope of the curve log mT (electrophoretic mobility) vs
gel density (%T) is proportional to molecular mass,
while the y-intercept (Y0) is a measure of surface
charge. Examples of these plots are shown in
Figure 4. In Figure 4A the two parallel lines indicate
charge isomers; in Figure 4B, the fanning out lines
indicate a family of constant charge and different
mass; in Figure 4C, the two crossing lines indicate
proteins differing in both charge and mass. Recently,
non-linear Ferguson plots have been reported
(Chrambach, 1988), related to the reptation mode of
DNA in sieving media.
Sodium Dodecyl Sulfate (SDS)
Electrophoresis
SDS electrophoresis fractionates polypeptide chains
essentially on the basis of their size. It is therefore
a simple, yet powerful and reliable, method for mo-
lecular mass determination. In 1967 Shapiro et al.
Rrst reported that electrophoretic migration in SDS is
proportional to the effective molecular radius and,
thus, to the Mr of the polypeptide chain. This means
that SDS must bind to proteins and cancel out differ-
ences in molecular charge, so that all components will
migrate solely according to size. Surprisingly, large
amounts of SDS appear to be bound (an average of
1.4 g SDS/g protein). This means that the number of
SDS molecules bound is of the order of half the
number of amino acid residues in a polypeptide
chain. This amount of highly charged surfactant mol-
ecules is sufRcient to overwhelm effectively the intrin-
sic charges of the polymer coil, so that their net
charge per unit mass becomes approximately con-
stant. If migration in SDS (and disulRde-reducing
agents, such as 2-mercaptoethanol, in the denaturing
step, for a proper unfolding of the proteins) is propor-
tional only to Mr, then, in addition to cancelling out
charge differences, SDS also equalizes molecular
shape differences (e.g. globular vs rod-shaped mol-
ecules). This seems to be the case for protein}SDS
mixed micelles. These complexes can be assumed to
behave as ellipsoids of constant minor axis (c.
1.8 nm) and a major axis proportional to the length
of the amino acid chain (i.e. to molecular mass) of the
protein. The rod length for the 1.4 g SDS/g protein
 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis
Figure 4 Ferguson plots (log Rm, relative mobility, vs %T, total
monomer concentration) in the case of: (A) lactic dehydrogenase
(LDH) 1 and 2 (isomers of charge, exhibiting the same mass);
(B) serum albumin (polymeric forms, from monomer to heptamer,
having constant charge and pure size difference, since all curves
meet in gel-free environment, at 2% T where polyacrylamide will
liquefy); (C) ferritin and ovalbumin, two totally unrelated proteins
differing in both size and charge. (Parts (A) and (C) reproduced
with permission from Hedrick and Smith, 1968 and Part (B) from
Thorun, 1971.)
complex is of the order of 0.074 nm per amino acid
residue. For further information on detergent proper-
ties, see Helenius and Simons (1975).
1305
In SDS electrophoresis, the proteins can be
prelabelled with dyes that covalently bind to their
}NH2 residues. The dyes can be conventional, like the
blue dye Remazol, or Suorescent, such as dansyl
chloride, Suorescamine, O-phthaldialdehyde, and
MDPF (2-methoxy-2,4-diphenyl-3[2H]-furanone).
Prelabelling is compatible with SDS electrophoresis,
as the size increase is minimal, but would be anath-
ema in disc electrophoresis or IEF, as it would gener-
ate a series of bands of slightly altered mobility or pI
from an otherwise homogeneous protein. Although at
its inception SDS electrophoresis used continuous
buffers, today the preferred set up is via discontinu-
ous buffers and matrices, simpliRed from the original
disc electrophoresis assembly (see Figure 5). This en-
sures much higher resolving power, due to formation
of ultrathin protein zones.
For treatment of data, the sample and Mr standards
are electrophoresed side-by-side in a gel slab. After
detection of the polypeptide zones, the migration
distance (or RF) is plotted against log Mr to produce
a calibration curve (Neff et al., 1981) from which the
Mr of the sample can be calculated (see Figure 6). It
should be noted that in a gel of constant %T, linearity
is obtained only in a certain range of molecular sizes.
Outside this limit a new gel matrix of appropriate
porosity should be used. Two classes of proteins show
anomalous behaviour in SDS electrophoresis: glyco-
proteins (because their hydrophilic oligosaccharide
units prevent hydrophobic binding of SDS micelles)
and strongly basic proteins, e.g. histones (because of
electrostatic binding of SDS micelles through their
sulfate groups). The Rrst anomaly can be partially
alleviated by using alkaline Tris/borate buffers,
which will increase the net negative charge on the
glycoprotein and thus produce migration rates well
correlated with molecular size. The migration of hi-
stones can be improved by using pore-gradient gels
and allowing the polypeptide chains to approach the
pore limit.
Porosity Gradient Gels
When macromolecules are electrophoresed in a con-
tinuously varying matrix concentration (which results
in a porosity gradient) rather than in a gel of uniform
concentration, the protein zones are compacted along
their track, as the band front is, at any given time, at
a gel concentration somewhat higher than that of the
rear of the band, so that the former is decelerated
continuously. A progressive band sharpening thus
results. There are other reasons for resorting to gels of
graded porosity. We have seen that disc electrophor-
esis separates macromolecules on the basis of both
size and charge differences. If the inSuence of molecu-
1306 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis
Figure 5 Typical set up of a gel slab for SDS electrophoresis in a continuous (A) or discontinuous (B) gel and buffer system. In both
cases, the sample is applied as a dense liquid layer in pockets precast in the gel slab by the teeth of the comb. Note that this mode of
sample deposition avoids the use of the third gel phase, the sample gel, as typically adopted in disc electrophoresis.
lar charge could be eliminated, then clearly the
method could be used with a suitable calibration
for measuring molecular size. This has been
Figure 6 Typical log Mr vs RF plot after an SDS-PAGE run. Note
that the plot is linear only in the Mr 15 000}60 000 Da range. The
Mr markers are: (1) myosin (194 000 Da); (2) RNA polymerase
(--subunit, 160 000 Da); (3) -galactosidase (116 000 Da); (4)
phosphorylase B (94 000 Da); (5) RNA polymerase (-subunit;
95 000 Da); (6) bovine serum albumin (68 000 Da); (7) ovalbumin
43 000 Da); (8) RNA polymerase (-subunit; 38 400 Da); (9) car-
bonic anhydrase (30 000 Da); (10) trypsinogen (24 500 Da); (11)
-lactoglobulin (17 500 Da); and (12) lysozyme (14 500 Da). (Re-
produced with permission from Neff et al., 1981.)
accomplished by overcoming charge effects in two
main ways. In one such way, a relatively large
amount of charged ligand, such as SDS, is bound
to the protein, effectively swamping the initial
charges present on the protein molecules and giving
a quasi-constant charge-to-mass ratio. However,
in SDS electrophoresis, proteins are generally
dissociated into their constituent polypeptide
subunits, and the concomitant loss of functional in-
tegrity and antigenic properties cannot be prevented.
Therefore, the size of the original, native molecule
must be evaluated in the absence of denaturing
substances.
In the second method for Mr measurement, this can
be done by relying on a mathematical cancelling of
charge effects, following measurement of the mobility
of native proteins in gels of different concentrations.
This is the so-called ‘Ferguson plot’ discussed above.
As a third method for molecular size measurements
one can use gels of graded porosity. This method is
characterized by high resolving power and relative
insensitivity to variability in experimental conditions.
See Figure 7 for a typical experimental set up for
casting porosity gradients in gel slabs. Under appro-
priate conditions (at least 10 kV;hours), the mobil-
ity of most proteins becomes constant and eventually
ceases as each constituent reaches a gel density region
in which the average pore size approaches the dia-
meter of the protein (pore limit) (Margolis and Ken-
rick, 1968). Thus, the ratio between the migration
distance of a protein to that of any other becomes
a constant after the proteins have all entered a gel
region in which they are subjected to drastic sieving
conditions. This causes the electrophoretic pattern to
become constant after prolonged migration in a gel
gradient. The gel concentration at which the migra-
 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis 1307

Figure 7 Scheme of the apparatus used for the simultaneous preparation of eight gradient gel slabs. (1) and (2) two-chamber mixer;
(3) stirrer; (4) reservoir for peroxosulfate; (5) reservoir for TEMED; (6) proportioning pump; (7) modified disposable syringe, used as
small chamber mixer; (8) magnetic bar; (9) tube connecting the stirrer to the gel casting apparatus (10); (11) gel cassettes; (12) wedge;
(13) distributor; (14) magnetic stirrer; I and II, 0.5 mm i.d. vinyl tubings; III and IV, 3.16 mm i.d. vinyl tubings. (Reproduced with
permission from Rothe and Purkhanbaba, 1982.)

tion rate for a given protein becomes constant is proteins, it is possible to correlate the migration
called the ‘pore limit’. If this porosity is properly distance to the molecular mass of any constituent in
mapped with the aid of a suitable set of marker the mixture.

Figure 8 Typical log Mr vs migration distance (D) or gel composition (%T) plots after pore-gradient electrophoresis. Note that these
plots are non-linear, whereas when log Mr is plotted against (D or (%T a linear relationship is obtained. (Reproduced with permission
from Rothe and Purkhanbaba, 1982.)
1308 II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis
After electrophoresis has Rnished, the experimental
data gathered can be handled in two ways: a two-step
or a one-step method. The most promising two-step
approach appears to be that of Lambin and Fine
(1979), who observed that there is a linear relation-
ship between the migration distance of proteins and
the square root of electrophoresis time, provided that
time is kept between 1 and 8 h. The slopes of the
regression lines of each protein in the above graph are
an indication of molecular size. When the slopes of
the various regression lines thus obtained are plotted
against the respective molecular masses, a good linear
Rt is obtained, which allows Mr measurements of
proteins of between 2;105 and 106 Da. The shape of
the proteins (globular or Rbrillar), their carbohydrate
content (up to 45%) and their free electrophoretic
mobilities (between 2.1 and 5.9;10\5 cm2 V\1 s\1)
do not seem to be critical for proper Mr measure-
ments by this procedure. One-step methods have been
described by Rothe and Purkhanbaba (1982). These
authors found that when log Mr is plotted against
either D (distance migrated) or %T (acryl-
amide#Bis), a nonlinear correlation is always ob-
tained. However, when log Mr is plotted vs (%T or
(D, a linear regression line is obtained, which
allows the accurate determination of Mr values of
proteins (standard deviation of $3.7%; see Fig-
ure 8. The correlations log Mr!(%T or log
Mr!(D are not signiRcantly altered by the dura-
tion of electrophoresis. Therefore, a constant
Mr value should be obtained for a stable protein, no
matter how long electrophoresis takes. More re-
cently, Rothe and Maurer (1986) have demonstrated
that the relationship log Mr vs (D is also applicable
to SDS electrophoresis in linear polyacrylamide gel
gradients.
Further Reading
Chrambach A (1988) Particle and gel Rber properties
derived from the mobilities in gel electrophoresis:
the dilectics of Ferguson plots. In: Schafer-Nielsen
(ed.) Electrophoresis’ 88, pp. 28}40. Weinheim:
VCH.
Davis BJ (1964) Disc electrophoresis. II: method and ap-
plication to human serum proteins. Ann. N.Y. Acad. Sci.
121: 404}427.
Ferguson KA (1964) Derivation of size and charge of pro-
teins from polyacrylamide gel electrophoresis at differ-
ent %T. Metabolism 13: 985}995.
Frederick JF (1964) Gel electrophoresis. Ann. N.Y. Acad.
Sci. 121: 307}650.
Grabar P and Williams CA (1953) Methode permettant
l’etude conjugeH e des proprietes electrophoretiques et im-
munochimiques d’un melange de proteins; application
au serum sanguine. Biochim. Biophys. Acta 10:
193}201.
Hedrick JL and Smith AJ (1968) Size and charge isomer
separation and estimation of Mr of proteins by disc gel
electrophoresis. Arch. Biochem. Biophys. 126: 155}163.
Helenius A and Simons K (1975) Solubilization of mem-
branes by detergents. Biochim. Biophys. Acta 415:
29}79.
Lambin P and Fine JM (1979) Mr estimation of proteins by
electrophoresis in linear polyacrylamide gradient gels in
the absence of denaturing agents. Anal. Biochem. 98:
160}168.
Margolis J and Kenrick KG (1968) Polyacrylamide gel
electrophoresis in a continuous molecular sieve gradient.
Anal. Biochem. 25: 347}358.
Neff JL, Muniz N, Colbourn JL and de Castro AF (1981)
Convenient procedures for SDS and conventional disc
electrophoresis. In: Allen RC and Arnauds P (eds) Elec-
trophoresis’ 81, pp. 49}63. Berlin: de Gruyter.
Ornstein L (1964) Disc electrophoresis. I: background and
theory. Ann. N.Y. Acad. Sci. 121: 321}349.
Raymond S and Weintraub L (1959) Polyacrylamide gel:
a new matrix for zone electrophoresis of proteins.
Science 130: 711}712.
Rothe GM and Maurer WD (1986) One dimensional PAA-
gel electrophoretic techniques to separate functional and
denatured proteins. In: Dunn MJ (ed.) Gel Electrophor-
esis of Proteins, pp. 37}140. Bristol: Wright.
Rothe GM and Purkhanbaba M (1982) Determination of
Mr and Stokes’ radii of non-denatured proteins by
PAGE. I: an equation relating total polymer concentra-
tion, the Mr of proteins in the range 104d106 and dura-
tion of electrophoresis. Electrophoresis 3: 33}42.
Shapiro AL, Vinuela E and Maizel JV Jr (1967) Mr estima-
tion of polypeptide chains by electrophoresis in SDS-
polyacrylamide gels. Biochem. Biophys. Res. Commun.
28: 815}820.
Smithies O (1955) Zone electrophoresis in starch gels:
group variations in the serum proteins of normal adults.
Biochem. J. 61: 629}636.
Thorun W (1971) Estimation of the size of albumin
oligomers via Ferguson plots. Z. Klin. Chem. Klin. Bio-
chem. 9: 3}13.
II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis
One-dimensional Sodium Dodecyl Sulfate Polyacrylamide
Gel Electrophoresis
G. L. Jones, University of New England, Armidale,
Australia
Copyright ^ 2000 Academic Press
Introduction
Technical and Developmental Details of
Basic Technique
Any ion will undergo electrophoresis to migrate in an
electric Reld. Proteins are complex polyions with a net
charge at all pH values other than its isoelectric point.
Problems associated with convective disturbance in
free solution led early researchers to consider various
supporting media for electrophoresis such as paper,
cellulose acetate and various thin layer materials
where the separation depends largely on the charge
density at a given pH. The reader should refer to the
separate articles on Electrophoresis Theory and Cel-
lulose Acetate Electrophoresis for further details.
Other early workers considered the properties of vari-
ous gels where the pore size approximates the size of
the protein molecules themselves leading to a separ-
ation based on both charge and molecular size. The
extent of molecular sieving depends on the pore size
of the gel being used. For example, the pore size of
agarose gels is sufRciently large that sieving of most
proteins is minimal, whereas larger DNA molecules
are sieved very well. Again, for a discussion of
this refer to the separate article on Agarose Elec-
trophoresis.
The pore size of polyacrylamide gels may be
changed in a systematic and reproducible fashion by
varying the percentage of monomer and crosslinker
to give a matrix which maximizes the molecular siev-
ing effect for a wide range of proteins and the reader
should refer to the separate article on Polyacrylamide
Gel Electrophoresis (PAGE) for a discussion on vary-
ing porosity in this medium.
Native proteins, however, often occur as multiple
supramolecular assemblies of many peptide sub-
units in different conRgurations affected by non-
covalent bonding, particularly in the case of mem-
brane proteins. Shapiro et al. (1967) Rrst demon-
strated the potential of the superior protein dissocia-
ting qualities of sodium dodecyl sulfate (SDS) in an
electrophoretic system designed to separate indi-
vidual polypeptides on the basis of their molecular
weight alone but the deRnitive publication in this area
is undoubtedly that of Laemmli (1970) who Rrst
1309
combined a discontinuous buffer system (see separate
article on Discontinuous Electrophoresis) with the
use of SDS in sample preparation and gel electrophor-
esis. Protein complexes are solubilized and disso-
ciated with such high efRciency in 2% SDS and 5%
mercaptoethanol that typically over 90% of the pro-
tein in a crude lysate will enter the gel matrix and be
resolved.
In the discontinuous system, proteins are dissolved
by denaturing treatment at 1003C with the dissocia-
ting agents SDS and mercaptoethanol in a Tris-HCl
buffer at pH 6.8. Gels are constructed in two stages
both containing 0.1% SDS. The separating gel in the
original Laemmli publication was formed using 30%
stock acrylamide monomer with 0.8% bisacrylamide
as a cross-linker. A Rnal solution was made to 8 or
10% acrylamide containing 0.375 M Tris-HCl pH
8.8. The resolving (separating) gel is polymerized
using tetramethylenediamine (TEMED) (catalyst)
and ammonium persulfate (free radical initiator).
A ‘stacking’ gel (at 3% acrylamide) is then cast on top
of the resolving gel in the same manner but contain-
ing 0.125 M Tris-HCl at the same pH as the buffer in
which the protein mixtures were dissociated (pH 6.8).
The electrode buffer contains 0.025 M Tris/0.192 M
glycine to a pH of 8.3 also with 0.1% SDS. Upon
electrophoresis, protein anions in the form of rod-
shaped SDS complexes are compressed in the stacking
gel between the leading chloride ions and the trailing
glycinate ions which, because of the pH difference
between buffer systems, progressively close the gap as
electrophoresis proceeds. (Again, see the separate
articles on Discontinuous Electrophoresis and Iso-
tachophoresis). The result is a concentration or stack-
ing of the SDS}protein anions as extremely sharp
bands (+5}10
m) behind the leading chloride ion
in strict order of mobility. These complexes then
enter the separating gel and since, supposedly, the
charge}mass ratio is invariant (see later for a caveat)
are separated by molecular sieving according to their
molecular size only. Gels of particular acrylamide
concentration and therefore pore size may be calib-
rated using standard proteins of known molecular
weight. By extrapolation, reliable molecular weight
estimates of large numbers of polypeptides in a com-
plex mixture may be obtained. Proteins are Rxed in
the gel after electrophoresis using a 50% trich-
loroacetic acid solution and stained in a Coomassie
blue solution. Radiolabelled proteins were also detec-
ted by autoradiography. For a more detailed
1310 II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis

discussion please refer to the separate article on De-
tection Techniques, Staining, Autoradiography and
Blotting.
In the original Laemmli procedure, gels were elec-
trophoresed in small glass cylinders and since the
absolute mobility of SDS-polypeptides varied slightly
from gel to gel the relative mobility of standard and
unknown bands was calculated as the ratio between
the mobility of the protein and the mobility of the
tracking dye, bromphenol blue (BPB), which travels
with the SDS-micelle front behind the leading buffer
front.
A major improvement on the original procedure
uses rectangular slab gels of uniform thickness (typi-
cally 0.5}1.5 mm) with sample wells (typically
10}30) set into the stacking gel. Electrophoretic ap-
paratus was originally constructed according to a
variety of ad hoc patterns, in house, from perspex,
using gels polymerized between notched glass plates,
although now most laboratories use commercially
available equipment. The use of such equipment (see
the separate article on Slab Gel Electrophoresis:
Equipment) has led to highly standardized reliable
separations since standards and unknowns may be
run under identical conditions. The apparent molecu-
lar weights of unknowns may be obtained by extra-
polation from a plot of log MWt vs. mobility of
standard proteins. Figure 1 shows how individual

Figure 1 (See Colour Plate 41). The inset shows the separation of human head hair proteins using 12% sodium dodecyl
sulfate}polyacrylamide gel electrophoresis with the point of migration of standard proteins to the left under the heading kDa: lanes 1}3
S-carboxymethylated proteins; lane 4 reduced non-S-carboxmethylated proteins: lane 1 14C Autoradiograph; lanes 2 and 4, silver stain;
lane 3, Coomassie stain. The main body of Figure 1 shows a densitometric profile of the separated lane 1 (autoradiograph
14
C-S-carboxymethylated, green). Lane 2 (silver stain, S-carboxymethylated, red); lane 4 (silver stain, non-S-carboxymethylated).
Molecular weights were extrapolated from a plot of log MWt vs. relative migration of standard proteins.
II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis 1311

zones in a complex mixture of human hair keratins

may all be assigned an apparent molecular weight
and how the entire molecular weight proRle is appar-
ently shifted 20}40 kDa higher after S-carboxy-
methylation.
Although several alternative sample extraction and
buffer systems are available (see later), the use of the
basic Laemmli system has proved so robust and re-
liable that it has revolutionized protein characteriza-
tion in complex mixtures to the extent that proteins
of widely differing function are routinely described
according to their apparent molecular mass on
Laemmli SDS-PAGE. The explosive increase in the
use of this technique across a comprehensive range of
the biological sciences is illustrated in Figure 2 show-
ing the number of citations each year of the Laemmli
(1970) publication.
I will now discuss subsequent major methodologi-
cal developments of the basic Laemmli protocol.

Other SDS-PAGE Systems

The SDS-PAGE technique may also be performed
with the simpler continuous phosphate buffer system
of Weber and Osbourne, although dilute samples are
not concentrated by stacking as they are in the discon-
tinuous system of Laemmli. In general, however, very
dilute protein samples may be concentrated by pre-
cipitation with trichloroacetic acid or acetone prior to
loading.
Separations of peptides and protein mixtures
ranging in molecular weight from 300 kDa down to
about 10}12 kDa may be optimally achieved using
resolving gels of different Rxed acrylamide concentra-
tions as, for instance, shown in Figure 3A (7.5%) or
B (10%) for the separation of phosphoproteins dur-
ing the invasion of human red blood cells by the
malarial parasite. Alternatively, a gradient mixer may
be used to produce resolving gels with a linear or
concave concentration gradient.
Peptides of less than 10 kDa are not resolved using
the normal Laemmli system even at maximal acryl-
amide concentration so a variety of modiRed buffer
systems have been introduced to allow the separation
of, for example, cyanogen bromide and proteolyti-
cally cleaved peptides for structural analysis of pro-
teins. The original Swank and Monkres system uses
SDS/urea and is capable of resolution down to 2 kDa,
although some undesirable modiRcations of amino
acids may take place in the presence of urea. Other
buffer systems designed to fractionate low molecular
weight peptides such as the tricine and the modiRed
Laemmli system have been more recently introduced
to address this problem. For a detailed technical dis-
cussion of the preparation of gels using these various
Figure 2 The number of SCI citation of the original Laemmli,
U.K. (1970) paper on sodium dodecyl sulfate}polyacrylamide gel
electrophoresis for each year from 1974 (when SCI started).
buffer systems one should refer to one or more of the
following generally excellent technical guides avail-
able (Current Protocols in Molecular Biology; Gel
Electrophoresis of Proteins } A Practical Approach;
Protein PuriTcation, Principles } High Resolution
Methods and Techniques).
Identi\cation of Resolved Protein
Zones
Protein bands may be directly visualized by scanning
unstained unRxed gels at A280 but most workers em-
ploy a protein stain after Rxation. Coomassie R-250
dye binds noncovalently to proteins giving deep blue
bands on a clear background after diffusion (see
Figure 5) or electrophoretic destaining. Alternatively,
Rxed proteins may be revealed by the precipitation of
silver granules from an alkaline silver nitrate solution.
The procedure is more complicated than the dye-
binding protocol but can be up to 100 times more
1312 II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis

before and after heat shock. 3H and 14C labelled

amino acids are also used as is 32P phosphate
(see Figure 3) to label phosphoamino acids. 35S,
14
C and 32P may be detected by direct autoradiogra-
phy by placing Rxed dried gel slabs against suitably
sensitized X-ray Rlm. The sensitivity of detection de-
pends, on the level of incorporation of the particular

Figure 3 Sodium dodecyl sulfate}polyacrylamide gel elec-

trophoresis showing protein phosphorylation in human blood cells
parasitized with the malarial organism Plasmodium falciparum
using continuous metabolic labelling with 32P phosphate. A, 7.5%
Resolving gel; B, 10% Resolving gel. Mobility of molecular weight
standards indicated in kiloDaltons to the right of gel A and gel B.
Lane 1, parasite pellet (mature shizonts) 36 h post label; lane 2,
infected red cell ghost 36 h post label; lane 3, parasite pellet
(new rings) 50 h post label. Lane 4, infected red cell ghost 50 h
post label.

sensitive. Several commercial kits are now available,

ensuring the robust reproducibility of the once no-
toriously Rckle silver staining technique. Because of
this and the added sensitivity the silver staining tech-
nique has now become standard in most laboratories.
Proteins vary markedly in their afRnity for either
Coomassie dye or silver. The insert to Figure 1 (lanes
2 and 3) shows a comparison of the proteins
separated from human hair after solubilization and
carboxymethylation using either silver stain or
Coomassie stain respectively. The differing sensitivi-
ties are obvious. Lane 1 shows the same proteins
labelled with 14C iodoacetic acid and revealed by
autoradiography. Note that the lower molecular
weight zones are very strongly labelled whereas they Figure 4 A 10% sodium dodecyl sulfate}polyacrylamide gel
electrophoresis separation of 35S methionine-labelled proteins
stain relatively weakly with silver (lane 2) and very
from human lymphocytes after different recovery periods from 1 h
poorly with Coomassie dye (lane 3). heat shock at 423C. Control cells from the same individual were
maintained at 373C. Lane 1, 2 h control; lane 2, 2 h after heat
shock; lane 3, 3-h control; lane 4, 3 h after heat shock; lane 5, 4-h
Detection of Radiolabelled Proteins control; lane 6, 4 h after heat shock. Numbers to the right of the
gel indicate the position of major heat shock proteins (hsp) in-
Proteins may be radiolabelled during synthesis in the
duced after heat shock. Hsps are routinely classified according to
presence of labelled amino acids. 35S methionine is their apparent molecular weight (105, 90, 70 kDa, etc.) after one-
commonly used in this regard. Figure 4, for instance, dimensional SDS-PAGE. (Thanks to my PhD student D. Visala
shows biosynthetic labelling in human lymphocytes Rao for this gel.)
II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis 1313

amino acid into the particular protein zone but also

on the isotope itself. For example after a 24-h expo-
sure about 300}500 d.p.m.cm2 of 32P will give a vis-
ible band whereas about tenfold this level of radioac-
tivity is required to produce a visible band using 35S or
14
C label. 3H label is not detected because the low

-emissions do not penetrate the gel matrix. A Suor
may be introduced into the gel matrix prior to gel
drying to detect 3H-labelled proteins as well as to
improve the sensitivity to 14C and 35S.
Proteins may also be labelled post synthetically
using such reagents as 14C iodoacetic acid which pref-
erentially labels available sulfhydryl groups (see
Figure 1, insert lane 1) and 125I which preferentially
labels available tyrosine groups. Many laboratories
use phosphoimaging whereby the radioactive pro-
teins in the gel excite a phosphorescent screen and the
number of excitation events is directly digitized. Al-
though this instrumentation is rather expensive it
obviates the need for X-ray Rlm and gives results in
hours rather than days. Effective concentrations are
said to be linear over six orders of magnitude whereas
the linear range for X-ray Rlm detection rarely covers
one order of magnitude. For further details see the
separate article on Detection Techniques: Staining,
Figure 5 Sodium dodecyl sulfate}polyacrylamide gel elec-
Autoradiography and Blotting. Radioactivity may trophoresis of purified human carbonic anhydrase I. The horizon-
also be quantiRed in gels after slicing and solubiliz- tal arrow indicates the interface between the 4% stacking gel and
ation. In this case, a generally useful labile crosslinker the 10% resolving gel. (1) Purified Caucasian enzyme; (2) purified
N,N1-diallyltartardiamide (DATD) is often used in normal component from isoelectric focusing; (3) purified variant
place of bisacrylamide. component from isoelectric focusing; (4) purified heterozygote
mixture before resolution of normal and variant components.
Reproduced with permission from Jones GL and Shaw DC (1982)
Identity of Individual Protein Zones Biochemical Genetics 20: 943.

The appearance of single discrete zones on SDS-

PAGE, sometimes used as an indicator of homogen-
eity in protein puriRcation (see Figure 5), does not
preclude the possibility of multiple comigrating pro-
tein species with differing independent functions.
Zones may be identiRed immunologically after West-
ern blotting onto a suitable matrix and again the
reader should refer to the separate article on Detec-
tion Techniques: Staining, Autoradiography and
Blotting in this series for further details.
Although the process of the SDS-PAGE results in
largely inactive proteins, some techniques allow in
situ renaturation followed by speciRc enzyme detec-
tion (for proteases, for example). A potentially
powerful development involves a second dimension
SDS-PAGE separation of the entire repertoire of pro-
tein zones separated in the Rrst dimension after co-
stacking with a protease that retains its activity in the
presence of SDS. In this way, a comprehensive
peptide map of the proteins may be obtained and
homologous proteins identiRed within or between
species. Proteolytic fragments, so separated, may also
be transferred onto a PVDF membrane and se-
quenced.
Glycoproteins may be distinguished using the spe-
ciRc periodic acid}Schiffs (PAS) stain whereas phos-
phoproteins may be detected by intact cell metabolic
labelling with 32P phosphate (see Figure 3) or with
[-32P] ATP in crude lysates.
Scanning and Quanti\cation of
Individual Protein Zones
Proteins identiRed by speciRc or nonspeciRc staining
techniques or by autoradiography or Suorography
may be quantiRed in a relative sense after densito-
metric scanning of stained gels, photographs or
autoradiographs. The inset to Figure 1 shows the
separation of human hair proteins before (lane 4) and
after (lanes 1}3) 14C-S-carboxymethylation. Proteins
are detected by 14C autoradiography (lane 1),
1314 II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulfate Polyacrylamide Gel Electrophoresis
Coomassie (lane 2) or silver stain (lanes 2 and 4). The
main body of Figure 1 shows a gel scan of lane 2 (red)
and lane 4 (blue) as well as the scan of the autoradio-
graph (lane 1, green) according to the apparent mo-
lecular weight of each component in the complex
densitometric scan. Readers should refer to the separ-
ate article on Instrumentation for Scanning gels for
further details in this Reld.
Anomalous Migration in SDS-PAGE
The precise structural relationship of SDS to protein
during SDS-PAGE is unknown but various studies
have indicated that a wide variety of proteins all bind
a relatively constant amount of SDS (+1.4 g SDS g\1
protein) and adopt a similar Sexible rod shape regard-
less of their native conRguration. It is this supposedly
constant very high and uniform charge}mass ratio
which allows for reliable molecular weight deter-
mination since migration rate therefore depends on
molecular sieving alone. Having said this, even the
original paper of Laemmli referred to the anomal-
ously slow migration of a bacteriophage point mu-
tant. The very basic proteins, histones, behave so
anomalously that special buffer conditions in the
presence of urea must be employed to determine
reliably their molecular weights. Glycoproteins also
show anomalous migration. In this laboratory we
have, over the years, seen direct evidence of discrete
changes in apparent migration in SDS-PAGE related
to single mutations which change the charge on a pro-
tein without signiRcantly affecting its calculated mo-
lecular weight. Figure 5, for instance, shows an SDS-
PAGE separation of puriRed carbonic anhydrase
1 (CA1) from a Caucasian blood donor (lane 1) or an
Australian Aboriginal donor (lane 4) who is hetero-
zygous for a polymorphic variant CA1-9. The normal
and the variant component of the heterozygote were
resolved by isoelectric focusing (lanes 2 and 3, respec-
tively). The variant component differed from the nor-
mal component at only one position (AspPGly) al-
though the apparent molecular weight after SDS-
PAGE was 27 kDa compared with 28.5 Da. From this
and other previous studies with deRned point muta-
tions of deRned proteins we propose that each extra
negative charge on a modiRed protein results in retar-
dation on SDS-PAGE such that its apparent molecu-
lar weight is greater by 1.5 kDa.We have applied this
rule of thumb to changes in the apparent molecular
weights of the keratins of human hair after S-car-
boxymethylation which substitutes an extra negative
charge for each thiol. See Figure 1 for general mo-
lecular weight shift of hair proteins after S-car-
boxymethylation } compare the proRle from lane
4 (red) with the proRle from lane 2 (blue). Given the
published molecular weights of the unmodiRed
keratins we were then able to calculate an apparent
thiol content in rough agreement with previous
estimations assuming that each thiol group was
substituted with an extra negative charge after
carboxymethylation (lane 2).
We are therefore convinced that electrophoretic
migration of SDS}protein complexes is not totally
independent of charge of the native protein and fur-
thermore that if a one-dimensional SDS-PAGE system
could be suitably calibrated, the relationship we have
described could become useful in studies where the
net charge on a given protein is changed genetically or
epigenetically in an incremental way.
Conclusions
Over the past 30 years, one-dimensional SDS-PAGE
has become a standard technique in most biological
and biomedical laboratories. It offers a powerful
combination of resolution (up to 200}300 compo-
nents in crude mixtures), reliability, versatility (pore
size, extraction conditions and buffer systems can be
tailored precisely to the conditions required for par-
ticular proteins) and reproducibility (different labor-
atories all refer conRdently to certain proteins by their
apparent molecular weight on SDS-PAGE, e.g. Hsp
70; see Figure 4). In addition, selectivity may be en-
hanced by using a variety of detection methods
ranging from nonspeciRc staining, metabolic and
postmetabolic radiolabelling to highly speciRc im-
munological detection after Western blotting. One
may conRdently expect that one-dimensional SDS-
PAGE will continue to play a central role in protein
separation in biological laboratories for many years
to come, even though Figure 2 suggests that the tech-
nology uptake is now stable in that the citation rate
peaked in 1990. It may be that improvements in
two-dimensional technology (see the article on Two-
dimensional SDS-PAGE) and the developing disci-
pline of proteomics involving ever more powerful
software to analyse complex two-dimensional protein
gels will see a slow erosion of the central position of
one-dimensional SDS-PAGE in isolation as method of
choice in the resolution of complex protein mixtures.
See Colour Plate 42.
Further Reading
Gallagher SR (1996) 1D SDS-PAGE. In: Ausubel et al. (eds)
Current Protocols in Molecular Biology, vol. 2, ch. 10.
Massachusetts: John Wiley and Son.
Hames BD and Rickwood D (eds) (1990) Gel Electrophor-
esis of Proteins: A Practical Approach, 2nd edn.
New York: Oxford University Press.
 II / ELECTROPHORESIS / Porosity Gradient Gels
Hunkapiller MW, Lujan E, Ostrander F and Hood LE
(1983) Isolation of microgram quantities of proteins
from polyacrylamide gels for amino acid sequence
analysis. Methods in Enzymology 91: 227.
Laas T (1989) Electrophoresis in gels. In: Janson J-C and
Ryden L (eds) Protein PuriTcation } Principles, High
Resolution Methods and Applications. New York,
Weinheim and Cambridge: VCH Publishers.
Laemmli UK (1970) Cleavage of structural proteins during
the assembly of the head of bacteriophage T4. Nature
227: 680.
Matsudaira PT and Burgess DR (1978) SDS microslab
linear gradient polyacrylamide gel electrophoresis.
Analytical Biochemistry 87: 386.
Schagger H and von Jagow G (1987) Tricine-sodium
dodecyl sulfate}polyacrylamide gel electrophoresis for
Polyacrylamide Gel Electrophoresis
See II / ELECTROPHORESIS / One-dimensional Polyacrylamide Gel Electrophoresis;
II / ELECTROPHORESIS / One-dimensional Sodium Dodecyl Sulphate Polyacrylamide
Gel Electrophoresis;
II / ELECTROPHORESIS / Two-dimentional Polyacrylamide Gel Electrophoresis
Porosity Gradient Gels
G. M. Rothe, Johannes Gutenberg-University, Mainz,
Germany
Copyright ^ 2000 Academic Press
Introduction
The high resolving power of polyacrylamide (PA) gels
for proteins, peptides and nucleic acids can be im-
proved by using gradient gels instead of homogene-
ous (i.e. single concentration) gels. However, a more
speciRc separation of polynucleotides in PA gels af-
fords separation by incorporating a 40}80% de-
naturant gradient (7 mol L\1 urea, 40% (v/v) form-
amide) into a homogeneous PA gel (of e.g. 6.5%
(w/v) total polymer concentration) or applying a tem-
perature gradient to a homogeneous PA gel.
In PA gradient gels the average pore radius de-
creases with increasing gel concentrations, i.e. in the
direction of the migrating protein (polynucleotide)
bands. This results in a sharpening of the bands be-
cause the molecules at the front of the moving band
are slower than those at the rear. Because of this
effect, gradient gels need not be covered by a stacking
gel, as in disc gel electrophoresis. In porosity gradient
1315
the separation of proteins in the range from 1 to
100 kDa. Analytical Biochemistry 166: 368.
Shapiro AL, Vinuela E and Maizzel Jr JV (1967) Molecular
weight estimation of polypeptide chains by electrophor-
esis in SDS-polyacrylamide gels. Biochem. Biophys. Res.
Commun. 28: 815.
Takano E, Maki M, Mori H, Hatanaka N, Marti T,
Titani K, Kannagi R, Ooi T and Murachi T (1988)
Pig heart calpastatin: identiRcation of repetitive
domain structures and anomalous behaviour in poly-
acrylamide gel electrophoresis. Biochemistry 27:
1964.
Weber K, Pringle JR and Osborn M (1972) Measurement of
molecular weights by electrophoresis on SDS-acrylam-
ide gel. Methods in Enzymology 26: 3.
gels with a steep increase of polymer concentration
(e.g. from 4 to 30% T (w/v) where %T"g acrylam-
ide#g Bis"N,N-methylenebisacrylamide (Bis) per
100 mL) proteins of a large size range (approximately
104}106 Da) can be separated. In shallow gradients
('4% T to (30% T), the separable size range of
proteins is limited but they still provide an improved
band sharpening.
There are two modes to run porosity gradient gels:
a Rxed-time mode, where electrophoresis is termin-
ated after a certain time, and a time-dependent mode,
which means that a number of consecutive electro-
phoretic mobilities are registered. Fixed-time elec-
trophoresis is performed if protein (polynucleotide)
patterns are to be screened, such as in population
genetics or when determining the molecular mass of
sodium dodecyl sulfate (SDS) denatured proteins.
Molecular size properties of nondenatured proteins,
however, cannot be elucidated that way, but afford
time-dependent investigation of protein mobilities.
On the other hand, time-dependent PA gradient
gel electrophoresis not only offers the possibility to
estimate the molecular mass of native proteins and
enzymes but also allows determination of their Stokes
radius, frictional coefRcient, free electrophoretic
1316 II / ELECTROPHORESIS / Porosity Gradient Gels
mobility and nett charge. A number of different
(iso)enzyme systems have been classiRed in this way
and comparisons between related species used to
study the evolution of enzyme systems.
Porosity gradient gels can be easily prepared using
one of the different devices on the market. Ready-to-
use pore gradient gels are commercially available
(Amersham Pharmacia Biotech, Freiburg, Germany;
Gradipore, 200 Harris Street, Pyrmont NSW 2009,
Sydney, Australia). Porosity gradient gels can be pre-
pared in casting glass cassettes either without any
further support or by adhering them to a silanized
glass plate or a reactive polyester Rlm. The latter two
methods are employed when ultra-thin gels are to be
used horizontally. Glass cassette cast PA gradient gels
without any further support are used vertically.
Porosity of Polyacrylamide Gradient
Gels
In 1962 Ornstein and Davis were the Rrst to suggest
a formula to estimate roughly the average pore dia-
meter of homogeneous PA gels:
pav (nm)"12.67;(%T)\1/2
where pav (nm) is the average pore diameter in
nanometres and %T is the total acrylamide concen-
tration (g acrylamide#g Bis in 100 mL).
Based on the Ogston model which describes dex-
tran gels as assembled from arbitrarily arranged gel
Figure 1 Plot of average pore radius (rav (nm)) against PA gel concentration (T (%)). Triangles, average pore radii calculated as
suggested by Ornstein and Davis (1962). Squares, average pore radii calculated as suggested by Raymond and Nakamichi (1962).
Circles, maximum pore radii as marked by native proteins of known radius: 1, thyroglobulin; 2, ferritin; 3, catalase; 4, lactate
dehydrogenase; 5, bovine serum albumin; 6, ovalbumin. Reproduced with permission from Rothe and Maurer (1986).
rods, Raymond and Nakamichi related the average
pore diameter of PA gels to the total polymer concen-
tration (T) as follows:
pav (nm)"K;d;(100;p)1/2;(%T)\1/2 [2]
where K is the factor resulting from the angle in
which the gel rods are linked together (1.5), d (nm) is
the diameter of a PA gel rod (0.5), p (g cm\1) is the
density of gel rod (1.2). This results in:
pav (nm)"8.216;(%T)\1/2 [3]
The largest pore diameter in a PA gel of a certain
concentration is, however, much larger than the aver-
age pore diameter (Figure 1). Moreover, the largest
pore diameter deviates increasingly from the average
pore diameter with decreasing gel concentration. The
pores therefore are statistically distributed, but the
standard deviations of the average pore radii and the
distribution function (Gaussian or logarithmic distri-
bution) are unknown.
The generally held assumption of a random mesh-
work of cross-linked individual PA rods could not be
conRrmed by electron microscope images. They re-
[1] vealed sponge-like structures in the submicron range.
Such structures are in accordance with the mode in
which gels polymerize. PA molecules Rrst arrange as
high molecular aggregates that are in the sol state and
not interconnected. Thereafter, cross-linkage to
a three-dimensional gel occurs: this is indicated by an
abrupt start of gelation.
 II / ELECTROPHORESIS / Porosity Gradient Gels
Analytical Separation of Native
Proteins in a Glass Cassette-Cast
Porosity Gradient
Gradient preparation is performed with acrylamide
solutions of high and low concentrations, usually by
using a two-chamber gradient mixer, although more
sophisticated gradient formers have been developed.
Linear PA gradients are usually prepared by the tech-
nique which was Rrst described by Martin and Ames
in 1961 for the preparation of linear sucrose gradi-
ents. Glass cassette-cast gels are mostly 82;82
(140) mm or 125;250 mm and a thickness of 3.0,
1.0, 0.8, 0.5 or 0.1 mm.
Preparation of a Batch of Unattached Gradient Gels
Polyacrylamide gradient gels cast in glass cassettes
may be prepared individually or simultaneously in
batches (Figure 2). The latter method saves time and,
although the gradients usually deviate slightly from
each other, they are well suited to determine protein
patterns, e.g. isozyme patterns as in population gen-
etics. Any form of gradient (linear, concave, convex)
may be prepared but linearly increasing gradients of
total polymer concentration are most commonly
used.
The device shown in Figure 3 can prepare six
gradient gels simultaneously. In each gel the PA con-
centration increases linearly from top to bottom from
approximately 5 to 25% T. The gels are encased in
glass cassettes of internal dimensions 172;82
Figure 2 Assembly of a glass cassette to cast a PA (gradient) gel slab. A, Slot former; B, front and D rear glass plate of cassette;
C, left and right distance bar. 1, Exploded view of cassette; 2, side view; 3, front view. Procedure according to Pharmacia, Uppsala,
Sweden. Reproduced with permission from Rothe (1991).
1317
;1.0 mm. Each cassette is Rtted with a slot former
and inserted in a gel-casting device. The linear PA
gradient is prepared by using a two-chamber gradient
mixer, a separate reservoir (for the catalyst solution),
a proportioning pump, a 1 mL mixing chamber (and
a reservoir Rlled with sucrose and a pump to lift the
gradient into the cassettes). The device shown in
Figure 3 is used as follows: The inner chamber (1)
and the connecting tube (3) to the outer chamber (2)
of the gradient mixer are Rlled with 57 mL of
Tmin solution. Then the tube (3) to chamber (2) of the
mixer is closed. Afterwards 57 mL of the Tmax solu-
tion is pipetted into chamber (2) of the gradient
mixer. Now 22.5
L of N,N,N
,N
-tetra-
methylethylenediamine (TEMED) is mixed with the
Tmin and the Tmax solution, respectively. A separate
reservoir (4) is Rlled with 35 mL of gel buffer contain-
ing 50 mg ammonium persulfate. The connection be-
tween chamber (1) and (2) of the gradient mixer is
opened, after which the stirrer (5) of the gradient
mixer and the stirrer (6) of the mixing chamber (7) as
well as the peristaltic pump (8) are switched on.
Immediately after chamber (1) is empty, the pump (8)
is switched off and a sufRcient amount of sucrose
solution (50% (w/v)) is pumped from the correspond-
ing reservoir (9) with the help of a separate pump (10)
underneath the gel cassettes (12) to lift the whole
gradient into the cassettes, which are in the gel-cast-
ing device (14).
The Tmin and Tmax solution contain acrylamide and
Bis at the same ratio (acrylamide}Bis"24 : 1). The
Tmin solution contains 4.205 g acrylamide and
1318 II / ELECTROPHORESIS / Porosity Gradient Gels

Figure 3 (A) Device for preparing a batch of six PA porosity gradient gels. (B) Scheme for preparing a batch of six porosity gradient
gels each encased in a glass cassette without further support. 1 and 2, chambers of the gradient mixer (1 with magnetic bar);
3, connecting tube between both chambers which can be closed by a stopcock (not shown); 4, reservoir to hold the catalyst solution
(ammonium persulfate); 5 and 6, stirrers; 7, mixing chamber (modified 1 mL syringe); 8, two-channel pump; 9, reservoir to hold sucrose
solution; 10, one-channel pump; 11, air trap; 12, gel cassettes with 13, inserted slot formers; 14, gel-casting apparatus (made of
perspex) with 15 removable front plate. Reproduced with permission from Rothe (1994).

0.175 g Bis per 100 mL gel buffer while the Tmax solu-
tion contains 31.54 g acrylamide and 1.314 g Bis in
100 mL gel buffer. The Tmin and Tmax solutions are
diluted upon gradient formation with catalyst solu-
tion by a factor of 1.255 (Figure 3) and the gel solu-
tion is pumped to about 5 mm above the slot tem-
plate. This results in a Rnal concentration range of
approximately 5}25% T. (Mixing both catalysts
into the Tmin and Tmax solution is also possible
but carries the danger that the gel may solidify
before being completely cast in the cassette). Prior
to use all solutions are brought to room temper-
ature and degassed. The ammonium persulfate solu-
tion should be prepared freshly each time.
90 mmol L\1 Tris, 45 mmol L\1 boric acid and
2.5 mmol L\1 EDTA}Na2, pH 8.4 is used as gel and
electrode buffer. Further buffer systems are given in
Table 1.
Vertical Electrophoresis
Glass cassette-cast gradient gels are mounted
vertically into an electrophoretic apparatus consisting
of an upper and lower electrode vessel (Figure 4).
The upper buffer tank has rubber gaskets into which
two or four cassettes can be inserted. The lower
electrode vessel is Rlled with cooled buffer; the upper
electrode vessel with the inserted gel cassettes is
mounted into the electrophoresis apparatus and Rlled
with electrode buffer. Then the samples (enriched
with 10% sucrose) are added to the slots (with
a Hamilton syringe). Afterwards the voltage is
switched on (440 V cm\1) for 15 min for the pro-
teins to migrate into the gel. Finally the buffer is
circulated from buffer tank to buffer tank at the same
voltage. The lower buffer tank is cooled to 53C
during electrophoresis by a cooling coil.
 II / ELECTROPHORESIS / Porosity Gradient Gels 1319

Table 1 Buffer systems used in porosity gradient gel electrophoresis to separate native proteins

Gel buffer Electrode buffer %T range Authors

0.35 mol L\1 Tris HCl, 0.06 mol L\1, Tris, 3}20 KopperschlaK ger et al. (1969)
pH 8.9 0.40 mol L\1 glycine, pH 8.3
0.09 mol L\1 Tris, Same as gel buffer 4}26 Anderson et al. (1972)
0.08 mol L\1 boric acid, Lasky (1978)
0.003 mol L\1 EDTA}Na2,
pH 8.3
0.01 mol L\1 Tris, Same as gel buffer 5}30 Slater (1969)
0.08 mol L\1 glycine, 5}15
pH 8.3
0.04 mol L\1 Veronal}Na, Same as gel buffer 5}30 Lambin and Fine (1979)
0.04 mol L\1 Tris,
0.01 mol L\1 glycine,
0.04 mol L\1 ethanolamine,
0.001 mol L\1 EDTA}Na2
pH 9.8
0.01 mol L\1 Na}phosphate, Same as gel buffer 5}30 Lambin and Fine (1979)
pH 7.2

References as given in Rothe and Maurer (1986). Reproduced with permission from Rothe and Maurer (1986).

PA gradient gel electrophoresis under nondenatur-
ing conditions has proved to be advantageous com-
pared to electrophoresis in homogeneous gels, e.g.
in plant population genetics. Figure 5 gives an
example.
Separation of Native Proteins in
an Ultra-Thin Support-Bound Porosity
Gradient
To prepare a thin gradient gel of the dimensions
120;250;0.5 mm Rxed to a derivatized clear and
Sexible polyester foil (e.g. manufactured by Gel
Bond, Marine Colloids, Rockland, MN, USA or
Serva, Heidelberg, Germany), the gel-forming devices
shown in Figure 6 may be used. (When the cassettes
are assembled the slot formers must not touch the
opposite glass wall but leave a space of 0.1 mm in
between). The following solutions may be used to
form a gradient ranging from 3 to 30% T:
1. gel buffer: 90 mmol L\1 Tris, 80 mmol L\1 boric
acid, 2.5 mmol L\1 EDTA}Na2 , pH 8.4;
2. electrode buffer: 1 in 2 diluted gel buffer;
3. stock acrylamide solution (30% T: 28.8 g acryl-
amide plus 1.2 g Bis plus 50 mL gel buffer, made
to 100 mL with distilled water);

Figure 4 Vertical electrophoretic apparatus in which up to four glass cassette-cast porosity gradient gels can be inserted. Upper
electrode vessel with 2 rubber gaskets to hold 2 to 4 glass cassettes, each containing a porosity gradient made of PA; #, !,
electrodes. Modified from an instruction leaflet published by Pharmacia, Uppsala, Sweden.
1320 II / ELECTROPHORESIS / Porosity Gradient Gels

Figure 5 Electrophoresis of plant diaphorase isoenzymes in a 4}20% T PA gradient gel of 0.8 mm thickness (length 175 mm,
height 75 mm). (A) Zymogram of diaphorase enzymes (numbers indicate genotypes of the tetrameric enzyme at locus B. (B) Schematic
representation of genotypes at locus DIA-A and DIA-B. Enzyme source: leaf buds of seven different trees of European beech (Fagus
sylvatica L.). Conditions of electrophoresis: gel and electrode buffer: 45 mmol L\1 Tris, 40 mmol L\1 boric acid, 1.25 mmol L\1
EDTA}Na2; pH 8.4; running time 4 h; voltage gradient 40 V cm\1; temperature 53C.
Enzyme extraction: 1.5 mL Eppendorf tubes containing 150 mg of green bud leaves, 50 mg of quartz sand and 600
L of extraction
medium were cooled from underneath with ice water. A motor-driven grinding cone adapted in the shape of the tube (rotating at
700 rpm) was used to homogenize the material. The extraction medium contained in 100 mL: 1.21 g Tris, 1.43 g Na2HPO4, 60 mg
L-cysteine, 210 mg ascorbic acid, 14 g sucrose, 40 mg NADP, 15 g polyclar AT (PVPP) and 1 g polyethylene glycol, pH 7.5 (with
H3PO4). The homogenate was centrifuged for 30 min at 43C and 10 000 g and the clear supernatant used as crude enzyme extract.
Samples of 8
L were applied per lane. Diaphorase isozymes were visualized histochemically (60 mL 25 mmol L\1 Tris-HCl, pH 8.5,
containing 24 mg NADH, 1.5 mg 2,5 dichlorphenolindophenol-Na;2H2O (DCPIP) and 1.8 mL MTT (500 mg 100 mL\1 aq. bidest.
water). Anode at bottom. A, Enzymes of gene locus DIA-A; B, locus DIA-B. In (A) not all genotypes indicated in (B) are shown.

4. dense acrylamide solution (30% T: to 6.5 mL
stock solution is added, shortly before use, 20
L
TEMED (1 in 10 with H2O diluted solution) and
5
L ammonium persulfate solution (40% w/v in
distilled water));
5. light acrylamide solution (3% T: 1 vol of stock
acrylamide solution is diluted with 4.5 vol distilled
water and 4.5 vol of gel buffer shortly before use
and 40
L TEMED (1 in 10 with distilled water
diluted solution) and 10
L ammonium persulfate
 II / ELECTROPHORESIS / Porosity Gradient Gels 1321

Figure 6 Preparation of an ultra-thin PA gradient gel fixed to a polyester foil. (A) Rolling the polyester foil (reactive side up, e.g. Gel
Bond) on to one of the glass plates used to build the casting glass cassette: 1, levelling table; 2, glass plate; 3, hydrophilic side of
polyester foil; 4, water layer; 5, rubber roller. (B) Trough template preparation. The bars are prepared from two layers of self-adhesive
tape with a scalpel. (C) Assembling the glass cassette to cast the PA gradient. (D) Casting the porosity gradient: two-chamber mixer
and glass cassette. Reproduced with permission from Rothe (1991).

solution (40% w/v in distilled water) is added. The
gradient is made of 6.5 mL of dense acrylamide
solution and 6.5 mL of light acrylamide solution.
After gradient formation, 2 mL of light acrylamide
solution is overlaid; the slots must be situated in
the middle of the 3% T range.
esis is performed at 1000 V (50 V cm\1). Then the
slots are Rlled with protein solution (or electrode
buffer) and the power is turned on again at a voltage
of 1000 V for approximately 2 h. Afterwards the gel,
Rxed on the polyester foil, may be stained for proteins
or (iso)enzymes (Figure 8).

Horizontal Electrophoresis
Before electrophoresis, the gel is taken out of the
cassette. A few drops of kerosene are put on the
cooling plate of the opened electrophoretic apparatus
(Figure 7) and the gel, Rrmly adhering to the polyes-
ter foil, is placed on it, carefully avoiding the inclu-
sion of air bubbles. Both ends of the gel are connected
with the buffer vessels by paper wicks or a household
sponge-like material. A 15}30 min pre-electrophor-
Determination of the Course and
Concentration of a Porosity
Gradient Gel
The course and %T range of laboratory-made PA
gradient gels can be controlled by densitometry if
a coloured dye such as p-nitrophenol is added to the
denser acrylamide solution prior to gradient forma-
tion. After polymerization, the increase in colour
1322 II / ELECTROPHORESIS / Porosity Gradient Gels

Figure 7 Horizontal electrophoretic apparatus with cooling

plate. 1, Cover lock; 2, gassing stud; 3, high voltage connection of
the lid; 4, flexible tube to the cooling plate; 7, with a cooling device
(not shown); 5, electrode bar (used in isoelectric focusing); 6,
electrode ledge; 8, support for cooling plate. For PA gradient gel
electrophoresis the electrode bars are replaced by two buffer
vessels (not shown) under the cooling plate and connected to the
electrode ledge. The gel is connected to the buffer reservoirs by
(paper) wicks (not shown). Reproduced with permission from
Rothe (1991).

intensity from top to bottom of the gel can be used to

measure the course of the gradient and its precise
concentration in polyacrylamide. For a 1 mm thick
gel 15 mg p-nitrophenol may be added to 100 mL of
the dense acrylamide solution. After gelation the col-
our intensity is quantiRed by densitometry at 405 nm.
Whilst the course of the gradient can be seen directly
on the densitogram, the %T range of the gradient can
be calculated with the formula:

T(%)"Ts;(E405!Ep);Mr;(c;d;)\1 [4]

where Ts (%) is PA concentration of stock solution,

E405 is absorbance of p-nitrophenol, Ep is absorbance
of empty cassette at 405 nm, c (g L\1) is concentra-
tion of p-nitrophenol in stock acrylamide solution
(c"0.150), Mr (g L\1) is mol mass of p-nitrophenol
(Mr"139.1), d (mm) is thickness of gel (e.g. 0.5),
E [L (mol mm)\1]"molar extinction coefRcient of
p-nitrophenol at 405 nm ("1728) and T (%), as in
eqn [1].
Cross-Linkers Other than Bis and
Mixed Polyacrylamide Gels
PA is normally cross-linked with Bis to obtain an
electrophoretic matrix. The use of N,N
-(1,2-dihy-
droxyethylene)bisacrylamide (DHEBA) instead of
Bis gives gels that can be solubilized in dilute periodic
Figure 8 Electrophoresis of plant (iso)enzymes on ultra-thin PA
gradient gels fixed on a polyester film. Gel dimensions:
240;120;0.5 (mm); PA gradient from 4 to 28% T. Enzyme
source: current-year (1989) needles of Norway spruce (Picea
abies L., Karst.) sampled from a variety of clones (clone numbers
indicated) of the multiple clone variety East Prussian Late Spruce
(Hessische Forstliche Versuchsanstalt, Hann. MuK nden, Ger-
many). Enzyme extraction: 2 g of fresh needles was homogen-
ized in 10 mL of homogenizing medium (0.1 mol L\1 Na}phos-
phate, pH 7.5, containing 5% w/v Polyclar AT and 0.5% w/v
Triton X-100. The crude extract was centrifuged for 30 min at
38 000 g and the supernatant concentrated by a factor of 4 using
the ultrafiltration system Centrisart I (Sartorius, GoK ttingen, Ger-
many). Samples of 10
L were applied per lane. Conditions of
electrophoresis: 1000 V for 90 min at 43C; gel and electrode
buffer: 45 mmol L\1 Tris, 40 mmol L\1 boric acid, 1.25 mmol L\1
EDTA}Na2, pH 8.4. Enzymes were stained histochemically. An-
ode at top. Reproduced with permission from Rothe (1991).
 II / ELECTROPHORESIS / Porosity Gradient Gels
acid or dilute aqueous solutions of bases to liberate
proteins after the electrophoretic separation. Gradi-
ent Sat gels (140;120;3 mm) with an increasing
acrylamide concentration but a constant ratio of
DHEBA have been used to separate protein mixtures
from fruit with radio-labelled amino acids. Following
electrophoresis, gel slices containing protein zones
are placed in a glass scintillation counting vial
Rtted with a TeSon-lined plastic cap, 1 mL of
0.025 mol L\1 periodic acid is added and the vials are
sealed. After incubation for 48 h at 503C, 10 mL of
Mix I scintillation Suid is added, and the vials cooled
overnight before counting. PA gels produced with
DHEBA may be used with the common alkaline buf-
fer systems except borate buffers, which form nega-
tively charged complexes with the cis-1, 2-diol struc-
ture of the cross-linker DHEBA.
To improve the retardation of PA gradient gels for
low molecular mass proteins, a mixture of acryl-
amide, Bis and N,N
,N
-triallylcitric triamide has
been suggested.
The use of N-substituted acrylamido derivatives,
such as N-acryloyltris(hydroxymethyl)aminomethane
(NAT) gives PA gels with larger pores, although the
pores are still smaller than those of agarose. Gels of
similar pore sizes can be made from allyl-activated
agarose and acrylamide or N-substituted acrylamido
derivatives. The mixed-bed gels of agarose}acrylam-
ide have average pore sizes which are about 30%
larger than those of a regular 3.3% Bis cross-linked
gel with the same %T.
Size Estimation of Native Proteins and
Enzymes
The size of native proteins can be deduced from their
migration behaviour in homogeneous or gradient
gels. Both methods have the advantage that crude
tissue or cell extracts can be used as the protein
source, provided a speciRc staining method exists
with which they can be located in the gel after elec-
trophoresis. The method with homogeneous gels uses
a number of gels of different PA concentration in the
range of 4}35% T and estimates the relative elec-
trophoretic mobility referred to Bromophenol blue
(RF value) of a set of marker proteins and the sample
protein(s). From these values the gel concentration is
estimated at which the electrophoretic mobility is
zero (or would become zero). This is achieved by
plotting the logarithm of the %T concentration
(log T) in which the mobility is measured against the
respective RF value. In the underlying linear function
(log T"!k;RF#log Tlim), the value of Tlim rep-
resents the exclusion limit, the %T concentration
at which protein mobility stops. The Tlim values cal-
1323
culated for a number of marker proteins can be corre-
lated to their corresponding Stokes radii (RS) to ob-
tain a calibration line. A linear function is obtained
when RS is plotted against the reciprocal of
Tlim (RS"a;1/Tlim#b). Into this equation the ex-
clusion limit of a sample protein is inserted and this
then allows calculation of the corresponding Stokes
radius.
Polyacrylamide gradient gel electrophoresis can also
be used to estimate the molecular size of nondenatured
proteins, provided it is performed in a time-dependent
way. The following physicochemical properties of na-
tive proteins (enzymes) are obtainable:
1. molecular mass (Mr);
2. hydrodynamic radius (Stokes radius (RS));
3. frictional coefRcient (f/fo) (molecular eccentricity,
considering the molecular shape as a rotational
ellipsoid and f/fo as the quotient of the
ratio of the two half axes of the rotational ellip-
soid, f"half axis of ellipsoid, fo"half axis of
circle);
4. isomeric nature of multiple protein forms (size
isomers or charge isomers);
5. free electrophoretic mobility (and nett negative
charge (valence Z, charge Q)) at the pH value of
the electrophoresis.
The mathematical procedures used to calculate
these parameters are bound by several preconditions:
1. The PA gradient increases linearly (at a constant
ratio of acrylamide to Bis). The gradient range
however, can be chosen freely.
2. The electrophoretic pH value and the voltage
gradient are chosen in a way that marker and
sample proteins migrate sufRciently.
3. The same buffer system has been used as gel and
electrode buffer, if net charges are to be obtained.
4. The sizes of the marker and sample proteins Rt the
pore range of the PA gradient.
5. Marker and sample proteins have migrated on the
same gel slab.
6. Parts of the gel slab which have been cut into two
or more parts and stained differently are re-equi-
librated to the original gel length before protein
migrations are measured.
7. Approximately 10 (or more) time-dependent mi-
gration distances of marker and sample proteins
are accurately measured.
Estimation of the Maximum Migration Distance and
Recognition of Size Isomers
With increasing times of electrophoresis under non-
denaturing conditions, the migration of proteins in
a PA gradient gel gradually decreases (Figures 9+11).
1324 II / ELECTROPHORESIS / Porosity Gradient Gels

Figure 9 Time-dependent migration patterns of marker proteins and carbonic anhydrase (EC 4.2.1.1) (iso)enzymes from mam-
malian erythrocytes. Lanes 1 and 6, marker proteins. Lanes 2}5; carbonic anhydrases from (2) bovine, (3) human, (4) rabbit and (5)
canine. Mol mass of marker proteins: ovalbumin (45 000), bovine serum albumin (67 000), lactate dehydrogenase (140 000), catalase
(232 000), ferritin (440 000) and thyroglobulin (669 000). Linear PA 4}30% T gradient (acrylamide}Bis"24 : 1), 300 V per 73 mm
of gel length, 53C. Running times: 2, 8 and 16 h. Gel and electrode buffer: 90 mmol L\1 Tris, 80 mmol L\1 boric acid, 1.25 mmol L\1
EDTA}Na2}H2O, pH 8.4. Protein staining with Coomassie brilliant blue. Enzyme preparations from Sigma, Munich, Germany.
Reproduced with permission from Rothe (1991).

Migration of globular proteins comes to an end becomes:

when the maximum pore size of a gel region equals
their own size. The corresponding migration distance
is called the maximum migration distance (Dmax
(mm)). The maximum migration distance can be ob-
tained from a number of time-dependent protein
migrations (D (mm)) (Figures 9 and 10) which are
directly measured on the gel after proteins have
been visualized following electrophoretic separation
(Table 2). To obtain the maximum migration dis-
tance of a certain protein, the following mathematical
approximation procedure can be applied: the migra-
tion distances are double-logarithmized (ln(ln D))
and plotted versus the reciprocal of the square root of
electrophoretic migration time, 1/t1/2 (t (h)). This
results in a straight line (Figure 11) whereby the
transformed migration values (ln(ln D)) and the
transformed times of electrophoresis (t\1/2) are inter-
related by the equation:
ln(ln D)"!a;t\1/2#b
ln(ln D)"ln(ln Dmax)"b [6]
and:
D"Dmax"exp(eb) [7]
A plot of ln(ln D) versus t\1/2 can also be used to
distinguish size isomers from charge isomers. Equally
sized but differently charged forms of an enzyme
or protein system are recognized by the fact that the
straight line of each enzyme form intersects at the
same point on the ln(ln D) axis as is for example the
case with mammalian carbonic anhydrase (cf. Fig-
ure 11) and mammalian lactate dehydrogenase. On
the other hand, migration of charge isomers should
result in lines of equal slope. Proteins differing in
charge and size, however, give straight lines with both
different slopes and intercepts.
[5]
Estimation of Stokes Radius and Molecular Mass

where a and b are the slope and the intercept of
the corresponding straight line. The equation pre-
dicts that at very high values of t, t\1/2 reaches
zero. This means that the maximum migration of
a protein (Dmax (mm)) can be taken from the intercept
of the straight line with the ordinate in a plot of
ln(ln D) versus t\1/2 provided protein migrations
were larger than 2 mm and a sufRcient number of
different migration distances are registered. Let-
ting t approximate to inRnity means that eqn [5]
The maximum migration distance of globular pro-
teins is related to the maximum gel pore radius at the
respective gel concentration (cf. Figure 1). Therefore,
the maximum migration distances (Dmax) of proteins
can be correlated to their Stokes radius (RS). A linear
relationship is obtained if the logarithm of the max-
imum migration distance (ln Dmax) of proteins is plot-
ted versus the logarithm of their Stokes radius (ln RS):
ln Dmax"!m;ln RS#b [8]

Figure 10 (A) Plot of migration distances (D (mm)) of marker
proteins and (B) of five different carbonic anhydrases versus
times of electrophoresis (t (h)) in a linear PA gradient gel of
4}30% T. Conditions of electrophoresis are given in Figure 9.
Migration distances and times of electrophoresis as listed in Table
2. OVA, Ovalbumin; BSA, bovine serum albumin; LDH, lactate
dehydrogenase; CAT, catalase; FER, ferritin; TYR, thyroglobulin.
Marker proteins and carbonic anhydrases were migrated on the
same gradient gel. Purified enzyme preparations (Sigma, Munich,
Germany) comprised carbonic anhydrases from bovine (I}III),
rabbit (III, IV), human (V) and canine (V) erythrocytes.
(Reproduced with permission from Chrambach et al. Advances in
Electrophoresis Vol 4: pp 351I358.)
where ln Dmax equals eb of eqn [7], and m and b rep-
resent the slope and intercept of the straight line
(Figure 12).
It has been shown that a similar equation correlates
the logarithm of the maximum migration distance
(ln Dmax) to the logarithm of the molecular mass
(ln Mr):
ln Dmax"!z;ln Mr#c
II / ELECTROPHORESIS / Porosity Gradient Gels 1325
where ln Dmax equals eb of eqn [7], and z and c rep-
resent the slope and intercept of the straight line
(Figure 12).
Knowing the maximum migration distance of any
native globular protein, the calibration line can be
used to calculate the molecular mass of the protein by
inserting the calculated ln Dmax value and the values
of the slope (z) and the intercept (c) of the calibration
line into the equation ln Dmax"!z;ln Mr#c
(Table 3) or inserting the ln Dmax value and the values
of the slope (m) and the intercept (b) of the cali-
bration line into the equation ln Dmax"!m;
ln RS#b (Table 4).
When using PA gradients of 4}30% T and a buffer
of pH 8.4 (45 mmol L\1 Tris, 40 mmol L\1 boric
acid, 1.25 mmol L\1 EDTA}Na2, pH 8.4) a number
of markers can be used, ranging from carbonic an-
hydrase (Mr 30 000, RS 3.05) to thyroglobulin (Mr
669 000, RS 8.50; Table 5). -Galactosidase (Mr
116 000, RS 4.23) and carbonic anhydratase (Sigma,
St Louis, MO, USA) are run in the same lane and the
other marker proteins are run in a separate one. The
marker proteins bovine serum albumin, lactate de-
hydrogenase, catalase, ferritin and thyroglobulin can
be obtained as a freeze-dried mixture (Amersham
Pharmacia Biotech, Freiburg, Germany) and dis-
solved in a solution of pure ovalbumin (Boehringer,
Mannheim, Germany). Separation times depend on
the voltage gradient and may range from 0.5 to more
than 20 h (Table 2).
Estimation of Frictional Coef\cient
The frictional coefRcient (f/fo) relates the hydro-
dynamic volume of a protein molecule to its molecu-
lar mass. According to Siegel and Monty, the Stokes
radius (RS) of a protein is related to its molecular
mass (Mr) by the following equation:
RS (m)"f/fo;(3;;Mr)1/3;(4;;NA)\1/3 [10]
where RS (m) is the Stokes radius, f/fo is the frictional
coefRcient (equivalent to the quotient of the half axes
of a rotational ellipsoid),  (m3 g\1) is the partial
speciRc volume (the reciprocal of the average density
of a protein, ("0.75;10\6 ), NA (mol\1) is
Avogadro’s number (NA"6.022;1023), and Mr
(Da"g mol\1) is the molecular mass of a protein.
By substituting the actual values one obtains:
RS (m)"f/fo;66.1;10\12;Mr1/3 [11]
The geometric mean radius of a molecular mass
[9] equivalent sphere is deRned as Rm (m). It is obtained
1326 II / ELECTROPHORESIS / Porosity Gradient Gels

Figure 11 Plot of transformed migration distances (ln(ln D)) against transformed migration times (t\1/2 ) of (A) marker proteins and
(B) five carbonic anhydrase variants. Migration distances and times of electrophoresis as listed in Table 2. Abbreviations as in Figure
10. The common point of intersection of the various straight lines marked I}V on the ln(ln D) axis indicates that the investigated
enzymes are size isomers. Reproduced with permission from Rothe (1991).

by setting f/fo"1 in eqn [11] to give eqn [12]: In eqn [11] the frictional coefRcient of native pro-
teins is assumed to be constant. However, when
RS (m)"66.1;10\12;Mr1/3 [12] analysing the molecular mass (Mr) and Stokes
radius (RS) of more than 60 native proteins it became
This means that RS and Rm are interrelated through
apparent that the frictional coefRcient increases with
the frictional coefRcient:
increasing protein size (see Further Reading). A more
RS"f/fo;Rm [13] precise equation relating RS and Mr is the following:

The frictional coefRcient can be obtained from the RS (m)"M0.0225

experimentally obtained Mr and RS values and eqn
[13].
Extremely high frictional ratios are to be expected
for molecules with rod-like or Rbrous structures,
which are characterized by a high axial ratio such as
Rbrinogen or myosin or by bulky and voluminous
globular molecules with normal axial ratios. Exam-
ples of the latter are the spider-like immunoglobulin
M, the shell-like apoferritin or the branched -macro-
globulin. Usually, native proteins and enzymes do not
belong to these groups of proteins.
r ;55.1;10\12;M0.0142
r ;Mr1/3 [14]
According to this expression the frictional coefRcient
of globular proteins equals f/fo"M0.0225 r and in-
creases with molecular masses of 103 to 9;106 from
f/fo"1.17 to f/fo"1.43 while the factor 66;10\12
of the expression of Siegel and Monty
(RS (nm)"f/fo;66.1;10\12;Mr1/3) increases from
61;10\12 to 67;10\12.
As an average, the frictional ratio of globular pro-
teins sized 45}100 kDa is f/fo"1.23, for those in the
range of 100}500 kDa f/fo"1.28 and in the range of
 II / ELECTROPHORESIS / Porosity Gradient Gels 1327

Table 2 Time-dependent migration distances of marker proteins and carbonic anhydrase (iso)enzymes from erythrocytes of four
mammalian species in a porosity gradient gel from 4 to 30%T

Protein Time t (h) of electrophoresis (1/(t given in brackets)

0.5 1 2 4 8 12 16 20
(1.41421) (1.00000) (0.70711) (0.50000) (0.35355) (0.28868) (0.25000) (0.22361)

Ovalbumin D (mm) 13.05 20.25 31.5 44.0 54.0 61.0 67.5

Bovine serum albumin D (mm) 11.7 17.8 26.5 36.3 43.5 47.5 50.5 53.2
L-lactate D (mm) 7.5 11.9 17.5 24.5 30.2 33.3 35.5 37.5
dehydrogenase
Catalase D (mm) 5.5 8.8 13.2 18.8 23.5 26.6 28.5 30.0
Ferritin D (mm) 3.7 6.5 9.0 12.0 14.3 16.6 17.9 18.9
Thyroglobulin D (mm) 1.9 3.5 4.7 6.5 8.0 10.0 10.8 11.6
Bovine I D (mm) 7.3 12.5 21.5 35.5 48.2 56.0 62.5
Bovine II D (mm) 6.3 11.0 19.0 32.5 45.0 52.3 58.8 68.0
Bovine, rabbit III D (mm) 5.0 8.8 15.5 27.5 40.1 47.5 52.0 58.0
Rabbit IV D (mm) 3.8 6.7 11.8 21.5 33.6 41.0 45.7 50.2
Canine, Human V D (mm) 3.5 6.3 11.2 20.0 32.5 39.8 44.5 48.5

D (mm), Time-dependent migration distances of marker proteins and carbonic anhydrase (EC 4.2.1.1) variants. Gel length (D (mm))
and gel concentration (T (%)) are interrelated by the equation T"D# where "0.3528$0.0054 and "4.1116$0.2344; the
correlation coefficient is r"0.9985. Reproduced with permission from Rothe (1991).

500}1000 kDa f/fo"1.43. From these data and the
Stokes radius of a globular protein its molecular mass
can be estimated:
Mr"(1/(f/fo))3;3463;R3S
with Mr, f/fo and RS as in eqn [10].
This can be exempliRed by mammalian liver alco-
hol dehydrogenase (EC 1.1.1.1), which has a molecu-
lar mass of 80 kDa and a Stokes radius of 3.5 nm; the
average frictional coefRcient of globular proteins in
that range is f/fo"1.23. By inserting these values into
eqn [15] one obtains: Mr (Da)"(1/1.23)3;3463;
3.53"79 791.
[15]
Determination of Migration Velocities
The migration velocity of a protein migrating in an
electrophoretic support medium can be obtained by
computing the quotient of the difference in the
distance migrated between two consecutive time

Figure 12 Calibration lines to calculate the molecular mass (M) and Stokes radius (RS) of five carbonic anhydrase isoenzymes. The
logarithm of the maximum migration distance (ln Dmax) correlates linearly to the logarithm of the mol mass (ln Mr) and the logarithm of
the Stokes radius (ln RS), respectively. CA, Carbonic anhydrase (average ln Dmax of isozymes I}V); OVA, ovalbumin; BSA; bovine
serum albumin; LDH, lactate dehydrogenase; CAT, catalase; FER, ferritin; THY, thyroglobulin. The calculated mol masses and Stokes
radii are listed in Tables 3 and 4.
1328 II / ELECTROPHORESIS / Porosity Gradient Gels

Table 3 Calculated molecular mass of marker proteins and mammalian carbonic anhydrase (iso)enzymes and calculation of
percentage of deviation of the calculated values from the literature

Protein Mol mass ln Mr Frictional Calculated mol mass (Mr)

(Mr (g mol\1))a coefficient
(f/fo) ln Mr M rb Deviation c ln Dmax
(%)

Ovalbumin 43 000 10.6690 1.18 10.8234 50 181 #16.7 4.6563

Bovine serum albumin 67 000 11.1125 1.34 11.2981 80 668 #20.4 4.3537
L-lactate dehydrogenase 140 000 11.8494 11.7695 129 249 !7.7 4.0532
Catalase 232 000 12.3545 1.27 12.0356 168 653 !27.3 3.8836
Ferritin 440 000 12.9945 1.40 12.7717 352 110 !20 3.4144
Thyroglobulin 669 000 13.4135 13.6949 886 379 #32 2.8259
Carbonic anhydrase
Bovine I 10.5233 37 171 4.8476
Bovine II 10.4904 35 968 4.8686
Bovine/rabbit III 38 000 10.5373 37 695 4.8387
Rabbit IV 10.5775 39 241 4.8131
Canine/human V 29 700 10.5462 38 032 4.8330
Arithmetic mean 10.5346 37 594 4.8404
a
Literature values.
b
The molecular mass of bovine carbonic anhydrase as estimated by sequence analysis was reported to be 28 980 while that of the
enzyme from mouse was found to be 29 068.
c
The molecular sizes calculated are compared with the literature data and the percentage deviation indicated.
Reproduced with permission from Rothe (1991).

intervals during electrophoresis, and the correspond- Eqn [16] summarizes this procedure:
ing time difference:
v (mm s\1)"(Dn!Dm);(tn!tm)\1"dD;dt\1
v (mm s\ )"(D1!D0);(t1!t0)\
1 1
[16]
v (mm s\1)"(D2!D1);(t2!t1)\1
v (mm s\1)"(D3!D2);(t3!t2)\1 where Dn (mm) equals the migration distance of a
 protein at a time tn (s) and Dm (mm) equals its
migration distance at a time tm (s) where tn'tm
v (mm s\1)"(DZ!DZ 1);(tZ!tZ 1)\1 (Figure 13).
\ \

Table 4 Calculated Stokes radius of marker proteins and mammalian carbonic anhydrase (iso)enzymes and calculation of percent-
age of deviation of calculated values from the literature

Protein Stokes radius ln RS Calculated Stokes radius (RS) ln Dmax

(RS (nm))a
ln RS RS Percentage
(nm) deviation b

Ovalbumin 3.05 !19.6081 !19.5992 3.08 #0.9 4.6563

Bovine serum albumin 3.55 !19.4563 !19.4285 3.65 #2.8 4.3537
L-Lactate dehydrogenase 4.20 !19.2881 !19.2590 4.32 #2.9 4.0532
Catalase 5.25 !19.0650 !19.1634 4.76 !9.3 3.8836
Ferritin 6.10 !18.9150 !18.8987 6.20 #1.6 3.4144
Thyroglobulin 8.50 !18.5832 !18.5668 8.64 #1.7 2.8259
Carbonic anhydrase
Bovine I !19.7070 2.76 4.8476
Bovine II !19.7189 2.73 4.8686
Bovine/rabbit III !19.7020 2.78 4.8387
Rabbit IV !19.6876 2.82 4.8131
Canine/human V !19.6988 2.78 4.8330
Arithmetic mean !19.7030 2.77 4.8404
a
Literature values.
b
The molecular sizes calculated are compared with the literature data and the percentage deviation indicated.
Reproduced with permission from Rothe (1991).
 II / ELECTROPHORESIS / Porosity Gradient Gels 1329

Table 5 Marker proteins that can be used to estimate the native Substituting eqns [19] and [20] into eqn [17] yields
molecular size of proteins the formula:
Marker protein Mr RS
v (mm s\1)"[((Tmax!);\1)!((T!);\1)]B
Carbonic anhydrase 30 000 2.43 [21]
Ovalbumin 45 000 3.05
Bovine serum albumin 67 000 3.55
-Galactosidase 116 000 4.23 which can be arranged to:
Lactate dehydrogenase 140 000 4.20
Catalase 232 000 5.25 v (mm s\1)";\B;(Tmax!T)B [22]
Ferritin 440 000 6.10
Thyroglobulin 669 000 8.50 and:

Mr (Da), Molecular mass; RS (nm), Stokes’ radius of proteins.
These markers can be taken when using PA gradients of 4}30%
T and a buffer of pH 8.4 (45 mmol L\1 Tris, 40 mmol L\1 boric
acid, 1.25 mmol L\1 EDTA}Na2, pH 8.4).
Correlating Migration Velocities and
Migration Distances
The migration velocities may be plotted against the
corresponding migration distances at the end of each
time interval to correlate migration velocities and
migration distances (Figure 13). The function by
which v and D are interrelated is best described by the
following exponential equation:
v (mm s\1)"(Dmax!D)B [17]
where , Dmax and  are constants, D (mm) is the
independent variable and v (mm s\1) the dependent
variable. Dmax represents the maximum migration
distance which a protein can cover, i.e. the migration
distance at which the migration velocity becomes
zero. If this point is reached then Dmax"D and:
v (mm s\1)"(D!D)B"0 [18]
Eqn [17] can be used to relate the apparent migration
velocity (v) of a protein to the PA concentration (T
(%)) that corresponds to the migration distance
travelled during a given period of electrophoresis.
When using a linear gel gradient, the PA concen-
tration and the gel length are interrelated by
eqn [19]:
D"\1 (T!) [19]
whilst Tmax (%), the stacking gel concentration, is
related to the maximum distance Dmax (mm) by eqn
[20]:
Dmax"\1 (Tmax!) [20]
v (mm s\1)"h;(Tmax!T)B [23]
where h";\B.
This derivation shows that, indeed, the apparent
migration velocity of a protein (v) is related by the
same function to the distance (D) as to the PA concen-
tration (T) it has reached in a linear pore gradient,
although the constants ( and Dmax, respectively,
h and Tmax) are different. The exponent  in both
equations, however, is the same.
Eqn [23] predicts that zero protein mobility (v"0)
results if the apparent gel concentration (T (%)) is
equal to the stacking gel concentration (Tmax (%)), i.e.
if T"Tmax. The apparent free electrophoretic mobil-
ity of a protein unhindered by the PA matrix (

(mm s\1)), can be calculated by simply extrapolating
its apparent mobility to zero T (%):
 (mm s\1)"h;(Tmax!0)B [24]
thus:
 (mm s\1)"h;TBmax [25]
This expression may be used to divide eqn [23] to
yield eqns [26] and [27]:
v;\1"(h;(Tmax!T)B);(h;TBmax)\1 [26]
which can be rewritten as:
1
v"[1!(T;T\
max)]B [27]
The value of the quotient (Tmax!T);T\ 1
max ranges
from one (T"0) to zero (T"Tmax) and thus the
value of v extends from the apparent free electro-
phoretic mobility () to zero.
This means that, in a linear PA gradient, the appar-
ent migration velocity (v) of a protein (migrating
under a constant electrical Reld strength) is equal to
its apparent free mobility () times a retardation
factor ([1!(T;Tmax)\1]B which depends on the PA
concentration (T) that the protein has just reached
and its exclusion limit (Tmax). This factor always takes
1330 II / ELECTROPHORESIS / Porosity Gradient Gels

Figure 13 (A) Estimation of the migration velocity of a protein (OVA, ovalbumin) in a linear PA gradient gel. Tn, migration distance at
a longer time of electrophoresis (tn); Tm, migration distance at a shorter time of electrophoresis (tm). (B) Plot of the resulting migration
velocities (v (mm s\1) versus the corresponding gel concentrations (T (%)) at the end of each time interval.

values between zero and one and increases exponenti-
ally with increasing gel concentrations.
In order to solve eqn [23] (v (mm s\1)"h;
(Tmax!T)B), the following sequence of calculations is
recommended:
1. determination of the maximum migration distance
of the protein under investigation from a plot of
ln (ln D) vs. t\1/2 (eqn [5])
2. computation of the maximum gel concentration
(Tmax) by use of eqn [20] (Dmax"\1(Tmax!));
3. calculation of the gel concentration equivalent
to the migration distances with eqn [19] (D"
\1(T!)), (the values of the constants  and
 may be obtained from a gel scan at 405 nm if
p-nitrophenol has been mixed into the more con-
centrated of the two solutions used to prepare the
gradient gel);
4. then the values of (Tmax!T) are calculated
5. Rnally the constants h and  in eqn [23] are cal-
culated by plotting ln v vs. ln (Tmax!T) and per-
forming a linear regression analysis with these data,
 II / ELECTROPHORESIS / Porosity Gradient Gels 1331

i.e. taking the logarithmized version of eqn [23]: At a Rrst approximation, the free electrophoretic
mobility, unhindered by a gel matrix (U
ln v";ln(Tmax!T)#ln h [28] (m2 V\1 s\1)), can be described by eqn [30]:

Calculation of the Free Electrophoretic Mobility U"(Z;);(6;; ;RS)\1

The free electrophoretic mobility (U (m2 V\1 s\1)) of
a protein results from its apparent free electrophotetic
mobility unhindered by the gel matrix ( (m s\1)) and
the electric Reld strength E (V m\1) acting on it:
U";E\1 (m s\1 (V m\1)\1"m2 V\1 s\1)
The apparent free electrophoretic mobility can be
obtained by applying eqn [25] ( (mm s\1)"
B ). The free electrophoretic mobilities of
h;Tmax
various marker proteins and Rve different mam-
malian carbonic anhydrases calculated by these pro-
cedures are listed in Table 6.
Computation of the Nett Charge
Estimation of the number of unit charges (Z) in
a nondenatured protein requires prior knowledge of
its Stokes radius (RS) and its apparent free electro-
phoretic mobility () or its free electrophoretic mobil-
ity (U). In addition to this, the ionic strength (I) and
viscosity ( ) of the buffer system used to estimate
Z and RS must be known. Time-dependent gradient
gel electrophoresis can be used to determine the
Stokes radius of a protein and its free electrophoretic
mobility.
(C (Pa s m)\1"m2 (V s\1) [30]
where Z is the number of unit charges (1);  is the unit
charge (protonic charge)"1.602;10\19 (C);
"3.142; is the dynamic viscosity of the medium
[29] (Pa s); RS is the Stokes radius (m) and the following
coherences 1 C"1 A s, 1 Pa"1 N m\2, 1 V A"
1 W and 1 W s"1 N m.
Since migration of proteins is studied in buffered
solutions, there are also positive and negative buffer
ions present, in addition to the protein ions. The
small ions of sign opposite to that of the protein, also
called counterions, are present in excess and to be
found in the vicinity of the protein molecules.
The electric Reld which drives the protein molecules
also acts on the counterions, but in the opposite
direction and since the migrating counterions drag
solvent along with them and the solvent in turn acts
on the protein, the nett effect is a secondary force
on the protein opposite in direction to the primary
force. The migration velocity of the protein molecules
towards the electric Reld may therefore be reduced
well below that predicted by eqn [30], an effect
known as the electrophoretic effect. This is why
eqn [30] must be corrected by a retardation factor (F),

Table 6 Free electrophoretic mobility (U ) and net negative charge (valence, Z ; charge, Q) of several marker proteins and carbonic
anhydrase (iso)enzymes from mammalia at pH 8.4

Protein U (m2 (V s)\1;10\9) Negative charge

I"0.529;103a
(mol m\3) I"0.1;103b Z Q (C molecule\1);10\19
(mol m\3)

Ovalbumin 3.45 5.99 13.06 20.92

Bovine serum albumin 4.40 7.85 22.42 35.92
Lactate dehydrogenase 3.27 6.00 22.63 35.25
Catalase 2.60 4.94 21.43 34.33
Ferritin 3.28 6.38 43.81 70.18
Thyroglobulin 2.78 5.62 68.46 109.67
CA I 1.58 2.69 4.93 7.90
CA II 1.17 1.99 3.58 5.74
CA III 1.05 1.79 3.32 5.32
CA IV 0.851 1.46 2.75 4.41
CA V 0.734 1.25 2.31 3.70

a
Ionic strength of electrophoretic buffer system.
b
Free electrophoretic mobility at ionic strength 0.1;103 (m2 (V s)\1).
CA, Carbonic anhydrase (iso)enzymes from mammalian erythrocytes: (bovine, I, II), bovine, rabbit (III), rabbit (IV) and canine, human
(V). Conditions of electrophoresis: linear polyacrylamide gradient from 4 to 27% T (acrylamide}Bis"24 : 1); gel length 73 mm; buffer
system 90 mmol L\1 Tris; 80 mmol L\1 boric acid; 1.25 mmol L\1 EDTA}Na2, pH 8.4 (I"529 (mol m\3); field strength: 41 V cm\1;
43C.
Reproduced with permission from Rothe (1991).
1332 II / ELECTROPHORESIS / Porosity Gradient Gels

the quantity of which depends on the composition
and strength of the small ions of the buffer used.
Henry proposed a method for computing this factor
using the formula:
F"(X1( ;RS));(1#( ;RS))\1
where X1 is a function of ;RS. Introducing this
factor into eqn [30] yields eqn [32]:
U"(Z;);(6;; ;RS)\1;(X1( ;RS))
;(1#( ;RS))\1 (m2 (V s\1))
The function X1( ;RS) is complicated but always
gives values between 1.0 and 1.5, as shown in
Figure 14. According to Henry, three different equa-
tions must be used to compute the values of the
function X1. If ;RS'24 then the Rrst of the three
equations indicated in Figure 14 must be used. When
;RS)5 the last of the three equations in Figure 14
is applied. In the range between the two border values
5 and 24, a linear equation is taken, which is also
given in Figure 14. It is somewhat difRcult to calcu-
late the X1 values when ;RS)5. Therefore,
Table 7 provides a number of values in the range of
;RS"0.01}5. Kappa ( (m\1)) represents the re-
ciprocal of the radius of the ion cloud, i.e. the radius
[31] of the cloud of counterions surrounding the protein.
Depending on the ionic composition, ionic strength
and temperature of the solution, acquires values
ranging from zero to inRnity, and at increasing ionic
strengths the value of increases whilst the radius of
the ionic cloud decreases and vice versa. In a salt-free
solution, "0 so that the electrophoretic mobility
[32] U is not inSuenced at all, whilst conversely it de-
creases permanently in solutions with increasing salt
concentrations. The value of kappa can be obtained
from the equation:
"[(2NA;2)
;(D0;D;k;T)\1]1/2;I1/2(m\1) [33]
where NA"6.025;1023 (mol\1);  is the unit charge
(protonic charge)"1.602;10\19 (C); D0 represents

Figure 14 Graphical representation of Henry’s function X1 ( RS). Depending on the value of RS three different equations must be
used to compute the values of X1. If RS'24 (case 1), the first of the three equations given is used. The second equation (case 2)
comes into use if 54 RS424 while the third equation (case 3) is applied if RS)5. In the latter case, Table 6 provides a number of
values. Reproduced with permission from Rothe (1991).
 II / ELECTROPHORESIS / Porosity Gradient Gels 1333

Table 7 Values of Henry’s function (X1( ;RS)) if ;RS (5 (cf. Figure 14)a

;RS log10 ( ;RS) X1 according to ;RS log10 ( ;RS) X1 according to

Overbeek’s modification Overbeek’s modification
of Henry’s equation of Henry’s equation

0.01 !2 1.0000062 1.95 0.2900346 1.0632127

0.05 !1.30103 1.0001452 2.00 0.30103 1.0651048
0.10 !1 1.0005451 2.05 0.3117539 1.0669887
0.15 !0.8239087 1.0011577 2.10 0.3222193 1.0688642
0.20 !0.69897 1.001951 2.15 0.3324385 1.0707308
0.25 !0.60206 1.0028994 2.20 0.3424227 1.0725882
0.30 !0.5228787 1.003982 2.25 0.3521825 1.0744361
0.35 !0.455932 1.005181 2.30 0.3617278 1.0762744
0.40 !0.39794 1.0064817 2.35 0.3710679 1.0781027
0.45 !0.3467875 1.0078712 2.40 0.3802112 1.0799208
0.50 !0.30103 1.0093387 2.45 0.3891661 1.0817286
0.55 !0.2596373 1.0108744 2.50 0.39794 1.0835259
0.60 !0.2218487 1.0124701 2.55 0.4065402 1.0853126
0.65 !0.1870866 1.0141185 2.60 0.4149733 1.0870886
0.70 !0.154902 1.0158129 2.65 0.4232459 1.0888537
0.75 !0.1249387 1.0175476 2.70 0.4313638 1.0906078
0.80 !0.09691 1.0193175 2.75 0.4393327 1.0923509
0.85 !0.0705811 1.0211181 2.80 0.447158 1.094083
0.90 !0.0457575 1.0229452 2.85 0.4548449 1.0958039
0.95 !0.0222764 1.0247952 2.90 0.462398 1.0975136
1.00 0.0 1.0266648 2.95 0.469822 1.0992121
1.05 0.0211893 1.028551 3.00 0.4771213 1.1008994
1.10 0.0413927 1.0304511 3.05 0.4842998 1.1025754
1.15 0.0606978 1.0323626 3.10 0.4913617 1.1042402
1.20 0.0791812 1.0342836 3.15 0.4983106 1.1058938
1.25 0.09691 1.0362118 3.20 0.50515 1.1075361
1.30 0.1139434 1.0381455 3.25 0.5118834 1.1091672
1.35 0.1303338 1.0400832 3.30 0.5185139 1.1107871
1.40 0.146128 1.0420233 3.35 0.5250448 1.1123958
1.45 0.161368 1.0439644 3.40 0.5314789 1.1139934
1.50 0.1760913 1.0459054 3.45 0.5378191 1.11558
1.55 0.1903317 1.0478451 3.50 0.544068 1.1171554
1.60 0.20412 1.0497825 3.55 0.5502284 1.1187199
1.65 0.2174839 1.0517167 3.60 0.5563025 1.1202734
1.70 0.2304489 1.0536469 3.65 0.5622929 1.1218159
1.75 0.243038 1.0555723 3.70 0.5682017 1.1233477
1.80 0.2552725 1.0574921 3.75 0.5740313 1.1248686
1.85 0.2671717 1.0594059 3.80 0.5797836 1.1263788
1.90 0.2787536 1.0613129 3.85 0.5854607 1.1278783
3.90 0.5910646 1.1293672 4.45 0.64836 1.1450647
3.95 0.5965971 1.1308456 4.50 0.6532125 1.1464318
4.00 0.60206 1.1323134 4.55 0.6580114 1.1477892
4.05 0.607455 1.1337709 4.60 0.6627578 1.1491371
4.10 0.6127839 1.135218 4.65 0.667453 1.1504754
4.15 0.6180481 1.1366549 4.70 0.6720979 1.1518043
4.20 0.6232493 1.1380816 4.75 0.6766936 1.1531238
4.25 0.6283889 1.1394981 4.80 0.6812412 1.1544341
4.30 0.6334685 1.1409047 4.85 0.6857417 1.1557352
4.35 0.6384893 1.1423012 4.90 0.6901961 1.1570272
4.40 0.6434527 1.1436879 4.95 0.6946052 1.1583101
5.00 0.69897 1.159584
a
Values were calculated using eqn [3] of Figure 14 (cf. Overbeek JTG (1950) Advances in Colloid Science, 3: 97}135). Tolerance of
values: 10\6, calculation of integral: 7 digits. Data from Rothe (1991).

the dielectric constant of vacuum"8.8542;10\12 stant"1.3805;10\23 (J K\1"N m K\1); T is ab-

(C V\1 m\1"C2 N\1 m\2); D is the temperature- solute temperature (K) and I is the ionic strength
dependent dielectric constant of water (without (mol m\3) of the buffer that was used for electrophor-
dimension, cf. Table 8), k is Boltzmann’s con- esis.
1334 II / ELECTROPHORESIS / Porosity Gradient Gels

Table 8 Dielectric constant (D) of water depending on the tem- or:

perature t (3C)
I"0.52875;103 (mol m\3) [41]
t (3C) D t (3C) D
Substituting this value into eqn [36] gives:
0 87.90 18 80.93
5 85.90 20 80.18
10 83.95 25 78.36 "1.02930525;108 (m mol\1)1/2
15 82.04 30 76.58
;(528.75)1/2 (mol m\3)1/2 [42]
Reproduced with permission from West (1976}1977).
which rearranges to:
By substituting these values into the equation one
obtains: "2.366842604;109 (m\1) [43]

"[(2;6.025;1023;(1.602;10\19)2);(8.8542 Taking ferritin as an example, with a Stokes radius

;10\12;1.3805;10\23)\1]1/2 (K m (mol)\1)1/2
;(I (D T)\1)1/2 (mol m\3 K\1)1/2
thus:
"1.590608013;1010 (K m mol\1)1/2
;(I (D T)\1)1/2 (mol m\3 K\1)1/2
of 6.20;10\9 (m), then ;RS"14.67. The log of
;RS equals 1.167 and using this value one obtains
[34] from the equation in case 2 shown in Figure 14,
a value of 1.293 for the function X1 ( ;RS).
Thus, inserting these values into eqn [31], it follows
that:
[35] F"(X1( ;RS));(1#( ;RS))\1

At a temperature of 53C (278 K), the dielectric "1.293;(1#14.67)\1"0.08251 [44]

constant of water is 85.90 (cf. Table 8). Inserting
both values into eqn [35] yields eqn [36]: To calculate the number of nett charges in ferritin,
eqn [32] must be solved for Z:
"1.02930525;108 (m mol\1)1/2
Z"((U;6;; ;RS);\1);((1#( ;RS))
;(I (mol m\ ) 3 1/2
[36]
;(X1( ;RS))\1) [45]
3
The ionic strength I (mol m\ ) is calculated using
the formula: From gradient gel electrophoresis results, the free
electrophoretic mobility of ferritin was calculated as
I"1/2ciZ2i (mol m\3) [37] U"3.28;10\9 (m2 V\1 s\1). Substituting this value,
that of factor F and the value for the temperature-
where ci (mol m\3) represents the concentrations of dependent dynamic viscosity ( (N s m\2)) of water
the ionic species of the buffer times their squared as taken from Table 9 into eqn [45], the number
charges (Zi). of unit charges that ferritin acquires under the
Taking , for example, a 90 mmol L\1 Tris, 80 mmol electrophoretic conditions indicated above can be
L\1 boric acid, 1.25 mmol L\1 EDTA-Na2 buffer of computed as:
pH 8.0, the ionic strength of this buffer is:
Z"(3.28;10\9;6;;1.519;10\3
2 3
I"1/2ciZ (mol dm\ )
i [38] ;6.20;10\9);(1.602;10\19)\1
thus: ;((1;0.08251)\1) [46]
2 2
I"1/2[(0.09;1 )#(3;0.08;1 )
which works out to:
#(0.08;(!3)2)#(2;0.00125;12)
#(0.00125;(!2)2)] [39] Z"44.05 [47]

which becomes: The actual charge on the molecule is given by Z; [C

molecule\1]"44.05;1.602;10\19"7.057;10\18
I"0.52875 (mol dm\3) [40] (Table 6).
 II / ELECTROPHORESIS / Porosity Gradient Gels 1335

Table 9 Dynamic viscosity ( (N s m\2)) of water depending on where U0.1 and U 0.1 (m2 V\1 s\1) represent the free
the temperature (t (3C))

electrophoretic mobilities at an ionic strength of
0.1 (mol m\3 ) and '0.1 (mol m\3 ) respectively; 0.1
t (3C ) (N s m\2 ) 10\3 t (3C ) (N s m\2 ) 10\3 
and 0.1 represent the reciprocal of the effective thick-
0 1.787 16 1.109 ness of the ionic cloud at ionic strength of
1 1.728 17 1.081 0.1 (mol m\3 ) and '0.1 (mol m\3 ) respectively and
2 1.671 18 1.053 RS (m) is the Stokes radius of the protein.
3 1.618 19 1.027
For the experimental conditions given earlier:
4 1.567 20 1.002
5 1.519 21 0.9779
6 1.472 22 0.9548 "1.02930525;108 (m mol\1)1/2
0.1
7 1.428 23 0.9325
8 1.386 24 0.9111 ;(0.1;103)1/2 (mol m\3)1/2 [49]
9 1.346 25 0.8904
10 1.307 26 0.8705 thus:
11 1.271 27 0.8513
12 1.235 28 0.8327 "1.02930525;109 (m\1) [50]
13 1.202 29 0.8148
14 1.169 30 0.7975 Taking ferritin as an example, for which U
15 1.139 (m2 V\1 s\1)"3.28;10\9 (at I"0.529;103
3
(mol m\ ), if "2.366842604;109 and RS"
Reproduced with permission from West (1976}1977). 9 0.1
6.20;10\ (m) are determined and substituted into eqn
[48], it follows that:

Evaluation of the Free Electrophoretic Mobility U0.1"((3.28;10\9 );(2.366842604;109;6.20

at an Ionic Strength of 0.1 mol L+1 ;10\9#2.4));(1.02930525;109
For reasons of comparability, the free electrophoretic ;6.20;10\9#2.4)\1 [51]
mobility obtained for a given set of experimental
conditions may be corrected to an effective mobility thus:
at an ionic strength of 0.1 mol L\1. This can be
achieved by substituting the relevant values into U0.1"6.377;10\9 (m2 V\1 s\1) [52]
Abramson’s equation:
The free electrophoretic mobilities at I"0.52875;103
(mol m\3) and at I"0.1;103 (mol m\3) of several
U0.1"(U 0.1( 0.1;RS#2.4));( ;RS#2.4)\1 marker proteins and some carbonic anhydrase
  0.1

[48] isozymes are listed in Table 10.

Table 10 Free electrophoretic mobility of ferritin in buffered solution

Experimental conditions Free mobility Reference

U (m2 V\1 s\1);10\9

Moving boundary method !6.1 Mazur et al. (1950)

I"0.1 (mol L\1), 0 (3C), pH 8.6
Agarose gel electrophoresis !10.5 Gosh et al. (1974)
I"0.05 (mol L\1), #20 (3C), pH 6.8
Disc electrophoresis Rodbard et al. (1971)
C"2%; 0 (3C), pH 8.88
I"0.0034 (mol L\1) !10.97
I"0.10 (mol L\1) !5.67
PA gradient gel electrophoresis Rothe (1991)
5}30 T (%), acrylamide}Bis"24 : 1; #4 (3C), pH 8.4a
I"0.529 (mol L\1) !3.28
I"0.10 (mol L\1) !6.38

a
Electrophoretic conditions; 90 mmol L\1 Tris, 80 mmol L\1 boric acid, 1.25 mmol L\1 EDTA}Na2, pH 8.4; separation distance 73 mm,
voltage gradient 41.1 (V cm\1)
References are given in Rothe (1991).
1336 II / ELECTROPHORESIS / Porosity Gradient Gels
The result of calculating the net protonic charge of
a protein of course remains unaffected whether the
ionic strength of the experiment or that of a buffer
strength of 0.1 mol L\1 is used.
Comprehensive Equation Describing
Electrophoretic Mobility of Proteins Migrating
in a Linear PA Gradient Gel
As explained above, the velocity with which a protein
migrates in a linear PA gradient gel depends on its
apparent free electrophoretic mobility times a retar-
dation factor (eqn [27]):
1 1
v";[1!(T;T\
max)]B (m s\ ) [27]
where v is the migration velocity (m s\1), T (%) is the
PA concentration which the migrating protein of mo-
bility v has reached and Tmax (%) represents the exclu-
sion limit of the migrating protein.
Since U";E\1 (m2 V\1 s\1), it follows that:
v;E\1"U;[1!(T;T\1 1 1
max)]B (m V\ s\ ) [53]
U is deRned by eqn [32] as equivalent to:
U"(Z;);(6;; ;RS)\1;(X1( ;RS))
;(1#( ;RS))\1 (m2 (V s)\1)
with the deRnitions given above.
Using all this information, a complete description
of the electrophoretic mobility of proteins migrating
in a linear PA gradient gel can then be given by the
equation:
v;E\1"(Z;);(6;; ;RS)\1;(X1( ;RS))
;(1#( ;RS))\1;[1!(T;T\ 1 2 1
max)]B (m (V s)\ )
[54]
with the deRnitions as given above.
Sodium Dodecyl sulfate Porosity
Gradient Gel Electrophoresis
Polyacrylamide gradient gels also offer greater possi-
bilities for the electrophoretic separation of proteins
in the presence of SDS. Porosity gradient gels have
a much higher resolving capacity, for example the
two chains of haemoglobin of Mr"15 126 and
15 866 Da, respectively, can be clearly separated in
a 3}30% T gradient gel. An 8% T continuous
PA}SDS gel does not exhibit this resolving capacity.
In SDS porosity gradient gel electrophoresis the
use of continuous buffer systems is recommended
(Table 11) since partial deloading of SDS}protein
complexes has been observed when the gel contained
SDS but not the electrode buffer. This results in a con-
fusing multitude of bands.
Estimation of Molecular Mass of Denatured
Proteins and Small Peptides
When SDS electrophoresis is performed in a linear PA
gradient gel of 3}30% T, a linear relationship can be
set up between the logarithm of the mol mass (log Mr)
and the log of the PA concentration (log T) reached
by proteins after a certain time of electrophoresis.
The validity of the corresponding relationship
log Mr"!a;log T#b has been conRrmed with
some 40 proteins between 14 and 950 kDa. In PA
gradient gels in the presence of SDS the molar mass
of both unreduced and 2-mercaptoethanol-reduced
proteins as well as the molar mass of glycoproteins
can be determined with the same accuracy ($5%,
Table 12). Ribonuclease and lysozyme binding
normal amounts of SDS migrate anomalously in ho-
mogeneous SDS gels but not in SDS PA gradient gels.
Papain and pepsin, which also bind only traces of
SDS, migrate regularly in SDS PA gradient gels.
The migration distance of proteins in linear SDS PA
gradient gels and their respective mol mass can also
be correlated by the equation:
log Mr"!a;(D#b [55]
where D (mm) is the migration distance. This rela-
tionship can be applied to SDS-complexed and re-
duced and to SDS-complexed nonreduced proteins, to
glycoproteins and to carbohydrate-free proteins
(Figure 15). The relationship is not affected by the
buffer system, the concentration of the cross-linker
within 1}8% C or the concentration range of the
gradient within 3}30% T at the commonly used gel
length of 8}15 cm. The value of the constants a and b,
on the other hand, are changed when the experi-
mental parameters are altered. If SDS electrophoresis
is performed in a linear gradient gel of
approximately 6}27% T, the relationship log Mr"
!a;(D#b is practically independent of the time
of electrophoresis. This means that the molecular
mass estimation can be made when the best resolu-
tion of a set of proteins has been obtained. It is not
necessary to wait until the proteins have reached their
exclusion pore size. On the contrary, under pro-
longed electrophoresis protein}SDS complexes can
reach a pore size where the complexing SDS is
stripped off the protein molecules which leads to
erroneous banding patterns. This is particularly
 II / ELECTROPHORESIS / Porosity Gradient Gels 1337

Table 11 Gel and buffer systems used in SDS PA gradient gel electrophoresis to separate denatured proteins

PA range (%T ) Gel shape Buffer systems Current or Running Correlation Notes Authors
(acrylamide}Bis) (dimensions voltage time (h) (Mr range
(mm)) Gel buffer Electrode buffer per gel (kDa))

3}30 Column 0.1 mol L\1 0.1 mol L\1 4 mA 24 * a Exposito and
(30 : 0.8) (150;6) Na}phosphate, Na}phosphate, (12}125) Obijeski
0.1% SDS, 0.1% SDS, pH (1976)
5}15% (v/v) 7.0
glycerol, pH 7.0
3}30 Slab gel 10.75 g Tris, 0.01 mol L\1 40 V 16 log Mr vs. b Lambin et al.
(9.62 : 0.38) (length: 80) 5.04 g boric acid Na-phosphate, log T (1976),
0.93 g 1% SDS, 1% 2- (13}950) Lambin
EDTA}Na2, mercaptoethanol, (1978)
pH 7.2 pH 7.2
1.5}40 Microcolumn 0.1 mol L\1 29 g glycine plus 60 V 2 log Mr vs. c RuK chel et al.
(12.57 : 1) i.d. 0.43, Na}phosphate, Tris to pH 8.4, 1 g RF (1974)
length 15 pH 7.2, SDS, (13}300)
0.1% SDS or H2O to 1000 mL
0.35 mol L\1
Tris-sulfate, 0.1%
SDS, pH 8.5, or
0.05 mol L\1
Tris}glycine,
0.1% SDS, pH
8.4, or
0.065 mol L\1
Tris}borate, 0.1%
SDS, pH 9.3
1.5}40 Microcolumn 4 g Tris and 29 g glycine plus 60 V 0.33 log Mr vs. c RuK chel et al.
(12.57 : 1) (i.d. 0.43, H2SO4 to pH 8.4, Tris to pH 8.4, 1 g RF (13}300) (1974)
length 15) H2O to 10 mL SDS, H2O to
1000 mL
3}30 Slab gel 0.04 mol L\1 Tris, Same as gel 150 V 0.5}8 log Mr vs. d Rothe (1982)
(28 : 1) (width 80, 0.02 mol L\1 buffer (D (13}950)
length 80, Na}acetate,
thickness 1) 0.02 mol L\1
Na}EDTA, pH
7.4, 0.2% SDS

a
Gels were stored at room temperature before use in a solution which contained 0.1 mol L\1 Na}phosphate, 0.01% SDS, 15%
glycerol, 2 mmol L\1 EDTA}Na2 and 0.01% NaN3. Samples were dissolved at 1003C for 3 min in 0.01 phosphate buffer, pH 7,
containing 2.5% SDS, 5% 2-mercaptoethanol, 10% glycerol and 0.005% Bromophenol blue. On each column 20}100
g protein was
loaded.
b
T(%) g acrylamide plus g Bis per 100 mL solvent. Protein samples (0.5 mg mL\1) were incubated in 0.01 mol L\1 phosphate buffer,
containing 1% SDS, pH 7.2 for 3 min in a 1003C bath; for cleavage of disulfide bridges 1% 2-mercaptoethanol was added. The
% T concentration reached by each protein after electrophoresis was determined and log T plotted versus log mol mass.
c
Resolution was found to be better in discontinuous than in continuous buffer systems. Samples (1 mg protein mL\1) were treated for
2 min at 1003C with 1% SDS and 1% 2-mercaptoethanol in 0.035 mol L\1 Tris}sulfate, pH 8.6, 0.35 mol L\1 Tris}sulfate, pH 8.6 or
0.1 mol L\1 phosphate. Complete removal of SDS from proteins can be achieved with SDS-free electrode buffers. The activity of
-galactosidase denatured with SDS and separated on an SDS-free PA gradient gel could be restored to 10%.
d
(D, square root of migration distance (D (mm)). Re-evaluation of the data from Lambin (1978), Lasky (1978) and Poduslo and
Rodbard (1980) confirmed the validity of the log Mr!(D relationship, found when evaluating time-dependent SDS-porosity gradient
gel electrophoresis using marker proteins in the range of 10}330 kDa.
References as given in Rothe and Maurer (1986). Reproduced with permission from Rothe and Maurer (1986).

true when in an alkaline buffer system the upper of 1.4}10 kDa cannot be resolved. Separation is
electrode buffer contains no SDS. possible, however, in 10}18% T gels in the presence
In SDS electrophoresis with linear PA gradients of 0.1% SDS and 7 mol L\1 urea (cf. Tables 13
ranging from 3 to 30% T, polypeptides in the range and 14).
1338 II / ELECTROPHORESIS / Porosity Gradient Gels

Table 12 Separation characteristics of some proteins in SDS PA gel electrophoresis and deviation of calculated mol masses from
those given in the literature
No. Protein Mr (Da) 3}30%T, C"8.4%a 3}30%T, C"3.8%b
(literature value)
D Mr c T Mr d D Mr c T Mr d
(mm) $% (%) $% (mm) $% (%) $%
1 Prealbumin 13 745 51.5 !0.4 20.7 !4.0 56 !1.7 22.4 !6.8
2 Lysozyme 14 314 53.5 !16.5 21.4 !20.4 51.5 #16.9 20.8 #13.6
3 Ribonuclease B 14 700 52 !10.0 20.9 !14.0 55.5 !6.0 22.2 !10.3
4 Haemoglobin 15 500 51 !8.6 20.5 !11.2 55 !8.7 22 !12.3
5 Avidin 16 000 49.5 !1.7 20 !4.2 51 #7.1 20.6 #4.8
6 Soybean trypsin inhibitor 20 095 47 !6.6 19.2 !9.2 50 #10.4 20.3 !12.7
7 Papain 23 426 44.5 !4.0 18.3 !4.8 48 #15.1 19.6 !16.5
8 -chain of IgG 23 500
9 Chymotrypsinogen A 25 666 43.5 !5.6 18 !7.0 44 #4.8 18.2 !4.8
10 Carbonic anhydrase B 28 739 41 #1.8 17.1 #2.2 42 #5.5 17.5 !4.7
11 Carboxypeptidase A 34 409 40 !8.1 16.7 !6.3 39.5 !9.5 16.7 !9.0
12 Pepsin 34 700 37.5 #11.0 15.9 #12.5 37.5 #0.4 16 #1.8
13 Glycerol-3-phosphate 35 700 37.5 #7.9 15.9 #9.4 36 #6.4 15.5 #7.9
dehydrogenase
14 Lactate dehydrogenase 36 180 37.5 #6.4 15.9 #7.9 37 !1.0 15.8 #1.0
15 Aldolase 38 994 36 #11.5 15.4 #13.1 35 #3.2 15.1 #6.0
16 Alcohol dehydrogenase 39 805 35.5 #13.8 15.2 #16.4 34.5 #4.2 14.9 #7.7
17 1-Acid glycoprotein 40 000 35 #18.0 15 #21.7 32 #20.6 14.1 #23.9
18 Ovalbumin 43 000 35.5 #5.4 15.2 #7.8 35.5 !9.1 15.3 !7.2
19 Fibrinogen chain 47 000
20 Glutamate oxalacetate 50 000 32.5 #16.7 14.2 #19.2 29 #16.6 13 #21.9
transaminase
21 Heavy chain IgG 50 000
22 Fibrinogen  chain 56 000
23 Catalase 57 500 32 #5.9 14 #9.1 29 #1.4 13 #6.0
24 Fibrinogen  chain 63 500
25 Albumin monomer 66 290 31.5 #10.7 13.8 #14.9 28 #8.2 12.7 !12.3
26 Heavy chain IgM 72 000
27 Transferrin 76 000 30.5 !8.6 13.5 !6.0 24 #7.6 11.3 #12.4
28 Plasminogen 81 000 29 !1.8 13 #0.7 23 #8.5 11 #12.2
29 Phosphorylase b 96 800 25.5 #14.2 11.8 #17.1 21 #5.3 10.3 #8.9
30 Ceruloplasmin 124 000 23.5 #8.8 11.1 #11.6 17 #13.2 8.9 #16.2
31 Albumin, dimer 132 580 24 !3.3 11.2 #1.4 17.5 #1.6 9 #6.2
32 Immunoglobulin G 150 000 21.5 #10.6 10.4 #13.4 15 #11.4 8.2 #13.4
33 Immunoglobulin A 160 000 22 #1.6 10.6 #0.1 14 #14.5 7.8 #17.2
34 Reduced 2-macroglobulin 190 000
35 Albumin, trimer 198 870 19.75 #0.8 9.8 #2.6 12 #11.8 7.1 #12.6
36 Immunoglobulin A 320 000 16 !3.1 8.5 !3.5 9 !4.0 6.1 !8.4
37 Thyroglobulin 330 000 16.5 !11.6 8.7 !12.4 11 !25.4 6.8 !26.6
38 Fibrinogen 340 000 15 #3.3 8.2 #0.4 8.5 !4.2 5.9 !8.8
39 2-Macroglobulin 380 000 15.5 !13.2 8.3 !13.1 8 !9.0 5.8 !16.0
40 Immunoglobulin A, trimer 480 000 13 !4.8 7.5 !9.5 6.5 !12.4 5.2 !20.6
41 Immunoglobulin M 950 000 8.75 !9.1 6 !20.1
Average % deviation $7.8 $9.6 $8.7 $11.2
Lambin (1978) $5.9 $7.4
a
The gel buffer contained no 2-mercaptoethanol (gel length 78.5 mm).
b
Gel buffer with 2-mercaptoethanol (gel length 81 mm), gel and electrode buffer as well as conditions of electrophoresis as given in
Table 8. Mr (Da), mol mass; D (mm), migration distance; T (%), g acrylamide plus g Bis per 100 mL, as reached by a protein.
c
Mr $%, %-deviation of calculated mol mass from the literature value using the relationship log Mr"a;(D#b.
d
Mr $%, %-deviation of calculated mol mass from the literature value using the relationship log Mr"a;(T#b.
Reproduced with permission from Rothe and Maurer (1986).

Separation of Urinary Proteins and Diagnosis
of Proteinurias
Diagnosis of pathological urinary proRles and estima-
tion of the molecular size of the corresponding
proteins is possible by SDS PA gradient gel electro-
phoresis under nonreducing conditions. Protein
patterns may be estimated in micro-sized
(43;50;0.45 mm) SDS gradient gels of 8}25%
T Rxed to a plastic backing (GelBond2+) as they are
 II / ELECTROPHORESIS / Porosity Gradient Gels 1339

Figure 15 Migration distances of denatured proteins and protein subunits obtained by SDS PA gradient gel electrophoresis. The
logarithm of the molecular mass of proteins (log Mr) is linearly correlated to the square root of the PA concentration ((% T ) which they
reached upon electrophoresis. Also, log Mr is linearly correlated to the square root of the migration distance ((D (D (mm)) which
proteins reached upon electrophoresis. Reproduced with permission from Rothe (1994).

commercially available together with a suitable hori-
zontal electrophoretic apparatus (Phast system) and
an automated silver staining device (Amersham Phar-
macia Biotech). The method has the advantage that
urine samples need not be concentrated or desalted
before electrophoresis. Samples may be stored frozen
at !203C after addition of sodium azide and after
particulate removal by centrifugation. Samples with
protein concentrations above 0.30 mg mL\1 must be
diluted. Proteins must not be reduced (e.g. with 2-
mercaptoethanol) since under SDS and nonreducing
conditions the quarternary structure of all major
serum proteins excreted in urine is unaffected, except
haemoglobin which is split into its monomers and
dimers. Figure 16 shows some selected protein pat-
terns of renal malfunctions.
1340 II / ELECTROPHORESIS / Porosity Gradient Gels

Table 13 Gel and buffer system used to separate small peptides in SDS PA gradient gel electrophoresis

Acrylamide Bis Gel buffer (pH) Electrode buffer (pH) Correlation Authors
(g 100 mL\1) (g 100 mL\1) (Mr (Da) range)

10}18 0.5}0.9 Stacking gel: 5% acrylamide, 0.05 mol L\1 Tris, log Mr vs. D Hashimoto et al. (1983),
0.13% Bis, 0.067 mol L\1 0.38 mol L\1 (1400}17 000) Laemmli (1970)
Tris}HCl, pH 6.8, 0.1% SDS, glycine, 0.1% SDS,
0.067% ammonium persulfate pH 8.5
and 0.067% TEMED; separation
gel: 0.45 mol L\1 Tris}HCl, pH
6.9, 0.1% SDS, 0.05%
ammonium persulfate, 0.05%
TEMED, 7 mol L\1 urea

Mr (Da), mol mass; D (mm), migration distance; TEMED, N,N,N
,N
-tetramethylethylenediamine. The buffer solution containing 10%
acrylamide (0.5% Bis) contains no sucrose while the buffer solution containing 18% acrylamide (0.9% Bis) contains 10% (w/v) of
sucrose. The PAA concentration and the sucrose concentration increase linearly from top to bottom. The system can also be used to
separate lipopolysaccharides and phospholipids. The addition of iodoacetamide to samples prior to electrophoresis eliminated artifacts
currently observed in silver staining of protein bands. Log Mr correlates linearly with migration distance (D (mm)) in the mon mass range
of 1.4 (kDa) to 17 (kDa). Flat gels of the dimensions 150;140 (height);1 (mm) were used. Gels were run for at least 15 h at 120 V.
Samples were heated for 2 min at 1003C in a sample buffer containing 10% sucrose, 0.0625 M Tris-HCl, pH 6.8, 2% SDS, 10 mM
dithiothreitol and 0.0025% Bromophenol blue (if necessary they were treated with iodoacetamide). Reproduced with permission from
Rothe and Maurer (1986).

Diagnosis of the following proteinurias is possible:

Table 14 Mol masses of polypeptides and peptides employed 1. Proteinuria in the normal range of total protein
for urea}SDS gel electrophoresis 2. Orthostatic (postural) proteinuria
3. Post-renal proteinurias
Protein Mol mass (Da)
(a) Post-renal haematuria
Literature value a Computed b (b) Local excretion of proteins
4. Bence}Jones proteinuria
Ovalbumin 46 000 5. Lower and upper urinary tract infections: cystitis
Carboxypeptidase A 34 500 c and pyelonephritis
Myoglobin 17 200
Myoglobin I#II 14 900 6. Diabetes mellitus
Cytochrome c 12 300 c
Myoglobin I 8270
Cytochrome c I 7760 Concluding Remarks
Myoglobin II 6420
Gradient gel electrophoresis has may advantages over
Bovine trypsin inhibitor 6160 c
Adrenocorticotrophic hormone 4550 6500 conventional gel electrophoresis. Gradient range and
Insulin 5700 course can be adapted to any individual separation
Insulin B chain 3400 problem, and protein bands are much sharper than in
Insulin A chain 2300 Cellogel or starch gel electrophoresis. So far, for
Glucagon 3460 1800
example, more than 20 enzyme bands of an enzyme
Cytochrome c II 2780
Myoglobin III 2550 system such as plant acid phosphatase have been
Cytochrome c III 1810 clearly resolved and genetically interpreted, and
Bacitracin 1400 crude enzyme extracts can be used as the enzyme
Polymyxin B 1225 2200 source, provided a speciRc detection (staining) system
a
is available. The disadvantages are few compared to
Values as cited by Swank and Munkres (1971).
b
Values calculated by Swank and Munkres (1971) using least- conventional gel electrophoresis, such as availability
squares regression analysis and assuming a linear correlation of gradient gel, a load of two to three times more
between log Mr (Mr, mol mass (Da)) and migration distance enzyme activity per cm2 of gel cross-section as com-
D (mm). pared to starch gels, and the exclusive migration of
c
The mol masses of these proteins also deviate considerably if
proteins towards the cathode (anode), whereas in
a straight line in a log Mr vs. D plot is drawn through the points of
carboxypeptidase A and bacitracin. Cellogel and starch gel electrophoresis both cathodi-
References as given in Rothe and Maurer (1986). Reproduced cally and anodically migrating proteins can be detected
with permission from Rothe and Maurer (1986). within the matrix.
 II / ELECTROPHORESIS / Porosity Gradient Gels 1341

Figure 16 Separation of urinary proteins by macro SDS PA gradient gel electrophoresis. Gel: 4}20% T. Running conditions: 3 h
at 350 V, 50 mA. Samples: up to 50
L urine. (A) and (B) show urine samples from paediatric patients with pyelonephritis at various
stages of follow-up. Series A: 1, acute phase: 2, and 3, urine taken at weekly intervals; 4, reinfection (acute phase); 5, 1 week follow-up.
Series B: 1, acute phase; 2}4, follow-up at weekly intervals (note blood contamination in 2 and 3, - and -globin chains at 16 kDa),
5, reinfection, 6, and 7, weekly follow-up. Black dot and vertical arrow on the right side of the gel represent the application point
and migration direction, respectively. # Gel polarity. Alb., albumin. Reproduced with permission from Bianchi-Bosisio et al.
(1991).

In addition to pure PA gels, matrices of mixed
polymers can be used for porosity gradients. How-
ever, this possibility has been rarely used, although it
could extend the separation possibilities.
PA gradients are widely used to determine the mo-
lecular mass of SDS-denatured proteins, because this
method offers a larger separation range and a much
better resolution of protein bands. However, native,
time-dependent PA gradient gel electrophoresis has
much more possibilities to offer, such as differenti-
ation between size and charge isomers, determination
of the molecular size of native proteins and (iso)en-
zymes (Mr, RS), estimation of the molecular excentric-
ity (f/fo), and calculation of the net negative charge at
a given pH value. Using these possibilities the evolu-
tion of homologous proteins in related animal and
plant species can be studied as well as the net charge
of isozymes from different cells compartments.
Further Reading
Abramson HA (1933) Electrokinetic phenomena. Journal
of General Physiology 16: 593}603.
Altland K and Altland A (1984) Forming reproducible
density and solute gradients by computer-controlled co-
operation of stepmotor-driven burettes. Electrophoresis
5: 143}147.
Bianchi-Bosisio A, D’Agrosa F, Gaboardi F et al. (1991)
Review, Sodium dodecyl sulphate electrophoresis of uri-
nary proteins. Journal of Chromatography 569: 243}260.
Chiari M, Campoleoni A, Conti P et al. (1996) Electro-
phoretic separation of biopolymers in a matrix of poly-
acrylamide covalently linked to agarose. Electrophoresis
17: 473}478.
Felgenhauer K (1974) Evaluation of molecular size by gel
electrophoretic techniques. Hoppe-Seyler’s Zeitung fu( r
Physiologische Chemie 355: 1281}1290.
Henry DC (1931) The cataphoresis of suspended particles.
Part I. The equation of cataphoresis. Proceedings of the
Royal Society A 133: 106}140.
Horvath ZS, Corthals GL, Wrigley CW and Margolis
J (1994) Multifunctional apparatus for electrokinetic
processing of proteins. Electrophoresis 15: 968}971.
Lasky M (1978) Protein molecular weight determination
using polyacrylamide gradient gels in the presence and
absence of sodium dodecyl sulfate. In: Catsimpoolas
N (ed.) Electrophoresis, pp. 195}210, Amsterdam:
North Holland.
Righetti PG and Tudor G (1981) Isoelectric points and
molecular weights of proteins } a new table. Journal of
Chromatography 220: 115}194.
Rothe GM (1982) Applicability of the log MM-(D rela-
tionship to linear polyacrylamide gradient gel elec-
trophoresis under a wide range of experimental condi-
tions. Electrophoresis 3: 255}263.
Rothe GM (1991) Determination of the size, isomeric na-
ture and net charge of enzymes by pore gradient gel
electrophoresis. In: Chrambach A, Dunn MJ and Radola
BJ (eds) Advances in Electrophoresis, vol. 4, pp.
251}358. Weinheim: VCH.
Rothe GM (1994) Electrophoresis of Enzymes: Laboratory
Methods, p. 307. Berlin: Springer Verlag.
1342 II / ELECTROPHORESIS / Proteins, Detection of
Rothe GM (1994) Molecular relationship and possible
evolution of 15 enzyme loci in Rve Pinaceae species. In:
Zin-Suh K and Hattener H (eds) Conservation and Ma-
nipulation of Genetic Resources in Forestry, pp.
161}141. Seoul: Kwang Moon Kag.
Rothe GM and Maurer WD (1986) One dimensional
PAA-gel electrophoretic techniques to separate func-
tional and denatured proteins. In: Dunn MJ (ed.)
Proteins, Detection of
M. J. Dunn, National Heart and Lung Institute,
Imperial College School of Medicine, Heart Science
Centre, Harefield Hospital, UK
Copyright ^ 2000 Academic Press
Introduction
After polyacrylamide gel electrophoresis, it is essen-
tial that separated protein zones be detected for sub-
sequent analysis, whether this is to be done by simple
visual inspection or by quantitative computerized
densitometry. In the early days of electrophoresis,
methods for the detection of separated zones (ultra-
violet absorption, Schlieren optics) were limited and
insensitive. The subsequent development of organic
dyes able to react with proteins made stains such as
Bromophenol Blue and Amido Black 10B popular. In
particular, Coomassie Brilliant Blue was for many
years the method of choice for protein detection fol-
lowing gel electrophoresis owing to its relatively high
sensitivity. However, the need for increased sensitiv-
ity resulted in the development of a group of staining
methods based on the use of silver (approximately
0.1 ng of protein per band). Recently, there has been
a renewed interest in the use of Suorescent methods of
protein detection as they provide high sensitivity
equivalent to silver staining combined with excellent
linearity and extended dynamic range. Detection
methods based on the use of radiolabelling also pro-
vide high sensitivity but cannot be applied in all
situations. Finally, methods are available for the de-
tection of groups of proteins with speciRc post-trans-
lational modiRcations, for example glycoproteins,
phosphoproteins and lipoproteins.
Fixation
After electrophoresis is complete, the gel is removed
from the apparatus for localization of the separated
zones. Procedures have been described for the direct
visualization of unRxed proteins within gels. How-
Gel Electrophoresis of Proteins, pp. 37}140. Bristol:
Wright.
Tanford C (1961) Physical Chemistry of Macromolecules,
p. 417, pp. 425}532. New York: John Wiley.
Wedler G (1982) Lehrbuch der physikalischen Chemie, pp.
172}212. Weinheim: Verlag Chemie.
West RC (eds) (1976}1977) Handbook of Chemistry and
Physics, 57th edn. Cleveland, Ohio: CRC Press.
ever, for the majority of protein detection methods it
is necessary to precipitate and immobilize (i.e. ‘Rx’)
the separated proteins within the gel and to remove
any nonprotein components which might interfere
with subsequent staining. Gels that are to be used for
visualization of enzymatic activity of the separated
proteins must not be Rxed. The best general purpose
Rxative is 20% w/v trichloroacetic acid (TCA) as it
gives effective precipitation of most proteins. Acid
methanol (or ethanol), typically a solution containing
10% v/v acetic acid, 45% v/v methanol, and 45%
deionized water, is often used for gel Rxation, but it
should be noted that this can be ineffective for small
proteins, basic proteins and glycoproteins. Aqueous
solutions of reagents such as 5% w/v formaldehyde
or 2% w/v glutaraldehyde can be used to cross-link
proteins covalently to the gel matrix, but this is not
a commonly used approach.
Coomassie Brilliant Blue
The most popular general protein-staining proced-
ures following gel electrophoresis are based on the
use of the non-polar, sulfated triphenylmethane
Coomassie stains, developed for the textile industry,
Coomassie Brilliant Blue (CBB) R-250 is most often
used and requires an acidic medium for electrostatic
interaction between the dye molecules and the amino
groups of proteins. Staining is usually carried out
using 0.1% w/v CBB R-250 in the same acid meth-
anol solution used for Rxation (10% acetic acid, 45%
methanol). Depending on gel thickness and poly-
acrylamide concentration, staining can take from
30 min to several hours. In practice, it is often conve-
nient to stain the gel overnight and then destain it by
several changes in the same acid methanol solution
until intense blue protein zones can be seen against
a clear background. This method is able to detect
a minimum of around 100 ng protein per band
(Figure 1), so that for complex mixtures containing
several hundred components, it is necessary to load
relatively high amounts of total protein ('50
g).
 II / ELECTROPHORESIS / Proteins, Detection of
Figure 1 SDS}PAGE separation of human heart proteins
(lanes b}g). Lane (a) contains the molecular weight marker pro-
teins and the scale at the left indicates protein size in kDa. The gel
has been stained with Coomassie Brilliant Blue R-250. The
sample protein loadings were (b) 1
g, (c) 5
g, (d) 10
g,
(e) 25
g, (f) 50
g, (g) 100
g.
More sensitive staining (down to 10 ng protein per
band) can be achieved using the dimethylated form of
the dye, CBB G-250, as a 0.1% w/v colloidal disper-
sion in 2% w/v phosphoric acid, 10% w/v am-
monium sulfate, and 20% v/v methanol. An addi-
tional advantage of this method is that the colloidal
dye only binds to the separated proteins as it is unable
to penetrate the gel matrix. This means that no de-
staining step is required and the intensity of staining
can be controlled by visual inspection during the
staining process. Related dyes such as Acid Violet 17,
Serva Violet 49 and Fast Green FCF also form col-
loids in strongly acidic solutions and stain proteins
in gels with low background.
Silver Staining
Silver has been known to be able to develop images
for over two hundred years, Rrst being usefully ex-
ploited in photography and then rapidly adopted for
use in histological staining procedures. The ability of
silver to detect proteins following their separation by
gel electrophoresis was Rrst recognized by Merril and
his colleagues in 1979. Subsequently, more than
a hundred silver-staining procedures have been de-
scribed and this group of methods has become the
standard approach for the sensitive detection of gel-
separated proteins. However, certain classes of pro-
teins, such as calcium-binding proteins and glycopro-
teins, stain rather poorly, with an inverse relationship
1343
between the intensity of silver staining and the pro-
portion of the molecule that is composed of carbohy-
drate. Pre-staining with cationic dyes prior to silver
staining can signiRcantly improve the sensitivity of
detection of glycoproteins.
Depending on the method, silver staining is be-
tween ten and a hundred times more sensitive than
staining with CBB R-250, and is able to detect low
nanogram amounts of protein. There can be prob-
lems in using silver staining as a quantitative proced-
ure as it is known to be non-stoichiometric. However,
staining intensity is linear over a 40-fold range, com-
paring well with the 20-fold linear range of CBB
R-250. Above this limit, the stain becomes non-linear,
resulting in saturation and even negative staining of
bands and spots at very high protein concentrations,
making quantitation of such protein zones imposs-
ible. In a two-dimensional electrophoresis study of
human leukocyte proteins, over 200 spots were ob-
served to have coefRcients of variation less than or
equal to 15% when data from replicate patterns were
analysed. In dilution experiments, the majority
('80%) of the proteins were found to have a linear
relationship between the amount of protein loaded
and the spot volume. An additional problem with the
quantitation of silver staining is that the relationship
between staining intensity and protein concentration
may be different for each protein. However, it is often
forgotten that this is also the case for staining with
CBB R-250.
All silver-staining procedures depend on the reduc-
tion of ionic silver to its metallic form, but the precise
mechanism involved in the staining of proteins has
not been fully established. It has been proposed that
silver cations complex with protein amino groups,
particularly the
-amino group of lysine, and with
sulfur residues of cysteine and methionine. However,
staining cannot be attributed exclusively to speciRc
amino groups suggesting that some other component
of protein structure is also responsible for differential
protein staining.
Procedures for silver staining can be grouped into
two main types depending on the chemical state of
the silver when used for impregnating the gel. The
Rrst group comprises alkaline methods based on the
use of ammoniacal silver or diamine solution, pre-
pared by adding silver nitrate to sodium-ammonium
hydroxide mixture. Copper can be included in these
diamine methods to give increased sensitivity, prob-
ably by a mechanism similar to the Biuret reaction.
The silver ions complexed to proteins in the gel are
then developed by reduction to metallic silver with
formaldehyde in an acidiRed environment, usually
using citric acid. In the alternative group of methods,
silver nitrate in a weakly acidic (around pH 6.0)
1344 II / ELECTROPHORESIS / Proteins, Detection of
solution is used for gel impregnation. Development is
subsequently achieved by the selective reduction of
ionic silver to metallic silver by formaldehyde made
alkaline with either sodium carbonate or sodium hy-
droxide. Any free silver is washed out of the gel prior
to development to prevent precipitation of silver ox-
ide that would result in high background staining.
The majority of silver staining procedures are
monochromatic, resulting in dark brown to black
protein zones. However, if the development time
is extended with saturation of the zones of highest
protein concentration, then colour effects can be pro-
duced. In a comparative study of several methods
based on both the silver diamine and silver nitrate
approaches, the most rapid procedures were found to
be generally less sensitive than those which were more
time-consuming. The use of glutaraldehyde pre-treat-
ment of the gel and silver diamine as the silvering
agent were found to be the most sensitive and
example of a gel stained with a method of this type is
shown in Figure 2.
Increasingly, proteins are being visualized in gels
for subsequent identiRcation and characterization by
techniques such as mass spectrometry. In this case,
glutaraldehyde cannot be used and silver-staining
protocols that omit this reagent must be used. How-
ever, this modiRcation results in a decrease in sensitiv-
ity and uniformity of staining as well as an increase in
background.
It is a common experience that silver-staining pro-
cedures can give rise to problems when based on the
Figure 2 SDS}PAGE separation of human heart proteins
(lanes b}g). Lane (a) contains the molecular weight marker pro-
teins and the scale at the left indicates protein size in kDa. The gel
has been silver stained. The sample protein loadings were
(b) 1
g, (c) 5
g, (d) 10
g, (e) 25
g, (f) 50
g, (g) 100
g.
use of laboratory-prepared reagents. If care is not
taken with the use of high-purity water, reagents and
glassware, then problems of high background stain-
ing, surface ‘mirror’ effects and poor reproducibility
can be experienced. Many of these problems can be
alleviated using one of the commercially available
silver-staining kits (for example from Amersham-
Pharmacia Biotech, Bio-Rad Laboratories, Rich-
mond, CA, USA).
Reverse Stains
One disadvantage of the standard protocols for stain-
ing with Coomassie Blue dyes and silver is that it is
essential to use a Rxation step prior to staining. Un-
fortunately, this can result in reduced recovery of
proteins from the gel for subsequent chemical charac-
terization. Reverse stains have been developed to spe-
ciRcally overcome this problem. The result of these
stains is a semi-opaque background on the gel sur-
face, while the proteins are visible as transparent
zones using back-lighting. The process of staining is
rapid, requiring generally between 5 and 15 min.
After staining, the proteins can be eluted after chela-
tion of the metal ions with agents such as EDTA. It
should be noted that reverse stains are not suitable for
quantitative applications.
A variety of reverse-stain methods suitable for vis-
ualizing proteins after SDS}PAGE have been de-
scribed. The most popular methods have been those
using potassium chloride, copper chloride and zinc
chloride, with the last being the most sensitive. The
zinc imidazole-staining method is quite sensitive,
with a limit of detection of around 10 ng protein per
band. In the presence of imidazole, free or weakly
bound zinc ions are readily precipitated as zinc
imidazole, while tightly bound ions associated with
proteins do not precipitate. This results in clear pro-
tein zones on an opaque background.
Fluorescent Detection Methods
Many of the problems inherent in the quantiRcation
of gel-separated proteins visualized by silver staining
can be overcome using detection methods based on
the use of Suorescent compounds. This group of
methods is highly sensitive and generally exhibits
excellent linearity and a high dynamic range, making
it possible to achieve good quantitative analysis, par-
ticularly if a suitable imaging device is used.
Two approaches can be used, the Rrst being to
couple the proteins with a Suorescent-labelled
compound prior to electrophoresis. Examples of
such compounds are: dansyl chloride; Suorescamine
(4-phenyl-[furan-2(3H)-1-phthalan]-3,3-dione);
 II / ELECTROPHORESIS / Proteins, Detection of 1345

o-pththaldialdehyde (OPA)#a thiol; and MDPF (2- electrophoretic Suorescence staining reagents,
methoxy-2,4-diphenyl-3(2H)-furanone). The last re- SYPRO orange and red (Molecular Probes, Eugene,
agent has a reported sensitivity of 1 ng protein per Oregon, USA), have been described. These stains
band and is linear over the range 1}500 ng protein have a very high sensitivity (1}2 ng protein per band)
per band. and excellent linearity with a high dynamic range.
The main disadvantage of pre-electrophoretic Using a Suorescence imaging device, the SYPRO dyes
staining procedures is that they can cause protein- have been shown to be linear over three orders of
charge modiRcations, for example Suorescamine con- magnitude in protein quantity. The other advantage
verts an amino group to a carboxyl group when it of this method is that staining can be achieved in only
reacts with proteins. Such modiRcations usually do 30 min, compared with staining with silver and CBB
not compromise SDS}PAGE unless a large number of R-250 which can take from 2 h to overnight. Gels can
additional charged groups are introduced into the be stained without Rxation so that they can be sub-
protein. However, they result in altered mobility jected to subsequent Western blotting procedures.
during other forms of electrophoresis, giving rise to However, staining with these reagents requires that
altered separations by native PAGE, IEF and two- the proteins are complexed with SDS, so that if the
dimensional electrophoresis. Recently, compounds gels are Rxed prior to staining or electrophoresis is
that react with cysteine or lysine residues have carried out in the absence of this detergent, then the
been described and used successfully for two- gels must be incubated in a solution of SDS prior to
dimensional electrophoresis separations. The cys- staining. An SDS}PAGE gel separation visualized
teine-reactive reagent monobromobimane has been using SYPRO red is shown in Figure 3.
used to label proteins prior to analysis by two-dimen-
sional electrophoresis. Using a cooled CCD camera to
measure Suorescence, the limit of detection was
Metal Chelate Stains
found to be 1 pg protein per spot. This recently developed group of stains have been
In an alternative approach, two amine-reactive developed speciRcally for compatibility with charac-
dyes (propyl Cy3 and methyl Cy5) have been syn- terization methods such as mass spectrometry as they
thesized and used to label Escherichia coli proteins do not use reagents such as glutaraldehyde or formal-
prior to electrophoresis. These cyanine dyes have an dehyde which reduce their efRcacy. Although a stain
inherent positive charge, which preserves the overall of this type, using the pink bathophenanthroline dis-
charge of the proteins after dye coupling. The two ulfonate/ferrous complex, was described over twenty
dyes have sufRciently different Suorescence spectra
that they can be distinguished when they are present
together. This allowed two different protein samples,
each labelled with one of the dyes, to be mixed to-
gether and subjected to two-dimensional electrophor-
esis on the same gel. This method, which has been
termed ‘difference gel electrophoresis (DIGE)’, has
great potential for improving the efRciency of detec-
tion of differences in two-dimensional electrophoresis
protein proRles between different samples.
For two-dimensional electrophoresis, one ap-
proach to overcoming the problems associated with
charge modiRcation during the IEF dimension is to
label the proteins while present in the Rrst dimension
gel after IEF, prior to the second dimension separ-
ation by SDS}PAGE. Two Suorescent labels that have
been used in this way are MDPF and a Suorescent
maleimide derivative.
The alternative approach, which also overcomes
the problem of protein-charge modiRcations, is to
label the proteins with Suorescent molecules such as Figure 3 SDS}PAGE separation of human heart proteins
(lanes b}g). Lane (a) contains the molecular weight marker pro-
1-aniline-8-naphthalenesulfonate (ANS) and OPA
teins and the scale at the left indicates protein size in kDa. The gel
after the electrophoretic separation has been com- has been stained with SYPRO red. The sample protein loadings
pleted. However, these two methods suffer the disad- were (b) 1
g, (c) 5
g, (d) 10
g, (e) 25
g, (f) 50
g,
vantage of relative insensitivity. Recently, two post- (g) 100
g.
1346 II / ELECTROPHORESIS / Proteins, Detection of
years ago, its insensitivity (600 ng protein) ensured its
lack of acceptance. Sensitivity can be increased by
introducing 59Fe into the complex, but only to a level
equivalent of colloidal Coomassie Blue staining.
The poor performance of these dyes resulted in
a recent investigation of luminescent metal chelate
stains. Such stains utilizing metal chelates complexed
to certain transition metal ions (e.g. europium, ter-
bium, and ruthenium) offer much greater sensitivity
compared to the previous colorimetric methods. Of
particular interest is SYPRO Ruby, a proprietary ru-
thenium-based metal chelate stain (Molecular Probes,
Eugene, Oregon, USA). This allows one-step, low
background staining of proteins in polyacrylamide
gels without resorting to lengthy destaining proced-
ures. The linear dynamic range of this dye extends
over three orders of magnitude, thus surpassing
silver and Coomassie stains in performance. Its
sensitivity is claimed to be up to thirty times more
sensitive than silver staining. Moreover, staining
times (unlike silver protocols) are not critical and
staining can be carried out overnight without over-
development.
Radioactive Detection Methods
Metabolic labelling of proteins with a radiolabelled
amino acid prior to their separation by gel elec-
trophoresis represents a very sensitive method for the
detection of proteins and is ideal for the analysis of
protein synthetic events occurring in response to an
experimental intervention. This approach is most
commonly used in combination with in vitro cell
culture systems, but it is also possible to radiolabel
synthetically the proteins of small pieces of fresh
tissue in this way. In this method, the cells or tissue
are incubated in the presence of the radiolabelled
amino acid for a period of time, normally between
3 and 24 h. It is important to use a tissue culture
medium that has been depleted of the amino acid
used for radiolabelling. The most commonly used
amino acids for radiolabelling are [35S]-methionine
and [14C]-leucine. [3H]-amino acids can be used, but
these are more difRcult to detect due to the weak
energy of their beta emissions. Methods are also
available for synthetic radiolabelling to detect speciRc
post-translational modiRcations of proteins.
Proteins can also be radiolabelled post-syntheti-
cally, prior to their separation by gel electrophoresis,
using a variety of methods such as radioiodination
with 125I or reductive methylation with [3H]-sodium
borohydride. However, most of these methods result
in signiRcant charge modiRcation of the target pro-
teins, generally precluding their use for elec-
trophoretic techniques other than SDS}PAGE.
Following electrophoresis of radiolabelled pro-
teins, the gel must normally be dried prior to detec-
tion of the radioactive zones. Thin gels cast on plastic
supports can be dried, after equilibration in 3% w/v
glycerol, in air or in an oven at 40}503C. It is also
possible to air-dry gels which have not been cast on
supports. These should be equilibrated in 3% w/v
glycerol and placed between two cellophane sheets
supported in a plastic frame. The gels are then dried
in hot air at 40}503C; the process usually taking 2 or
3 h. The best method for drying gels which are not on
supports is by heating them under vacuum. Gels
should be soaked in 3% w/v glycerol prior to drying.
Gradient polyacrylamide gels are particularly prone
to cracking and these can be protected by soaking in
a solution containing 1% w/v glycerol and 2% v/v
dimethyl sulfoxide (DMSO). Gels can be dried down
onto Rlter paper or onto cellophane. A temperature of
803C is normally used, but it is better to use a lower
temperature (40}603C) for gels at risk of cracking
(i.e. thick, high percentage or gradient gels).
Radiolabelled proteins are most easily detected by
direct autoradiography, in which the dried gel is
placed in contact with X-ray Rlm and exposed for the
appropriate time. This method works satisfactorily
for isotopes such as 14C, 35S, 32P and 125I, but is not
suitable for 3H owing to its low-energy beta-emis-
sions which are not able to penetrate the gel matrix.
Much sensitive detection can be achieved using
Sourography in which the gel is impregnated with
a scintillant, such that low-energy beta particles ex-
cite the Suor molecules to emit photons which can be
detected on a suitable (usually blue-sensitive) X-ray
Rlm. In the original procedure, 2,5-diphenyloxazole
(PPO) which must be dissolved in DMSO, was used.
However, Suorography with commercially available
enhancers is simpler and less tedious than the original
PO}DMSO method, and produces equivalent results.
Pre-exposure of the X-ray Rlm to a brief Sash of light
(approximately 1 ms) increases the sensitivity of
Suorography by two- or threefold. The use of an
intensifying screen and exposure at low temperature
(!703C) also result in a signiRcant increase in sensi-
tivity.
Techniques of autoradiography and Suorography
are simple and require little specialized equipment,
apart from the access to darkroom facilities. How-
ever, prolonged exposure times are often required to
achieve the desired level of sensitivity of protein de-
tection. Moreover, the nonlinear response of X-ray
Rlm and its limited dynamic range present severe
problems to accurate quantitation. To overcome
these problems several devices for detecting
radiolabelled proteins directly in gels have been
described. The best and most practical of these
 II / ELECTROPHORESIS / Proteins, Detection of
approaches are imaging devices based on the use of
photostimulable storage phosphor-imaging screens.
Detection of Speci\c Biological
Compounds
Detection of Glycoproteins
Proteins with limited glycosylation can be detected
following gel electrophoresis with the general protein
stains such as CBB R-250 and silver. However, such
staining gives no direct indication that these proteins
are glycosylated and the methods are much less sensi-
tive if the proteins are more highly glycosylated. Pro-
teoglycans are usually stained with cationic dyes,
such as Alcian Blue or Toluidine Blue, which bind to
the negatively charged glycosaminoglycan side
chains. Glycoproteins have generally been detected
using variations of the Schiff base reaction, involving
oxidation with periodic acid followed by staining
with Schiff reagent, Alcian Blue or a hydrazine deriv-
ative. A twofold increase in sensitivity can be
achieved with methods in which Alcian Blue is used
as the primary staining agent followed by subsequent
enhancement using a neutral silver-staining protocol.
An alternative approach to the analysis of
glycosylated proteins is to radiolabel then in vitro,
followed by gel electrophoretic separation of the
radiolabelled proteins and their detection. N-linked
sugar labelling can be achieved using [3H]-mannose
and terminal O-linked N-acetylglucosamine can be
labelled by galactosyltransferase and UDP-[3H]-
galactose.
Probably the most versatile reagents for the charac-
terization of glycosylated proteins following their
separation by electrophoresis are radiolabelled, Suor-
escent or enzyme-conjugated lectins. Although it is
possible to use these directly in the gel matrix, much
better results are achieved using Western blotting
techniques.
Detection of Phosphoproteins
The most commonly used approach to the analysis of
protein phosphorylation is to radiolabel cells in cul-
ture with either [32P]-orthophosphate or [-32P]-ATP.
An alternative approach, which avoids the use of
radioactive materials, is to use antibodies which
are speciRc to phosphotyrosine, phosphothreonine
and phosphoserine in combination with Western
immunoblotting.
Detection of Lipoproteins
Lipoproteins can be stained following electrophoresis
with Sudan black B. Prenylated proteins can be
1347
radiolabelled prior to electrophoresis with [3H]-
mevalonolactone, while fatty acylated proteins can be
radiolabelled with [3H]-palmitic or [3H]-myristic acid.
Detection of Enzymes
It is generally considered that speciRc enzyme activ-
ities can only be visualized following gel electrophor-
esis if native conditions have been used. However,
there are several reports demonstrating that SDS-de-
natured proteins can also be visualized provided that
it is possible to achieve at least partial renaturation of
their spatial conRguration. Such renaturation is most
effective if disulRde bonds are not essential for en-
zymic activity and if the native protein is not com-
posed of subunits of different molecular weights. Pre-
electrophoresis of gels is usually recommended to
remove unreacted acrylamide monomers and cata-
lysts.
Enzyme staining can be achieved by incubating the
unRxed gel in a solution of the appropriate reagents
using either Suorogenic or chromogenic substrates.
This method works well if the Rnal reaction product
is insoluble. However, a soluble reaction product will
rapidly diffuse resulting in loss of resolution. It is
generally preferable to use a print or gel overlay
technique. In this approach, the substrates and other
reagents are either impregnated into a Rlter or in-
cluded in a thin layer of agarose or polyacrylamide
gel cast on a glass or plastic support. The overlay is
then placed in direct contact with the surface of the
separation gel and following a suitable period of
incubation, the enzymic activity is visualized on the
overlay. Methods are available for the visualization
of a large number of enzyme activities following gel
electrophoresis.
See also: III/Proteins: Capillary Electrophoresis; Elec-
trophoresis. Electrophoresis: Detection Techniques:
Staining; Autoradiography and Blotting
Further Reading
Fernandez-Patron C, Castellanos-Serra L, Hardy E, Guerra
M, Estevez E, Mehl E and Frank R (1998) Understand-
ing the mechanism of the zinc-ion stains of biomac-
romolecules in electrophoresis gels: Generalization of
the reverse-staining technique. Electrophoresis 19:
2398}2406.
Laskey RA (1980) The use of intensifying screens or organic
scintillators for visualizing radioactive molecules re-
solved by gel electrophoresis. Methods in Enzymology
65: 363}371.
Merril CR (1987) Detection of proteins separated by elec-
trophoresis. In: Chrambach A, Dunn MJ and Radola BJ
(eds). Advances in Electrophoresis, Vol. 1, p 111}139.
Weinheim: VCH.
1348 II / ELECTROPHORESIS / Theory of Electrophoresis
Neuhoff V, Arold N, Taube D and Ehrhardt W (1988)
Improved staining of proteins in polyacrylamide gels
including isoelectric focusing gels with clear background
at nanogram sensitivity using Coomassie Brilliant Blue
G-250 and R-250. Electrophoresis 9: 255}262.
Rabilloud T (1990) Mechanisms of protein silver staining
in polyacrylamide gels: a 10-year synthesis. Electrophor-
esis 11: 785}794.
Rabilloud T (1992) A comparison between low back-
ground silver diamine and silver nitrate protein stains.
Electrophoresis 13: 429}439.
Rothe GM (1994) Electrophoresis of Enzymes: Laboratory
Methods. Berlin: Springer.
Shevchenko A, Wilm M, Worm O and Mann M (1996)
Mass spectrometric sequencing of proteins from silver-
stained polyacrylamide gels. Analytical Chemistry 68:
850}858.
Steinberg TH, Jones LJ, Haugland RP and Singer VL (1996)
SYPRO orange and SYPRO red protein gel stains: One-
Staining
See II / ELECTROPHORESIS / Detection Techniques: Staining, Autoradiography and Blotting
Theory of Electrophoresis
K.S. Pitre, Dr. Harisingh Gour University, Sagar, India
Copyright ^ 2000 Academic Press
This article is reproduced from Encyclopedia of Analyti-
cal Science, Copyright ^ 1995 Academic Press
Principles
Electrophoresis is a very separation technique which
involves the separation of charged species (molecules)
on the basis of their movement under the inSuence of
an applied electric Reld. It is widely used by chemists
and biochemists in studies related to medical techno-
logy, environmental research, food and water analy-
sis, pollution control and forensic investigations. The
development and applications of electrophoretic sep-
aration methods are an example of the fruitfulness of
using physical methods in tackling biological and
biochemical problems.
The migration of charged colloidal particles in an
electric Reld was originally given the name cataphor-
esis or electrophoresis. Because there has been some
diversity of opinion about the deRnition of a colloid,
and thus about the distinction between colloidal and
molecular systems, there has also been some differ-
step Suorescent staining of denaturing gels for detection
of nanogram levels of protein. Analytical Biochemistry
239: 223}237.
Sutherland JC (1993) Electronic imaging of electrophoretic
gels and blots. In: Chrambach A, Dunn MJ and Radola
BJ (eds). Advances in Electrophoresis, Vol. 6, pp. 3}42.
Weinheim: VCH.
Unlu M, Morgan ME and Minden JS (1997). Difference gel
electrophoresis: A single gel method for detecting cha-
nges in protein extracts. Electrophoresis 18: 2071}2077.
Urwin VE and Jackson P (1993) Two-dimensional polyac-
rylamide gel electrophoresis of proteins labelled with the
Suorophore monobromobimane prior to Rrst-dimen-
sional isoelectric focusing: Imaging of the Suorescent
protein spot patterns using a charge-coupled device.
Analytical Biochemistry 209: 57}62.
Wirth P and Ramano A (1995) Staining methods in gel
electrophoresis, including the use of multiple detection
methods. Journal of Chromatography A 698: 123}143.
ence of opinion as to how widely the term ‘elec-
trophoresis’ should be used. Some authors prefer the
term ionophoresis to describe the movement of rela-
tively small molecules or ions under such conditions.
The 1940s and 1950s witnessed very rapid devel-
opments in the applications of methods making use of
the migration of particles in an electric Reld. These
applications covered the whole range of particle sizes
from the largest protein molecules to small molecules
like amino acids, sugars (at high pH) and even simple
inorganic ions, using the sample types of procedures
and apparatus. Although it is not a form of chromato-
graphy, the differences in the rates of migration of the
charged particles provide a powerful means of separ-
ating biocolloids such as proteins, polysaccharides
and nucleic acids, as well as for the characterization
of their components. For these reasons, and also for
historical reasons, it is now general practice to use the
term ‘electrophoresis’ to refer to all these procedures.
Electrophoresis pertains to the transport of electri-
cally charged particles } ions, colloids, macromolecu-
lar ions or particulate matter } in an electric Reld.
 II / ELECTROPHORESIS / Theory of Electrophoresis
Electrophoresis experiments are usually carried out
to obtain information on the electrical double layers
surrounding the mobile particles, to analyse a mix-
ture, or to separate it into components. Interpretation
of experimental results requires a theory connecting
the electrophoretic mobility with the fundamental
quantities relating to the electric double layer } the
electrical potential, charge and structure.
The electrical double layer is not restricted to the
interface between electrically conducting phases. For
example, if a glass rod is immersed in an aqueous
electrolyte, then it will carry a double layer of ions
wholly within the electrolyte phase. This double layer
originates from the speciRc adsorption of a Helm-
holtz layer of anions or cations from solution onto
the glass surface. The resulting excess of charge is
neutralized by a diffuse or Gouy layer dispersed fur-
ther out in the solution. If we consider the case of two
insulating phases, namely glass and oil, the double
layer at the interface may be considered to arise either
from the speciRc adsorption of ions generated by very
weak electrolytes or from the orientation of dipolar
molecules. The behaviour of the diffuse or mobile
component of the double layer may be correlated
with a class of phenomena which includes elec-
trokinetic effects.
Electrokinetic effects are associated with the rela-
tionship between the relative motion of two phases
(generally a liquid and a solid) and the electrical
properties of the interface between them. Electro-
kinetic phenomena arise in microheterogeneous sys-
tems, i.e. in cases when one phase is dispersed in
another. Electrokinetic effects may be classiRed
into four groups: (1) electroosmosis, (2) electrophor-
esis, (3) streaming potential and (4) sedimentation
potential.
1. Electroosmosis is the movement of a liquid along
a capillary, a system of capillaries or a porous plug
under the inSuence of an externally applied elec-
tric Reld.
2. Electrophoresis is the movement of solid particles
under the inSuence of an electric Reld applied to
the medium in which the particles are suspended.
In this case the disturbance of the double layers
attached to the solid moving particles produces the
effect. It may be regarded as the reverse of elec-
troosmosis, in which the solid phase is Rxed and it
is the movement of the liquid phase that is induced
by the applied electric Reld. In both electrophor-
esis and electroosmosis the applied potential dif-
ference sets up a mechanical force which results in
the movement of one phase.
3. Streaming potential is the building up of potential
difference between the upstream and downstream
1349
ends of liquid Sow. This is caused by friction
between the moving liquid layer and the wall of
the capillary, the system of capillaries or the por-
ous plug.
4. Sedimentation potential (Dorn effect) is the con-
verse of electrophoresis and results in the building
up of potential difference between the top and the
bottom of a vessel in which dispersed solid par-
ticles are suspended in a liquid.
The theoretical treatment of the electrical double
layer depends on its geometry. The double layer at
a Sat interface constitutes the most simple case, which
we can analyse to explain many of the facts connected
with double layers. The boundary between two
phases is a layer of Rnite dimensions. The properties
of the two adjacent phases change gradually over
a certain distance. These changes depend both on
geometrical factors and on the forces between the
molecules. The density and orientation of the molecu-
les, even in a one-component system, undergo a grad-
ual change when going from one phase to another,
e.g. from the liquid to the gas phase. In multicompo-
nent systems the boundary layer concentrations are
different from those in the bulk, leading to what is
called adsorption. Though these changes near phase
boundaries are limited to only a very few layers of
molecules, all the properties of the phases are
changed in this transition layer.
When one or both phases contain ions, the
transition layer may be much more extended. In such
a case, one type of ion is strongly concentrated at the
phase boundary by short-range forces. When ions of
one sign are adsorbed at the phase boundary, ions of
the opposite sign will be attracted by the resulting
electric Reld and will accumulate near the phase
boundary. This accumulation is opposed by their
Brownain movement. As a result an electrically
neutral double layer is formed which may extend
to a considerable thickness (a few tens of nano-
metres).
In order to apply simple mathematical treatment to
electrokinetic phenomena it is assumed that the dif-
fused double layer acts as a parallel plate electric
capacitor whose plates are d cm apart, each carrying
a charge e per cm2. The zeta potential, i.e.
, the
potential difference between the plates, is given by
eqn [1]:

"4ed/D [1]
where D is the dielectric constant of the medium
between the hypothetical plates. This is the funda-
mental equation for the quantitative treatment of all
types of electrokinetic phenomena.
1350 II / ELECTROPHORESIS / Theory of Electrophoresis
When a liquid is forced by electroosmosis through
the Rne capillaries of a porous diaphragm, two oppos-
ing factors will determine the Sow, namely the force
of electroosmosis and the frictional force between the
moving liquid layers and the capillary wall. When the
two forces become equal, the Sow attains a uniform
rate. If u is the uniform velocity so obtained and d is
the effective thickness of the double layer across
which the Sow takes place, then the velocity gradient
in the double layer may be taken as equal to u/d. Since
the velocity at one side, i.e. the wall, is zero, and the
average value on the other side, i.e. in the moving
liquid, is u, the force due to frictional effects is equal
to u/d, where  is the coefRcient of viscosity of the
liquid. If E is the potential gradient across the mem-
brane and e is the charge per cm2 at the boundary of
movement, then the electrical force causing electroos-
mosis is equal to Ee. Hence at the steady state eqn [2]
applies:
Ee"u/d [2]
Substituting the value for d from eqn [2] in eqn [1]
we obtain eqn [3]:

"4u/DE [3]
Following on from this discussion of electrokinetic
phenomena, electrophoresis takes place due to the
presence of an electrical double layer at the interface
between the dispersed phase and the dispersion me-
dium, and the consequent presence of a zeta poten-
tial. On applying an external electromotive force,
positive and negative portions of the double layer are
displaced relative to each other. Since the particles in
a solution are free to move, they will migrate under
the inSuence of the applied electric Reld. As has been
noted previously, the double layer surrounding a par-
ticle may be treated as a capacitor. We can therefore
derive a relationship for the observed velocity u of
the particle from eqn [3], namely eqn [4]:
u"
DE/4 [4]
Here the quantity u/E"U represents the particle’s
mobility, i.e. the velocity for a potential gradient of
1 V cm\1.
Consider the case of a comparatively large spheri-
cal particle of radius R carrying a charge q in a me-
dium of dielectric constant D. According to elec-
trokinetic theory the potential of the particle may be
given by q/DR. If the charge is identiRed with that
present in the diffused double layer only, then the
potential is
and since the thickness of the Helm-
holtz double layer is negligible compared with the
radius of large particles, R may be taken as equal to
the radius of particle plus its Helmholtz layer. This
can be written as shown in eqn [5]:

"q/DR [5]
From eqns [4] and [5] we get eqns [6] and [7]:
u"qE/4R [6]
U"q/4R [7]
where U is the electrophoretic mobility. If the sur-
rounding medium is an electrolyte, the interaction
between the charged and migrating particles will re-
duce the zeta potential,
, of the particle. The magni-
tude of this effect has been evaluated by Debye and
HuK ckel, who observed that
is reduced by a factor of
1/(1#KR), where 1/K is ionic length. K is of the
order of 10\7}10\8 cm and can be calculated in
terms of ionic charges in the electrolyte, the concen-
tration and dielectric constant of the electrolyte and
the radius at which the ionic atmosphere would need
to be concentrated to obtain the potential of the ion.
The electrophoretic mobility for a comparatively
large spherical particle in an electrolyte may be given
by eqn [8]:



q 1
U" [8]
4R 1#KR
Eqn [8] is not applicable to small spherical particles
where the curvature of the double layer is too large
for streaming to take place entirely in the direction of
the applied Reld. In such a case the electrical force on
the particle is equal to the viscous drag as given by
Stokes’ law. Considering the effect of the electrolyte
on the zeta potential (as in eqn [8]) the result for
a small spherical particle is given by eqn [9]:



q 1
U" [9]
6R 1#KR
Debye and HuK ckel made an exact treatment and
found that the factor 4 in eqn [8] is strictly only
applicable for cylindrical particles, and it should be
replaced by the factor 6 (as in eqn [9]) for spherical
particles. Eqns [8] and [9] are thus the special cases of
a general expression covering all sizes of particle.
A slight modiRcation can be made to eqn [9] to give
the electrophoretic mobility according to:



q 1
U" f(KR) [10]
4R 1#KR
 II / ELECTROPHORESIS / Theory of Electrophoresis
where f(KR) is a complicated function of K and R.
More elaborate equations for electrophoretic mobil-
ity have been derived by taking into account the Rnite
sizes of ions in the double layer attached to the par-
ticle. Gorin has also given a treatment for cylindrical
particles.
Looking at the consequences of the complications
associated with the theoretical deviation of the rela-
tionship between the electrophoretic mobility and the
shape of the particle, attempts have been made to
solve the problem experimentally. However, the ex-
perimental results are equally inconclusive. Abram-
son established that the electrophoretic mobility is
independent of the shape of the moving particles by
performing experiments on the movement of spheri-
cal particles of some oils and of needles of asbestos
and m-aminobenzoic acid coated with the same pro-
tein.
The applicability of eqns [8] and [9] (for spherical
particles) can also be tested by the comparison of
electroosmotic and electrophoretic mobilities using
a microelectrophoresis cell made of the same material
as the suspended particle, e.g. glass or quartz. If
eqn [8] is correct, then the ratio of the two mobilities
should be unity; on the other hand if eqn [9] is cor-
rect, the ratio should be 1.5. Experiments with spheri-
cal particles and surfaces, all coated with adsorbed
protein to ensure that the surfaces were the same,
indicated that the mobility ratio was approximately
unity, as required by eqn [8]. Objections have been
raised to this conclusion on the grounds that the
independence of electrophoretic mobility with respect
to the shape of particles and the value of unity for the
mobility ratio were due to the use of a liquid medium
containing a relatively high concentration of the elec-
trolyte. The objectors stated that if the electrolyte
concentration is less than 1 mmol L\1 then the ratio
of electrophoretic to electoosmotic mobility is unity.
Other workers also questioned the conclusion, and as
such the situation is somewhat uncertain. However,
eqn [8] may be regarded as reasonably adequate for
particles of any shape.
Factors Affecting Electrophoretic
Mobilities
Several factors have a deRnite inSuence on the elec-
trophoretic mobilities of charged molecules or ions.
Nature of the Charged Molecule
The nett charge, size, shape and relative mass of
particles has a great inSuence on their electrophoretic
mobilities. The charge-to-size ratio of molecules is an
important parameter. The higher its charge-to-size
1351
ratio (e/r), the faster a molecule will migrate under
given conditions.
Nature of the Electrophoretic System
As well as the characteristics of the substances to be
separated, there are several parameters relating to the
electrophoretic system itself that have a pronounced
effect on the electrophoretic mobilities of the molecu-
les or ions. These parameters are as follows.
1. The ionic composition of the electrophoresis buf-
fer.
2. The temperature.
3. The pH of the electrophoresis buffer.
4. The applied voltage.
5. In the case of zone electrophoresis, the type of
support medium chosen, and if the support
medium is gel, its pore size.
Ionic composition of the electrophoresis buffer
A charged macromolecule becomes surrounded by an
ionic atmosphere of opposite charges because of in-
teractions between ionizable groups on the surface of
the charged molecule and ions in the electrophoresis
buffer. As a result, both its net charge and its elec-
trophoretic mobility are decreased. This effect is quite
pronounced in the electrophoretic separation of pro-
teins, since different proteins have different amino
acid side chains which interact to varying degrees
with the ions in the solutions used.
In order to minimize these ‘counterion’ effects it is
advisable to use an electrophoresis buffer with as low
an ionic strength as possible. However in some cases,
such as with polypeptides and polynucleotides, elec-
trophoresis has to be carried out in solutions of high
ionic strength, otherwise these macromolecules will
not be soluble. It therefore becomes necessary to
choose a suitable salt concentration.
Temperature Temperature plays a pronounced ef-
fect on electrophoresis. In an electrophoretic run,
heat (Joule’s heat) is generated and may affect the
electrophoresis in a number of ways.
1. Diffusion. An increase in temperature causes an
increase in the diffusion of migration zones of
charged molecules. If the electrophoresis takes
a long time (several hours) diffusion effects be-
come more signiRcant.
2. Evaporation. It is customary to perform elec-
trophoresis in a closed system to avoid loss of
water by evaporation, which increases with tem-
perature. This evaporation results in the drying
out of the supporting medium and also leads to an
1352 II / ELECTROPHORESIS / Theory of Electrophoresis
increase in the ionic strength of the buffer during
the analysis.
3. Viscosity. In gel electrophoresis an increase in tem-
perature can change the viscosity of the medium.
Since this takes place during the electrophoretic
run, the interpretation of the results may become
complex.
4. Distortion of zones. During an electrophoretic
run, particularly in column gels if cooling is inad-
equate, the portions of the migration zones in the
warmer parts of the gel move faster than those in
the cooler parts. This difference in migration
speeds produces curved bands. This may result in
the overlap of neighbouring zones and conse-
quently in poor resolution.
5. Convection currents. During an electrophoretic
run the warmer solution in the centre of the appar-
atus has a lower density than the cooler solution
close to the walls. This density gradient induces
convection currents in the solution. Since water
has its maximum density at 43C, and the smallest
variations in density of aqueous solutions are ob-
served around this temperature, it is advisable to
perform electrophoresis at a temperature as close
as possible to 43C. However, the viscosity of an
aqueous solution increases as the temperature is
lowered, which may result in an increase in the
frictional resistance to the migration of the
charged molecules. If the temperature is main-
tained at 43C the electrophoretic mobilities of the
charged molecules will be relatively low. It is
therefore necessary to choose an optimal temper-
ature for a particular electrophoretic run and
maintain it throughout the course of analysis.
pH of the electrophoresis buffer pH has a marked
effect on the nett charge on a protein molecule. At
a deRnite pH value, i.e. at the isoelectric point, the
nett charge on the molecular is zero. Molecules ac-
quire a nett positive charge at pH values below their
isoelectric points. At pH values above their isoelectric
points, they acquire a nett negative charge. Thus
different molecules (e.g. proteins) at any particular
pH value will have different nett charges. To optimize
the separation of a mixture of (protein) molecules the
buffer pH must be chosed on the basis of the nett
molecular charges. For example in an electrophoretic
run two or more proteins may migrate together to
give only one band. If the analysis is done at a differ-
ent buffer pH value it may result in the appearance of
extra protein bands, indicating the presence of other
proteins in the sample.
Applied voltage In electrophoresis the applied volt-
age plays an important role. The migration velocity of
a molecule is proportional to the Reld strength across
the medium. The higher the applied voltage, the larger
the Reld strength across the medium, and the faster
a molecule will migrate. Thus, the charged molecules
will migrate more quickly with increasing voltage.
This saves time and reduces the diffusion of migrating
molecules. However, with increasing voltage the cur-
rent also increases, resulting in greater power genera-
tion (the power increases as the square of the current).
Some of this power is dissipated as heat (Joule’s
heat). The heat generated can have serious effects on
an electrophoretic analysis, as discussed previously. It
is therefore necessary to select a deRnite value of ap-
plied voltage. The voltage (or current) should be large
enough to allow rapid migration of charged molecules,
but not so large as to generate excessive heat.
Support medium In zone electrophoresis different
types of support media are used. The selection of the
most suitable support medium for a particular zone
electrophoretic analysis is based on the following
considerations.
1. Sample quality.
2. Size of the molecules } whether they are small or
large. If a sieving gel has to be used its pore size
(concentration) is chosen so as to suit the molecu-
lar size under study.
3. The time required for analysis.
Types of Electrophoresis
Electrophoresis analysis can be divided into three
main types, as listed here.
1. moving-boundary electrophoresis
2. zone electrophoresis
3. steady-state electrophoresis.
Moving-Boundary Electrophoresis
Moving-boundary or free solution electrophoresis
was Rrst proposed by Picton and Linde in 1892 and
was fully developed by Tiselius in 1930, Rnding wide
application between 1935 and the 1950s. Its principle
use was in protein research, where it provided invalu-
able results.
The apparatus for moving-boundary electrophor-
esis, in its simplest form, consists of a U-tube with
a rectangular cross-section (Figure 1). It is partially
Rlled with the solution to be analysed (protein solu-
tion) and a buffer solution is layered over it. Elec-
trodes are immersed in the buffer solution.
When an electromotive force is applied across these
electrodes the charged molecules move towards the
appropriate electrode. If the solution under study is
 II / ELECTROPHORESIS / Theory of Electrophoresis
Figure 1 Apparatus for moving-boundary electrophoresis.
coloured and contains several components with dif-
ferent electrophoretic mobilities, then their migration
can be observed as multiple moving boundaries in the
system. Where the migration boundary is not visible
to the naked eye it may be made visible by causing it
to Suoresce as a result of exposure to ultraviolet (UV)
light. In such a case the apparatus should be made of
quartz. In the case of protein solutions which have
a refractive index slightly greater than that of the
buffer, there will be a charge in refractive index at the
protein boundary in the apparatus, which can be
detected by optical methods.
After completion of the electrophoretic analysis the
different fractions containing the components separ-
ated from the original sample can be analysed using
appropriate methods.
Moving-boundary electrophoresis has proved to be
useful in the determination of complex heterogeneous
samples. However, it has a drawback in that only the
slowest and fastest moving components of a sample
can be obtained in pure form.
Zone Electrophoresis
Zone electrophoresis results in the complete separ-
ation of the components of a mixture in the form of
discrete zones. The separated zones may be stabilized
Figure 2 Apparatus for zone electrophoresis.
1353
in a number of ways which include the use of support
media, density gradients and free zone techniques.
Use of support media An important development in
electrophoresis was the use of support media for
stabilizing zones of electrophoretically separated
mixtures. A porous medium such as Rlter paper, cel-
lulose acetate Rlm, a gel, glass beads, granular starch
or polyvinyl particles, etc., is used as the support
medium. The separatory power of some of the gels
exceeds that of other media or free solution.
The most common apparatus for zone electrophor-
esis using support media is depicted in Figure 2.
The sample is placed on the support medium in the
form of a spot or narrow band at the sample origin.
On subjecting it to electrophoresis the components of
the sample separate into bands which are kept dis-
tinct by the presence of the support medium.
Zone electrophoresis in density gradients An impor-
tant method that allows complete separation of com-
plex mixtures involves carrying out electrophoresis in
a density gradient column. The density gradients are
usually solutions of substances such as ethylene
glycol, glycerol, sucrose, etc., which do not ionize or
interact with the materials under examination.
Free zone electrophoresis Although it is a complex
technique, by using this method it is reasonably simple
to recover the separated components at the end of
analysis. The electrophoretic analysis may be per-
formed in two ways, by rotation of the electrophoresis
vessel during the run or by streaming a continuous Rlm
of buffer across the electrophoretic system, in a direc-
tion perpendicular to the applied electric Reld. These
techniques are not in common use because they require
elaborate equipment and are expensive.
Steady-State Electrophoresis
The fundamental basis of this method is that, after
electrophoresis has proceeded for a certain length of
1354 II / ELECTROPHORESIS / Theory of Electrophoresis
time, a state known as the steady-state is reached in
which no further changes to either the position or the
width of the zone of the separated components occur
with time. Steady-state electrophoresis involves two
types of techniques, isoelectric focusing and iso-
tachophoresis.
Isoelectric focusing In this technique the charged
components of the sample move through the medium
under the inSuence of an applied electric Reld until
they ultimately reach a position in the pH gradient
where their nett charge is zero and hence they do not
migrate any more. This pH is their isoelectric
point. The individual bands may be collected from
the apparatus, and examined chemically or bio-
logically.
This technique has very high resolving power, and
may be used preparatively. It is not applicable to
some compounds (proteins) that precipitate at
their isoelectric point, which makes their recovery
difRcult.
Isotachophoresis This technique is particularly ap-
plicable to the separation of charged components
with small relative molecular masses, such as some
drugs and polypeptides of medical interest. In this
technique all the charged components become
stacked one behind the other, depending upon their
electrophoretic mobilities.
To achieve a good separation of two charged com-
ponents from one another by isotachophoresis, their
electrophoretic mobilities must differ at least by 10%.
Comparison of Moving-Boundary
and Zone Electrophoresis
In moving-boundary electrophoresis a differential
movement of the charged particles towards one or the
other of the electrodes is observed. Separation takes
place as a result of differences in mobilities. The
mobility of a particle is approximately proportional
to its charge-to-mass ratio. However, this technique
suffers from several disadvantages, one of the most
serious being the tendency of the separated compo-
nents to mix by convection as a consequence of ther-
mal and density gradients and mechanical vibrations.
Thus, careful thermal regulation and isolation from
mechanical vibrations is essential. Also, elaborate op-
tical systems are often required for locating and
measuring the fraction, making this method quite
expensive.
Many of these experimental difRculties associated
with moving-boundary or free solution electrophor-
esis may be avoided if separations are carried out in
a supporting medium (zone electrophoresis), such as
on paper. This prevents convection currents from
distorting the electrophoretic pattern. This technique
closely resembles the various chromatographic
methods, with the additional parameter of a super-
imposed electric Reld. The separations mainly depend
upon the properties of the medium and may result
primarily from the electrophoretic effect or from
a combination of electrophoresis and adsorption, ion
exchange, or other distribution equilibria.
Practical Problems
The Joule’s heat generated in an electrophoretic run
causes various problems, including changes in the pH
of the buffer medium, diffusion and distortion of the
zones, and evaporation and convection currents. It
should be remembered that these effects may occur
simultaneously, which may complicate the results. It
is therefore necessary to optimize the temperature for
a particular analysis and keep it constant.
In paper electrophoresis there is liable to be varia-
bility in the quality of paper from batch to batch, and
also perhaps within one sheet. Some papers have
adsorptive properties that may affect the elec-
trophoretic mobility of the molecules under study.
Ionizable groups may produce an electroosmotic ef-
fect on the paper, which in turn may distort the
migration characteristics of the separating compo-
nents. The physical and chemical inhomogeneity of
the support medium also has a pronounced effect on
the migration of substances which in turn affects the
results.
The method used for the detection of sample com-
ponents separated by electrophoresis depends on the
type of electrophoresis, the nature of the molecules to
be detected, and the purpose of analysis. Some detec-
tion methods are listed in Table 1.
Separations using electrophoretic methods may not
be adequate for very complex mixtures. Resolutions
may be improved substantially by combining elec-
trophoresis with some other separation technique.
The sample subjected to electrophoresis in one direc-
tion while the other separation analysis is carried out
in a perpendicular direction (so-called two-dimen-
sional techniques).
Applications
Electrophoresis involves the separation of charged
species on the basis of their movement under the
inSuence of an applied electric Reld. It has found wide
applications in the characterization of biological mol-
ecules (proteins and nucleic acids). The main applica-
tions of electrophoresis have been in the separation of
biological molecules, which includes molecules with
 II / ELECTROPHORESIS / Theory of Electrophoresis
Table 1 Methods of detection for quantitative analysis of
sample components separated by electrophoresis
Optical methods UV absorption
Staining
Fluorescence
Raidiochemical methods Liquid scintillation counting
Autoradiography
Biological assay and Immunoelectrophoresis
immuno-methods Rocket electrophoresis
relatively lower relative molecular masses such as
amino acids, and also molecules of higher relative
molecular masses such as proteins and polynucleot-
ides (including RNA and DNA molecules). An
example of the use of paper electrophoresis follows.
Paper electrophoresis has been extensively used in
almost all laboratories where proteins and other sim-
ilar macromelecular electrolytes are investigated. The
apparatus (Figure 2) consists of two electrode cham-
bers placed 15 cm apart. There is also a device which
can support up to six (30 cm) Rlter paper strips be-
tween the electrodes. A d.c. supply source (0}250 V)
is used to apply the desired voltage across the elec-
trodes. The two electrode chambers are Rlled to equal
heights with the buffer solution. The buffers com-
monly used for this purpose are (1) Aronsson and
Gronwall buffer, i.e. dimethyl barbiturate buffer,
which is a mixture of 20.60 g sodium dimethyl bar-
biturate and 2.80 g barbituric acid with a pH of 8.6,
and (2) Consden and Powell buffer, i.e. borate buffer,
which is a mixture of 1.77 g sodium hydroxide and
9.60 g orthoboric acid with a pH of 8.6.
Whatman paper (M540) strips (about 30 cm long)
are cut and dipped in a container of buffer until they
are thoroughly wet. The excess buffer is then re-
moved by laying them out on a large sheet of Rlter
paper. The strips are then immersed into the electrode
chambers so that the ends of the strips dip in the
buffer solutions. The sample is applied at the centre
of the paper. The paper strips are allowed to stand for
about 1 h, to equilibrate the bed with the liquid even-
ly throughout the paper. The power supply is then
switched on and the voltage adjusted to about 75 V.
Excellent sharply deRned separations of serum pro-
teins into Rve fractions within a span of 2 cm can be
obtained within 1 h. If the run is extended to 16 h,
a pattern approximately 12 cm long with Rve frac-
tions may be obtained. On completing the run the
fractions are measured by staining. The dye most
commonly employed for this purpose is amido black
10B. The paper strip is dried and developed in a dye
bath containing a saturated solution of the dye in
a mixture of methanol and glacial acetic acid (9 : 1
1355
v/v). Staining is allowed for 10 min with constant
shaking.
After the electrophoretogram has been stained, the
excess of dye is removed (destaining) by dipping the
stained paper in baths of methanol}glacial acetic acid
(9 : 1 v/v) several times. This destaining procedure is
a slow process and can be made more efRcient and
faster by opting for electrophoretic destaining. After
destaining, the paper strip is dried and scanned in
a densitometer, i.e. the strip is illuminated and moved
along the light source, the transmittance showing the
distribution of the separated compounds. On plotting
the reciprocal of transmittance against the wave-
length one or more maxima are observed, depending
upon the number of components. The amount of each
component can be estimated by measuring the area
under each peak. The estimation of serum proteins
can also be done by an elution method. The destained
paper strip is cut transversely into small pieces 5 mm
wide, one unstained small strip at the end of the paper
providing the blank value. The elution is done in
1.0 mol L\1 NaOH in 50% ethanol#0.25% (0.25 g
per 100 mL) ethylenediaminetetraacetic acid
(EDTA). After elution is over, the optical density is
measured using a colorimeter.
A more rapid separation of serum proteins can be
achieved using polyacrylamide gel electrophoresis.
However, paper electrophoresis is still of particular
interest where small amounts of protein need to be
isolated for further analysis or testing.
Further Reading
Antropov LI (1975) Theoretical Electrochemistry. Mos-
cow: Mir Publishers.
Bier M, ed. (1956) Electrophoresis: Theory Methods and
Applications. New York: Academic Press.
Glasstone S (1956) An Introduction to Electrochemistry.
New York: D van Nostrand.
Longsworth LG (1943) A differential moving boundary
method for transference numbers. Journal of the Ameri-
can Chemical Society 65: 1755}1765.
Longsworth LG (1939) A modiRcation of the Schlieren
method for use in electrophoretic analysis. Journal of the
American Chemical Society 61: 529}530.
Melvin M (1987) Analytical chemistry by open learning. In:
Kealey D (ed.) Electrophoresis. New York: Wiley.
Scott ND and Svedberg T (1924) Measurements of the
mobility of egg albumin at different acidities. Journal of
the American Chemical Society 46: 2700}2707.
Svedberg T and Tiselius A (1926) A new method for the
determination of proteins. Journal of the American
Chemical Society 48: 2272}2278.
Svedberg T and Jette ER (1923) The cataphoresis of pro-
teins. Journal of the American Chemical Society 45:
954}957.
1356 II / ELECTROPHORESIS / Two-dimensional Electrophoresis
Two-dimensional Electrophoresis
M. Fountoulakis, F. Hoffman-La Roche Ltd.,
Pharmaceutical Research-Gene Technology, Basel,
Switzerland
Copyright ^ 2000 Academic Press
Introduction
Two-dimensional (2D) polyacrylamide gel electro-
phoresis is a classical technique for the separation of
proteins. It Rrst appeared in the mid-1970s but for
a long time it only found limited applications. Re-
cently it has enjoyed an impressive renaissance. The
major reasons for this are the introduction of the
immobilized pH gradient (IPG) strips and the devel-
opment of analytical methods capable of identifying
proteins present in very low quantities. 2D elec-
trophoresis represents the core methodology of the
new, technology-driven science proteomics. Pro-
teomics Rnds a wide application, in both clinical
diagnosis and in pharmaceutical research, for the
detection of novel drug targets. Figure 1 demon-
strates the application of 2D electrophoresis for the
detection of variable protein levels between diseased
and healthy brain tissue. In a sample from the parietal
lobe of the brain of a patient with Alzheimer’s dis-
ease, a strong spot representing glial Rbrillary acidic
protein (GFAP), a marker for neuronal loss, is pres-
Figure 1 Partial 2D gel images showing human brain proteins from a control (A) and a patient with Alzheimer’s disease (B). The
proteins were separated on pH 3}10 nonlinear strips, followed by 9}16% SDS gels, stained with colloidal Coomassie blue. The spots
representing glial fibrillary acidic protein (GFAP) in (B), a known marker of neuronal loss, are indicated.
ent, whereas in the control sample, the corresponding
spot is very weak.
The aim of proteomics is the high throughput anal-
ysis of the proteome (protein complement expressed
by a genome) of various organisms or tissues. It con-
sists of two steps: (1) the separation of protein mix-
tures by 2D electrophoresis, and (2) the identiRcation
of the separated proteins by analytical techniques,
such as mass spectrometry and amino acid composi-
tion analysis. The process is facilitated by the use of
highly sophisticated software for advanced image
analysis and the high reproducibility of images in
intra- and inter-laboratory studies. The 2D elec-
trophoresis itself involves: (1) separation of the pro-
teins on the basis of differences in their net charge,
called isoelectric focusing (IEF), and (2) separation of
the focused proteins on the basis of differences in
their molecular masses. Table 1 gives a summary of
the most signiRcant chronological events in the devel-
opment of 2D electrophoresis. The state-of-the-art
technology will be discussed without entering into
extensive technical details that can be found in the
literature provided.
First-Dimensional Separation (IEF)
Proteins carry positively and negatively charged side
groups and are, therefore, amphoteric molecules. The
 II / ELECTROPHORESIS / Two-dimensional Electrophoresis 1357

Table 1 Important advances in 2D electrophoresis technology

1970 Introduction of sodium dodecyl sulfate in 1D gel electrophoresis to efficiently separate complex protein mixtures
(Laemmli UK, Nature 227: 680)
1975 Separation of protein mixtures by 2D gel electrophoresis using tube gels and pH gradients formed with carrier
ampholytes (O’Farrell PH, Journal of Biological Chemistry 250: 4007)
1980}1990 Pioneering work to improve pH gradient stability; synthesis of Immobilines and preparation of IPG strips (Bjellqvist B,
Journal of Biochemistry Biophysics Methods 6: 317); electrotransfer of proteins from gels to PVDF membranes
1990}today IPG strips became commercially available; introduction of sigmoidal strips, efficient separation of basic proteins,
improvement of gel staining and protein solubilization techniques (Bjellqvist, Dunn, Goerg, Hochstrasser, Rabilloud,
Righetti and others); development of high throughput protein analytical techniques (MALDI-MS, amino acid
analysis); development of software for protein identification and image comparison; establishment of databases
accessible via the WorldWideWeb; sequencing of the complete genome of microorganisms; preparation of 2D
protein maps for organs, cell lines, organisms; new terms Proteome, Proteomics were introduced

protein charge depends on the pH value of the solu- narrow bands. This type of IEF is usually performed
tion. IEF is an equilibrium process, during which, in strips of acrylamide gel Rxed on a plastic sheet.
under the inSuence of a high voltage Reld the proteins
move along a pH gradient, according to their net Carrier Ampholytes
charge, to a position, where they have no net charge
The pH gradient formed by the carrier ampholytes
and consequently stop moving. This pH value is
during IEF can be affected by the amount of total
called the isoelectric point (pI). The resolving power
protein loaded. Proteins when applied in large quant-
of IEF is deRned by the equation of Svensson:
ities, having themselves a buffering capacity, can af-
fect the focusing position along the pH gradient and

pI"[D[d(pH)/dx]:E[!du/d(pH)]]1/2
consequently the reproducibility. Therefore, factors
such as protein quantity, temperature, voltage and
where
pI"resolution capacity (pI difference re-
chemicals can strongly affect performance. Only
quired to resolve neighbouring spots), D"diffusion
small quantities of protein (of the order of 0.1 mg)
coefRcient of the protein, E"Reld strength (V cm\1),
should be applied for IEF with the carrier ampholytes
d(pH)/dx pH gradient, du/d(pH) mobility slope at pI.
approach. The difRculties in controlling the various
According to this equation, the resolution capacity
factors which affect reproducibility together with the
is inSuenced by the pore size of the gel, which affects
difRculties of preparing and transferring the tube gels
the diffusion of the protein, the slope of the pH
to the second dimensional separation, contribute to
gradient and the voltage value.
the reasons why carrier ampholyte-based IEF re-
An efRcient and reproducible protein separation
mained a scientiRc speciality of only a few laborator-
during IEF requires a stable pH gradient. There are
ies. Nevertheless, these laboratories are able to con-
two pH gradient systems in use. In the Rrst one, the
trol the conditions affecting reproducibility, so that
pH gradient is created by an excess of carrier am-
IEF with carrier ampholytes is still used. This ap-
pholytes during the IEF run. Ampholytes are am-
proach allows a very reliable protein quantiRcation of
photeric compounds of low molecular mass with
complex mixtures. Because only a small amount of
closely related pI values. Upon application of an
protein can be applied, 2D gels made following the
electric Reld, the ampholyte molecules move and
carrier ampholyte approach are more suitable for
align themselves between the electrodes, forming
analytical purposes. IEF based on carrier ampholytes
a pH gradient, which increases from anode to cath-
can efRciently separate proteins with pIs within the
ode. This type of IEF is usually performed in tube
pH range of 3}8. Proteins with higher pIs separate
acrylamide gels.
poorly due to cathodic drift during isofocusing. The
In the second pH gradient system, the pH gradient
cathodic drift is the result of a high electroosmotic
is immobilized and has been formed prior to IEF run.
Sow, caused by the charged groups on the glass walls
IPGs are formed by acrylamide derivatives, called
of the gel tubes.
immobilines, which are weak acids and bases with
a buffering capacity. Immobilines are copolymerized
IPG Strips
in a polyacrylamide gel, such that a pH gradient is
formed between basic and acidic molecules. When an The increased application of two-dimensional gel
electric Reld is applied, the pH gradient does not move. electrophoresis today is to a large extent due to
Only the charged molecules of the protein sample the introduction of IPG strips. The major advantage
move and are focused according to their pIs into of using IPG strips is the ability to maintain high
1358 II / ELECTROPHORESIS / Two-dimensional Electrophoresis

Figure 2 Partial 2D gel images showing the high fidelity in the reproducibility of separation of bacterial (A, B) and mammalian (C, D)
proteins. The proteins were separated as stated in the legend to Figure 1. (A, B). Separation of proteins of Haemophilus influenzae
were 1.5 and 3.0 mg, respectively. (C, D) Separation of rat brain proteins, 1.5 and 2.0 mg, respectively.

reproducibility. The increase in reproducibility has
allowed a high throughput analysis of proteomes and
the application of larger sample quantities, a require-
ment for protein spot analysis. Figure 2 provides
examples of the reproducibility of separation of bac-
terial and mammalian proteins, following IEF on pH
3}10 nonlinear IPG strips. Although minor differ-
ences can be detected, the reproducibility concerning
both the position and the intensity of the protein
spots can be considered as satisfactory.
As mentioned earlier, immobilines are polymerized
in a gradient, in a polyacrylamide gel, and the gel is
then dried on a plastic sheet. Before the IEF run, the
dry strips are rehydrated in a speciRc rehydration
solution, containing a reducing agent, ampholites and
high concentrations of urea and a zwitterionic or
nonionic detergent (usually CHAPS). Rehydration can
also be performed in a solution containing the protein
sample to be analysed (see sample application). Large
numbers of strips can be rehydrated at a time and this
represents an advantage of the method regarding high
throughput analysis and performance. Today, IPG dry
strips are commercially available from Amersham
Pharmacia Biotechnology in two lengths } 11 and
18 cm and in three pH ranges } 3}10, 4}7 and 6}11.
The pH 3}10 strips are available in a linear and non-
linear (sigmoidal) form. The latter allow a more efR-
cient focusing of proteins with pIs between 4.5 and 7.
A large percentage of proteins from various organisms
have pI values within this range. The dry strips can be
kept frozen at !203C for a long time (the expiration
date is indicated on the packaging).
 II / ELECTROPHORESIS / Two-dimensional Electrophoresis
The use of narrow pH range strips (i.e. of 1 pH
unit) provides a higher resolution and allows the
detection of protein isoforms; this is an additional
advantage of IEF using IPG strips. Strips of more
narrow pH ranges are not currently commercially
available and have to be prepared by the user. IPG
strips can be made in any biochemical laboratory
using a gradient marker and Immobilines of various
pK values which can be purchased. Recipes for the
preparation of IPG strips can be found in handbooks,
for example in Electrophoresis in Practice (see Fur-
ther Reading). On the narrow pH range the spots
appear stretched compared to the wide range strips.
Figure 3 shows an example of a protein which ap-
pears as one spot following IEF after focusing on
a pH 3-10 strip and as Rve spots after IEF on a pH
4}7 strip. IEF on strips of an even narrower pH range
would result in the detection of additional spots re-
sulting from further isoforms of the protein. Follow-
ing IEF, the IPG strips can be either immediately used
for the second dimensional separation or stored
frozen at !203C for long periods (for example, in
petri dishes sealed with paraRlm). Strips stored for
4 months have been used at !203C without any
Figure 3 Partial 2D gel images showing examples of proteins
represented by multiple spots. (A, B) -enolase from human brain.
The protein is represented by one spot when IEF was performed
on pH 3}10 nonlinear IPG strips (A), and by five spots when IEF
was performed on pH 4}7 strips (B). (C) Dihydropyrimidinase-
related protein 2 from human brain shows a high heterogeneity,
represented by five spots, localized into two regions on the gel.
(D) 1-antitrypsin from human cerebrospinal fluid is represented
by many spots, most likely denoting different glycoforms of the
protein.
1359
effect on the spot resolution. Longer storage times of
up to 1 year have been reported.
Sample Preparation
Careful sample preparation is a prerequisite of a suc-
cessful analysis. Most proteins are soluble and are
easily recovered in the sample preparation solution,
which includes urea, CHAPS, a reductant and, op-
tionally, protease inhibitors. Recovery of the proteins
that are insoluble in this solution is often a problem.
A centrifugation step is necessary for removal of
nondissolved material. The addition of thiourea and
of a noncharged reducing agent, such as tributyl
phosphine, to the sample buffer increases protein
solubility during IEF. It would appear that hydropho-
bic interactions between proteins and the acrylamide
gel of the IPG strips are responsible for protein losses
during IEF. Nucleic acids present in the sample can
also seriously affect spot resolution. Enzymatic diges-
tion with an endonuclease prior to sample application
is usually recommended.
Sample Application
The protein application mode can affect the amount
of protein entering the IPG strip during IEF. There are
several ways of applying the sample. In the system
supplied by Amersham Pharmacia Biotechnology
(Multiphor II), the sample is usually loaded into ap-
plication cups (also supplied by Amersham Pharma-
cia Biotechnology). Up to 150
L can be applied in
one cup. The cups are Rxed in special ‘cup accommo-
dating bridges’ which are placed near the basic or
acidic end of the strip. It seems that sample applica-
tion at the basic end of the strip is more advantageous
compared to the application at the acidic end. We
have, however, found that simultaneous sample ap-
plication at both the basic and the acidic ends of the
strip can result in the detection of more and stronger
protein spots compared to sample application at only
one end. It also allows the simultaneous application
of sample volumes larger than 150
L. From a tech-
nical point of view, sample application using the cups
is the most difRcult operation to perform. The cups
should touch the polyacrylamide gel on the strip,
otherwise the sample will leak; they should also not
damage the gel at the contact point, otherwise the
proteins will not enter the strip.
An alternative method of sample application is the
rehydration of the strip in a solution containing the
protein sample. This approach is convenient to per-
form and theoretically it should result in the detection
of all proteins present in the sample. However, more
comparative studies are required to prove that this
approach is more efRcient than the loading of sample
1360 II / ELECTROPHORESIS / Two-dimensional Electrophoresis

into cups. Amersham Pharmacia Biotechnology has rium step, the proteins are negatively charged
recently introduced a new IEF apparatus (IPGphor) by addition of the anionic detergent SDS. Upon
in which sample application and IEF can be per- application of an electric Reld, the charged proteins
formed. The strips are placed in special ceramic move along a porous polyacrylamide gel and are
strip holders and rehydrated for the desired time separated according to their size. A reducing agent
in a solution containing the protein sample. Each is also included to disrupt disulRde bonds. In
strip holder holds a single IPG strip throughout comparison with IEF, SDS-PAGE is relatively easy
rehydration and IEF. IEF starts automatically after to control. The terms ISO-DALT and IPG-DALT
rehydration according to the conditions pro- are often used to mean 2D gel electrophoresis
grammed. Whether the performance of IEF will be employing IEF with carrier ampholytes or IPG strips,
improved with the use of this instrument is not clear respectively.
at present. Horizontal or more usually vertical slab gels, run-
The quantity of protein to be applied on the strip ning in a discontinuous buffer system are employed.
naturally depends on the goal of the analysis. If the A high throughput analysis is facilitated by the use of
identiRcation of protein spots is intended, the amount tanks accommodating 6}20 gels running in parallel.
loaded should be in the order of 1 mg or higher, An efRcient separation of thousands of proteins pres-
depending on the number of proteins in the mixture. ent in complex mixtures, can only be performed on
A 1D gel analysis of the sample prior to 2D elec- gels of a large format (18;20 or 25;25 cm). Either
trophoresis may provide helpful information as to gradient gels or gels of a constant acrylamide concen-
deRning the right protein quantity. If large amounts tration can be used. Because of the complexity of the
of protein are applied, a percentage of the proteins
may not enter the strip. Presently, this constitutes
a drawback of the IPG strip approach. Because cer-
tain proteins in the sample (mainly major compo-
nents) only partially enter the strip, this can result
in an unreliable quantiRcation of a particular
protein in a given mixture. While the application of
15 mg or more of protein sample has been reported,
we consider that 2}4 mg is the limit for a productive
separation, using the strips that are presently avail-
able.
IEF using IPG strips can separate basic proteins
efRciently with pIs up to about 12. The introduction
of low concentrations of isopropanol in the rehydra-
tion buffer improves focusing of basic proteins. Hy-
drophobic proteins probably precipitate at the point
of application and efRcient separation has not yet
been reported. Hydrophobic proteins can be analysed
in a different 2D electrophoresis system, which uses
the interaction of the proteins with a cationic deter-
gent in the Rrst dimension rather than pI. The second
dimension is, as described below, dependent on the
molecular mass. The separated proteins form approx-
imately a diagonal line. Relatively, only a small num-
ber of proteins can be successfully separated using
this approach.

Second-Dimensional Separation
(Sodium Dodecyl Sulfate-Polyacryl-
amide Gel Electrophoresis,
SDS-PAGE)
Following IEF, the proteins are separated according
to their molecular masses. During this nonequilib-
Figure 4 Partial 2D gel images showing an improved spot res-
olution by using different acrylamide concentrations. Separation
of rat brain proteins on a 9}16% SDS gel (A) and on a 7.5}16%
SDS gel (B). Separation of low molecular mass soluble proteins
from H. influenzae on 9I16% SDS gel (C) and on a 10% SDS
Tricine gel (D). (B, D) The gel parts comprising the corresponding
proteins shown in A and B, respectively, are longer on account of
the different acrylamide concentrations.
 II / ELECTROPHORESIS / Two-dimensional Electrophoresis 1361

technology and the large diversity of the samples to
be analysed, and in order for the data to be useful to
a broad research community, 2D PAGE has been to
a large extent standardized. In the second dimension,
for example, we usually use 9}16% linear gradient
gels. This gel system represents a good compromise,
as it separates proteins between 5 and 200 kDa.
However, efRcient separation is limited to a range of
approximately 15}40 kDa. Outside this range, in
particular above 50 and below 10 kDa, the separ-
ation is often suboptimal. For more efRcient separ-
ation, gels of a different acrylamide concentration
should be tried. Figure 4 gives examples of the im-
proved separation of high molecular mass brain pro-
teins using gels of lower acrylamide concentration
and of low molecular mass proteins from Haemo-
philus inUuenzae, using gels with Tricine as the trail-
ing ion instead of Tris.
For spot visualization, the gels can be stained with
either silver or Coomassie blue (usually colloidal
Coomassie blue), depending on the quantity of pro-
tein sample applied and the aim of analysis. Silver
stain may be preferentially used for gel comparison
studies, whereas staining with Coomassie is preferred
when the spots are intended for protein identiRcation.
Colloidal Coomassie blue has the advantage that the
stain is sensitive enough and the gels can be easily
destained with water. The simultaneous staining of
many gels in one tank substantially increases the
throughput. Apart from silver and Coomassie blue,
several other protein detection methods exist, such as
staining with various metals, labelling with Suor-
escent agents or detection of radiolabelled com-
pounds, after gel drying and exposure, for example to
a Rlm.
Proteome Analysis
An essential step of proteomics is the identiRcation
and mapping of the proteins separated by 2D elec-
trophoresis. From a mammalian organism, compris-
ing approximately 100 000 possible gene products,

Figure 5 Partial 2D gel images showing soluble proteins of (A) H. influenzae, (B) E. coli and (C) B. subtilis. The proteins were
separated as stated in the legend to Figure 1. Note the similarity in the distribution of the major proteins in the three bacterial organisms.
1362 II / ELECTROPHORESIS / Two-dimensional Electrophoresis
approximately 1000}2000 protein spots can be vis-
ualized on one 2D-gel, using Coomassie blue. Higher
numbers can be detected, following staining with
silver or after radiolabelling. Approximately one-half
of the visible spots are available in sufRcient quantit-
ies to be analysed for identiRcation. Figure 5 shows
the analysis by 2D electrophoresis of the proteomes
of three bacteria, H. inUuenzae, Escherichia coli and
Bacillus subtilis. The genomes of the three microor-
ganisms have been completely sequenced, so that
theoretically all expressed proteins can be mapped.
This has however not yet been accomplished. The
largest 2D proteome maps, such as that of H. inUu-
enzae prepared at F. Hoffmann-La Roche, Basel, in-
clude approximately 500 mapped proteins. Many of
the unidentiRed proteins are not expressed in sufR-
cient amounts to be visualized.
For the mapping of proteomes of the various or-
ganisms, protein enrichment steps need to be intro-
duced before analysis. We have used several
chromatographic steps, such as heparin chromatogra-
phy, hydrophobic interaction chromatography,
chromatofocusing, hydroxyapatite chromatography
and several other approaches, to enrich the low-
abundance gene products of H. inUuenzae and E. coli.
Figure 6 Partial 2D gel images showing the enrichment of enolase by hydrophobic interaction chromatography. (A) Total extract;
(B) proteins from a fraction collected from the column.
Additional enrichment steps are required for an efR-
cient mapping of proteins present at low abundance,
such as cytokines or transcription factors. Figure 6
shows an example of protein enrichment by hydro-
phobic interaction chromatography. One protein
(enolase), represented by a strong spot in the 2D
map of the total protein extract, is highly enriched
after chromatography. Another example of protein
enrichment, this time using heparin chromato-
graphy is shown in Figure 7. In two fractions col-
lected from the column, proteins which are not visible
in the 2D gel image of the total extract can also be
detected.
On a 2D map, proteins are often represented by
more than one spot. Figure 3C shows an example of
a brain protein represented by Rve spots, in two
locations, with different pI and M values. Presently,
P
we do not know the reasons and the biological signiR-
cance for most of these cases of observed heterogen-
eity. It may be the consequence of post-translational
modiRcations, such as deamidation, phosphorylation
or glycosylation, which result in the alteration of the
pI of the molecule and its focusing position. Another
reason may be the carbamylation of the protein upon
contact of the sample with urea. In Figure 3D an
 II / ELECTROPHORESIS / Two-dimensional Electrophoresis 1363

Figure 7 Partial 2D gel images showing the enrichment of low abundance proteins of H. influenzae by heparin chromatography.
(A) Total extract; (B, C) proteins from fractions collected from the heparin column. The arrowheads indicate spots representing two
proteins (B, topoisomerase I; C, ATP-dependent RNA helicase) which are not visible in total extract (A).

example of protein heterogeneity, most likely due to
glycosylation, is presented.
Following 2D electrophoresis, proteins can be iden-
tiRed by mass spectrometric analysis of the peptides
resulting from the in-gel digestion with a speciRc
protease, such as trypsin. In another approach, the
proteins can be electrotransferred onto membranes
and identiRed by immunoreaction with speciRc anti-
bodies, by N-terminal sequencing or amino acid com-
position analysis. For those proteins for which the

Table 2 Steps in the preparation of 2D electrophoresis

IPG strip rehydration and sample preparation

Protein extraction, centrifugation, recovery in sample solution
Sample application
Application in cups at either or at both ends of the strip or strip rehydration in a solution containing the protein sample
First dimensional separation (isoelectric focusing)
Start at 200 V and increase gradually to 5000 V; keep 5000 V for 6}48 h, depending on sample, quantity and strip range; narrow pH
range strips require longer focusing times
Reduction and alkylation of proteins on IPG strip
Equilibration of strip with reducing and alkylating agents or freeze until use
Second dimensional separation (SDS-PAGE)
Preparation of gel of the desired acrylamide concentration; gels should carry a label to identify them afterwards; establishment of
contact between strip and gel with agarose solution; run at 40 mA/gel
Protein \xing and staining or blotting
Fixation of proteins within the gel and staining with silver or Coomassie blue or drying of the gel and exposure to a film or
phosphorimager for detection of radiolabelled proteins or electrotransfer of proteins to membranes for immunoblot, MS or amino acid
analysis
Gel scanning
Storage of image in a database
Gel comparison
Gel comparison and protein quantification using specific software; comparison with database master gels via the WorldWideWeb
Identi\cation of proteins
Identification of protein spots from gels by mass spectrometry or from membranes by N-terminal sequencing, amino acid composition
analysis, MS or immunoblots
1364 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis
genomic sequence is in a database, the most efRcient
identiRcation method presently available is matrix-
associated laser desorption ionization mass spectro-
metry (MALDI-MS) with which about 500 spots can
be analysed daily by one person. The method toler-
ates small amounts of salt in the sample, so that no
time-consuming desalting steps are required after di-
gestion. Several approaches using a combination of
protein digestion on membranes and MS have also
been reported. Table 2 summarizes the essential steps
of 2D electrophoresis and protein analysis.
Future Developments
2D electrophoresis is still in a developmental stage.
Several technical improvements, mainly concerning
further simpliRcation of the technology and possible
automation, an increase in reproducibility and sensi-
tivity, and expansion of the pH detection spectrum,
have to be achieved in order for the method to be-
come routine in any biochemical laboratory. Gel grin-
ding techniques, together with sophisticated software
using the mass spectroscopic data, may be developed
to produce a gel image without previous staining of
the gel. Such a development could be decisive as to
whether the technology will reach its major goal, i.e.
the investigation of biological problems by a faithful
comparison of protein expression levels. The comple-
tion of the sequencing of more genomes together with
improvements in the analytical techniques will also
Two-dimensional Polyacrylamide Gel Electrophoresis
J.-D. Tissot and P. Schneider,
Service Re& gional Vaudois de Transfusion Sanguine,
Lausanne, Switzerland
M. A. Duchosal, Centre Hospitalier Universitaire Vaudois,
Lausanne, Switzerland
Copyright ^ 2000 Academic Press
Introduction
The evolution of tools utilized in biology and medi-
cine, together with the exponential progress accomp-
lished recently in the area of bioinformation, enables
analysis of whole organism constituents. Such
analyses are best exempliRed by complete genomic
sequences of different microorganisms, and by the
recent development in techniques permitting
dissection of the whole protein repertoire of an indi-
vidual, namely its proteome. Furthermore, the dra-
matic growth in the number of genome projects as
lead to a more widespread application of the tech-
nology.
See Colour Plate 43.
Further Reading
Anderson L (1991) Two-dimensional Electrophoresis: Op-
eration of the ISO-DALT System. Rockville: Large Scale
Biology Press.
Fountoulakis M and Lahm H-W (1998) Hydrolysis and
amino acid composition analysis of proteins. Journal of
Chromatography 826: 109.
Hames BD and Rickwood D (1990) Electrophoresis of
Proteins: A Practical Approach. Oxford: IRL Press.
Humphery-Smith I, Cordwell SJ and Blackstock WP (1997)
Proteome research: Complementarity and limitations
with respect to the RNA and DNA worlds. Electrophor-
esis 18: 1217.
Kleinert T (1990) Elektrophoretische Methoden in der Pro-
teinanalytik. Stuttgart: Georg Thieme Verlag.
Righetti PG (1990) Immobilized pH Gradients: Theory and
Methodology. Amsterdam: Elsevier.
Smith BJ (1997) Methods in Molecular Biology: Protein
Sequencing Protocols, vol. 64. Totowa: Humana Press.
Walsh BJ and Herbert B (1998) Setting up Two-dimensional
Gel Electrophoresis for Proteome Projects.
http://rbams3115/Pages/2DPAGE/ABRFNews}2dpage.html
Westermeier R (1993) Electrophoresis in Practice. Wein-
heim: VCH Verlagsgesellschaft.
Wilkins MR, Williams KL, Appel RD and Hochstrasser DF
(1997) Proteome Research: New Frontiers in Functional
Genomics. Berlin: Springer.
well as the speed with which genome sequences are
determined has generated huge amounts of informa-
tion. This progress has boosted techniques, notably
two-dimension polyacrylamide gel electrophoresis
(2D-PAGE), enabling the analysis of a proteome con-
sisting of all the proteins expressed by a genome. Such
analyses give information on the effector molecule
itself, namely the protein, and take into account high-
ly sophisticated mechanisms regulating gene expres-
sion. 2D-PAGE is the most powerful tool to
separate a multitude of polypeptides that are con-
tained in a single biological sample. Various pro-
cedures have been described to separate proteins
according to biophysical parameters. In 1975,
O’Farrel, Klose and Scheele described optimized 2D
procedures in which proteins were denatured and
separated by electrophoresis on polyacrylamide gel.
The Rrst gel dimension comprised a separation ac-
cording to the protein charge by isoelectric focusing,
 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis 1365

and the second gel dimension separated proteins ac-
cording to their sizes. Thus, peptides are separated
from one another according to two independent bio-
chemical properties. 2D-PAGE was shown to be par-
ticularly valuable in the study, as well as in the identi-
Rcation of thousands of cellular or secreted proteins,
including many of those present in human
plasma/serum (Figure 1).
During the past few years, tremendous progress has
been made in the Reld of 2D-PAGE. The 2D tech-
nique has been simpliRed, and, more importantly,
made reproducible. Commercially manufactured im-
mobilized pH gradients (IPGs), with both acidic and
basic high resolution power and precast sodium
dodecyl sulfate (SDS) PAGE gels are now available. In
addition, progress in protein solubilization and in the
development of systems allowing high loading capa-
cities has been made.
More than 20 years after its birth, 2D-PAGE is now
a major technique in protein sciences. Over the past
few years there has been an exponential increase in
the creation and expansion of protein databases such
as the SWISS-2DPAGE, the HEART-2DPAGE and the HSC-
2DPAGE. Furthermore, tools have been developed to

Figure 1 The normal human plasma 2D map. Polypeptides (0.3
L of plasma) were separated by pH 3.5}10 carrier ampholyte
gradient, followed by gradient 9}16% T polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate (SDS). The
ammoniacal silver-stained gel was photographed with the higher molecular weight at the top and the acidic side on the left. 1,

2-macroglobulin; 2, ceruloplasmin; 3, glu-plasminogen; 4, lys-plasminogen; 5, complement factor B; 6, complement C1s; 7, protrans-
ferrin; 8, prothrombin; 9,
1-B-glycoprotein; 10, transferrin; 11, hemopexin; 12,
2-antiplasmin; 13,
1-antichymotrypsin; 14, fibrinogen

chain; 15,
2-HS-glycoprotein; 16,
1-antitrypsin; 17, antithrombin III; 18, fibrinogen  chain; 19, extended fibrinogen  chain; 20,
Ge-globulin; 21, lysin-rich glycoprotein; 22, fibrinogen  chain; 23, apolipoprotein A-IV; 24, haptoglobin  chain; 25, Zn-
-glycoprotein;
26, apolipoprotein J; 27, cleaved haptoglobin  chain; 28, apolipoprotein E (phenotype E 3/4); 29,  chain of complement C4; 30,

1-microglobulin; 31, apolipoprotein D; 32, apo A-I; 33, proapolipoprotein A-I; 34, retinol-binding protein; 35, haptoglobin
1 chain; 36;
transthyretin (prealbumin); 37, haptoglobin
2 chain; 38, haemoglobin  chain; 39, apolipoproteins C-II and C-III; a, albumin; ,
polyclonal heavy chains of IgM;
, polyclonal heavy chains of IgA; , polyclonal heavy chains of IgG; }, polyclonal Ig light chains.
Reproduced with permission from Tissot and Spertini (1995).
1366 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis
compare 2D gels images across the Internet. Most
importantly, methods for the analysis of 2D gels are
continually improving. The high sensitivity and
throughput of these techniques now enable the char-
acterization of hundreds of proteins from a whole
cell, tissue or body Suid. Proteins can be identiRed
according to primary parameters such as their isoelec-
tric points, apparent molecular mass, real mass and
protein N- or C-terminal sequence tag, but also ac-
cording to secondary attributes such as peptide mass
Rngerprint, peptide fragmentation data or amino acid
composition. Interfacing and integrating databases
from 2D gels such as SWISS-2DPAGE, SWISS-PROT, Gen-
Bank, EMBL nucleotide sequence database, dbest,
GSBD and the NLM’s MEDLINE bibliographical refer-
ence database provide researchers with invaluable
tools to study both genome and proteome. The conti-
nuing progress accomplished in both proteome re-
search and bio-information will contribute to the
implement of the Cyber-Encyclopedia of the Pro-
teome, as suggested by R.D. Appel. This development
increases the need for simple protocols to run repro-
ducible 2D gels and is an important step for investiga-
tors involved in proteomics. Several well-written pro-
tocols and reviews are accessible in the literature or
through the World Wide Web (http://expasy.hcuge.
ch/ch2d/technical-info.html; http://www.abdn.ac.
uk/&mmb023/2dhome.htm). We will review here
some of the important features that must be evaluated
in order to implement a new 2D-gel laboratory.
Sample Preparation and Protein
Solubilization
2D-PAGE is an ideal separation tool. Nevertheless,
sample preparation and protein solubilization are still
a key step that is frequently ignored. In addition,
there is no universal and ideal sample buffer for
2D-PAGE. The goal of sample preparation is to maxi-
mize solubilization and disaggregation of the tissues
while preventing protein degradation. For these rea-
sons, and because samples have various character-
istics, protocols have to be adapted according to the
origin of the samples. Sample preparation is quite
easy with soluble proteins such as those of plasma or
cerebrospinal Suid, but presents major difRculties in
the presence of membrane or nuclear proteins. Tissue
samples are usually mechanically disrupted (ultra-
sonication, rapid agitation in the presence of glass or
zirconium microspheres), washed in a low salt buffer
and then chemically (lysis buffer) solubilized. Con-
stituents such as nucleic acids, lipids or salts can
interfere with both solubilization and with the elec-
trophoretic properties of the proteins. Before loading,
nonsolubilized material must be eliminated by high
speed centrifugation. To break down interpolypept-
ide disulRde bonds and to maintain reducing condi-
tions, components such as dithiothreitol, dithioeryth-
reitol or -mercaptoethanol are used. The solubility
problem can be alleviated to a large extent by the
proper use of a combination of chaotropes and deter-
gents in combination. Urea is a common constituent
of protein sample preparations, and in its presence
care must be taken to avoid heating above 323C,
which causes carbamylation of peptides. Impressive
improvements in protein solubilization have been ob-
tained with a denaturing solution containing urea,
thiourea and detergents (both nonionic and zwit-
terionic). The ideal conditions would combine the
highest chaotropic power (i.e. 2 mol L\1 thiourea
and 7}8 mol L\1 urea) with a detergent cocktail (3-
[(3-cholamidopropyl)dimethylammonio]-1-propane
sulfonate, Triton X-100).
The main problem encountered with thiourea is the
strong inhibition of acrylamide polymerization. An-
other problem complicating protein separation by
2D-PAGE is that, at high concentrations, many pro-
teins are prone to precipitation, resulting in poor
resolution after isoelectric focusing. In order to load
large amounts of proteins, investigators have over-
come the problem by using ‘in gel’ application of the
samples, avoiding the use of sample cups and elimin-
ating precipitation at the same application site.
Isoelectric Focusing
In 1964, synthetic carrier ampholytes (aliphatic
oligoamino and oligocarboxylic acids) were syn-
thesized allowing separation of peptides according to
their charges. For 2D-PAGE, carrier ampholyte
isoelectric focusing is usually performed in an am-
pholyte}polyacrylamide matrix that is polymerized in
a glass tube. After sample loading, polypeptides are
concentrated into narrow bands within a continuous
pH gradient in the polyacrylamide gel matrix. Pro-
teins migrate in an electric Reld until they arrive at
a position in which they have no net charge, i.e. their
isoelectric point (pI). Isoelectric focusing is useful
because: (i) no diffusion of proteins occurs because
of the focusing effect; (ii) it offers a resolution allowing
the separation of microheterogeneous populations of
proteins; and (iii) the pI of the protein can be esti-
mated. In the past, pH gradients were generated by
carrier ampholytes or amphoteric buffer moving free-
ly within an acrylamide matrix.
Practically, many factors can affect measurement
of the apparent pI of the proteins during isoelectric
focusing with carrier ampholytes: (i) since the pro-
teins have their own inherent charges, they can act as
ampholytes themselves and affect pH during focus-
 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis
ing; (ii) the carrier ampholytes have a higher mobility
than the proteins; (iii) some proteins may never reach
steady state due to polyacrylamide gel matrix restric-
tion; (iv) ampholyte}protein interactions may alter
the observed pI of the proteins. Temperature, time,
voltage and salt concentration are other parameters
that may dramatically inSuence the determination of
the pI. Finally, the basic proteins are not detected
without using nonequilibrium pH gradient elec-
trophoresis. Nowadays, many of these problems have
been resolved by the development of isoelectric focus-
ing with an IPG. The pH gradient is created by
copolymerization of acrylamide/N,N-methylene-
bisacrylamide with acrylamido derivatives, contain-
ing either carboxyl or tertiary amino groups as buf-
fers and sulfate groups (acidic) or quaternary am-
monium (basic) as strong titrants (Immobilines). This
method is a true equilibrium method, which signiR-
cantly improves the feasibility of the 2D-PAGE.
Recently, highly reproducible, commercially avail-
able, wide-range as well as narrow-range IPGs have
been produced. The latter gradients allow a pH scale
that enables comparison of several 2D gel maps gen-
erated with many IPGs in the Rrst dimension and with
various biological samples. IPGs also offer the possi-
bility of determining pI without major differences
from the calculated pI values, unless there are signiR-
cant post-translational protein modiRcations. Now-
adays, with the improvements of IPG production, it is
possible to detect proteins with pIs up to pH 12 in
a single IPG gel with highly reproducible protein
patterns.
Finally, as mentioned in the previous section, entire
IPG gels can be used for sample application, with the
protein entering the gels during their rehydration.
This approach is useful because it eliminates precipi-
tation at the sample application site, it improves the
resolution over the entire pH range of the gels, and it
allows precise control of protein amounts and sample
volumes loaded on to the IPG gels. Up to 5 mg of
proteins can be loaded on wide IPG gels and up to
15 mg on narrow pH range gels. Contrarily to iso-
electric focusing with carrier ampholytes, electroen-
dosmosis (transport of water towards the cathode at
low pH values or towards the anode at high pH
values) is generally not a problem with IPGs. Current-
ly, isoelectric focusing using carrier ampholytes still
has a place in a 2D laboratory, because the resolution
of particular proteins is sometimes better.
From the First to the Second
Dimension
Transfer of the Rrst-dimension spaghetti-like gels
after isoelectric focusing in the presence of am-
1367
pholytes used to be a manipulation challenge for
beginners in the Reld of 2D-PAGE. Extrusion of the
gels from the glass tubes was difRcult, and the gels
frequently broke into several pieces. With the use of
IPG strips, which are deposited on a plastic backing
material, the transfer is easy. Practically, after the Rrst
dimension run, the strips are equilibrated in buffers
containing SDS in order to maintain proteins in solu-
tion and to reduce }S}S} bonds. Subsequently, strips
are placed over SDS gels that may or may not contain
a stacking gel.
SDS-PAGE
In the second dimension, polypeptides are separated
according to their size in a gel matrix. Practically,
after electric focusing in reducing conditions, proteins
are separated into their polypeptide components. The
latter are mixed with SDS-containing buffers. SDS
binds to polypeptides at a constant mass ratio (1.4 g
SDS per gram of protein). As a consequence, poly-
peptides organize as rod-like molecules, with a dia-
meter of 1.8 nm, and their lengths depend on their
molecular weight. The bound SDS molecules contrib-
ute to a strong negative charge, which effectively
swamps the intrinsic charge of the polypeptides.
Thus, in general the SDS}polypeptide complexes
have the same mass/charge ratio and, in a sieving
polyacrylamide matrix, they will migrate according
to their molecular weight. Glycoproteins and lipop-
roteins can migrate abnormally as they are not easily
saturated with SDS.
The gel matrix is most frequently composed of
polycrylamide generated by polymerization of acryl-
amide monomers into long chains that are cross-
linked. Usually, cross-linkers are bifunctional compo-
nents such as N,N-methylenebisacrylamide (bis) or
diacrylpiperazine. Polymerization of acrylamide is in-
itiated either by the use of ammonium persulfate or
riboSavin, and is accelerated by the use of N,N,N,N-
tetramethylethylenediamine (TEMED). Oxygen inhi-
bits polymerization, and thus gel mixtures are usually
degassed. The composition of a polyacrylamide gel is
deRned by two parameters: % T and % C. The
% T (w/v) is the total concentration of the monomer
(acrylamide plus cross-linker), whereas % C corres-
ponds to the ratio (w/w) of the cross-linker to the
acrylamide. The pore size of the polymerized acryl-
amide will depend on these two parameters, but since
pore formation is random, pore sizes will never be
totally uniform. The choice of the mean pore size will
depend on the molecular weight of the proteins to be
studied. The second-dimension SDS-PAGE can be
performed with home-made vertical or horizontal
systems, using linear or gradient polyacrylamide
1368 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis

(9}16%) gels. Commercially manufactured gels are
also available. The advantages of the latter reside
in their reproducibility and safety (polymerized
acrylamide being clearly less neurotoxic as compared
with monomeric acrylamide). The sensitivity as well as
the resolution power of the protein detection must be
kept in mind before choosing optimal conditions for
SDS-PAGE. Gels polymerized with the photoinitiator
system, composed of methylene blue, toluene sulRnate
and diphenyliodonium chloride, lead to low resolution
power after silver staining. Resolution can be restored
if methylene blue is replaced by riboSavin. Gels poly-
merized with the riboSavin/sulRnate/iodonium system
yield better results upon N-terminal microsequencing
after blotting than gels polymerized with the standard
TEMED/ammonium persulfate system.
Protein Visualization
Several methods have been described to detect protein
spots after 2D-PAGE. These methods use reactants
such as Coomassie Brilliant Blue R-250, Amido
Black, Ponceau S, Fast Green, negative staining, silver
staining, Suorescein and radioisotopes. The two most
popular approaches are Coomassie Brilliant Blue
R-250 and silver staining. A good stoichiometric rela-
tionship has been documented between protein
abundance and integrated optical density of protein
spots for Coomassie Brilliant Blue R-250. The silver
staining methods are more sensitive than those using
Coomassie Blue and can detect as little as 1}4 ng of
proteins. Several methods of silver staining of pro-
teins have been described, with the most rapid ones
being usually less sensitive and less reproducible than
the more time-consuming ones. Among the latter
methods, those using silver}diamine complex give the
most uniform sensitivity. However, they require
special home-made gels and cannot be applied to
several electrophoretic systems. For these reasons,
protocols based on silver nitrate are of more general
use and are favoured. A variety of systems using
different metal cations (K#, Cu2#, Zn2#) has been

Figure 2 Microheterogeneity of proteins. 2D gel of a cryoprecipitate containing fibrinogen (cryofibrinogen). A, albumin; AT,
1-
antitrypsin; T, transferrin; H, haptoglobin  chain; A-1, apolipoprotein A-1; F
, fibrinogen
chain; F, fibrinogen  chain; F, fibrinogen
 chain; F, extend fibrinogen  chain. Unknown protein spots are shown by arrowheads. All major identified proteins present charge
microheterogeneities. First dimension: immobilized pH gradient.
 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis 1369

developed to stain SDS-PAGE separated proteins

without the need for Rxative, organic dye or chemical
modiRer. SDS proteins stain negatively upon gel treat-
ment with solutions of heavy metal salts. The zinc
imidazolate reverse-staining method offers the ad-
vantage of combining good sensitivity, rapidity and
reversible interaction. Furthermore, the zinc
imidazolate reverse-staining method can be used in
situations where Coomassie Brilliant Blue R-250 or
silver staining is inappropriate or fail to produce
detection of the polypeptides of interest.

Protein Microheterogeneity
Polypeptides separated by 2D-PAGE rarely appear as
single spots, and most are resolved as multiple spots
characterized by charge and size microheterogeneities
(Figure 2).
Microheterogeneity is due to several factors that
frequently occur together. The Rrst cause of micro-
heterogeneity is genetic polymorphism where hetero-
zygote individuals express both forms of the gene
(Figure 3); the second is related to protein co- and
post-translational modiRcations. These modiRcations
are multiple and have all the potential of modifying
a protein’s charge, hydrophobicity, conformation
and/or stability. Furthermore, the ‘one gene one poly-
Figure 3 Genetic heterogeneity of plasma proteins. Close-up
peptide’ paradigm is challenged by the alternative
of the human plasma 2D map corresponding to the haptoglobin
splicing of many genes responsible for synthesizing  chain (pH 3.5}10 carrier ampholyte gradient). (A) L form
several proteins from a single gene. An important (homozygous); (B) L}R forms (heterozygous); (C) R form
feature of 2D gel analysis is that various protein (homozygous). A reference spot is shown by the arrowhead. This
isoforms generated by co- and/or post-translational genetic polymorphism is less detectable with immobilized pH
gradients. Note the microheterogeneity of haptoglobin  chain
modiRcations can be separated by isoelectric focusing
spots, which present both charge and size heterogeneities.
and/or by SDS-PAGE. Among the modiRcations
which lead to a charge-dependent change to a pro-
tein, acylation, alkylation, carboxymethylation, acid analysis, peptide mass Rngerprinting and/or
phosphorylation, sulfation, carboxylation, sialylation mass spectrometry. The development of automated,
and proteolytic processing are involved. Finally, high throughput technologies for the rapid identiRca-
glycosylation of proteins may lead to both charge tion of proteins is in progress. Automation already
and size modiRcations, and microheterogeneity of exists in several stages of the protein identiRcation
a protein may reSect the presence of several glyco- process. This includes automated acquisition of
forms. matrix-assisted laser desorption ionization-time-of-
Sight mass spectra and peptide mass Rngerprinting.
Bioinformation allows identiRcation of proteins by
Protein Identi\cation mixing several databases (a classiRed index can be
Several approaches have been used to identify pro- found at the following address: http://expasy.
teins after 2D-PAGE. Co-migration with puriRed hcuge.ch/alinks.htmlCProteins).
known proteins and Western blotting were employed
by the pioneers of the 2D Reld. The use of speciRc
antibodies and the recent developments of anti-
Data Management
gen}antibody interactions with enhanced chemi- Many investigators have analysed their 2D gels by
luminescence allow detection and identiRcation of holding two, sometimes three gels together towards
traces of proteins. However, monoclonal antibodies a light source, and tried to identify differences be-
may not detect denatured polypeptides. Nowadays, tween them (Figure 4). However, analysis of a multi-
proteins are identiRed by microsequencing, amino tude of 2D gels, with its bulk of information is greatly
1370 II / ELECTROPHORESIS / Two-dimensional Polyacrylamide Gel Electrophoresis

Figure 4 Analysis of 2D gels. 2D gels of human platelets from a single blood donor (pH 3.5}10 carrier ampholyte gradient). (A)
Platelets stored with leukocytes, 1 day after collection; (B) platelets stored in the presence of leukocytes, 7 days after collection; (C)
platelets stored in the absence of leukocytes, 7 days after collection. Without computerized analysis, it is not possible to draw
a definitive conclusion from the comparison of this set of gels. Reproduced with permission from Sarraj-Reguieg A, Tissot JD,
Hochstrasser DF, von Fliedner V, Bachmann F and Schneider P (1993) Effect of prestorage leukocyte reduction on proteins of platelets
obtained by apheresis. Vox Sanguinis 65: 279.

facilitated by the use of computer-based data process-
ing. The improvements in image acquisition and im-
age analysis allow clear spot detection, background
subtraction, spot matching and database construc-
tion. Furthermore, interpretation of 2D-PAGE im-
ages is facilitated by statistical methods, artiRcial in-
telligence and machine-learning programs. Ascendant
hierarchical classiRcation sorts the image into mean-
ingful groups. The use of correspondence analysis
and ascendant hierarchical clustering allows identi-
Rcation of new, potentially important proteins. Many
database of 2D gel master images are accessible
through the World Wide Web (Internet sites can be
found at the following addresses: http://www.ex-
pasy.ch/ch2d/2d-index.html or http://www-lmmb.
ncifcrf.gov/ABRF97//abrf3.html). It is also possible
to compare 2D gels from various laboratories, or 2D
gels with masters, on the World Wide Web by using
the Sicker created by P.F. Lemkin (accessible at
http://www-lecb.ncifcrf.gov/Sicker/).
Protein Functions
After 2D-PAGE analysis of cells, thousands of spots
are observed. Such an observation is frequently im-
pressive, but not very useful. IdentiRcation of the
polypeptide sequence corresponding to a spot is al-
ready undergoing major progress. However, under-
standing the protein’s function(s) remains the Rnal
goal of any analysis. It is also relevant to study the
expression level, the phosphorylation state, the sub-
cellular location, the association with other proteins
and the rate of synthesis or degradation of the
proteins of interest. Combination of all this informa-
tion will make possible the study of a functional
proteome.
Concluding Remarks
Amino acids are like letters. Amino acids make pro-
teins; letters make words. Proteins are like words.
Some are known, others are unknown. Proteome
databases are like dictionaries; they contain a lot of
information and are very useful. Organization of the
words makes the texts; organization and regulation
of protein production make the cells. 2D-PAGE is
a major proteomics tool. The technique should be
applied to resolve speciRc biological problems. It
should not be used for random investigations.
Further Reading
Bjellqvist B, Basse B, Olsen E and Celis JE (1994) Reference
points for comparisons of two-dimensional maps of pro-
teins from different human cell types deRned in a pH
scale where isoelectric points correlate with polypeptide
compositions. Electrophoresis 15: 529.
Chevallet M, Santoni V, Poinas A et al. (1998) New zwit-
terionic detergents improve the analysis of membrane
proteins by two-dimensional electrophoresis. Elec-
trophoresis 19: 1901.
Corbett JM, Dunn MJ, Posch A and GoK rg A (1994) Posi-
tional reproducibility of protein spots in two-dimen-
sional polyacrylamide gel electrophoresis using im-
mobilised pH gradient isoelectric focusing in the Rrst
dimension: an interlaboratory comparison. Electrophor-
esis 15: 1205.

GoK rg A, Boguth G, Obermaier C, Posch A and Weiss
W (1995) Two-dimensional polyacrylamide gel elec-
trophoresis with immobilized pH gradients in the Rrst
dimension (IPG-Dalt): the state of the art and the con-
troversy of vertical versus horizontal systems. Elec-
trophoresis 16: 1079.
GoK rg A, Boguth G, Obermaier C and Weiss W (1998)
Two-dimensional electrophoresis of proteins in an im-
mobilized pH 4}12 gradient. Electrophoresis 19: 1516.
Humphrey-Smith I, Cordwell SJ and Blackstock WP (1997)
Proteome research: complementarity and limitations
with respect to the RNBA and DNA words. Elec-
trophoresis 18: 1217.
Rabilloud T, Vuillard L, Gilly C and Lawrence JJ (1994)
Silver-staining of proteins in polyacrylamide gels: a gen-
eral overview. Cellular and Molecular Biology 40: 57.
Rabilloud T, Adessi C, Giraudel A and Lunardi J (1997)
Improvement of the solubilization of proteins in two-
EXTRACTION
Analytical Extractions
M. K. L. Bicking, ACCTA Inc., Woodbury, MN,
USA
Copyright ^ 2000 Academic Press
Introduction
The process of generating analytical data involves
some combination of planning, sampling, sample
preparation, quantiRcation, data review and report-
ing. Initially, each step in the method required com-
parable effort. Sample preparation, generally involv-
ing some form of extraction followed by analyte
enrichment, has in the past been a laborious process,
with only a few tools available. Likewise, quantiRca-
tion usually consisted of a ‘wet’ chemistry process
such as titration or precipitation. Before the develop-
ment of personal computers, the planning, sampling,
data review and reporting steps also required con-
siderable effort. Since each step presented formidable
challenges to the analytical scientist, the relative im-
portance of each step remained about the same.
Modern techniques, particularly chromatography,
have changed the situation. The rapid and successful
development of gas and liquid chromatography dra-
matically reduced quantiRcation steps from hours or
days to a matter of minutes, often with better accu-
racy and precision. The other steps in the method,
II / EXTRACTION / Analytical Extractions 1371
dimensional electrophoresis with immobilized pH gradi-
ents. Electrophoresis 18: 307.
Rabilloud T (1998) Use of thiourea to increase the solubil-
ity of membrane proteins in two-dimensional elec-
trophoresis. Electrophoresis 19: 758}760.
Sanchez JC, Rouge V, Pisteur M et al. (1997) Improved
and simpliRed in-gel sample application using reswelling
of dry immobilized pH gradients. Electrophoresis 18:
324.
Tissot JD and Spertini F (1995) Analysis of immuno-
globulins by two-dimensional gel electrophoresis. Jour-
nal of Chromatography A 698: 225.
Traini M, Gooley AA, Ou K et al. (1998) Towards an
automated approach for protein identiRcation in pro-
teome projects. Electrophoresis 19: 1941.
Wilkins MR, Williams KL, Appel RD and Hochstrasser DF
(eds) (1997) Proteome Research: New Frontiers in Func-
tional Genomics. Berlin: Springer-Verlag.
especially sample preparation, were regarded as of
secondary importance, serving only to support the
ultimate (i.e. chromatographic) step in the method.
Since most of the creative } and Rnancial } resources
Sowed into chromatography development, research
in the other areas slowed and sample preparation
came to be viewed as the ‘low tech’ part of the
method.
Chromatography is now considered a mature
science, being an integral part of nearly every analyti-
cal laboratory. The slower pace of chromatographic
research, coupled with outside pressures to improve
the efRciency of the entire analytical method, has
Rnally resulted in an increased interest in sample
preparation. These efforts have produced a number
of advances that improve efRciency, selectivity and
time required. The discussion will provide an
overview of some of the many sample preparation
principles and techniques available, focusing on the
analytical extraction part of the process. The goal is
to provide the reader with a more balanced view of
this important part of analytical methodology.
Principles of Extraction
Developing a successful extraction as part of an ana-
lytical method requires an understanding of the
chemical and physical principles involved. Thus, we
will begin this discussion of analytical extraction by
focusing on the underlying principles that make the
techniques work. Only with understanding and
1372 II / EXTRACTION / Analytical Extractions
appreciation of these principles can full advantage be
taken of them.
De\nition of Analytical Extraction
Extraction is the process of moving one or more
compounds of interest (analytes) from their original
location (usually referred to as the sample or matrix)
to a physically separate location where further pro-
cessing and analysis occur. The sample may be
a solid, liquid or gas. The separate location is usually
a Suid (an extracting solvent), but extractions into the
gas phase and on to solid sorbents are also common.
Finally, the word analytical implies that this process
involves small amounts of analyte (as opposed to
preparative extraction). Most analytical methods aim
at complete extraction although situations frequently
exist where good analytical results are possible with
only partial extraction.
Thermodynamics and Kinetics
These two terms are often interchanged, when in fact
they have very different chemical meanings. Thermo-
dynamics is the study of energy, in this case the energy
associated with the chemical process of extraction.
Through this study of energy, we can determine if the
process is favourable or unfavourable. That is, will
this extraction give a good result or a bad one? Even if
the process is favourable, it may not happen quickly
because of kinetic factors. Kinetics is the study of the
rate at which these processes occur.
It is important to realize that these two principles
are completely independent of each other. Complete
extraction is not necessarily a fast process and devel-
Figure 1 Solubility of amino acids in water as a function of structure.
opment of a successful extraction method requires
that consideration to be given to both aspects.
Like Dissolves Like
A compound will be soluble in, or mix with, another
compound that is chemically like it. That is, the two
compounds must be from the same, or similar, chem-
ical families. This simple principle is an implied re-
quirement in every analytical extraction. The concept
of moving analytes from the matrix to some other
location requires them to be transported using some
medium in which they are soluble. Therefore, we
must carefully consider how the like-dissolves-like
concept can help to achieve the desired result: extract-
ing the desired compounds and not extracting the
undesired ones.
Figure 1 illustrates how simple changes in molecu-
lar structure can have a profound inSuence on solu-
bility behaviour. This plot shows the solubility of
three related amino acids in water. Amino acids are
generally considered to be polar, so their solubility in
the polar solvent, water, is generally high. As nonpo-
lar functionality is added to the molecule, in the
sequence from glycine to phenylalanine, the nonpolar
character of the entire molecule increases (i.e. it be-
comes less like water). The result is a signiRcant
reduction in water solubility.
The same situation exists when considering the
relative solubility of any compound in a series of
solvents. A higher solubility will be observed when
the solvent is most like the compound in question.
The reader is referred to the Further Reading section
for additional examples.

Temperature Effects
Temperature has an effect on three important phe-
nomena: solubility, vapour pressure and kinetics.
While increasing temperature generally increases the
magnitude of each effect, there are some aspects of
this principle that are particularly relevant to analyti-
cal extractions.
Effect of temperature on solubility In most cases
involving organic analytes, increasing the temper-
ature of a liquid results in increased analyte solubility.
Figure 2 illustrates that for the same three amino
acids as shown in Figure 1 a temperature increase
from 0 to 753C results in a three- to fourfold increase
in solubility. Even marginally soluble compounds
show a dramatic improvement from this simple
change in conditions. Indeed, as will be seen later, this
principle is used in most extraction procedures.
Effect of temperature on vapour pressure Increasing
the temperature of a liquid will result in an increase in
vapour pressure. Boiling occurs when the vapour
pressure above the liquid equals the applied (usually
atmospheric) pressure.
Figure 3 shows calculated vapour pressures for sev-
eral common solvents. Note that the vapour pressure
is relatively large at temperatures as much as 203C
below the boiling point of the solvent. Simple evapor-
ation in a stream of nitrogen at room temperature
uses this fact to evaporate a solvent rapidly without
boiling. If the applied pressure is raised, the boiling
point is also raised, so that the solvent can be main-
tained in its liquid state at higher temperatures. Press-
Figure 2 Solubility of amino acids in water as a function of temperature.
II / EXTRACTION / Analytical Extractions 1373
urized Suid extraction uses this phenomenon to ad-
vantage. Similarly, lowering the applied pressure, as
in a rotary evaporator, reduces the boiling point,
allowing faster evaporation at lower temperatures.
Finally, at any given temperature, the relative vapour
pressure of each compound above the liquid phase
provides an estimate of the relative evaporation rates
of the liquids. Such knowledge is essential when per-
forming critical steps such as solvent evaporation or
solvent exchange.
Effect of temperature on kinetics All chemical pro-
cesses are affected by the temperature at which the
process is occurring, although the exact change in
reaction rate with temperature is unique for any pro-
cess. However, many reaction rates will approxim-
ately double for each 103C increase in temperature,
and this rule of thumb can be a helpful guide
in understanding the effects of temperature
changes. These changes can be either positive or nega-
tive, depending on whether the temperature change is
increasing or decreasing. For example, storing sam-
ples and solutions at low temperatures slows down
evaporation and degradation. These processes are
about four times slower if the solution is stored at 43C
compared to room temperature.
Effect of pH
The pH of an aqueous sample will inSuence the
success or failure of an extraction for acidic and basis
analytes. Acids and bases involve an equilibrium be-
tween two forms, one neutral and one ionic. Each
form has signiRcantly different chemical and physical
properties, as noted in Table 1.
1374 II / EXTRACTION / Analytical Extractions

Figure 3 Vapour pressaure of common solvents as a function of temperature. (Calculated from data in Handbook of Chemistry and
Physics (1971).)

Extraction of organic acids from water is only likely to change, often in an unpredictable way. Any
practical at pH values more than two units below the required pH adjustments and measurements must,
pKa of the acid. Only at this pH is most of the therefore, be made before the organic component is
compound in the neutral form and amenable to ex- added.
traction with an organic solvent. Similarly, to keep
a base, such as an aromatic amine, in the neutral Two-phase Distribution Equilibria
(extractable) form, the pH of the solution must be DeVnitions The process of extraction, by deRnition,
adjusted to at least two units above the pKb of the requires that the analyte be transferred from the
base. Readings with a pH meter are likely to be matrix to a different phase. When the extracting
unstable and/or unreliable in the presence of organic medium Rrst contacts the matrix, the analytes will
solvents, and the equilibrium constant, Ka, is also become distributed between the two phases in a well-
deRned ratio. Since the matrix is usually a liquid or
solid, and the extracting medium can be a solid,
Table 1 Properties of individual forms in acid}base equilibria
liquid or Suid, this usually refers to liquid}liquid and
Neutral form Ionic form liquid}solid distribution equilibria.
These distribution equilibria can be described by
HA#H2O 8 A\#H3O# several important equations. First, the distribution
Acid Conjugate base ratio, D, for extracting from phase 1 into phase 2 is
B#H2O 8 BH##OH\
Base Conjugate acid
deRned as:
More soluble in organic solvents Less soluble in organic solvents C2
Insoluble in water Soluble in water D"
More volatile Nonvolatile C1
Sour/bitter taste, bad odour Little odour
where C is the stoichiometric concentration of the
Reproduced with permission of ACCTA, Inc. analyte in each of the phases. (Actually, D is related
 II / EXTRACTION / Analytical Extractions 1375

to the ratio of activities rather than concentration,
but in dilute solution the difference is negligible.)
This ratio is a constant that depends on the analyte,
the two phases, the composition of the phases (pH,
ionic strength, etc.) and the temperature.
The fraction extracted,
, in any one equilibration
is deRned as:
D

"
1#D
where  is the phase ratio, the ratio of the volumes of
the two phases ("V2/V1). The fraction remaining in
the initial phase (V1) is, of course 1!
.
The amount extracted depends on the physico-
chemical interactions between the two phases and the
analyte, and the volume of each phase. A change in
these variables will cause a change in the extraction
result.
Effect of analyte structure on D Actual values of
D in Table 2 show how simple changes in molecular
structure have a profound inSuence on the success of
an extraction.
The addition of nonpolar functional groups
(methyl- and chloro-) to benzene make the molecule
more nonpolar, so that the new molecule favours the
hexane phase (larger value for D). Conversely, adding
polar groups (amine, hydroxy, carboxylic acid)
makes the molecule more like the water phase (small-
er D). It is important to keep these general principles
in mind when developing an extraction method and
understanding the results.
Table 3 Equations used for multiple extractions
Extraction number Fraction extracted Fraction remaining
into phase 2 in phase 1
1
1!

2
(1!
) (1!
)2
3
(1!
)2 (1!
)3
n
(1!
)n\1 (1!
)n
volume of solvent, although it is seldom worth carry-
ing out more than three extractions.
Effect of variations in D and  The effects of
variations in D and  on extraction results are shown
in Table 4. The total recovery after multiple extrac-
tions is calculated for various combinations of D and
. These calculations show the importance of all three
variables: phase ratio, distribution ratio and number
of extractions.
In summary, two-phase distribution equilibria are
an important part of every analytical extraction, and
the laboratory scientist must ensure that all critical
variables are controlled in order to generate reliable
results.
Other Principles
The preceding principles do not represent an exhaust-
ive list. Certainly, there are other chemical principles
that contribute to the extraction process, but play
a more minor role. Some of these are discussed brieSy
below.

Multiple extractions When multiple extractions are
performed on the same sample, the amount extracted
into phase 2 and the amount remaining in phase 1 are
calculated using the equations shown in Table 3.
In general, several extractions with the same total
volume of extracting solvent will always produce
better recovery than a single extraction with the same
Time A longer extraction time will usually produce
better recoveries, but this effect assumes that the
Table 4 Total per cent recovery as a function of , D and
number of extractions
Phase ratio D After 1st After 2nd After 3rd After 4th
"V2 / V1 extraction extraction extraction extraction

1/1 1 50 75 88 94
Table 2 Distribution ratios for extraction from water into hexane 2 67 89 96
10 91 99
Analyte Added functional Functional group D (25 3C) 100 99
group category
1/4 1 20 36 49 59
2 33 56 70 80
Benzene 275
10 71 92 98 99
Toluene }CH3 Nonplar 970
100 96 99
Chlorobenzene }Cl Nonpolar 950
Nitrobenzene }NO2 Moderately polar 31.2 1/10 1 9 17 25 32
Aniline }NH2 Polar 0.90 2 17 31 42 52
Phenol }OH Polar 0.13 10 50 75
Benzoic acid }COOH Very polar 0.051 100 91 99

Reproduced with permission from Sekine and Hasegawa (1977) Reproduced with permission of ACCTA, Inc.
1376 II / EXTRACTION / Analytical Extractions

analytes, reagents and solvents are nonvolatile, stable
and do not react with each other. If these assumptions
are not valid, longer extraction times may actually
produce poorer recoveries.
Ionic strength The addition of ionic species to an
aqueous solution results in a ‘salting-out’ effect. This
procedure often enhances extraction of neutral or-
ganic analytes from water by increasing D.
Surface area Reducing the particle size of a solid
matrix, thereby increasing contact areas between
phases, can cause a dramatic increase in extraction
rates.
Stirring/mixing Adequate stirring enhances the rate
of procedures that are otherwise limited by diffusion
processes.
Analytical Extraction Techniques
This discussion will focus on the most popular tradi-
tional techniques, and provide an introduction to
some of the newer extraction technologies. In each
case, the principles involved will be considered to-
gether with some practical operating tips.
wasted draining off one layer, except after the Rnal
equilibration.
E Shaking: the most important variable is the time
spent shaking the two layers, not the intensity of
the shaking. Because of this, automated shakers
provide adequate extraction, even though the
intensity of mixing may be quite low.
Continuous liquid+liquid extractors These systems
are usually reserved for larger water samples and/or
situations where a long extraction time is required.
There are two basic design types, depending on
whether the extracting solvent is more dense
(Figure 4) or less dense (Figure 5) than water. In each
design, the solvent in the Sask is heated to boiling,
causing solvent vapours to collect in the condenser.
The condensed solvent then passes through the
sample in the main chamber.
The principles are the same as for separatory funnel
extractions. However, each drop of solvent represents
a separate two-phase distribution system with a small
phase ratio but high surface area and fast extraction
kinetics. Since each drop represents an equilibration
step, the extraction consists of thousands of multiple
extractions. The result is generally a high analyte

Liquid+Liquid Extraction Techniques

Separatory funnel techniques There are few limita-
tions on what size or type of liquid samples can be
extracted, except that the two liquid phases must be
immiscible and unreactive with each other. Separ-
atory funnels are available to handle samples from as
small as a few millilitres to several litres. Extraction
times vary from 1 to 15 min, depending on the speci-
Rc requirements of the method, but equilibration is
usually fast in all but the most viscous liquids. As
noted in the section on principles of extraction, mul-
tiple extractions with a smaller volume of extracting
solvent are preferred over a single extraction with
a larger volume. The primary disadvantages of separ-
atory funnels are the labour necessary, the need to
evaporate an often large volume of solvent and the
formation of emulsions.
The following practical points should be con-
sidered:

E Funnel size: to allow adequate mixing, the Sask

size should be chosen so that at least 25% of the
funnel volume is free space.
E Venting: regular venting is a required safety pro-
Figure 4 Continuous liquid}liquid extractor for use with extract-
cedure, especially at the start of the extraction ing solvents that are denser than water. (Reproduced from Bur-
process. ford and Hawthorne (1994) Journal of Chromatography A 65:
E Draining layers: too much time should not be 75}94, with permission of ACCTA, Inc.)
 II / EXTRACTION / Analytical Extractions 1377

Israel) and the VectaSep CLE威 system (Whatman,

Inc., Clifton, NJ, USA).
The Mixxor2+ system (Figure 6) consists of a re-
ceiver and piston assembly. The aqueous sample is
placed in the receiver (B) with a small quantity of
immiscible organic solvent (D). The sample is extrac-
ted by moving piston (A) up and down a number of
times. After extraction, the plunger is moved to the
bottom and the separated organic solvent is forced
into the axial chamber (C), where it is easily removed.
The entire extraction and separation process is com-
pleted in less than 5 min and can provide a concentra-
tion factor of 30 or more. Extractors are available for
samples with volumes ranging from 2 to 50 mL. This
system is more convenient than separatory funnels,
although at the expense of some Sexibility in sample
and extraction solvent volumes. Also, the design pre-
cludes the use of heavier-than-water solvents.
The VectaSep CLE威 system (Figure 7) is parti-
cularly useful for smaller samples. The extraction
solvent is placed in the larger tube. The sample
(1.5 mL) is added to the sample dispenser, which is
then placed in the extraction tube and centrifuged,
typically for 10 min at 3500 rpm. Centrifugal force
pushes the sample through a dispersion membrane
in the bottom of the sample dispenser, causing the
sample to emerge as small droplets which travel along
Figure 5 Continuous liquid}liquid extractor for use with extract-
ing solvents that are less dense than water. (Reproduced with
permission of ACCTA, Inc.)

recovery. The primary disadvantages are the set-up

time, lengthy extraction time (6}24 h), and large
quantities of solvent. The latter problem has been
solved somewhat by integrated extraction systems
that allow extraction, evaporation and concentration
of solvent in one apparatus.
For continuous liquid}liquid extractors the follow-
ing considerations are important:

E The extractor only works properly if the con-

densed solvent passes through the bulk of the
sample, rather than along the sides of the Sask.
E The reSux (boiling) rate determines the overall
extraction rate, and some minimum rate must be
maintained to ensure complete extraction.
E Emulsions can be a problem. See below for ways to
deal with them.

Other liquid+liquid extraction devices While the

chemistry and mechanics of liquid}liquid extraction
have not changed, many practical variations have
improved the speed and convenience of the technique.
Two examples of these improvements are the Mix-
xor2+ extractor (New Biology Systems Ltd, Haifa,
Figure 6 MixxorTM extraction device. (A) Upper chamber;
(B) sample reservoir; (C) axial chamber; (D) organic solvent.
(Reproduced with permission of New Biology Systems Ltd.)
1378 II / EXTRACTION / Analytical Extractions
Figure 7 VectaSep CLE extraction system. (Reproduced with
permission of Whatman, Inc.)
the inside of the tube to the bottom. After centrifu-
gation, the sample dispenser is removed and the sep-
arating cup is added to locate the phase separation
boundary. Evaporation of the upper organic layer
then deposits extracted analytes in the separating cup,
where they can be readily redissolved in an appropri-
ate solvent. Although sample size and extraction sol-
vent choices are somewhat limited, this system makes
clever use of the extraction principles discussed
earlier. In this case, the sample is passed through the
extraction solvent (a reverse of the other methods) in
small droplets, increasing surface area (kinetics) and
offering a very favourable phase ratio.
These systems offer advantages in terms of sample
handling, safety and efRciency considerations. So,
despite their limitations, both of these systems, and
others like them, merit consideration as a replace-
ment for the more traditional techniques.
Emulsions No discussion of liquid}liquid extrac-
tion would be complete without mention of the emul-
sion problem. Emulsions are a mixture of two nor-
mally immiscible phases that won’t separate in prac-
tice. This problem is often caused by the presence of
surfactants or solids at the phase interface, high vis-
cosity of one of the phases, or a small phase ratio (not
enough organic phase). Although each emulsion is
unique, one of the following remedies will often result
in separation into two distinct layers:
E Wait: many emulsions will disappear with sufR-
cient time.
E Gentle mechanical agitation/stirring with a glass
rod or spatula.
E Immersion in an ultrasonic cleaning bath.
E Add ‘a salt’: this makes the aqueous phase less like
the organic phase.
E Increase the phase ratio: add more organic phase.
E Pass through a bed packed with glass wool or
diatomaceous earth.
E Centrifuge.
E Freeze the aqueous layer with dry ice/acetone or
liquid nitrogen, then simply pour off the organic
layer.
Liquid+Solid Extraction Techniques
Soxhlet techniques The Soxhlet extractor (Figure 8)
is one of the oldest extraction systems available but is
still very common. A solid sample is placed in an
extraction thimble inside the middle chamber.
Upon boiling, the solvent vapours from the bottom
Sask travel up to the condenser and then drip through
the sample. The sample is soaked in solvent (a
two-phase distribution equilibrium), which then
returns to the Sask when the liquid reaches the top
of the siphon. The sample is exposed to fresh solvent
after every siphon cycle, usually at a rate of about
six cycles per hour. Typical extraction times are
6}24 h. Once assembled and operating, there is
little that can go wrong with this system. However,
operators must be aware of the following general
hints:
E Proper cycling is required: the rate (cycles per hour)
is usually speciRed in the method, and the operator
must ensure that the unit siphons in distinct events
rather than continuously draining.
E Solvent level in the thimble: if too high, sample
may be lost from the thimble, contaminating the
extract.
E Moisture content: dry samples work best; add
a drying agent to remove free moisture.
The system requires a large volume of organic
solvent, and extraction time is long. Despite these
limitations, the Soxhlet extractor is still in widespread

Figure 8 Soxhlet extractor. (Reproduced with permission of
ACCTA, Inc.)
use, primarily because of its excellent reputation for
providing complete extraction. Indeed, Soxhlet
values are often used as the standard against which
other extraction methodologies are compared.
ModiVed Soxhlet extractors The lengthy Soxhlet
extraction times have prompted the development of
modiRed extractors, such as the Soxtec威 system (Foss
Tecator AB, HoK ganaK s, Sweden). The sample is placed
in an extraction thimble, but the thimble is then
directly immersed in boiling solvent, rather than
bathed in cooler condensed solvent. The increased
temperature means faster extraction kinetics. After
about 1 h equilibration, the sample is removed from
the solvent and Sushed with fresh condensed solvent
for an additional hour. The apparatus even allows
II / EXTRACTION / Analytical Extractions 1379
the evaporation and collection of solvent, further
improving efRciency. These alternatives offer
considerable advantages in terms of time and solvent
use, and results are generally comparable to the tradi-
tional method.
Solid-phase extraction Solid-phase extraction (SPE)
is an alternative to liquid}liquid extraction where the
extraction solvent is replaced with a solid sorbent.
The sorbent is usually packed into a cartridge
(Figure 9) that can vary in size from about 1 mL to
more than 50 mL. The quantity of sorbent can range
from about 500 mg to 10 g. Extraction is accomp-
lished by forcing the aqueous sample past the sorbent
(via vacuum or pressure), causing analytes in the
sample to be sorbed. This two-phase distribution is
similar to the partitioning that occurs in chromatog-
raphy. After the sample has passed through the
sorbent bed, the sorbed analytes are eluted with
a strong solvent, such as methanol, acetonitrile or
carbon disulRde.
The SPE process involves the following sequential
steps:
E Conditioning/cleaning of the sorbent with an or-
ganic solvent such as methanol
E Extraction of the sample.
E Air drying or rinsing to remove any remaining
sample.
E Elution of analytes using a strong organic solvent.
SPE offers three primary advantages over conven-
tional liquid}liquid extraction: reduced solvent
usage, extraction speed and selective chemistry. In an
ideal method, only a few millilitres of organic solvent
may be necessary for an extraction and it is possible
to extract and elute 10 100-mL samples or more in as
little as 15}20 min. Finally, by varying the nature of
the sorbent, it is possible to achieve selective extrac-
tion and/or selective elution. For example, a minor
change in bonded phase from a C18 phase to
a C8 phase can actually result in a signiRcant change
in selectivity. The shorter chain C8 phase is less reten-
tive towards more hydrophobic molecules and ex-
poses somewhat more of the polar character from the
underlying silica. This trend can be extended using
even shorter aliphatic bonded phases or by adding
a polar functional group to the chain (e.g. cyano- or
phenyl-). There are no analogous series in liquid}
liquid systems.
There are a host of sorbents available, including
more polar functional groups, polymer-based, ion
exchange, afRnity and chelating materials. Nearly
every liquid}liquid extraction method has an SPE
counterpart, and almost all provide equal if not better
1380 II / EXTRACTION / Analytical Extractions
Figure 9 Solid-phase extraction (SPE) cartridge design. (Reproduced with permission of ACCTA, Inc.)
results, with considerably less effort. It should also be
noted that SPE can be performed on nonaqueous
samples using a polar sorbent, but this application is
usually used for sample clean-up rather than extrac-
tion.
Finally, it is important to note SPE’s limitations:
E High particulate samples will often plug the frits.
E Extracting capacity (total extractable mass) is
more limited than with conventional solvent ex-
traction.
E Reproducibility (batch-to-batch) can be a problem,
although this is less of a concern now than during
early development of the technique.
Membrane disc extraction Membrane extraction
discs, Rrst sold under the brand name Empore威 (3M,
St Paul, MN, USA), are an alternative SPE system. In
the membrane discs the sorbent is enclosed in a sup-
port network rather than simply being packed into
a cartridge. The unique Empore威 design consists of
90% (w/w) sorbent particles (8}10
m diameter), in
a network of polytetraSuoroethylene Rbrils, in a disc
format that resembles a thicker version of conven-
tional synthetic membrane Rlters. A typical disc is
about 0.5 mm thick with diameters ranging from 1 to
90 cm.
A membrane disc extraction method would typi-
cally consist of the following steps:
E Pre-washing the disc with the Rnal eluting solvent.
E Pre-wetting the disc with methanol or some other
solvent that is miscible with the sample (which is
usually aqueous).
E Extraction, i.e. drawing the sample through the
disc.
E Elution of analytes, which involves a soak with the
elution solvent for a period of time, followed by
elution with the aid of a vacuum. This elution
step may be repeated with different solvents if
necessary.
Membrane discs have the same advantages over
liquid}liquid extraction as SPE but are superior to
conventional SPE because the extraction rate is faster;
Sow rates of 100}200 mL min\1 are typical. The
small particles also provide greater capacity and uni-
formity of packing. Unfortunately, the discs are more
sensitive to the presence of particulates, so a pre-Rlter
is often necessary.
Early applications of membrane discs focused on
environmental analysis, where large sample sizes
made the fast extraction rates attractive. Membrane
discs can also be formulated into SPE-like car-
tridges, allowing the efRcient processing of small cli-
nical samples (e.g. serum, urine, etc.). Use of
membrane disc applications continues to grow, al-
though the number of reported applications is not as
high as SPE, due to the relative age of the two tech-
niques.
Solid-phase microextraction Solid-phase microex-
traction (SPME) is another version of liquid}solid
extraction techniques. In this system, the extraction

Figure 10 Solid-phase microextraction system. (Reproduced with permission of ACCTA, Inc.)
phase consists of a fused silica Rbre coated with
a sorbent (e.g. dimethylsilicone or other immobilized
polymer) with thicknesses ranging from about 10 to
100
m (Figure 10). The Rbre is placed in contact
with the liquid or gas sample and analytes are sorbed
on to the phase, from which they are directly desor-
bed into a chromatograph.
Unlike the other techniques discussed here, SPME
is a completely solvent-free extraction method. The
extraction step tends to be rapid, usually requiring
10}20 min, and desorption can take only a few sec-
onds. Thus, a fast analysis with a good lower limit of
detection is possible, since the entire extract is ana-
lysed. Furthermore, the selectivity and extractability
can be affected by changes in Rbre chemistry as well
as solution pH, ionic strength, etc. Sampling by im-
mersion in the sample or extraction from the head-
space above the liquid (often a faster extraction)
provides additional Sexibility. However, SPME, by
its nature, is not applicable to as wide a range of
samples as other techniques, and there are currently
a more limited number of sorbents available
compared to SPE. However, SPME offers some
unique advantages that make it an attractive alterna-
tive for many applications.
Pressurized Fluid Extraction
So far, each extraction medium discussed has been
either a conventional liquid or solid, each performing
its function at or near room pressure and at or below
the boiling point of the liquid phase. Experimentally,
these conditions are easiest to attain in the laboratory
and require relatively unsophisticated equipment.
Unfortunately, in some cases these conditions can
also result in slow extraction kinetics and/or incom-
plete extraction because of the (relatively) mild condi-
tions employed. Such problems can often be solved
II / EXTRACTION / Analytical Extractions 1381
simply by maintaining the extraction Suid at a
higher pressure so that higher temperatures can be
used. This can result in a dramatic improvement in
extraction efRciency. In addition to the pressurization
of conventional solvents in a closed vessel, supercriti-
cal Suids may also be used at high temperature and
pressure.
Accelerated liquid extraction As noted earlier, in-
creased temperature improves solubility and extrac-
tion kinetics, and increases the vapour pressure. In
addition, an increase in applied pressure causes an
increase in the boiling point of a liquid. Logically,
then, one would expect improvement in extraction
results at higher applied pressure, where the increased
boiling point would then allow liquid extractions
above the normal boiling point of the solvent. This
approach has been successfully applied in two differ-
ent ways.
In the Rrst method, called microwave-assisted sol-
vent extraction (MASE), the sample and extraction
solvent are placed in a sealed vessel, usually construc-
ted of polytetraSuoroethylene or other inert polymer.
When placed in a microwave Reld, polar materials
(e.g. water) absorb energy and the sample heats up.
Since the vessel is closed, the pressure also increases,
resulting in a signiRcantly elevated boiling point. For
example, hexane}acetone mixtures can be used at
1153C, which is more than 403C above the boiling
point of either solvent. The increased temperature has
several beneRcial effects on the extraction, such as
increased solubility, faster diffusion, reduced viscos-
ity and reduced surface tension (increased wetta-
bility).
This approach is also used for sample digestion in
inorganic analysis, and succeeds for the same reasons
} temperature-related improvements in reaction
rates.
1382 II / EXTRACTION / Analytical Extractions
The second approach uses conventional electrical
conduction heating in a sealed stainless steel vessel
(Dionex Corporation, Salt Lake City, UT, USA) to
accomplish the same effect. With this equipment, the
pressure and temperature can be set independently,
whereas in the MASE process the pressure increase
results from the temperature increase.
Both methods allow Soxhlet-type extractions to be
completed in 30 min or less and require small solvent
volumes. This approach has received widespread ac-
ceptance because it draws from existing experience
with organic extraction solvents. The basic chemistry
of the extraction does not change signiRcantly, only
the rate. In theory, then, any liquid solvent-based
extraction method could be adapted for accelerated
liquid extraction.
The primary disadvantages involve the safety as-
pects associated with the use of organic solvents at
high temperatures and pressures. In addition, the
same process that enhances analyte extraction may
also cause extraction of other unwanted components
from the matrix. However, the reductions in solvent
use and extraction time make accelerated liquid ex-
traction an attractive alternative to unpressurized
techniques.
Supercritical Wuid extraction As the temperature
and pressure on a compound are raised, a point is
reached, called the critical point, where the substance
is no longer a gas or liquid, but has properties inter-
mediate between these two states. Supercritical Suids
are good solvents with gas-like viscosities and dif-
fusivities and no surface tension.
Carbon dioxide is the most popular choice for
a supercritical Suid, because of its relatively low criti-
cal point (313C, 73 atm). Supercritical Suid extrac-
tion (SFE) then involves placing the sample in a high
pressure vessel and contacting the sample with the
supercritical Suid. Extraction temperature can be var-
ied from about 403C to more than 1503C while pres-
sures may be adjusted between 100 and as high as
680 atm or more. Since carbon dioxide is actually
a nonpolar Suid, 10}20% of polar modiRers such as
methanol can be added to improve the range of
solubilities. A typical extraction will be complete in
less than 20 min.
SFE receives much attention because the extrac-
tions are fast and, with carbon dioxide as the extrac-
tion Suid, evaporation of the extracting medium is
spontaneous upon decompression to atmospheric
conditions. The ability to control extracting power,
through changes in temperature, pressure and modi-
Rer, offers more selectivity and Sexibility than with
liquid solvents. While SFE is not a universal replace-
ment for liquid solvent-based methods, it is clearly
the best choice for many speciRc applications, espe-
cially foods, natural products, polymers and environ-
mental samples.
Final Comments
Analytical extraction (and sample preparation in gen-
eral) has returned to its rightful place as an equally
important part of the analytical method. There are
many options available to achieve extraction, de-
pending on the type and size of sample as well as
other more practical considerations. The laboratory
worker can choose from 100-year-old techniques
that still provide excellent results, or instrumental-
based technologies that offer faster extractions on
smaller samples.
The extraction step, as a distinct part of the analyti-
cal method, will retain its importance as long as
chromatographic procedures are used for quantiRca-
tion. The objective of moving the analytes from the
sample to the point of quantiRcation will still be
required. Future research is likely to focus on better
ways of accomplishing this movement, resulting in
reduced solvent usage and sample size, automation
and online transfer of extracts to subsequent process-
ing and quantiRcation steps. But throughout these
changes, it will be important to remember that, al-
though the names may change, the chemistry will
remain the same.
Further Reading
Freiser H (1973) Solvent extraction. In: Karger BL, Snyder
LR and Horvath C (eds). An Introduction to Separation
Science, Ch. 9. New York: Wiley-Interscience.
Lopez-Avila V, Young R and Teplitsky N (1996)
Microwave-assisted extraction as an alternative to Sox-
hlet, sonication, and supercritical Suid extraction. Jour-
nal of the Association of OfTcial Analytical Chemists
International 79: 142}156.
Peleg I and Vromen S (1983) An efRcient novel device for
solvent extraction. Chemistry and Industry 61:
615}616.
Sekine Y and Hasegawa Y (1977) Solvent Extraction
Chemistry, p. 105. New York: Marcel Dekker.
Weast RC (ed.) (1971) Handbood of Chemistry and Phys-
ics, p. C-743, D-151. Cleveland, OH: Chemical Rubber
Co.
Zhang Z, Yang MJ and Pawliszyn J (1994) Solid phase
micro-extraction. Analytical Chemistry 66(17):
844A}853A.
 II / EXTRACTION / Analytical Inorganic Extractions 1383

Analytical Inorganic Extractions

K. A. Anderson, Oregon State University, Corvallis, Hydrometallurgical applications of liquid}liquid
OR, USA inorganic extractions are numerous and remain the
Copyright ^ 2000 Academic Press contemporary choice of separation for many pro-
cesses today. In addition, as environmental regula-
tions develop, increased interest in recovery methods
for metals from a variety of waste streams will
Historical Development no doubt renew interest in metal separation
techniques.
Trace elemental analysis is under constant develop- The prevailing industrial use of inorganic solvent
ment and the challenging analyses of today become extraction includes the separation of the lanthanide
the routine of tomorrow. Despite recent and rapid (III) ions. Individual lanthanides are widely used in
advances in analytical instrumentation, it is still ne- many of today’s ‘high-technology’ applications for
cessary in many applications to use separation and example, lasers (neodymium in yttrium}aluminum
preconcentration techniques prior to the analytical garnet), superconducting materials, specialty ceram-
determination. Typically, the reason for performing ics, catalyst, the nuclear industry and colour video
a separation and/or preconcentration step is to bring phosphors.
the concentration of the trace element to a detectable
level and/or separate it from interfering substances in
the sample matrix. Separation and preconcentration Inorganic Processes
are therefore a frequent component of an analytical
Solvent extraction, ion exchange, volatilization and
scheme. Inorganic solvent extractions are also used
precipitation are the most commonly used separation
extensively in industrial applications. Moreover a
approaches for trace elemental analysis. Inorganic
rapidly developing Reld, elemental speciation, will
preparation schemes generally follow a Sow diagram,
depend in part on sophisticated separation techniques
as shown in Figure 1.
such as liquid}liquid inorganic extractions.
The processes by which extraction of inorganic
Inorganic solvent extractions have been known and
compounds occur using organic solvents are varied
performed since the nineteenth century. The extrac-
and may be relatively involved. Consequently, at-
tion of uranyl nitrate into diethyl ether was reported
tempts to classify inorganic extraction processes
in the 1840s, but it was some time later before quant-
are difRcult. Attempts have been made, based on
itative understanding of the inorganic liquid}liquid
the identity of the extracted compound, or of the
extraction distribution equilibria was forthcoming.
extracting agent, or pH of the extraction solution.
Nernst presented the thermodynamical explanation
For the purposes of this chapter, a simple sub-
of the distribution in the 1890s. Chelate extraction
division will be adopted based on the extraction
was also developing at this time, most notable
reagent used.
was the use of 1,5-diphenylcarbohydrazide which
chelated with chromium. The work of Fisher in the
Extraction Considerations
1920s with dithizonates is noteworthy; this group
studied the distribution of the elements as a function The need for separation and/or preconcentration
of reagent concentration, metal, complexing agents in trace metal analyses are fundamentally related
and pH. During this time, a wide array of solvent to available instrumentation and instrumental
extraction methods was developed.
Rapid and distinguished progress in inorganic
extractions occurred during World War II, most
as part of the ‘Manhattan Project’ research in
atomic energy. One of the most signiRcant applica-
tions of liquid}liquid extraction in inorganic
chemical technology was the separation of uranium
and plutonium from nuclear reaction Rssion products
in the late 1940s. Later, inorganic extractions re-
placed ion exchange at the beginning of the nuclear
fuel cycle for separating uranium from other leach
liquors. Figure 1 Inorganic preparation flow diagram.
1384 II / EXTRACTION / Analytical Inorganic Extractions

Table 1 Comparison of a select list of analytical techniquesa

Comparison ASV Flame AAS ET-AAS ICP-AES NAA ICP-MS

General 0.1}0.01 mg L\1 1}5000 mg L\1 0.1}0.001 mg L\1 1}100 mg L\1 1}1000 mg L\1 0.01}1 mg L\1
detection (ppb) (ppb) (ppb) (ppb) (ppb) (ppb)
limits
General Excellent, select Moderate, Excellent, Moderate, Moderate Excellent
sensitivity elements refractories poor refractories limited refractories
excellent
Instrument Well established Well established Well established Established Established New and growing
maturity and growing
Interferences Some Few, well Many, controllable Spectral Few Moderate, mass
understood overlap
Instrument Readily Readily Readily Readily Specialized Specialized
availability laboratory laboratory
Instrument- Developed Numerous, Well developed Developed Well developed Undeveloped
specific well developed
inorganic
extraction

a
Abbrevations: ASV, anodic stripping voltametry; Flame AAS, flame atomic absorption spectrometry; ET-AAS electrothermal atomic
absorption spectrometry; ICP-AES, inductively coupled plasma atomic emission spectrometry; NAA, neutron activation analysis; ICP-MS:
inductively coupled plasma mass spectrometry.

capabilities. Basically, separation and/or preconcen-
tration are needed when one of the following situ-
ations occurs: concentration of analyte is below the
sensitivity of the instrumental method; interferences
exist in the sample (relative to the instrument to be
used); or physical or chemical states of the sample are
not appropriate for the instrument. Sensitivities for
elements varies with the instrumental method and are
relative to matrix type; however, a general listing of
sensitivities of commonly used analytical equipment
is given in Table 1.
The impetus for doing an extraction will therefore
depend on the availability of instruments and the
capability of the instrument relative to the matrix
type (i.e. interferences). Inorganic extraction schemes
are typically instrument speciRc. Although instru-
ment development has signiRcantly reduced detection
limits, availability of some of the more state-of-the-
art equipment is still limited to specialized or well-
equipped laboratories. The need for separation and
preconcentration therefore still exist. Speciation stud-
ies will also continue to support development and
research into inorganic separations from complex
matrices.
Because atomic absorption spectrometry (AAS) is
readily available, but the detection limits are high in
relation to today’s needs, there are numerous solvent
extraction methods available for metals in AAS anal-
ysis. Several excellent sources are listed in Further
Reading; these have lengthy tables of inorganic ex-
traction schemes.
Theory and Equations of Inorganic Solvent
Extraction
The solvent extraction process to separate and/or
preconcentrate an analyte of interest is performed by
using two immiscible solvents. A complex (typically
neutral in charge) is formed with the element of
interest, typically in an aqueous solution and will
partition into a mutually insoluble (organic solvent)
phase. The Nernst partition (or distribution law)
states that at equilibrium a given solute will be dis-
tributed between two essentially immiscible liquids
according to the following equation:
KD"
o[A]o/
aq[A]aq
where KD is the distribution coefRcient (also called
the partition coefRcient) and [A] is the concentration
of the analyte,
are activity coefRcients, subscript ‘o’
denotes organic phase and the ‘aq’ subscript denotes
aqueous phase. The above equation holds true in only
the most rigorously well-deRned thermodynamic sys-
tems. For simplicity the relationship assumes that no
side reactions occur in either the aqueous or organic
phase and that no stable intermediates are formed
with the analyte (e.g. metal) of interest. From a prac-
titioner’s standpoint, the total amount of analyte (e.g.
metal) transferred from one phase to the other is of
most interest. An empirical distribution ratio, D, is
deRned by the simpliRed relationship given below:
D"[AT]o/[AT]aq
 II / EXTRACTION / Analytical Inorganic Extractions
where [AT]o includes all (T"total) complexes of the
analyte of interest in the organic phase and activity
coefRcients are assumed unity. The assumption of

"1 for chelating extractions is reasonable. How-
ever, for ion-pair extractions where the electrolyte
concentration is high, to assume unity for the activity
coefRcients is a poor assumption. The simplifying
relationship is still often employed, however, with
the assumption that the (
o/
aq) ratio will remain
constant.
The extraction efRciency, %E, which deRnes the
amount of analyte transferred from the aqueous
phase to the organic phase, is deRned as follows:
% Extraction"100D/D#(Vaq/Vo)
where Vaq is the volume of the aqueous phase and
Vo is the volume of the organic phase. An important
property of the above relationship is that the extrac-
tion efRciency is independent of the initial analyte
concentration. High extraction efRciencies can be
achieved when the Vaq/Vo ratio is small (that is, small
aqueous volumes used with large organic volumes).
There is of course a practical limit to this approach.
Multiple extractions with reasonable volumes per-
form better than a single extraction with one large
volume. Large values of D, distribution ratio, corres-
pond to high extraction efRciencies (e.g. D"100
then %E"99%, D"0.1 then %E"10%, for
a 1 : 1 volume ratio).
An extraction reaction may be described by the
general chemical equation given below:
#
aq #nHXAXno#nH
An#
where A is the analyte of interest (e.g. metal ion) with
charge n#, HX is the extracting agent (e.g. chelating
agent). Note, that extracting agents are often acidic.
From the following extraction reaction the equilib-
rium constant, Kex is:
Kex"[AXn]o[H#]n/[An]aq[HX]no
By substituting the distribution ratio, D, the equation
simpliRes to:
Kex"D[H#]n/[HX]no
hence:
D"Kex[HX]no/[H#]n
The logarithmic form for the distribution coefRcient
is then:
log D"log Kex#npH#n log[HX]o
1385
From this model, for a given system, the degree of
extraction increases as the concentration of the
chelate [HX]o increases. Extraction increases with
increasing pH (decreasing hydrogen concentration) in
the aqueous phase. A one unit increase in pH results
in a factor of 10 increase in the distribution
coefRcient for n"1; for n"2, the distribution co-
efRcient increases by a factor of 100. Hydrolysis of
the metal ion and decreased solubility of the chelate
occur at high pH limiting this general approach.
Plots of log D versus pH (or %E versus pH or
versus pH1/2) are often used to deRne extraction sys-
tems. These types of plots produce sigmoidal curves,
with the overall position relative to the pH axis de-
pendent on Kex with the slope"n. For purposes of
comparison, if D"1 (i.e. E"50%) and [HX]o"1,
the pH is constant and equal to log Kex/n. This term is
referred to as pH1/2, and is characteristic of the ex-
traction process. Analyte/chelate agent values of
pH1/2 are often cited and are used as a measure of the
feasibility of separating two analytes.
Further theoretical discussion is beyond the scope
of this chapter but includes topics on solvent proper-
ties, such as the solvent Hildebrand parameter, sol-
vent dielectric constant, and complex properties such
as the complex size, polarity and polarizability, as
well as pH, temperature and reaction kinetics.
Inorganic Solvent Extractions
The essential prerequisite for an element to be extrac-
ted from an aqueous solution is that it be part of
a neutral complex. Charge neutrality reduces the elec-
trostatic interactions between the element (analyte of
interest) complex and water and therefore lowers the
solubility of the complex in water. Consequently, the
neutral complex can be extracted into the less polar
organic solvent. General attributes and chemical
properties that can facilitate separation are: charge
neutrality, increase size of complex formed and incor-
poration of hydrophobic or organophillic properties.
These general attributes can be accomplished by sev-
eral mechanisms including the element associated
with naturally occurring complexants, chelate com-
plexes formed with analyte and ion-associated (ion-
paired) complexes. All three mechanisms can and will
form a neutral complex with the analyte of interest.
Depending on the ligands (complex associated with
the element) other chemical properties such as com-
plex size and hydrophobic/organophillic properties
can be incorporated.
Extraction Schemes
ClassiRcation schemes are numerous and no one
scheme covers all systems. The most common
1386 II / EXTRACTION / Analytical Inorganic Extractions
schemes are based on the form of the extracted ele-
ment that transfers into the organic phase. This sim-
pliRed classiRcation scheme is adequate for the dis-
cussion here. However, even with simple schemes the
categories are not exclusive, and some extraction
agents could be classiRed into other categories. The
experimental process of inorganic extraction of
a neutral complex, regardless of the type of complex,
is essentially the same. The neutral complex’s interac-
tion with the aqueous phase, including but not lim-
ited to the solubility, depends on the charge and
polarity of the overall complex. The Rrst step is to
generate a neutral complex with the analyte of inter-
est through one of the mechanisms listed above.
A small volume of organic solvent is added to the
sample mixture. For example, a 1 L aqueous sample
may be extracted into 40 mL of organic solvent. Ex-
traction can be performed in a separatory funnel or
by using a mechanical shaker table. The pH of the
mixture may need to be manipulated, depending on
the exact extraction scheme used. In addition, mask-
ing agents may be used to obtain speciRcity (see
below). After mixing, the two phases are separated
and the procedure is generally repeated several times.
The organic phase is combined from each extraction.
The concentration of the elements in the sample is
increased by 1}3 orders of magnitude in the organic
phase. The extract can then be further pre-concen-
trated if needed (back-extraction, evaporation, etc.)
or analysed directly, for example by Same AAS.
Naturally occurring complexants Elements that can
form neutral complexes can already exist in naturally
occurring water systems. These complexes are formed
essentially with covalent bonding between the ele-
ment and naturally occurring ligand(s). Ligands are
molecules or ions bonded to a central metal ion and
tend to be Lewis bases; also included in this category
would be undissociated covalent species. Examples
of this category would include I2 and B2, the halides
of some metals (GeCl4 HgCl2, AsCl3) and oxides of
some metals (OsO4). The extraction of these types of
compounds would proceed in the same manner as for
chelates and ion-associated complexes.
Chelates A chelate is a type of ligand. A multiden-
tate (dentate is Latin for tooth) ligand that uses more
than one atom to bind to a metal in a coordination
complex, see Figure 2. The metal is the electron-pair
acceptor and the chelating agent the electron-pair
donor. When binding to the metal ion, the chelate
(ligand) forms a ring of atoms, of which the metal is
one member. The chelate complex charge exactly
neutralizes the charge on the metal ion. Most rings
contain '4 and (8 atom members; the most stable
Figure 2 Chemical structures of typical chelate}metal com-
plexes.
typically is a 5-membered ring. Bidendate describes
a chelate where two atoms from the chelate complex
bond to the metal and tridentate would indicate three
coordinating atoms. Many chelating extractants are
weak acids, therefore, control of pH is important in
many extracting schemes.
An exhaustive treatment of every chelate system is
beyond the scope of this chapter. Table 2 lists a selec-
tion of chelate types with one or two speciRc chelate
agents listed below these. To describe the selectivity
of each is not possible in a brief chapter, a sense of the
ability of each chelate reagent is given by listing the
wide range of complex-forming metals that are poss-
ible. Detailed information about the selectivity, sol-
vent and other experimental conditions can be found
in the references listed in Further Reading. The list in
Table 2 include inorganic extraction procedures for
a wide array of instrumental methods including:
Same AAS, electrothermal (ET)-AAS, inductively
coupled plasma-atomic emission spectroscopy, (ICP-
AES), neutron activation analysis (NAA), spectro-
photometric, chromatographic, Same photometry,
and polarography. In addition, most of the chelate
groups listed in Table 2 are compatible with more
than one organic solvent. Solvent Sexibility in an ana-
lytical scheme allows an extended range of instrumen-
tal methods which can be used for the determination.
Inorganic extraction, utilizing chelates, for analyti-
cal separation and/or preconcentration has been ex-
ploited for many instrumental systems. For Same
AAS analysis, often the inorganic solvent extraction is
designed to increase the concentration of elements of
interest and, most importantly, reduce the concentra-
tion of alkali and alkaline earth elements (i.e. leave
most of them in the aqueous phase). This separation
 II / EXTRACTION / Analytical Inorganic Extractions 1387

Table 2 A select list of inorganic extraction systems Ion association (ion pair) Neutral complexes can be
formed through ion association (ion-pair) and extrac-
Metals ted from an aqueous solution into an organic solvent.
extracted
Ion association inorganic extracts encompass a wide
Chelating agents range of extraction schemes. General sub-groupings
Oxines include chelated ion pairs, nonchelated ion pairs and
-8-Hydroxyquinoline '50 metals halide}cation ion-pairs. The halide}cation pairs are
-(and derivatives)
typically extracted into oxygen-containing solvents,
-Dioximes such as methyl isobutyl ketone, diethyl ether and
-Dimethylglyoxime Ni, Pd, Co alcohols. A select list of ion pair extracting agents is
Dithizones given in Table 2.
-Diphenyldithiocarbazone '30 metals Maximizing the coulombic forces of attraction be-
Dithiocarbamates tween the ion pairs facilitates extraction of the ion
-sodium diethyldithiocarbamate '50 metals pairs. The dielectric constant of the solvent is a large
-Sodium N N -phenylacetyldithiocarbamate '50 metals contributor to the overall extractability of a scheme.
-Diketones Enhancement of the extraction of ion-associated
-Acetylacetone '50 metals complexes is increased by the addition of electrolytes,
-Thenoyltrifluoroacetone '50 metals
called ‘salting-out’. The salting-out effect may be at-
Nitrosoarylhydroxylamines tributed to the increase in anion concentration, as
-Ammonium N-Nitroso- '30 metals well as the decrease of the dielectric constant of the
N-phenylthydroxylamine
(cupferron)
aqueous phase. Complexes can be formed by ligands
coordinated to the metal and an appropriate counter
Organophosphorus acids
anion that neutralizes the total charge. One of the
-di-n-butylphosphoric acid '30 metals
-Di(2-ethylhexyl)phosphoric acid '30 metals ions (either the complexed ligand or the anion) typi-
cally contains a large hydrophobic group(s) which
1-Nitroso-2-naphthol Co(II)
further enhances extraction of the ion pair into the
1-(2-Pyridylzao)-2-naphthol (PAN) '50 metals organic phase.
Ion-pair agents
Factors Affecting Inorganic Solvent Extractions
Chelated ion-pairs
-ethylenediaminetetraacetic '30 metals Control of pH is critical to ensure conditions are
acid (EDTA)/halide
favourable for the formation of the desired com-
-1,10-phenoanthroline/perchlorate Fe (II)
plexes. The extraction speciRcity needed inSuences
Non-chelated ion-pairs the acceptable range of pH. Many inorganic ex-
-tetraalkylammonium salts '50 metals
-tetraphenylarsonium salts '50 metals
traction schemes use buffers. The lack of a buffer in
an inorganic extraction should be viewed with suspi-
Halide ion pairs
cion, since the quantity of metal extracted is strongly
-HCI '50 metals
-HF '50 metals pH dependent. In addition, chelating agents will alter
-HI '50 metals the pH of the solution. Several buffers have been used
for inorganic extractions for AAS determination, in-
cluding borate, phosphate, citrate, acetate and forma-
te. Acetate should not be used if lead or silver or other
is especially necessary for many natural water sam-
stable metal acetates are to be determined. Buffers
ples such as seawater, brines, etc. Trace element ana-
can be a signiRcant source of contamination, as can
lyses of clinical samples such as blood, urine, etc.,
any unpuriRed reagent added to a sample.
also beneRt from inorganic extraction for Same AAS
The nature of the solvent is of special importance
analysis as well as other determination techniques
for inorganic extractions. There are several criteria
(ET-AAS, ICP-AES, etc.). Radiochemistry separ-
which should be evaluated when choosing a solvent
ations for NAA also often use inorganic extraction
for an inorganic extraction. The solvent should have
techniques that utilize chelating schemes.
the following characteristics:
Extraction schemes have also been developed
which leave the analyte of interest in the aqueous E Extracts the desired metal chelates
phase and remove interferences through the organic E Immiscible with aqueous solution (i.e. low solu-
phase. This technique has limited applicability owing bility in water); for convenience, density'water if
to the limited solubility of the (starting reagent) the sample is drawn off
chelate in organic solvents. E Does not form emulsions
1388 II / EXTRACTION / Analytical Inorganic Extractions
E Compatible with the analytical determination tech-
nique
E Environmentally safe and nontoxic
E Available in an acceptably uncontaminated
state.
In the case of Same AAS, ketones or esters are com-
monly used extraction solvents. The list of organic
solvents used in inorganic extractions is extensive.
Masking Agents
When the desired speciRcity for separation cannot be
controlled sufRciently by pH modiRcation, addition
of a masking agent will frequently be used. Masking
agents are complexing agents that form water-soluble
complexes which then compete with the extracting
agent. The masking agents prevent the extraction of
the metals they react with, by forming water-soluble
complexes (strong polar complexes) which remain in
the aqueous phase. Some extracting agents are speci-
Rc and many others can be made speciRc using pH
control and/or a masking agent. The most commonly
used masking agents include cyanide, thiocyanate,
thiosulfate, tartrate, carbonate, citrate, Suoride, bro-
mide, iodide and ethylenediaminetetraacetic acid
(EDTA). The effectiveness of masking agents is pH
speciRc, ranging from acidic to basic conditions. In
some schemes, more than one masking agent may be
used.
Applications
Inorganic extractions are used in both analytical and
industrial Relds. Chelates form an important part of
inorganic extractions and have extensive application
in many areas of science and industry.
As discussed above, analytical applications include
the separation and/or preconcentration of an analyte
for determination. Another analytical application is
the use of inorganic extraction techniques for reagent
puriRcation by removing trace metals (e.g. puriRca-
tion of aqueous buffers). Extraction of metals into
nonpolar organic phases crosses many scientiRc disci-
plines. For example, crown ethers are used extensive-
ly as phase transfer catalysts. Crown ethers extract an
element (e.g. K#) from the aqueous phase into an
organic phase. The K# ion is engulfed (chelated)
in the centre of the crown ether. A class of anti-
biotics, the ionophores (e.g. nonactin, valinomycin,
gramicidin, etc.) work much like crown ethers:
they alter the permeability (distribution) of bacterial
cells to metal ions and thereby disrupt their
metabolism.
An example of a large-scale inorganic metallurgical
extraction, is the Purex process used to extract ura-
nium isotopes and plutonium. Once the uranium
ore (e.g. carnotite) is crushed, it is concentrated
by physical means; the uranium is then further
concentrated by Sotation methods. The ore is then
roasted and leached with sulfuric acid (often with
an oxidizing agent) and precipitates as sodium
diuranate, a bright yellow solid called ‘yellowcake’.
This solid dissolves in nitric acid producing uranyl
nitrate. The inorganic solvent extraction (Purex pro-
cess) extracts the uranyl nitrate from the aqueous
solution into tributyl phosphate in an inert hydrocar-
bon diluent: the impurities remain in the aqueous
phase.
The aim of speciation studies is to identify and
quantify all species that together combine to comprise
the total element concentration. This is typically
achieved by physicochemical techniques. A range of
physicochemical separation techniques has been ap-
plied to speciation studies, including inorganic sol-
vent extraction.
Inorganic chelate extractions are used extensively
in industrial applications. A brief listing includes the
following applications:
E Metallurgical extraction
E The chelate: (EDTA)
䡩 used in water softeners
䡩 boiler scale removal
䡩 industrial cleaning
䡩 soil metal micronutrient transport
䡩 food preservation
E The chelate: nitrilotriacetic acid (NTA)
䡩 similar applications to those listed for EDTA
E Ion exchange resins can be chelates
䡩 water puriRcation processes
E Zeolites are a type of chelating ion exchange resin
䡩 water puriRcation processes
Future Developments
Although instrument development has had a signiR-
cant impact on inorganic extractions and the direc-
tion of research on separation and preconcentration
techniques, there remains an extensive need and inter-
est in inorganic solvent extraction techniques. One
area which is currently under intense investigation is
speciation. Chemical}physical methods of separation
incorporating inorganic extractions remain an impor-
tant part of this Reld. Another area of development is
the recovery and removal of metals from industrial
waste streams.
See also: II / Chromatography: Liquid: Ion Pair Liquid
Chromatography. Ion Exchange: Theory of Ion Ex-
change. III / Ion Analysis: Liquid Chromatography.
 II / EXTRACTION / Microwave-Assisted Extraction
Further Reading
Batley GE (1989) Trace Element Speciation: Analytical
Methods and Problems. Boca Raton: CRC
Press.
Howard AG and Statham PJ (1993) Inorganic Trace Analy-
sis Philosophy and Practice. New York: John Wiley and
Sons, Inc.
Minczewski J, Chwastowska J and Dybczynski R (1982)
Separation and Preconcentration Methods in Inor-
ganic Trace Analysis. New York: John Wiley and Sons
Inc.
Mizuike A (1983) Enrichment Techniques for Inorganic
Trace Analysis. New York: Springer-Verlag.
Extraction With Supercritical Fluid
See II / EXTRACTION / Supercritical Fluid Extraction
Inorganic Extractions
See II / EXTRACTION / Analytical Inorganic Extractions
Microwave-Assisted Extraction
V. Lopez-Avila, Midwest Research Institute,
Cupertino, CA, USA
Copyright ^ 2000 Academic Press
Introduction
Common extraction techniques for solid matrices in-
clude Soxhlet extraction, sonication extraction,
supercritical Suid extraction (SFE), microwave-assist-
ed extraction (MAE), and accelerated-solvent extrac-
tion (ASE).
Soxhlet extraction allows use of large amount of
sample (e.g. 10}30 g), no Rltration is required after
the extraction, the technique is not matrix dependent,
and many Soxhlet extractors can be set up to perform
in unattended operation. The most signiRcant draw-
backs of Soxhlet extraction are: long extraction times
(e.g. up to 24}48 h), large amount of solvent usage
(300}500 mL per sample), and the need for evapor-
ation after sample extraction.
Sonication extraction is faster than Soxhlet extrac-
tion (30}60 min per sample) and allows extraction of
1389
Ruthven DM (ed) (1997) Encyclopedia of Separation Tech-
nology. New York: John Wiley and Sons Inc.
Swaddle TW (1990) Applied Inorganic Chemistry. Calgary:
University of Calgary Press.
Tatsuya S and Yuko H (1977) Solvent Extraction Chem-
istry: Fundamentals and Applications. New York: Mar-
cel Dekker, Inc.
Vandecasteel C and Block CB (1993) Modern Methods for
Trace Element Determination. New York: John Wiley
and Sons Inc.
VanLoon JC (1985) Selected Methods of Trace Metal Anal-
ysis. New York: John Wiley and Sons, Inc.
Zolotov YA (1970) Extraction of Chelate Compounds.
Ann Arbor: Humphrey Science Publishers.
large amount of sample with a relatively low cost, but
it still uses about as much solvent as Soxhlet extrac-
tion, is labour intensive, and Rltration is required
after extraction.
The newer extraction techniques such as SFE,
MAE, and ASE are very attractive because they are
a lot faster, use much smaller amounts of solvents,
and are environmentally friendly techniques. For
example, SFE uses carbon dioxide or modiRed
carbon dioxide (e.g., carbon dioxide contain-
ing a small amount of an organic solvent known
as modiRer) for extraction. Carbon dioxide is a
nontoxic, nonSammable, and environmentally
friendly solvent. Furthermore, the extraction selectiv-
ity can be controlled by varying the pressure and
temperature of the supercritical Suid and by the addi-
tion of modiRers.
MAE uses microwaves that can easily penetrate
into the sample pores causing the solvent trapped in
the pores to heat evenly and rapidly. In contrast to
conventional heating where it takes a long time for
the vessel to heat and then transfer its energy to the
1390 II / EXTRACTION / Microwave-Assisted Extraction
solvent, MAE is very fast since the heat is transferred
directly to the solvent (provided that the solvent ab-
sorbs microwaves). MAE is promising because: it is
fast (e.g. 20}30 min per batch of as many as 12
samples); MAE uses small amounts of solvents as
compared to Soxhlet and sonication extraction
(30 mL in MAE versus 300}500 mL in Soxhlet ex-
traction); it allows full control of extraction para-
meters (time, power, temperature); stirring of the
sample is possible in MAE; allows high temperature
extraction; and no drying agents are needed in MAE
since water absorbs microwaves very fast and
thus can be used to heat up the matrix. MAE
has several drawbacks that contributed to its slow
acceptance such as: extracts must be Rltered after
extraction, which slows down the operation; polar
solvents are needed; cleanup of extracts is needed
because MAE is very efRcient (e.g. ‘everything’
gets extracted); and the equipment is moderately
expensive.
Accelerated solvent extraction is a fairly new ex-
traction method that was approved recently by the
U.S. Environmental Protection Agency (EPA) as
Method 3545. The extraction is done in a closed-
vessel at elevated temperatures (503 to 2003C) and
pressures (1500}2000 psi). This technique is attract-
ive because it is fast (e.g. extraction time is approxim-
ately 15 min per sample), uses minimal solvent
(15}40 mL), no Rltration is required after the extrac-
tion, and the instrumentation allows extraction in
unattended operation. At least 24 samples can be
processed sequentially and different sample sizes can
be accommodated (e.g. 11, 22, and 33-mL vessels are
available).
Theoretical Considerations in MAE
Microwaves are high-frequency electromagnetic
waves placed between radio frequency and the in-
frared regions of the electromagnetic spectrum (their
frequency range from 0.3 to 300 GHz corresponding
to wavelengths of 1 m to 1 mm). In contrast to con-
ventional heating where the heat penetrates slowly
from the outside to the inside of an object, in MAE
the heating appears right in the core of the body that
is being heated, and the heat spreads from the inside
to the outside of that body. The microwave energy
affects molecules by ionic conduction and dipole ro-
tation. In ionic conduction, the ions in solution will
migrate when an electromagnetic Reld is applied. The
resistance of solution to this Sow of ions will result in
friction and, thus, heating of the solution. Dipole
rotation means realignment of the dipoles with the
applied Reld. At 2450 MHz, the dipoles align and
randomize 4.9;109 times per second; this forced
molecular movement results in molecular ‘friction’
and, thus, heating of the solution.
Selection of proper solvents is the key to a success-
ful extraction. In selecting solvents, consideration
should be given to the microwave-absorbing proper-
ties of the solvent, the interaction of the solvent with
the matrix, and the analyte solubility in the solvent
(the principle of ‘like dissolves like’ is still applicable
in MAE). The larger the dipole moment of the solvent
the faster the solvent will heat under microwave ir-
radiation. For example, hexane (dipole moment is
(0.1 Debye) will not heat, whereas acetone with
a dipole moment of 2.69 Debye will heat in a matter
of seconds. Thus, a mixture of hexane and acetone is
an ideal solvent for compounds of environmental
signiRcance, and many applications described here
use hexane}acetone (1 : 1).
Other important factors under considerations in-
clude: 1. the compatibility between the extraction
solvent and the analytical method used in the analysis
of the extract (the less polar solvents seem to be
preferred for gas chromatographic analysis, whereas
the more polar ones for liquid chromatographic anal-
ysis and immunoassay techniques) and 2. the selectiv-
ity of the solvent. Little has been reported in the
literature on the selectivity of MAE because the tech-
nique is so efRcient that it can not be regarded as
a selective extraction technique. ‘Everything gets ex-
tracted’ so a cleanup step after the extraction is
needed in almost all cases.
When MAE is conducted in closed vessels, the
temperature achieved during the extraction will be
greater than the boiling points of the solvents. For
most of the solvents (e.g. acetone, acetone}hexane,
dichloromethane}acetone), the temperature inside
the vessel is two to three times the boiling point of the
solvent. These elevated temperatures result in im-
proved extraction efRciencies of the analyte from the
sample matrix. The reader should refer to Table 1 for
a listing of solvents and their maximum closed-vessel
temperatures achieved at 175 psi.
Instrumentation for MAE
The features of commercially available MAE systems
are identiRed in Table 2. The equipment (Figure 1)
used for closed-vessel MAE consists of a magnetron
tube, an oven where the individual extraction vessels
(closed vessels) are set up on a turntable or rotor,
monitoring devices for temperature and pressure, and
electronic components. It usually includes speciRc
safety features such as rupture membranes for the
extraction vessels, an exhaust fan to evacuate air
from the instrument cavity, a solvent vapour detector
(monitors the presence of solvent vapour in the
 II / EXTRACTION / Microwave-Assisted Extraction 1391

Table 1 Solvent boiling point and closed vessel temperaturea moved through the rupture vent tube), and an isolator
located in the wave guide that diverts reSected micro-
Solvent Boiling Closed vessel wave energy into a dummy load to reduce the micro-
point (3C) temperature
(3C) at 175 psi wave energy within the cavity. One manufacturer of
microwave equipment uses resealable vessels. In this
Dichloromethane 39.8 140 case, vessels are placed on a sample rotor and secured
Acetone 56.2 164 with a calibrated torque wrench for uniform pressure.
Methanol 64.7 151
If the pressure exceeds the vessel limits, a spring
Ethanol 78.3 164
Acetonitrile 81.6 194 device (Milestone’s patented technology) allows the
2-Propanol 82.4 145 vessel to open and close quickly, thus releasing the
Acetone}hexane (1 : 1) 52# 156 excess pressure. These sample rotors are available
Acetone}cyclohexane (70 : 30) 52# 160 with (perSuoroalkoxy)polymer (PFATM ) and (tetra-
Acetone}petroleum ether (1 : 1) 39# 147
b
Suoroalkoxy)polymer (TFMTM ) liners with pressure
DichloromethaneIacetone (1 : 1) 160c
Toluene}methanol (10 : 1) b
110I112c ratings of 435 psi to 1450 psi. Another safety feature
TolueneImethanol (1 : 10) b
146 c which was added to the microwave system is the
‘movable wall’. To prevent the door from being
a
Adapted from Kingston and Haswell. blown away, a door frame on spring-loaded, high-
b
Information not available.
C
impact steel bars was added such that the door moves
Taken from Reference 2.
out and in to release pressure from the microwave
cavity.
Typical pressures reached with most closed-vessel
microwave cavity and shuts off the microwave energy systems (Rrst-generation) were 105 psi, but today’s
whenever solvent vapour is detected in the instrument technology can handle pressures as high as
cavity), an expansion container (the extraction vessels 1500}1600 psi. A special rotor, which houses six
are connected to this expansion container through thick-walled vessels capable of working at 1600 psi,
vent tubing; in case the membrane ruptures, due to is available commercially on several systems, includ-
increased pressure in the vessel, then vapour is re- ing the CEM’s MARS-5, Milestone’s Ethos-1600,

Table 2 Features of commercially available MAE systemsa

Model/ Power Sensors Max. pressure Vessel volume Vessel Number of Max. temp.
manufacturer (watts) (bar) (mL) material vessels (3C)

Multiwave/ 1000 Pressure 70 100 TFM/ceramics 12 230

Anton Paar control in 70 100 TFM/ceramics 6 260
GmbH, Austria all vessels 130 50 TFM/ceramics 6 260
Infrared 130 50 Quartz 6 300
temperatue 130 20 Quartz 6 300
measurement
in all vessels
MARS-6/CEM, 1500 Infrared 36 100 TFM 14 300
USA temperatue 100 100 TFM 12 300
measurement
in all vessels
Ethos 900/1600, 1600 Pressure 30 120 TFM or PFA 10 240
Milestone, USA control in 100 120 TFM 6 280
all vessels
Temperature 30 120 TFM or PFA 12 240
control in 100 120 TFM 10 280
all vessels
Model 7195/ 950 13 90 TFM 12 200
O.l. Corp. USA 40 90 TFM 12 200
Soxwave 100/ 250 Temperature Open vessel 250 Quartz 1
3.6 Prolabo, control Open vessel 100 or 250 Quartz 6
France

a
Lopez-Avila V (1999) Critical Reviews in Analytical Chemistry 29: 195, reprinted with permission of CRC Press, Boca Raton, FL.
1392 II / EXTRACTION / Microwave-Assisted Extraction
Figure 1 Schematic diagram of a closed vessel MAE system.
and Plazmatronika’s UniClever system. In the Mile-
stone system, for example, if the operating pressure
inside the vessel exceeds the vessel limits, a special
spring device will allow the vessel to open and close,
thus reducing the pressure.
The vessels are typically made of microwave trans-
parent materials (e.g. polyetherimide, or TFM)
and are lined with perSuoroalkoxy or TeSon2+ liners.
A new microwave system introduced recently by one
Figure 2 (A) Standard lined extraction vessel and (B) lined extraction vessel with pressure temperature control.
manufacturer uses magnetic stir bars, which allow
extraction with polar and nonpolar solvents while
agitating the sample and solvent to achieve efRcient
mixing and improve analyte recoveries.
Figure 2 shows a schematic of CEM’s lined
digestion vessel with and without temperature
and pressure control. Vessel body and cap are
made of UltemTM, a polyetherimide. The cap and
cover of the control vessel are modiRed to allow
 II / EXTRACTION / Microwave-Assisted Extraction
a pressure-sensing tube and a Rbre optic temperature
probe. The Rbre optic probe is microwave transpar-
ent and is positioned in the control vessel using a glass
thermal well. Infrared temperature sensors are also
used to monitor the temperature inside the vessel. As
the turntable revolves, the infrared sensor measures
the temperature of each vessel. More detail on the
pressure and temperature feedback control can be
found elsewhere.
Additional features such as magnetic stirring of the
extraction solvent inside multiple sample vessels is
possible, at least on one commerical system (Ethos
1600 Labstation from Milestone, Inc.), Moreover,
nonpolar solvents, such as hexane, can now be heated
at elevated temperatures by use of magnetic stir bars
made of Milestone’s proprietary Suoropolymer Wef-
lonTM . (This polymer absorbs the microwave energy
and subsequently transfers heat to the surrounding
medium.)
All closed vessel systems that are available com-
mercially are multivessel systems which evenly space
the vessels on a carousel or rotor and rotate them
through a pattern on 3603 oscillating turntable.
Speci\c Applications for MAE
Selected MAE applications are identiRed in Table 3.
Polycyclic Aromatic Hydrocarbons (PAHs)
Work done by V. Lopez-Avila et al. indicated that
PAHs, with the exception of more volatile com-
pounds such as naphthalene, can be extracted quant-
itatively (recovery'80%) from soil and sediment
matrices with hexane}acetone (1 : 1) at temperatures
of 1153C. Typical extraction times for batches of up
to 12 samples (5 g each) are 10 min at 100% power
(1000 watts). The lower recoveries of naphthalene,
acenaphthene, and acenaphthylene were attributed to
the presence of water in the soil matrix (to prepare
a representative aged soil sample, water was added to
the soil matrix to bring its water content to 30%).
Other successful microwave-assisted extractions of
PAHs from soils, sediments, and Sy ash have been
reported with hexane}acetone (1 : 1), acetone alone,
dichloromethane alone, dichloromethane}toluene
(50 : 50), acetone}petroleum ether (1 : 1), methanol}
toluene (9 : 1), and toluene}water.
Dean et al. reported on a direct comparison be-
tween Soxhlet, MAE, and SFE for PAHs and con-
cluded that the major advantage of MAE is the speed
of extraction, but they also acknowledged that with-
out additional cooling after extraction it takes ap-
proximately 30 min until the vessels can be opened
and extracts processed. Barnabas, Dean and
1393
coworkers also investigated the effects of pressure,
temperature, extraction time, and percent of meth-
anol modiRer added to the extraction solvent in order
to optimize the extraction.
Chee et al. reported a 5-min heating at 1153C
with 30 mL hexane-acetone (1 : 1) as the optimum
extraction conditions for a 5 g sample, conditions
which are very similar to those reported by V. Lopez-
Avila et al.
Optimization of MAE of PAHs using open-vessel
technology was conducted by Budzinski et al., who
reported that the optimum conditions are 30% water,
30 mL dichloromethane, and 10 min heating at 30 W
power. When considering that the time needed to
reach the boiling point is about 2 min (for dich-
loromethane), a heating time of 10 min is more than
sufRcient to extract PAHs quantitatively from the
matrix, especially when adding water which is sup-
posed to cause swelling of the matrix.
Organochlorine pesticides (OCPs)
Onuska and Terry extracted aldrin, dieldrin, and
DDT from soils and sediments using acetonitrile,
isooctane, or a mixture of isooctane}acetonitrile
(1 : 1, v/v) and achieved quantitative recoveries using
Rve or seven 30-s irradiations with microwave en-
ergy. They also reported that MAE recoveries in-
creases as the moisture content of the soil increases up
to 15%. Fish and Revesz used hexane}acetone as
extraction solvent and reported that OCP recoveries
improved when changing from 1 : 1 hexane}acetone
to 2 : 3 hexane}acetone. The latter solvent has a com-
position similar to the azeotropic vapour in the Sox-
hlet extractor.
Lopez-Avila et al. extracted 45 OCPs from freshly
spiked and 24-h aged soil samples with
hexane}acetone (1 : 1, v/v). For the freshly spiked
soil, 38 compounds had recoveries between 80 and
120%, six compounds had recoveries between 50 and
80%, and the recovery of captafol was above 120%.
For the spiked soil samples aged for 24 h, 28 com-
pounds had recoveries between 80 and 120%; 12
compounds had recoveries between 50 and 80%;
three compounds including captafol, captan, and
dichlone were poorly recovered; and chloroneb and
4,4-DDT had recoveries above 120%.
When recoveries from freshly spiked soil were com-
pared to those from aged spiked soil, it was found
that the recovery of captafol dropped from 122% to
36%, the recovery of captan dropped from 106% to
21%, and the recovery of dichlone dropped from
78% to 10%. Captafol and captan appear to be quite
stable upon irradiation of soil/solvent suspensions,
but dichlone was found to disappear upon irradiation
1394 II / EXTRACTION / Microwave-Assisted Extraction

Table 3 Selected MAE applications reported in the literature

Analyte Matrix Solvent MAE conditions Ref.

17 PAHs, 14 phenols, 3 Reference marine Hexane}acetone (1 : 1) Closed-vessel 2, 3, 4,

20 organochlorine, 13 sediments extraction at 803C, 41
miscellaneous compounds 3 Reference soils 1153C for 5, 10, 20 min
(e.g. chlorinated benzenes Topsoil
nitroaromatic compounds
and phthalate esters)
PAHs Soil Acetone} 29 min at 1203C in closed 6
dichloromethane vessel
PAHs Marine sediments Dichloromethane 5 to 40 min irradiation at 30 to 7
90 W in open vessel, 10 min
Mussel tissue Dichloromethane}toluene irradiation at 30 W in open
Air particles (50 : 50) vessel
Acetone}hexane (50 : 50)

PAHs Reference marine sediments Hexane}acetone (1 : 1) 5 min at 1153C in closed 8

vessel
PAHs Reference marine sediments Dichloromethane 5 to 10 min at 353C in open 9, 10
vessel
PAHs Fly ash Hexane}acetone (90 : 10) 703C in closed vessel 11
PAHs Soil Acetone 20 min at 1203C, closed 12
vessel
PAHs Marine sediments Dichloromethane 5 and 15 min at 1153 and 13
Acetone}hexane (1 : 1) 1353C, closed vessel
PAHs Reference marine sediment Dichloromethane 10 min, 30 watts, open 14
Reference soil Dichloromethane}toluene vessel
Reference (50 : 50)
river sediment Acetone}hexane
Reference sewage sludge (50 : 50, 60 : 40)
Industrial soil Acetone
Marine sediment
Organochlorine pesticides Sediment saturated with dis- Acetonitrile 30 s irradiation in open 15
tilled water (1 g sample and Isooctane vessel; repeat up to five
2 mL water) Isooctane}acetonitrile times
(1 : 1)
16 Phenols, 20 organo- Topsoil Hexane}acetone (1 : 1) Closed-vessel extraction at 16
chlorine pesticides Clay soil 1153C for 10 min
Sand
Reference soil
16 PAHs Water samples preconcen- Acetone 1, 3, 5, 10 min at 803C, 17
10 Organochlorine trated on C18 membrane Dichloromethane 1003C, 1203C, closed vessel
pesticides discs
4 Aroclors
6 Phthalate esters
7 Organophosphorus
pesticides
5 Fungicides/herbicides
PCB 153 Seal Blubber n-Hexane Several 30 s extractions 18
at 1000 W
PCB 180 Pork fat Ethyl acetate}cyclohexane Several irradiations at 250 to 19
PCB 138 Cold liver (1 : 1) 1000 W in increments of
p,p-DDE 100 W
Hexachlorocyclohexane
Hexachlorobenzene

PCBs Municipal sewage sludge Hexane}acetone (1 : 1) 10 min, 30 W, open vessel 20

PCBs River sediments Hexane}acetone (1 : 1) 15 min, closed vessel 21
 II / EXTRACTION / Microwave-Assisted Extraction 1395

Table 3 Continued
Analyte Matrix Solvent MAE conditions Ref.

C16-C32 hydrocarbons Marine sediments Toluene}water 6 min, closed vessel 22

20 PAHs (1 : 5 to 1 : 2)
4 Organochlorine pesticides
PCBs
Phenol Soils Hexane and 1303C in closed vessel 23, 24
hexane}acetone (2 : 8) with
pyridine and acetic anhydride
for in-situ derivatization
Methyl phenols
Nonyl phenol Water samples preconcen- Dichloromethane 5 and 15 min at 1003C to 25
trated on C18-packed Acetone}petroleum ether 1203C, closed vessel
cartridge, C18-packed disc (1 : 1)
Sediments
Phenol Soil Acetone}hexane (various Closed vessel 26
2-Chlorphenol ratios)
2-Methylphenol
2-Nitrophenol
2,4-Dichlorophenol
Imidazolinone herbicides Soil 0.1 M ammonium 3 to 10 min irradiation at 27I29
acetate/ammonium 1253C in closed vessel
hydroxide (pH 9}10)
Atrazine and degradation Lupin seeds Water followed by 0.35 N Closed vessel, 95}983C 30
products Rat feces HCI
Atrazine Sandy loam Methanol 31
Simazine Clay Acetone}hexane (1 : 1)
Prometryne Bentonite Dichloromethane
Florisil Water
Atrazine Sand Dichloromethane with 5 to 45 min at 303C to 32, 33
Simazine Peat water, methanol, and 1303C, 20 min at 1153C
Metazachlor Clay acetonitrite
Desisopropyl atrazine Acetonitrite}0.5%
Desethyl atrazine ammonia in water (70 : 30)
Atrazine Soil Water 3, 4 and 5 min closed 34
vessel
Organotin compounds 2 Reference sediments 50% acetic acid 1 to 7 min irradiation in 35
(mono-, di- and tributyltin; Isooctane open vessel, up to 160 W
mono-, di- and triphenyltin) Methanol
Water
Artificial sea water
Organotin compounds Sediments 0.5 M ethanoic acid in 3 min, open vessel 36
methanol
Butyl and phenyl organotin Reference marine biological 25% tetramethyl} 3 min at 903C, 1153C and 37
matrix ammonium hydroxide in 1303C, closed vessel
Tuna tissue water
Mussel tissue
Organotin compounds Sediments 11 M acetic acid 3 min at 50 to 60 W, 38
NaBEt4 open vessel
Organomercury compounds Sediments 2 M nitric acid 3 min at 60 W, open 38
2 M hydrochloric acid vessel
Reference biological materials 25% tetramethyl} 2 to 4 min at 40 to 60 W,
ammonium hydroxide open vessel
Methylmercury Aquatic sediments Digestion with 6 M 10 min at 1203C, closed 39
Certified reference sediments HCI (methylmercury is vessel
extracted at room
temperature by complexa-
tion with cysteine acetate
and toluene)
1396 II / EXTRACTION / Microwave-Assisted Extraction
of the solvent. (The recovery of dichlone from solvent
was only 5.5% after heating at 1453C for 5 min and
2.6% after 20 min at the same temperature.) Micro-
bial degradation may be responsible for the low re-
coveries of captafol and captan, whereas in the case
of dichlone, it is quite likely that this compound is not
stable under the conditions used. Nonetheless, these
recoveries are higher than those obtained by Soxhlet
or sonication extraction.
Water samples can also be extracted by MAE;
however, they have to be preconcentrated Rrst on
a membrane disc or some adsorbent material. Chee
et al. used C18-membrane discs and then extracted the
discs with 20 mL solvent (acetone and dichloro-
methane) in a closed-vessel MAE system at 803C,
1003C and 1203C for 1, 3, 5 and 10 min. Acetone was
found to give higher recoveries than dichloro-
methane. This approach would allow extremely low
detection limits since several discs generated by pro-
cessing a large volume of sample can be extracted in
one vessel.
Vetter and coworkers extracted OCPs from fatty
tissues (e.g. seal blubber) with solvents such as
hexane and ethyl acetate (1 : 1). To transfer heat to
hexane, which is microwave transparent, discs of
WeSonTM (2.5 cm in diameter ;0.3 cm thickness)
were used in the extraction vessel. The yield of ex-
tractable fat and recoveries of OCPs after seven ir-
radiation cycles were comparable to those obtained
by Soxhlet extraction. Since ethyl acetate}cyclo-
hexane (1 : 1, v/v) seems to extract more fat than
hexane, a gel permeation chromatography step after
extraction is a must.
Polychlorinated Biphenyls (PCBs)
MAE of PCBs was reported by Lopez-Avila et al.
Onuska and Terri, Chee et al., Pastor et al., Dupont
et al. and Kodba and Marsel. Lopez-Avila et al. used
hexane}acetone (1 : 1, v/v) and reported that the
average recoveries from typical soil matrices were
greater than 70% for the Aroclors 1016 and 1260
and the method precision was better than 7%. Fur-
thermore, there was no degradation of PCBs upon
heating of solvent/soil suspensions with microwave
energy. Three reference materials and 24 soils from
a Superfund site, most of which contained Aroclors,
were extracted by MAE and analysed by both
GC/ECD and enzyme-linked immunosorbent assay
(ELISA). Because ELISA is very sensitive and its de-
tection range is quite narrow, the hexane}acetone
extracts were Rrst diluted with methanol and sub-
sequently with the assay buffer (which contained
50% methanol) to bring the Aroclor concentrations
to less than 5 ng mL\1. These data indicate excellent
agreement between the certiRed Soxhlet/GC/ECD
data and the MAE}ELISA data (correlation co-
efRcient 0.9986; slope 1.0168) and the MAE}
GC/ECD data and the MAE}ELISA data (correlation
coefRcient 0.9793; slope 1.0468).
Other solvents used successfully to extract PCBs
from environmental samples include isooctane,
acetone and dichloromethane, and toluene}water.
Phenols
MAE of phenolic compounds was reported by Lopez-
Avila et al., Llompart et al. Chee et al. and Egizabal et
al. Acetone}hexane seems to be the preferred
solvent for 16 phenolic compounds and dichloro-
methane, acetone}petroleum ether (1 : 1) were re-
ported to work well for extraction of nonylphenol.
The only compounds found to degrade during MAE
are 2,4-dinitrophenol and 4,6-dinitro-2-methyl-
phenol. MAE recoveries for phenolic compounds are
usually higher than the classical extraction method
recoveries, and the method precision is signiRcantly
better for MAE (e.g. coefRcient of variation of 3% for
MAE as compared to 15% for Soxhlet and 20% for
sonication).
Herbicides
Imidazolinones (e.g. imazapyr, imazmetapyr,
imazethapyr, imazaquin, etc.) are extracted from soil
with 0.1 M ammonium acetate/ammonium hydroxide
(pH 9}10) in a 10-min extraction. A variety of soil
samples fortiRed at 1 to 50 p.p.b. exhibited an aver-
age recovery of 92% (standard deviation 13%).
Triazine herbicides have been successfully extrac-
ted from soil by MAE with water, methanol,
acetone}hexane (1 : 1), dichloromethane, acetonitriled
0.5% ammonia in water (70 : 30), dich-
loromethane}water (50 : 50), methanol}dich-
loromethane (10 : 90). Water seems to be preferred
since it is very polar solvent and can interact strongly
with polar matter in soils to enhance the desorption
of triazines; it is a cheap, safe, and environmentally
friendly solvent; and it heats up very quickly when
irradiated with microwave energy microwave energy.
Xiong et al. reported that direct heating of soil with
water gave a 73.4% recovery for atrazine from soil
and, therefore, stated that ‘MAE is not only a simple
heating’.
Organotin and Organomercury Compounds
Methods reported in the literature for the determina-
tion of organotin compounds in soils use extraction
with organic solvents in the presence of complexing
agent, or leaching with acetic or hydrochloric acid
assisted by sonication or some sort of shaking.
 II / EXTRACTION / Microwave-Assisted Extraction
Open-vessel MAE was recommended to accelerate
the leaching with 50% acetic acid aqueous solution,
and the data showed that a 3-min irradiation at 60 W
was sufRcient to recover tributyl tin from certiRed
reference sediments. Ethanoic acid (0.5 M in meth-
anol) was also reported. When dealing with bio-
logical matrices (e.g., tuna tissue, mussel tissue),
solubilization with tetramethylammonium hydroxide
(TMAH) for a 3 min at 903C, 1153C, and 1303C in
a closed vessel was demonstrated to be as efRcient as
the hot-plate procedure. Schmitt et al. reported on the
integration of the solubilization step with the derivat-
ization/extraction step by using 11 M acetic acid for
solubilizationm and NaBEt4 for derivatization using
an open vessel MAE system.
Organomercury compounds can be extracted from
sediments with 6 M hydrochloric acid at 1203C for
10 min in closed vessel or 2 M nitric acid and 2 M
hydrochloric acid after 3 min irradiation at 60 W in
open vessel. Pure acetic acid and 1 M sulfuric acid
could only extract 85% and 55%, respectively.
Microwave-assisted digestion of the biological tissue
with 25% TMAH for 2}4 min at 40}60 W gave
quantitative recovery of both organomercury and in-
organic mercury.
Additives in Polymers
Antioxidants such as the Irganox 1010, Irganox
1076, and Irgaphos 168, which are added to poly-
mers to protect them during end-use applications,
can be extracted with '95% efRciency by MAE
with n-heptane}acetone in a few minutes. Higher
temperatures (e.g. 1403C) were used by Jordi
et al. with cyclohexane}chloroform}triethylamine
(45 : 45 : 10) to dissolve polyethylene and extract
compounds such as Tinuvin 770, Tinuvin 622,
Tinuvin 144, and Chimasorb 81.
Natural Products
Extraction of oils from mint leaves and other mater-
ials of biological origin is a patented process known
as the ‘microwave-assisted process’. Other reports on
MAE of natural products include that of Young,
Bichi et al. and Mattina et al. Young extracted ergo-
sterol from fungi and spores by MAE with methanol
and 2 M sodium hydoxide. Bichi et al. extracted pyr-
rolizidine alkaloids from Senecio palvadosos and
Senecio cordatus dried plants by MAE with methanol
at 65 to 1003C for 20 to 30 min. Mattina et al.
reported on the extraction of taxanes from Taxus
biomass by MAE with ethanol. Using 5 g of freshly
harvested needles (moisture content 55 to 65%)
soaked in 5 mL of water prior to MAE and 10 mL
ethanol at 853C for 9 min resulted in about 90%
1397
recovery. This procedure would signiRcantly reduce
the costs of the extraction of taxanes from biomass
with no reduction in the extraction yields.
See also: II/Extraction: Supercritical Fluid Extraction;
Ultrasound Extractions. III/Environmental Applications:
Soxhlet Extraction. Solid-Phase Extraction with
Disks.
Further Reading
1. For a comprehensive text on MAE, see Microwave-
Enhanced Chemistry edited by H.M. (Skip) Kingston
and Stephen J. Haswell, Amertican Chemical Society,
1997.
2. Lopez-Avila V, Young R and Beckert WF (1994)
Analytical Chemistry 66: 1097}1106.
3. Lopez-Avila V, Young R, Benedicto J, Ho P, Kim R
and Beckert WF (1995) Analytical Chemistry 67:
2096}2102.
4. Lopez-Avila V, Young R and Teplitsky NL (1995)
Journal of AOAC International 79: 142}156.
5. Fish JR and Revesz R (1996) LC-GC 14: 230}234.
6. Dean JR, Barnabas IJ and Fowlis IA (1995) Analytical
Proceedings Including Analytical Communications 32:
305}308.
7. Budzinski H, Baumard P, Papineau A, Wise S and
Garriques P (1995) Presented at the 15th PAC Sympo-
sium, Belyirule, Italy, 1995.
8. Chee KK, Wong MK and Lee HK (1996) Journal of
Chromatography A 723: 259}271.
9. Budzinski H, Papineau A, Baumard P and Garrigues P
(1995) Analytical Chemistry 321: 69}76.
10. Letellier M, Budzinski H, Garrigues P and Wise PS
(1996/7) Spectroscopy 13: 71}80.
11. Hsu TB and Chen YS (1996) Organohalogen Com-
pounds 27: 450}454.
12. Barnabas IJ, Dean JR, Fowlis IA and Owen SP (1995)
Analyst 120: 1897}1904.
13. Chee KK, Wong MK and Lee HK (1996) Journal of
Chromatography A 723: 259}271.
14. Budzinski H, Letellier M, Garrigues P and
Le Menach K (1999) Journal of Chromatography
A 837: 187}200.
15. Onuska FE and Terry KA (1993) Chromatographia 36:
191}194.
16. Lopez-Avila V, Young R, Kim R and Beckert WF
(1995) Journal of Chromatographic Science 33:
481}484.
17. Chee KK, Wong MK and Lee HK (1996) Analytica
Chimica Acta 330: 217}227.
18. Hummert K, Vetter W and Luckas B (1996)
Chromatographia 42: 300}304.
19. Vetter W, Weichbrodt M, Hummert K, Glotz D and
Luckas B (1998) Chemosphere 37: 2439}2449.
20. Dupont G, Delteil C, Camel V and Bermond A (1999)
Analyst 124: 453}458.
21. Kodba ZC and Marsel J (1999) Chromatographia 49:
21}27.
1398 II / EXTRACTION / Multistage Countercurrent Distribution
22. Pastor A, Vasquez E, Ciscar R and de la Guardia M
(1997) Analytica Chimica Acta 344: 241}249.
23. Llompart MP, Lorenzo RA, Cela R, PareH JRJ, BeH langer
JMR and Li K (1997) Journal of Chromatography
A 757: 153}164.
24. Llompart MP, Lorenzo RA, Rosa A, Cela R, Li K,
BeH langer JMR and Pare JRJ (1997) Journal of
Chromatography A 774: 243}251.
25. Chee KK, Wong MK and Lee HK (1996) Journal of
Liquid Chromatography and Related Technology 19:
259}275.
26. Egizabal A, Zuloaga O, Etxebarria N, Fernandez LA
and Madariaga JM (1998) Analyst 123: 1679}1684.
27. Stout SJ, da Cunha AR and Allardice DG (1996)
Analytical Chemistry 68: 653}658.
28. Stout SJ, da Cunha AR, Picard GL and Safarpour MM
(1996) Journal of Agriculture and Food Chemistry 44:
3548}3553.
29. Stout SJ, da Cunha AR and Safarpour MM (1997)
Journal of AOAC International 80: 426}432.
30. Steinheimer TR (1993) Journal of Agriculture and
Food Chemistry 41: 588}595.
31. Xiong G, Tang B, He X, Zhao M, Zhang Z and Zhang Z
(1999) Talanta 48: 333}339.
32. Hoogerbrugge R, Molins C and Baumann BA (1997)
Analytica Chimica Acta 348: 247}253.
33. Molins C, Hogendoom EA, Heusinkveld HAG,
Van Harten DC, Van Zoonen P and Baumann RA
(1996) Chromatographia 43: 527}532.
Multistage Countercurrent Distribution
G. Johansson, Chemical Center, Lund, Sweden
This article is reproduced from Encyclopedia of
Analytical Science, Copyright ^ 1995 Academic Press.
Theory
The separation of chemical compounds by partition-
ing between two liquid phases, so-called liquid}liquid
extraction, can be made more effective by using it as
a cascade process. One way in which this can be
carried out is by multiplicative partitioning, also
called countercurrent distribution (CCD). This pro-
cess, in which complete partition equilibrium is
achieved in each step, is presented schematically in
Figure 1. The principle is that two sets of liquid
phases, the upper and lower phase, come into contact
with each other stepwise. The bottom phases are
numbered 0, 1, 2 and so on. The sample to be ana-
lysed (fractionated) is included in the Rrst system
(containing bottom phase number 0). Before each
transfer of the upper phases (to the right in Figure 1)
the two-phase systems are equilibrated by mixing and
34. Xiong G, Liang J, Zou S and Zhang Z (1998) Analytica
Chimica Acta 371: 97}103.
35. Donard O, Lalere B, Martin F and Lobinski R (1995)
Analytical Chemistry 67: 4250}4254.
36. Lalere B, Szpunar J, Budzinski H, Garrigues P and
Donard OFX (1995) Analyst 120: 2665}2673.
37. Rodriguez I, Santamarina M, Bollain MH, Mejuto MC
and Cela R (1997) Journal of Chromatography A
774: 379}387.
38. Schmitt VO, de Diego A, Cosnier A, Tseng CM, Moreau J
and Donard OFX (1996/7) Spectroscopy 13: 99}111.
39. Vazquez MJ, Carro AM, Lorenzo RA and Cela R
(1997) Analytical Chemistry 69(2): 221}225.
40. Lopez-Avila V, Benedicto J, Charan C, Young R and
Beckert WF (1995) Environmental Science and Tech-
nology 29: 2709}2712.
41. Onuska FI and Terry KA (1995) Journal of High Res-
olution Chromatography 18: 417}421.
42. Freitag W and Angew JO (1990) Makromolecular
Chemistry 175: 181}185.
43. Jordi HC, Savage W and Bichard F (1995) Paper pre-
sented at Pittcon ’95, Paper No. 1209.
44. Pare JRJ (1991) US Patent 5,002,784.
45. Young JC (1995) Journal of Agriculture and Food
Chemistry 43: 2904.
46. Bichi C, Beliarab FF and Rubiolo P (1992) Lab 2000 6:
36
47. Mattina MJI, Berger WAI and Denson CL (1997)
Journal of Agriculture and Food Chemistry 45: 4691.
the sample components are distributed between each
pair of phases (each full two-phase system). The par-
titioning of a pure substance between the phases of
a two-phase system can be expressed either by a parti-
tion coefRcient, K, deRned as the ratio of the concen-
trations (C) of the component in the phases:
C (in phase I)
K" [1]
C (in phase II)
or by a partition ratio, G, deRned as the ratio of the
masses (m) of the components in the phase:
m (in phase I)
G" [2]
m (in phase II)
K and G are related by eqn [3]:
Volume (phase I)
G"K [3]
Volume (phase II)
In the following the upper phase is chosen as phase
I. A convenient way of analysing the CCD process is
to calculate the amounts (in fractions) of a pure
 II / EXTRACTION / Multistage Countercurrent Distribution 1399

Figure 1 Principle of CCD. The upper phases are transferred stepwise from left to right. Between each transfer the two-phase
systems are equilibrated by shaking and the phases are allowed to settle. Filled circle, substance with small G value; open circle,
substance with large G value.

component in the various phases based on its G value.
In the initial two-phase system (number 0) the com-
ponent is distributed with the fractions p in the upper
phase (phase I) and q in the lower phase (phase II). By
deRnition p#q"1. When the upper phase of system
number 0 is combined with the pure lower phase
number 1 equilibration of this system will make the
transferred fraction p distribute as p/q ("G) giving
the fraction p2 in the upper phase and the fraction pq
in the lower phase. In the new system number 0, with
a new upper phase, the remaining fraction q will
partition likewise and the equilibration results in the
fraction pq in the upper phase and q2 in the lower
phase. The resulting distribution after 10 transfers in
the 11 tubes 0}10 is shown in Figure 2. In the same
Rgure the material in each tube (upper plus lower
phase) has been calculated. These values are the same
as the terms obtained when (q#p)10 is written as
a polynomial. More generally, the amount of material
Tn,i (in fractions), in tube number i after a CCD with
n transfers is given by eqn [4]:
n!
Tn,i" piqn\i [4]
i!(n!i)!
The volume ratio is kept constant during the process.
By using the relations Gi"pi/qi and (1#G)n"

Figure 2 Distribution of a component which partitions in the ratio p/q between the upper and lower phase } after a CCD with 10
transfers. The amounts are given in fractions, i.e. p#q"1.
1400 II / EXTRACTION / Multistage Countercurrent Distribution

Figure 3 The resulting CCD profiles after (A) 10 transfers, (B) 100 transfers and (C) 1000 transfers for substances with the indicated
G values.

(1/q)n, eqn [5] is obtained: This relationship has been used to calculate the distri-
bution proRles in Figure 3 for three cases: i.e. with the
n! Gi number of transfers (n) equal to 10, 100 and 1000. In
Tn,i" [5]
i!(n!i)! (1#G)n each case the distributions of components with
 II / EXTRACTION / Multistage Countercurrent Distribution 1401

G values of 0.05, 0.1, 0.25, 0.5, 1, 2, 4, 10 and 20 are Table 2 Number of transfers, n, in a CCD necessary to separ-
shown. The resolution of components with various ate two substances, 1 and 2, with a given separation factor
"K1/K2 . The volume ratio is chosen in such a way (eqn [7]) that
G values increases with the increase in the number of
G1;G2"1
transfers. This is due to the fact that the difference in
the position (tube number) of the peaks (with given  G1 G2 n
G values) is proportional to the number of transfers,
while the peak width only increases with the square 2 0.709 1.414 264
3 0.577 1.732 110
root of the number of transfers.
4 0.500 2.00 70
The width of a peak
r, covering 99.7% of the 9 0.333 3.00 30
compound, can be approximately calculated using 16 0.250 4.00 20
eqn [6]: 25 0.200 5.00 16
36 0.167 6.00 14
6(nG 49 0.143 7.00 12

r" [6] 81 0.111 9.00 10
(1#G) 400 0.050 20.0 6
2500 0.020 50.0 4
The relative width of a peak,
rrel"
r/n, is given in
Table 1 for G"1 and a number of n values.
The most effective separation of two components,
with partition coefRcients K1 and K2, is achieved The G value can also be obtained by comparing the
when their distribution peaks are oriented symmetric- amount of a substance in two consecutive tubes,
ally around the middle of the tube train, i.e. tube numbers i and i#1. By combining the expressions
i"n/2. This, as can be seen in Figure 3, is equal to for Tn,i and Tn,i#1 obtained from eqn [5] the G value
the relation G1"1/G2 which can be written as is obtained as in eqn [9]:
G1;G2"1. By combining this relation with eqn [3]
the volume ratio, V, to be used for optimal separation Tn,i#1 i#1
can be calculated as shown in eqn [7]: G" [9]
Tn,i n!i

1
V" [7] By following the apparent G values over a distribu-
(K1K2 tion peak the presence of several substances differing
slightly in their G values can be detected.
The ratio between the K values of two substances
1 and 2 is called the separation factor, , and is
a measure of the separability of the substances. In Various Modes of Multiplicative
Table 2 the number of transfers necessary for vir- Partitioning
tually complete (more than 99.5%) separation has
been calculated for various  values. The basic process of CCD has been described, but it
The G value of a component which has its max- can be varied in a number of ways.
imum peak value in tube number nL after n transfers is
Single Withdrawal Procedure
given by the approximate eqn [8]:
In this process the addition of fresh upper phases is
G"nL /(n!nL ) [8] not stopped after n transfers, but continues from the
left while the overSowing upper phases (to the right)
are collected. This process is therefore similar to
Table 1 The absolute,
r, and relative,
rrel , peak width (per-
centage of the total number of tubes) of a substance with G"1
a chromatographic system with the lower phases cor-
after a CCD with n transfers responding to the stationary phase of the column and
the upper phases corresponding to the elution liquid.
n
r (no. of tubes)
rrel (%) Substances with high G values are easily eluted in this
way while those with low G values need a great
20 13 67
50 21 42
number of steps to be carried through the CCD train.
100 30 30 The greater the number of transfers the more diluted
200 42 21 the substance will be when leaving with the overSow-
500 67 13 ing upper phases. Likewise the upper phases can be
1000 95 9.5 chosen as the stationary phase and the lower phases
2000 134 6.7
5000 212 4.2
can be used as the eluting phases. In this case sub-
stances with low G values rapidly eluted.
1402 II / EXTRACTION / Multistage Countercurrent Distribution

O’Keefe Partitioning
The multiplicative partitioning can be carried in such
a way that a new portion of the solute mixture to be
separated is added after each transfer step. This
method of carrying out the partitioning is called
O’Keefe partitioning. Several two-phase systems are
arranged in a row and a solute sample is included in
the centre system. After equilibration and settling all
the upper phases are moved one step to the right and
the overSowing upper phase is collected. Then all the
lower phases are moved one step to the left and the
one to the extreme left is collected. The remaining set
is completed with one fresh lower phase (to the right)
ane one fresh upper phase (to the left) to restore the
original number of systems. After addition of a new
portion of solute to the centre system the next cycle
takes place. Substances with G'1 will be recovered
in the upper phases collected to the right and substan-
ces with G(1 in the lower phases removed to the
left. This process is used when large amounts of
substances are to be separated. The number of sys-
tems is usually small, e.g. 7, 9 or 11.
A similar process, called Watanabe}Morikawa
partitioning, differs in that the ongoing addition of
solute mixture is done in the Rrst system of the row.
This is useful when only components with high or low
G values are to be isolated.

Description of Some CCD Apparatus

CCD with only a few transfers (up to 10) can easily be
carried out by using a set of separating funnels for
mixing, settling and phase separation. For small
phase volumes, 0.5}5 mL, the separating funnels can
be replaced by test tubes and the upper phases can be
transferred with the aid of a pipette. For CCDs with
more than 10 transfers it is strongly advised that some
kind of automatic apparatus is used. Some examples
of such apparatuses are presented next.
Figure 4 Section of the all-glass apparatus for CCD construc-
ted by Craig. (A) Two segments (or tubes) are shown from side
and top views. Normally 50}1000 tubes are connected and placed
on a horizontal rack which allows simultaneous mixing of all
tubes. (B) By turning the axis the upper phases can be decanted.
When returning to the original position the upper phases are
transferred to the neighbouring tube.

All-Glass CCD According to Craig Thin-Layer CCD According to Albertsson

Several glass units, allowing mixing, settling and
phase transfers (Figure 4) are arranged in batteries on
a horizontal axis. Movement about this axis can be
used to gently mix the phases, to put the glass units in
position for settling of the phases and for decanting
the upper phases, respectively. The times for each
part of the CCD cycle as well as the number of
transfers can be programmed. Standard types of such
machines allow 50 transfers while more advanced
machines may be used for up to 1000 transfers. The
glass units can be obtained in various sizes but stan-
dard tubes have space for c. 2 mL (Rxed amount) of
lower phase and up to 5 mL of upper phase.
This CCD apparatus is designed to be used with
aqueous two-phase systems which normally need
a long time for settling. By using systems with low
heights, only a few millimetres, the settling time will
be acceptable (5}20 min). The operating unit
(Figure 5) consists of two circular plates of poly-
acrylic plastic, one on top of each other: the lower
plate (stator) is Rxed while the upper one (rotor) can
be rotated stepwise. Both plates have 60 (or 120)
radially oriented cavities which pair-wise form con-
tainers for the two-phase systems. The cavities of the
stator contain the lower phases (normally 0.8 mL)
while the upper phases (0.2}2 mL) are situated in the
 II / EXTRACTION / Multistage Countercurrent Distribution 1403

Figure 5 Thin-layer CCD apparatus according to Albertsson. (A) The two discs of polyacrylic plastic contain matching cavities for the
upper and lower phases and by rotation the upper plate can move all upper phases relative to the lower ones. See text for further
details. (Adapted from Albertsson P-A> (1986) Partition of Cell Particles and Macromolecules, 3rd edn, p. 126. New York: Wiley
(interscience), by permission.) (B) A thin-layer CCD machine constructed by Albertsson.

cavities of the rotor and therefore can be transferred
relative to the lower phases by rotating the upper
plate. The two plates are placed on the shaking table
of a machine which also contains a drive for the upper
plate rotation. The phase systems are added to each
container (chamber) via openings in the upper side of
the rotor. These openings, during the run, are covered
with a ring-formed lid. The shaking and settling peri-
ods and number of transfers are controlled by an
automatic unit. After the CCD run is completed the
two-phase systems can be collected by use of a frac-
tion collector. This is a circular rack with the same
number of test tubes (4 mL) as chambers. The rack is
placed over the inlet holes and is inverted together
with the plates which are then emptied.
Centrifugal CCD According to A> kerlund
This kind of CCD machine uses centrifugation to
speed up the settling of the phases. It consists of an
outer ring with cavities for the lower phases (attached
to a bottom plate) and an inner plate with cavities for
the upper phases (Figure 6.) The inner plate can ro-
tate relative to the outer ring. When the chambers
have been loaded with systems they are covered with
a lid. As in the case of the thin-layer machine the
functional unit is placed on a (round) table which can
be shaken. In this case, however, the table can also
rotate to allow centrifugation of the mixed systems
which speeds up the settling of the phases. The upper
phases are then transferred to the neighbouring lower
1404 II / EXTRACTION / Multistage Countercurrent Distribution

by adjusting the pH value. More selective adjustment

of the partitioning has been carried out by linking an
afRnity ligand to one of the phase-forming polymers
which then is concentrated in one of the phases.

Figure 6 Centrifugal CCD according to A> kerlund. A section of

the circular separation unit is shown. It is composed of four units:
c, the outer ring with cavities for the lower phase: d, the inner plate
with cavities for the upper phase; e, the lid; and f, an O-ring for
sealing. The position of the two-phase system during centrifu-
gation is shown, with the upper phase, a, and the lower phase, b.
(From Johansson G, A> kerlund H-E and Olde B (1984) Journal of
Chromatography 311: 277, by permission.)

phases by rotation of the inner plate when the whole

system is still spinning and the phases are in a vertical
position. After the transfer the centrifugation is stop-
ped and a new cycle begins with the shaking process.
After the run the phases are collected by pipette.
Continuous apparatus, such as extraction columns
and coil planet centrifuges, also corresponds to a
CCD process but differs in that no discrete steps are
used and equilibrium is not reached.

Analytical Applications
Many metal ions can be separated by using various
kinds of water/organic solvent systems for counter-
current distribution. Metal chelators, such as
dithizone, 8-quinolinol, cupferron, dimethylglyoxime
and acetylacetone, are used speciRcally to extract
metal ions into the organic phase. Likewise certain
anions such as halides, thiocyanate or nitrate can be
used. For example, uranium and plutonium have
been separated by using an aqueous phase containing
8 mol L\1 nitric acid.
Biochemical Applications
CCD has been used for fractionation of a number of
biochemical substances and cellular particles as well
as cells and viruses. Peptides, proteins and nucleic
acids have been fractionated by using aqueous two-
phase systems composed of water and the two poly-
mers dextran and poly(ethylene glycol) (PEG). The
partition coefRcients of the solutes can be adjusted by
addition of various salts to the two-phase system and
Figure 7 Countercurrent distribution of proteins and cell or-
ganelles. (A) Distribution of the enzymes hexokinase (0),
3-phosphoglyceratekinase (- - -), and phosphofructokinase
(- ) - ) - ) ) when an extract of baker’s yeast was applied to a CCD
with 55 transfers. The textile dyes Procion olive MX-3G or Procion
yellow HE-3G were used as PEG-bound affinity ligands enriched
in the mobile upper phase. Composition of two-phase system:
88% (w/w) water, 7% (w/w) dextran 500, 5% PEG with
Mr"8000, including dye-PEG, 1% of total PEG 50 mmol L\1
sodium phosphate buffer, pH 7.0, 0.2 mmol L\1 ethylenediamine-
tetraacetic acid and 5 mmol L\1 2-mercaptoethanol. Volume ra-
tio, 1.5. Temperature, 33C. A centrifugal CCD apparatus was
used with 5 min shaking and 3 min settling (centrifugation).
(Adapted from Johansson G, Joelsson M and A> kerlund H-E (1984)
Journal of Chromatography 298: 483, by permission.) (B) Frac-
tionation of photosynthetic particles, chloroplasts, from spinach
using a thin-layer CCD apparatus with 56 transfers and a dex-
tran}PEG two-phase system. Peak I, intact chloroplasts sur-
rounded by their envelope; peak II, naked thylakoid membranes
(class II chloroplasts); peak III, choloroplasts surrounded by a ‘bag’
of plasma membrane also containing cytoplasm, mitochondria and
peroxisomes. (From Larsson C, Collin C and Albertsson P-A>
(1971) Biochimica Biophysica Acta 245: 425, by permission.)
 II / EXTRACTION / Solid-Phase Extraction 1405

Biological membranes, cell organelles, whole cells
and viruses can be fractionated by CCD in the same
kind of systems. In this case, however, the particles
partition between the two liquid phases and the inter-
face between them. The CCD is therefore usually
carried out using a stationary interface. This is
achieved by using a smaller volume of the lower
phase than is needed to Rll the lower cavities. There-
fore, a portion of the upper phases will also be sta-
tionary. The G value satisfying eqn [5] is in this case
deRned as the amount of a pure compound, at equi-
librium, in the mobile part of the upper phase divided
by the amount of the compound in the rest of the
system (stationary upper phase, interface and lower
phase). Examples of CCD of proteins and of chloro-
plasts, the photosynthetic organelle in green plant
cells, are given in Figure 7.
See also: II/Chromatography: Countercurrent Chroma-
tography and High-Speed Countercurrent Chromato-
graphy: Instrumentation.
Further Reading
A> kerlund H-E and Albertsson P-A> (1994) Thin-layer
countercurrent distribution and centrifugal countercur-
rent distribution apparatus. Methods in Enzymology
228: 87}99.
Craig LC (1962) Countercurrent distribution. In: Florkin
M and Stotz EH (eds) Comprehensive Biochemistry,
vol. 4, pp. 1}31. Amsterdam: Elsevier.
Hecker E (1995) Verteilungsverfahren im Laboratorium.
Weinheim: Verlag Chemie.
Morris CJOR and Morris P (1976) Separation Methods
in Biochemistry, 2nd edn, pp. 638}702. London:
Pitman.

Solid-Phase Extraction

C. F. Poole, Wayne State University, Detroit, MI, biological Suids. These sorbents had reasonable
USA mechanical strength compared with gels, a large sur-
Copyright ^ 2000 Academic Press face area and sample capacity, low water retention,
and gave high sample recoveries by solvent desorp-
Solid-phase extraction is a method used to isolate tion. Compared with carbon the overall analyte re-
analytes from a gas, Suid or liquid by their transfer to covery was generally better and irreversible adsorp-
and retention on a solid-phase sorbent. After separ- tion and catalytic activity greatly diminished. These
ation of the sorbent from the sample the analytes are
recovered by elution using a liquid or Suid, or by
thermal desorption into the gas phase. If the analytes
are recovered from the sorbent in a Rnal volume that
is only a fraction of the sample volume, then concen-
tration as well as isolation is achieved. In addition, if
the sorption step, any subsequent rinse steps, and the
elution conditions are selective for retention and re-
covery of the analyte, then matrix simpliRcation is
achieved. Isolation, concentration and matrix simpli-
Rcation are the primary goals of solid-phase extrac-
tion.
Probably the earliest application of solid-phase ex-
traction was the use of charcoal-Rlled columns in the
1950s to isolate organic contaminants from surface
waters for toxicity evaluation. The large volume of
water generally sampled (more than 1000 L over sev-
eral days) precluded the use of liquid}liquid extrac-
tion techniques. The subsequent evolution of solid-
phase extraction techniques is summarized in
Figure 1.
The introduction of macroreticular porous poly-
mers in the early 1970s was responsible for rekindling
interest in solid-phase extraction and extending its Figure 1 Time line showing the general evolution of solid-
scope to air sampling and the isolation of drugs from phase extraction techniques.
1406 II / EXTRACTION / Solid-Phase Extraction
Figure 2 Schematic diagram showing the typical construction of a solid-phase extraction cartridge and a vacuum manifold for parallel
sample processing.
properties, together with a reduction in the amount of
material needed for identiRcation due to improved
instrumentation, resulted in the general use of small
columns, similar in size to those in use today. Porous
polymers with high thermal stability and low water
retention revolutionized the room temperature
sorbent extraction of volatile organic compounds
from air or purge gas from water samples. Trapped
compounds were thermally desorbed directly into
a gas chromatograph for analysis. Automated sys-
tems based on the above process are used for routine
analysis today.
Solid-phase extraction for liquid samples became
a widely used laboratory technique with the intro-
duction of disposable sorbent cartridges containing
porous, siloxane-bonded silica particles, sized
to allow sample processing by gentle suction
(Figure 2). A typical solid-phase extraction cartridge
consists of a short column (generally an open
syringe barrel) containing a sorbent with a nominal
particle size of 50}60
m, packed between porous
metal or plastic frits. A large number of sorbents are
in use today corresponding to the desire for general
purpose, class-speciRc and even compound-speciRc
extractions.
Slow sample processing rates for large sample vol-
umes, low tolerance to blockage by particles and
sorbed matrix components, and problems arising
from the low and variable packing density of car-
tridge devices spawned the development of alterna-
tive sampling formats based on disc technology. At
least three different designs for solid-phase extraction
discs are offered commercially today. The particle-
loaded membranes consist of a web of polytetraf-
luoroethylene (PTFE) microRbrils, suspended in
which are sorbent particles of about 8}10
m dia-
meter. The membranes are Sexible with a homogene-
ous structure containing 80% (w/w) or more of
sorbent particles formed into circular discs 0.5 mm
thick with diameters from 4 to 96 mm. For general
use they are supported on a sintered glass disc (or
other support) in a standard Rltration apparatus using
suction to generate the desired Sow through the mem-
brane (Figure 3). Particle-embedded glass Rbre discs
contain 10}30-
m sorbent particles woven into
a glass Rbre matrix. The small diameter discs are rigid
and self-supporting, while the larger diameter discs
require a supporting structure. Speediscs] (Figure 4)
consist of a sandwich of 10-
m sorbent particles held
between two glass-Rbre Rlters, with a screen to hold
the Rlters in place. Disc technology has contributed
directly to the automation of solid-phase extraction
through the development of the multiwell extraction
plate (Figure 5), which is used for the clean-up of
samples in high-throughput screening techniques for
drug development. Direct coupling of solid-phase ex-
traction and high pressure liquid chromatography for
on-line sample processing and analysis is now routine
and the direct coupling of solid-phase extraction
and gas chromatography for the analysis of liquid
samples has moved beyond the research phase.
Several research groups have demonstrated the direct
coupling of solid-phase extraction and electrophor-
etic and thin-layer chromatographic separation
techniques.
Replacement for Liquid+Liquid
Extraction
Solid-phase extraction was introduced as a replace-
ment for liquid}liquid extraction to give a practical
and economic solution to the real and perceived
problems associated with solvent extraction
techniques. Liquid}liquid extractions are labour

Figure 3 Typical cartridge and vacuum filtration formats for solid-phase extraction using discs.
intensive, difRcult to automate, and frequently
plagued by practical problems, such as emulsion
formation. Liquid}liquid extractions also tend to
consume large volumes of high purity solvents, which
may have signiRcant health hazards and disposal
costs associated with their use. In contrast, solid-
phase extraction beneRts from lower intrinsic costs,
reduced processing times, low solvent consumption
and simpler processing procedures. Solid-phase ex-
traction procedures are easily automated using ro-
botics, or special purpose Sow processing units that
Figure 4 Exploded-view of the Speedisc] used for solid-phase,
extraction.
II / EXTRACTION / Solid-Phase Extraction 1407
simultaneously extract samples and prepare them for
automatic injection, or by using centrifugal analysers,
which can batchwise process multiple samples. Solid-
phase extraction is convenient for Reld sampling since
it minimizes the transport and storage problems of
bulk samples, which have to be returned to the labor-
atory for processing.
Solid-phase extraction techniques have their own,
although different, problems to those of liquid}liquid
Figure 5 Multiwell plate for automated solid-phase extraction.
(Reproduced with permission from Plumb RS, Gray RDM, and
Jones CM (1997). Use of reduced sorbent bed and disc mem-
brane solid-phase extraction for the analysis of pharmaceutical
compounds in biological fluids, with applications in the 96-well
format. Journal of Chromatography B 694:123}133.)
1408 II / EXTRACTION / Solid-Phase Extraction
extraction. The sorption properties of manufactured
sorbents are not as reproducible as solvent properties.
Basic drugs, for example, are often retained on silica-
based, chemically bonded sorbents by a mixed reten-
tion mechanism involving non-speciRc sorption by
the bonded phase and ion exchange interactions with
accessible, dissociated silanol groups. The mixed re-
tention mechanism can interfere in the recovery of
analytes since solvent elution may be ineffective for
removing ionically bound analytes, and the extent of
binding through ionic sites can vary for different
sorbent lots. Sorbents tend to have a higher level of
contamination by manufacturing and packaging ma-
terials than do solvents. The chemical background
from impurities may interfere in the subsequent anal-
ysis of the sample. Solvent rinsing of cartridges and
discs and the running of blanks to establish back-
ground contamination levels diminishes sample
throughput and adds signiRcantly to solvent con-
sumption and processing costs. Sample processing
problems, such as column overloading, displacement
and blocking of sorbent pores, easily go unnoticed,
resulting in changes in analyte recovery. Sample over-
load and displacement are more important for extrac-
tion based on adsorption than for extraction based on
absorption.
When choosing between liquid}liquid or solid-
phase extraction for a particular problem, economic,
as well as technical features, should be taken into
consideration. In this sense, liquid}liquid and solid-
phase extraction techniques should be considered
complementary approaches, and although the general
trend is towards the replacement of liquid}liquid ex-
traction methods by solid-phase extraction, this is
never likely to be a complete replacement.
Disc Versus Cartridge Format
Cartridges have a small cross-sectional area, a slow
sample processing rate, and a low tolerance to block-
age by particles and adsorbed matrix components.
For large sample volumes containing suspended par-
ticles, discs are likely to function better. Discs provide
shorter sample processing times due to their larger
cross-sectional area and decreased pressure drop, en-
abling higher sample Sow rates to be used. The larger
cross-sectional area also reduces problems with plug-
ging. For example with a high particle burden, discs
with integral or separate depth Rlters are available, as
well as different materials that can be added to the
surface of the disc as Rlter aids.
Because of the low packing density of typical car-
tridge devices, longer sorbent beds than are needed
for extraction are used to compensate for reduced
retention resulting from channelling. Increased bed
mass results in increased non-speciRc matrix adsorp-
tion and dirtier extracts. The use of smaller particles
and the greater mechanical stability of discs reduces
channelling, and the optimized use of bed mass re-
sults in a cleaner background and lower interferences
due to reduced matrix adsorption. For small sample
sizes it is easier to miniaturize discs than cartridges,
and several disc devices (e.g. microdiscs, pipette tips,
etc.) that contain only a few milligrams of sorbent for
processing small samples are available. Immoblilized
analytes on microdiscs facilitate integrated sample
processing techniques such as in-vial extraction and
on-disc derivatization.
Inorganic Oxide Adsorbents and
their Applications
The most important adsorbents for extraction and
matrix simpliRcation are silica gel, alumina, Florisil
and diatomaceous earths. Silica gel, prepared from
sodium silicate using the sol-gel procedure, is the
most widely used general-purpose adsorbent. Silica
gels used for solid-phase extraction have surface areas
of about 300}800 m2 g\1, pore sizes from 4}10 nm,
and an apparent pH of 5.5}7.5. The apparent sorbent
pH is characterized as the observed pH of a 5% (w/w)
aqueous suspension. Alumina is prepared by the low
temperature dehydration ((7003C) of alumina
trihydrate and is a mixture of -alumina with small
amounts of -alumina (less active form) and sodium
carbonate. Depending on processing conditions,
alumina is available as neutral (pH 7.5$0.5), weak-
ly acidic (pH 6.0$0.5), acidic (pH 4.5$0.5) and
basic (pH 9.5$0.5) forms. Adsorbents used for ex-
traction and matrix simpliRcation have a surface
area of about 150 m2 g\1 and a pore size of
6 nm. Florisil is a magnesium silicate prepared
by precipitation from a mixture of magnesium sulfate
and sodium silicate solutions followed by calcining
at about 12003C. It has a surface area of about
250}300 m2 g\1 and an apparent pH of about
8.5. Diatomaceous earths are Sux-calcined forms of
natural silica with very small surface areas. They are
used as a Rlter aid and as a dispersant for liquid
extraction using matrix dispersion techniques (see
matrix dispersion).
The general extraction mechanism and applica-
tions of the inorganic oxide adsorbents are sum-
marized in Table 1. Adsorbent properties that in-
crease retention are a larger surface area and a high
activity. Adsorbent activity is controlled by the inten-
tional addition of water to the dried adsorbent prior
to use and by drying extracts with anhydrous sodium
sulfate, or a similar drying agent, prior to applying
the extract to the adsorbent. A small column of
 II / EXTRACTION / Solid-Phase Extraction 1409

Table 1 General applications of solid-phase extraction

(1) Inorganic oxide adsorbents

E Isolation of low and medium polarity analytes from non-aqueous solutions
E Isolation of cations (alumina and silica) and anions (alumina) from buffered aqueous solutions
E Matrix simplification by fractionation into groups containing a similar number and type of functional group
Examples
N Isolation of organochlorine pesticides and polychlorinated biphenyls from transformer oil, animal fats and oils, etc. using Florisil.
N Isolation of lipids by chromatography over silica gel using chloroform to elute simple lipids, acetone to elute glycolipids and
methanol to elute phospholipids.
N Group fractionation of polycyclic aromatic compounds (hydrocarbons, N-containing and OH-containing) in synthetic fuels over
alumina using a step solvent gradient.
N Isolation of paraquat and diquat from high moisture crops in a pH 9 aqueous extract using silica gel
N Mycotoxins in feeds using silica gel
N Pesticides in foods, feeds and soil extracts; alkaloids, pigments and flavour compounds from plants; sugars and caffeine in cola
beverages, inorganic anions and organic acids in aqueous solution using alumina; steroids and vitamins from creams and
oil-based suspensions.

(2) Low speci\city sorbents (aqueous solutions)

E Isolation of neutral and ionizable analytes from aqueous solution. Weak acids and bases by ion suppression. Strong acids and
bases using ion pair extraction (alternative to ion exchange)
E Retention increases with solute size and is reduced by polar interactions (particularly hydrogen-bonding) and ionization
E Polar bonded phases provide only weak retention and are not particularly useful unless elution of the analyte is a problem from non-
polar sorbents
Examples
N Isolation of agricultural and industrial chemicals from surface waters using C18, carbon or poly(styrene-divinylbenzene)(PS-DVB)
N Isolation of drugs from biofluids using C18, C8, PS-DVB or cyanopropyl (CN)
N Isolation of macromolecules from biofluids and fermentation broth using C4
N Isolation of pigments and colouring materials from beverages and food extracts using C18
N Isolation of carbohydrates and nucleosides from biofluids using AMINO
N Isolation of proteins, peptides and surfactants using DIOL

(3) Low speci\city sorbents (organic solvents)

E Retention depends on the type and number of functional groups. Solute size is not important
CN Strong dipole-type interactions and weak hydrogen-bond acidity
AMINO Strong hydrogen-bond base and weak hydrogen-bond acid. Weak dipole interactions
DIOL Strong hydrogen-bond acid and weak hydrogen-bond base with significant capacity for dipole-type interactions
Examples
N Isolation of polar pesticides from fats and oils
N Isolation of polycyclic aromatic compounds from fuel oils
N Active ingredients from ointments and suppositories

(4) Ion-exchange sorbents

E In general strong ion exchangers are used to isolate weak acid/bases of opposite charge and weak ion exchangers strong
acid/bases
E Retention selectivity can be adjusted by manipulating the sample pH and ionic strength
E Choice of competing ion, its concentration and eluent pH controls selectivity for matrix simplification and elution
E Isolation of macromolecules in an active form may require special non-denaturing sorbents based on cellulose, agarose or dextran
Examples
N Isolation of carboxylic, sulfonic and phosphoric acids, phenols, amines and inorganic ions from water
N Isolation of amino acids, organic acids, nucleosides and nucleotides from biofluids
N Isolation of organic acids and bases from coal-derived and synthetic fuels
N Isolation of organic acids, phenols and amines from wine, fruit juices and food extracts

sodium sulfate connected in tandem with the adsor-
bent cartridge can be used as an additional pre-
caution. The Brockmann scale (based on the relative
retention of test dyes, see Table 2) provides a widely
used standardized scale of adsorbent activity. Adsor-
bents of deRned activity are prepared by adding
a known amount of water to the adsorbent, shaking
to avoid clumping, and then allowing the adsorbent
to equilibrate overnight in a closed container. Analyte
properties that increase retention depend on the num-
ber and type of functional groups present. Hydrogen-
bonding functional groups are strongly retained,
those with a signiRcant dipole-character are retained
to a lesser extent, and polarizable functional groups
are the least retained. Irreversible adsorption and
catalytic degradation of sensitive analytes can occur
on all inorganic oxide adsorbents and is a source of
low recovery for some analytes. Alumina and silica
can function as selective ion exchange sorbents with
buffered aqueous samples (see Table 1).
1410 II / EXTRACTION / Solid-Phase Extraction
Table 2 Standardization of adsorbent activity
Brockmann activity grade Percentage of water (w/w)
Alumina Silica gel
I1 0 0
II 3 5
III 6 15
IV 10 25
V 15 38
1
Activate sorbents by heating alumina at 4003C for 8}12 h, silica
gel at 1803C for 8}12 h, and Florisil at 1303C for 8}12 h.
Coating silica or alumina with chemical reagents,
such as sulfuric acid, sodium hydroxide, alkaline po-
tassium permanganate, silver nitrate, etc., is used to
improve the selectivity of the isolation of some
analytes from their matrix. Silver nitrate, for
example, improves the isolation of oleRns from hy-
drocarbons due to formation of charge-transfer com-
plexes. Acids can be used for the selective isolation of
bases and vice versa. Silica impregnated with 2,4-
dinitrophenylhydrazine is widely used for the selec-
tive isolation of volatile ketones and aldehydes
from air for analysis by high pressure liquid
chromatography.
Low Speci\city Sorbents and
Their Applications
Low speciRcity sorbents include silica-based, chemic-
ally bonded sorbents, macroreticular porous poly-
mers and various forms of carbon. Silica-based,
chemically bonded sorbents are derived from mater-
ials developed for high pressure liquid chromatogra-
phy. They are generally prepared by reaction of
monofunctional or trifunctional silanes with silica gel
followed by end-capping in some cases. Trifunctional
reagents result in sorbents with a polymeric-bonded
layer of higher carbon loading and greater acid stabil-
ity and are the more common type of sorbent in
general use. Chemically bonded sorbents can be pre-
pared with a wide range of bonding densities, pore
sizes and functional group types. Some common
examples are given in Table 3. Chemically bonded
sorbents with large surface areas, long alkyl chains
and high phase loading maximize retention of small
analytes from aqueous solution, while wide-pore ma-
terials with low phase loading and short alkyl chains
are generally used to isolate macromolecules. Chem-
ically bonded sorbents with immobilized polar func-
tional groups are used to isolate analytes from or-
ganic solutions, based on their selective interactions
with analyte polar functional groups (see Table 1).
The macroreticular porous polymers are co-
polymers of styrene-divinylbenzene or acrylic esters,
prepared by suspension polymerization to yield par-
Florisil ticles consisting of agglomerates of randomly packed
microspheres permeated by a network of holes and
0 channels (Table 4). They are used exclusively for
7 extraction from aqueous solution and are more reten-
15
tive than most chemically bonded phases. They pos-
25
35 sess a high sample capacity and are frequently used in
large-scale isolation studies and for the puriRcation of
industrial products. Tenax] , a polymer based on 2,6-
diphenyl-4-phenylene oxide, revolutionized the
sorbent trapping of volatile organic compounds from
air and the purge-and-trap analysis of volatile organic
compounds in water. It exhibits strong retention of
semivolatile organic compounds ('C7) at room tem-
perature with little adsorption of water vapour and
can be rapidly heated to high temperatures, without
thermal breakdown, for the recovery of analytes by
thermal desorption. Since no single adsorbent is ideal
for trapping all analytes it is common practice to use
cartridges packed with several adsorbent beds in
series, so that a broad range of compounds with
different molecular weight and polarity can be trap-
ped on a single cartridge. Besides Tenax, different
forms of carbon, silica gel and liquid-coated sorbents
are used. In a multiple bed cartridge, each bed pro-
tects the next, increasingly active bed, by preventing
compounds from being held so strongly that they
cannot be desorbed quickly without decomposition.
During thermal desorption the carrier gas passes
through the trap in the reverse direction to the sample
Sow and the desorbed compounds are swept onto the
separation column in a gas chromatograph.
A cryogenic interface may be used to refocus the
desorbed sample to improve the chromatographic
separation. The complete processes of desorption and
separation can be automated for sample cartridges
stored in an autosampler.
The common forms of carbon used in solid-phase
extraction are granular activated carbon, graphitized
carbon blacks and carbon molecular sieves. Granular
activated carbons are prepared by the low temper-
ature oxidation of vegetable charcoals. They have
large surface areas (300}2000 m2 g\1), a wide pore
size distribution, and a heterogeneous surface con-
taining active functional groups. Their use in solid-
phase extraction is largely conRned to the isolation of
dissolved organic compounds in surface waters, and
as the sorbent material in personal monitors for
sampling workplace atmospheres. The most common
form of personal monitor makes use of a sorbent
cartridge Rlled with activated charcoal in conjunction
with a small pump to maintain a Rxed Sow of air
through the cartridge. Trapped volatile compounds
 II / EXTRACTION / Solid-Phase Extraction 1411

Table 3 Structures of silica-based chemically bonded sorbents

Type Functional group Structure

C18 Octadecyl

C8 Octyl

C2 Ethyl

CH Cyclohexyl

PH Phenyl

CN Cyanopropyl

NH2 Aminopropyl

DIOL 2,3-Dihydroxypropoxypropyl

SAX Trimethylaminopropyl (quaternary amine)

CBA Carboxypropyl

SCX Benzenesulfonic acid

PRS Propylsulfonic acid

are then eluted with carbon disulRde or another sol-
vent, or can be thermally desorbed by microwave
heating, for separation by gas chromatography. Poor
reproducibility of activated carbons and their vari-
able chemical and catalytic activity result in limited
laboratory use. Graphitized carbon blacks are more
reRned and generally nonporous, with surface areas
between about 5}100 m2 g\1. They are used prim-
arily for the room temperature trapping of volatile
organic compounds ('C4), either separately or in
combination with Tenax] . Carbon molecular sieves
have small pores and large surface areas ('500 m2 g\1
with some '1200 m2 g\1). They are used primarily
for the room temperature trapping of volatile organic
compounds (C1 and C2), usually as a component of
a multiple-bed sorbent trap for air sampling and
purge-and-trap analysis.
Foamed polyurethanes, composed of agglomerated
spherical micrometer-sized particles bonded to one
another in a rigid and highly permeable structure, are
suitable for sampling semivolatile organic com-
pounds (e.g. airborne pesticides and polychlorinated
biphenyls) at high Sow rates. They are frequently
used in conjunction with high-volume air samplers on
1412 II / EXTRACTION / Solid-Phase Extraction
Table 4 Characteristic properties of some macroreticular porous polymer sorbents
Amberlite sorbents Mean pore diameter Specific surface area
(nm) (m2 g\1)
XAD-2 (STY-DVB) 9 300
XAD-4 (STY-DVB) 4 725
XAD-7 (MMA) 9 450
XAD-16 (STY-DVB) 10 800
XAD-2010 (STY-DVB) 28 660
DAX-8 (MMA) 22.5 160
STY-DVB, styrene-divinylbenzene; MMA, methylmethacrylate.
account of their low pressure drop compared with
standard sorbent cartridges. They are used less fre-
quently for water analysis where macroreticular por-
ous polymers are considered a better choice.
Compound and Class-speci\c
Sorbents and their Applications
Various forms of selective sorbents for solid-phase
extraction based on ion exchange, bioafRnity, mo-
lecular recognition, and restricted access are used to
supplement the general class of sorbents discussed
above. Ion exchange is used to isolate ionizable com-
pounds (usually) in aqueous solution with sorbents
containing Rxed ionic sites of opposite charge to the
analytes of interest. Ion-exchange sorbents are usu-
ally classiRed as weak or strong depending on the
identity of the ionic group and whether its charge is
independent of the sample pH (strong ion exchanger)
or can be manipulated by changing the pH (weak ion
exchanger). Some examples of typical silica-based
ion-exchange sorbents are indicated in Table 3. Ion-
exchange sorbents with a porous polymer backbone
are also commonly used and have a higher exchange
capacity and a wider pH-operating range than silica-
based sorbents. For many applications either silica-
based or porous polymer ion-exchange sorbents with
the same immobilized ionic groups can be used inter-
changeably, although, because of non-speciRc ad-
sorption of matrix components, the chemical back-
ground of the extracts might be different. Ion-ex-
change sorbents are particularly attractive for the
isolation of ionizable substances since the neutral
molecules, which may interfere in the Rnal chromato-
graphic analysis, are easily rinsed from the sorbent
without affecting the recovery of the ionized compo-
nents. Mixed-mode sorbents containing ion-ex-
change sites and alkyl groups co-bonded to silica in
either cartridge or disc format are popular in clinical
and pharmaceutical laboratories, where they are used
for the isolation of ionized drugs and their metab-
olites from biological Suids. Standard protocols using
Pore volume Sample molecular
(mL g\1) weight limit
0.65 20 000
0.98
1.14 60 000
1.82 40 000
1.80
0.79 150 000
mixed-mode sorbents have been developed for the
isolation of most drugs of abuse (e.g. amphetamines,
barbiturates, cocaine, opiates, etc.). The strong reten-
tion and the use of efRcient rinse solvents results in
cleaner extracts compared with single-mode sorbents,
suitable for screening by thin-layer chromatography
and conRrmation by gas chromatography}mass spec-
trometry.
Resin-bound phenylboronic acids are used for the
isolation of compounds with vicinal diol groups such
as steroids, catecholamines and nucleotides. Surface-
bonded macrocyclic ligands, cryptands, can be used
for the selective isolation of metal ions. The cryptands
can be synthesized with a variety of cavity sizes suit-
able for the isolation of different metal ions. The
metal ion is sorbed in the cavity of the cryptand until
released by elution with a solution of a complexing
agent with a high binding constant for the metal.
Immunosorbents have been used for a long time for
sample pretreatment in medicine and biology, but
more general applications, such as to environmental
analysis, are relatively recent. In part, this is due to
the difRculty of making antibodies selective to small
molecules, as well as a lack of familiarity among
analytical chemists of the procedures used to make
speciRc antibodies. Immunosorbents are prepared by
covalently bonding a suitable antibody to an appro-
priate sorbent. A high degree of molecular selectivity
is obtained based on the speciRcity of the anti-
body}antigen (analyte) interaction. Because speciR-
city is high, immunosorbents are able to isolate target
analytes from complex matrices in a single step with
minimal co-extraction of matrix interferences. By
taking advantage of cross-reactivity, class-speciRc im-
munosorbents for the isolation of mycotoxins,
phenylurea herbicides and polycyclic aromatic hydro-
carbons have been developed. Manufactured im-
munosorbents have been available for only a short
time and the range of products is still narrow. A la-
boratory familiar with the techniques for raising and
isolating antibodies is required.
Molecular imprinting is a technique used for
preparing polymers with synthetic recognition sites

having a predetermined selectivity for a speciRed
analyte. The imprint is obtained by arranging poly-
merizable functional monomers around a template
(the analyte). Template}monomer complexes are for-
med in solution through molecular interactions and
subsequently Rxed in place by cross-linking. Removal
of the template from the resulting polymer matrix
creates vacant recognition sites that exhibit afRnity
for the analyte. For the time being, it is impossible to
predetermine the experimental conditions for suc-
cessful imprinting of target analytes. The template
molecule may be difRcult to leach from the imprinted
polymer, reducing the binding capacity of the poly-
mer, but more seriously, it may lead to contamination
of sample extracts. Only a few practical applications
using molecularly imprinted polymers for solid-phase
extraction have been demonstrated so far, most of
which are for the isolation of drugs from biological
Suids, but the future for this technology looks very
promising. Molecularly imprinted polymers should
be easier and cheaper to produce in chemical laborat-
ories than antibodies while, at least in theory, they
should be capable of similar speciRcity.
Restricted access sorbents have been developed for
the isolation of low molecular weight compounds,
generally drugs, directly from biological Suids with
minimum sample pretreatment. They work by pre-
venting access of macromolecules (proteins) to those
regions of the sorbent where retention of the analyte
occurs. Restricted access to the retentive part of the
sorbent is provided by either a physical diffusion
barrier, such as a pore diameter, or by a chemical
diffusion barrier, such as a polymer network at the
outer surface of the particle. In addition, the outer
surface of the particles must be non-adsorptive and
protein-compatible. Restricted access sorbents are
commonly used for automated on-line sample pro-
cessing in liquid chromatography. In this case, a short
precolumn packed with the restricted access material
is interfaced to a separation column by a six-port
switching valve. The bioSuid is injected directly onto
the precolumn, which retains the analytes of interest.
Potentially interfering sample constituents are then
Sushed to waste. Macromolecules (proteins) pass
through the precolumn unretained and do not inter-
fere in the subsequent separation of the analytes. The
analytes retained on the precolumn are eluted on-line
to the separation column and detected. Simulta-
neously the precolumn is reconditioned (or ex-
changed) before processing the next sample. An im-
portant consideration for automated sample process-
ing is the ability of the restricted access sorbent to
repeatedly extract the analyte without change in
properties or accumulation of sample matrix compo-
nents.
II / EXTRACTION / Solid-Phase Extraction 1413
Sample Processing Considerations
Solid-phase extraction cartridges are available in
a range of sizes containing from about 35 mg to 10 g
of sorbent, with the 100 mg and 500 mg sorbent
cartridges (or discs) being the most widely used for
extraction and the larger cartridge sizes for sample
clean up. As a rough guide, the sorbed sample
capacity of a solid-phase extraction device is about
1}5% of the sorbent mass. The sample volume
that can be processed depends primarily on the break-
through volume of the analyte, the concentration of
the analyte matrix, sample Sow rate, and the sorbent
mass. The sample volume is often selected to conform
to the needs of the instrumental detection step, and as
instrumental methods of determination have im-
proved in sensitivity, sample volumes have decreased
in size. Regulatory authorities often indicate action
levels in concentration units, which can also be used
to deRne an adequate sample volume for analysis.
Sorbent selection is based on the considerations
summarized in Figure 6. The sample solvent (aqueous
or organic), the analyte type (non-polar, polar or
ionized), and whether it is ionized (strong or weak,
acid or base) provides a logical guide for method
selection. Organic compounds soluble in polar or-
ganic solvents but difRcult to dissolve in solvents of
intermediate polarity, can be extracted in the rever-
sed-phase mode if they can be reconstituted in aque-
ous solution.
Sample processing involves four distinct steps. Ini-
tially, the sorbent is conditioned with solvent to im-
prove the reproducibility of analyte retention and to
reduce the carrythrough of sorbent impurities at the
elution stage. The conditioning solvent is then re-
placed with the same solvent as the sample solvent
and the sample passed through the sampling device at
a controlled Sow rate. Optionally, after the sample
has been processed, the sorbent is rinsed with a weak
solvent to displace undesired matrix components
from the sorbent without displacing the analytes.
Finally, the analytes of interest are eluted from the
sorbent in a small volume of strong solvent for sub-
sequent determination. Hidden in the above descrip-
tion of events are a number of sub-steps that can
dramatically inSuence analyte recovery if not ad-
equately optimized (Table 5). The conditioning step
is critically important for processing aqueous samples
using particle-loaded membranes. The high surface
tension of water combined with the microporosity of
the discs results in slow and uneven Sow through the
discs and low analyte recovery if the discs are not Rrst
conditioned with an organic solvent. For large sample
volumes, a small amount of the same organic solvent
is usually added to the sample to maintain a constant
1414 II / EXTRACTION / Solid-Phase Extraction
Figure 6 Method development guide for the isolation of organic compounds from liquid samples. SAX, strong anion exchanger; SCX,
strong cation exchanger; WCX, weak cation exchanger; RP, reversed-phase sampling mode; NP, normal-phase sampling mode; IE,
ion exchange sampling mode.
sample Sow rate. The drying step between processing
aqueous samples and eluting the retained analytes
with a water-miscible organic solvent is also impor-
tant. The purpose of the drying step is to reduce the
volume of water co-eluting from the sampling device
permitting further concentration of the eluent by the
gas-blow down method. Drying, by suction or stor-
age in a vacuum desiccator, should be sufRcient to
remove water trapped in the pores, but excessive
drying can result in low analyte recovery from evap-
oration or inefRcient elution. A new porous polymer
sorbent, prepared by copolymerization of divinylben-
zene and N-vinylpyrrolidone (Oasis HLB), and sol-
vated by water alone, has been suggested as a solution
to this problem. Recently, it was shown that all
sample-processing steps are amenable to computer-
aided method development, replacing the traditional
experimental trial-and-error approach by fast com-
puter simulations.
Automation
Automation provides a better utilization of laborat-
ory resources, unattended and out-of-hours operation
and improved precision compared with manual
methods. Common approaches to automation differ
signiRcantly. Using robotics, samples are processed
(usually) in a similar manner to manual methods.
Using Sow-processing schemes, samples are extracted
in parallel with computer or microprocessor control
of solvent management. Sorbent conditioning,
sample condition, solvent selection, rinse and elution
steps are performed automatically and can be varied
for method development. Positive displacement in-
stead of suction is used for solvent control, and ad-
vanced units can be programmed to replace sorbent
cartridges to increase sample throughput and inject
extracts into different chromatographic instruments.
On-line analysers with a direct coupling to chromato-
graphic instruments are widely used. Solid-phase ex-
traction using short precolumns and a switching valve
interface is a routine method for analysis by liquid
chromatography. Advanced systems even allow pro-
grammed replacement of the sorbent cartridges and
unattended 24-hour operation. The recovery and sep-
aration steps of purge-and-trap and sorbent trapping
of volatile organic compounds from air are easily
automated using thermal desorption with cold trap-
ping, if required, for the direct injection of analytes
into a gas chromatograph. Major strides have been
made in the on-line solid-phase extraction of water
samples with solvent desorption directly into a gas
chromatograph. This method is not far from becom-
ing routine today.
 II / EXTRACTION / Solid-Phase Extraction 1415

Table 5 Experimental variables that influence recovery of analytes by solid-phase extraction

E Conditioning solvent (typically 3}5 bed volumes)

N Ensures reproducible retention and flow. Critical step for particle-loaded membranes
N Helps to minimize contamination of extracts by sorbent impurities
N Replace by sample solvent before processing sample

E Flow rates (typical range 0.2}1.5 mm s\1)

N More critical for cartridges than discs due to their variable and heterogeneous packing density (channelling)
N More critical when the sample volume exceeds the breakthrough volume as typical sampling devices provide too few
theoretical plates for flow-independent retention

E Sample properties
N Dilute viscous samples with a weak low viscosity solvent to reduce sample processing time
N Remove excessive particle matter by filtration or centrifugation to maintain a constant sample-processing rate. Concentrated
hydrochloric acid is effective for dissolving inorganic particles in water samples
N Add small volume of organic solvent (1}3% v/v) to large volume water samples to ensure sorbent remains solvated and to
maintain a constant (fast) sample-processing rate. Important for particle-loaded membranes
N Adjust pH to reduce ionization of weak acids and bases for reversed-phase sampling
N Maintain ionic strength approximately constant for samples and standards with reversed-phase sampling conditions. Ionic
strength is a critical parameter for ion-exchange extraction
N Deproteination of biofluids may be required for acceptable recovery of low molecular weight analytes for reversed-phase
sampling
N Precipitation of inorganic acids (sulfate, phosphate, etc.) by barium hydroxide is sometimes required for acceptable recovery
of organic acids from biofluids using ion-exchange extraction

E Drying time (typically 1}5 min, but sometimes considerably longer)

N Sufficient to remove all sample solvent trapped in the sorbent pores
N Excessive drying may result in low recovery of analytes from evaporation or retention in poorly solvated regions of the sorbent

E Rinse solvent (optional)

N Small volume of intermediate strength solvent to elute matrix components. Analytes remain immobilized on the sorbent
N Biological fluids, plant extracts and soil extracts often require a rinse step but surface waters may not

E Eluting solvent (ideally 2}3 bed volumes but often larger)

N Should be a strong solvent able to displace all analyte from the sorbent in a small volume
N Should normally be volatile and miscible with the sample solvent

Future developments Extraction. Restricted-Access Media: Solid-Phase Ex-

Solid-phase extraction is approaching maturity and is
a familiar laboratory operation for many analytical
chemists. Advances are expected in the area of speci-
Rc sorbents based on molecular imprinting or bioaf-
Rnity designed for the convenient isolation of target
compounds in complex matrices. Advances are also
expected in the use of computer-aided method devel-
opment for the prediction of sampling and recovery
conditions by simulation to replace tedious experi-
mental trial-and-error approaches. A wider use and
further development of automated solid-phase ex-
traction systems can be expected, particularly in those
industries where high sample throughput or round-
the-clock process monitoring are important.
See also: II/Affinity Separation: Immobilised Boronates
and Lectins; Imprint Polymers. Extraction: Solvent Based
Separation. III/Airborne Samples: Solid Phase Extrac-
tion. Immunoaffinity Extraction. Immobilised Boronic
Acids: Extraction. Molecular Imprints for Solid-Phase
traction. Solid-Phase Extraction with Cartridges.
Solid-Phase Extraction with Disks. Solid-Phase Matrix
Dispersion: Extraction. Sorbent Selection for Solid-
Phase Extraction. Appendix 2 / Essential Guides to
Method Development in Affinity Chromatography.
Further Reading
Dean JR (1998) Extraction Methods For Environmental
Analysis. Chichester: John Wiley.
Hennion M-C and Pichon V (1994) Solid-phase extraction
of polar organic pollutants from water. Environmental
Science and Technology 28: 576A}583A.
Masque N, Marce RM and Borrull F (1998) New polymeric
and other types of sorbents for solid-phase extraction of
polar organic micropollutants from environmental water.
Trends in Analytical Chemistry 17: 384}394.
Mayer ML and Poole CF (1994) IdentiRcation of the pro-
cedural steps that affect recovery of semi-volatile com-
pounds by solid-phase extraction using cartridge and
particle-loaded membrane (disk) devices. Analytica
Chimica Acta 294: 113}126.
1416 II / EXTRACTION / Solid-Phase Microextraction
Pinchon V, Bouzige M, Miege C and Hennion M-C (1999)
Immunosorbents: natural molecular recognition mater-
ials for sample preparation of complex environmental
matrices. Trends in Analytical Chemistry 18: 219}235.
Poole CF, Poole SK, Seibert DS and Chapman CM (1997)
Determination of kinetic and retention properties of
cartridge and disk devices for solid-phase extraction.
Journal of Chromatography B 689: 245}260.
Poole CF and Poole SK (1991) Chromatography Today.
Amsterdam: Elsevier.
Seibert DS and Poole CF (1998) A general model for the
optimization of sample processing conditions by solid-
Solid-Phase Microextraction
J. Pawliszyn, University of Waterloo, Waterloo,
Canada
Copyright ^ 2000 Academic Press
Introduction
Solid-phase microextraction (SPME) was introduced
as a solvent-free sample preparation technique in
1990. The basic principle of this approach is to use
a small amount of the extracting phase (usually less
than 1
L) compared to the sample matrix. Sample
volume can be very large, when the investigated sys-
tem, for example air or lake water, is sampled
directly. The extracting phase can be either a high
molecular weight polymeric liquid, similar in nature
to chromatographic stationary phases, or it can be
a solid sorbent, typically of a high porosity to increase
the surface area available for adsorption.
To date the most practical geometric conRguration
of SPME utilizes a small fused silica Rbre, usually
coated with a thin Rlm of polymeric phase. The Rbre
is mounted for protection in a syringe-like device
(Figure 1A). The analytes are absorbed or adsorbed
by the Rbre coating (depending on the nature of the
coating) until an equilibrium is reached in the system.
The amount of an analyte extracted by the coating at
equilibrium is determined by the magnitude of the
partition coefRcient (distribution ratio) of the analyte
between the sample matrix and the coating material.
In SPME, analytes typically are not exhaustively
extracted from the matrix. However, equilibrium
methods are more selective because they take full
advantage of the differences in extracting phase/
matrix distribution constants to separate target
analytes from interferences. Exhaustive extraction
can be achieved in SPME when the distribution con-
stants are large enough. This can be accomplished for
most compounds by cooling the Rbre coating. This
phase extraction applied to the isolation of estrogens
from urine. Journal of High Resolution Chromatogra-
phy 21: 481}490.
Sellergren B (1999) Polymer- and template-related factors
inSuencing the efRciency in molecularly imprinted
solid-phase extractions. Trends Analytical Chemistry
18: 164}174.
Simpson NJK (2000) Solid Phase Extraction: Principles,
Strategies, and Applications. New York: Marcel
Dekker.
Thurman EH and Mills MS (1998) Solid-Phase Extraction,
Principles and Practice. New York: John Wiley.
concept was tested using a piece of microtubing
coated on the outside instead of a solid rod and
supplying liquid carbon dioxide into the tube to
achieve an internally cooled Rbre. In exhaustive ex-
traction, selectivity is sacriRced to obtain quantitative
transfer to target analytes into the extracting phase.
One advantage of this approach is that, in principle, it
does not require calibration, since all the analytes of
interest are transferred to the extracting phase. On
the other hand, the equilibrium approach usually
requires calibration through the use of surrogates or
standard addition to quantify the analytes and com-
pensate for matrix-to-matrix variations and their ef-
fect on distribution constants.
Since equilibrium rather than exhaustive extraction
occurs in microextraction methods, SPME is ideal for
Reld monitoring. It is unnecessary to measure the
volume of the extracted sample and therefore the
SPME device can be exposed directly to the investi-
gated system for quantitation of target analytes. Thin
coatings of extracting phase result in fast separations.
In addition, extracted analytes are introduced to the
analytical instrument inlet system by simply placing
the Rbre in the desorption unit (Figure 1B and 1C).
This convenient, solvent-free sample introduction
process facilitates sharp injection bands and rapid
separations. These features of SPME result in the
integration of the Rrst steps in the analytical process:
sampling, sample preparation and introduction of
extracted mixture to the analytical instrument. For
example, total analysis time in Reld applications can
be as low as a few minutes when portable instrumen-
tation is used.
The equilibrium nature of the technique also
facilitates speciation in natural systems since the
presence of a minute Rbre, which removes small
amounts of target analytes, is not likely to disturb
the system. Because of the small size, coated Rbres can
 II / EXTRACTION / Solid-Phase Microextraction 1417

Figure 1 (A) Design of a commercial SPME device. (B) SPME}HPLC interface: (a) stainless steel (SS) 1/16
tee; (b) 1/16
SS
tubing; (c) 1/16
polyetheretherketone (PEEK) tubing (0.02
i.d.); (d) two-piece finger-tight PEEK union; (e) PEEK tubing (0.005
i.d.)
with a one-piece PEEK union. (C) SPME}GC interface.

be used to extract analytes from very small samples.
For example, SPME has been used to probe for sub-
stances emitted by a single Sower bloom during its
lifespan.
Figure 1A illustrates the commercial SPME device
manufactured by Supelco, Inc. (Bellefonte, PA, USA).
The Rbre, glued into a piece of stainless steel tubing, is
mounted in a special holder. The holder is equipped
with an adjustable depth gauge, which makes it pos-
sible to control repeatably how far the needle of the
device is allowed to penetrate the sample container (if
any) or the injector. This is important, as the Rbre can
1418 II / EXTRACTION / Solid-Phase Microextraction
be easily broken when it hits an obstacle. The move-
ment of the plunger is limited by a small screw mov-
ing in the z-shaped slot of the device. For protection
during storage or septum piercing, the Rbre is with-
drawn into the needle of the device, with the screw in
the uppermost position. During extraction or desorp-
tion, the Rbre is exposed by depressing the plunger,
which can be locked in the lowered (middle) position
by turning it clockwise (the position depicted in
Figure 1A). The plunger is moved to its lowermost
position only for replacement of the Rbre assembly.
Each type of Rbre has a hub of a different colour. The
hub-viewing window permits a quick check to be
made of the type of Rbre mounted in the device.
If the sample is placed in a vial, the septum of the
vial is Rrst pierced with the needle (with the Rbre in
the retracted position) and the plunger is lowered,
which exposes the Rbre to the sample. The analytes
are allowed to partition into the coating for a prede-
termined time, and the Rbre is then retracted back
into the needle. When gas chromatography (GC) is
used for analyte separation and quantitation, the Rbre
is inserted into a hot injector, where thermal desorp-
tion of the trapped analytes takes place (Figure 1C).
All extracted compounds are introduced to the ana-
lytical instrument facilitating high sensitivity of deter-
minations. The Rbre desorption process can be auto-
mated by using an appropriately modiRed, commer-
cially available syringe autosampler. For high perfor-
mance liquid chromatography (HPLC) applications,
a simple interface mounted in place of the injection
loop can be used to re-extract analytes into the de-
sorption solvent (Figure 1B). The extraction phase
Figure 2 Modes of SPME operation: (A) direct extraction, (B) headspace extraction and (C) membrane-protected SPME.
can also coat the inner wall of the capillary. This
approach to microextraction can be automated using
a number of commercially available autosamplers,
but it is limited to extraction of relatively clean sam-
ples, which do not plug capillaries.
The SPME device is suitable for both spot and
time-averaged sampling. As described above, for
spot sampling, the Rbre is exposed to a sample matrix
until equilibrium is reached between the sample
matrix and the coating material on the Rbre. In the
time-average approach, on the other hand, the Rbre
remains in the needle during the exposure of the
SPME device to the sample. The coating works as
a trap for analytes that diffuse into the needle, result-
ing in the integration of concentration over given
time.
SPME sampling can be performed in three basic
modes: direct extraction, headspace extraction and
extraction with membrane protection. Figure 2 illus-
trates the differences between these modes. In direct
extraction mode (Figure 2A), the coated Rbre is in-
serted into the sample and the analytes are trans-
ported directly from the sample matrix to the extract-
ing phase. To facilitate rapid extraction, some level of
agitation is required to transport the analytes from
the bulk of the sample to the vicinity of the Rbre. For
gaseous samples, natural Sow (e.g. convection) is
frequently sufRcient to facilitate rapid equilibration,
but for aqueous matrices, more efRcient agitation
techniques, such as fast sample Sow, rapid Rbre or
vial movement, stirring or sonication are required to
reduce the effect of the depletion zone produced close
to the Rbre as a result of slow diffusional analyte

transport through the otherwise static layer of liquid
surrounding the Rbre.
In the headspace mode (Figure 2B), the analytes are
extracted from the gas phase equilibrated with the
sample. The primary reason for this modiRcation is to
protect the Rbre from adverse effects caused by non-
volatile, high molecular weight substances present in
the sample matrix (e.g. humic acids or proteins). The
headspace mode also allows matrix modiRcations,
including pH adjustment, without affecting the Rbre.
In a closed system consisting of a liquid sample and its
headspace, the amount of an analyte extracted by the
Rbre coating does not depend on the location of the
Rbre, therefore the sensitivity of headspace sampling
is the same as the sensitivity of direct sampling as long
as the volumes of the two phases are the same in both
sampling modes. Even when headspace is not used in
direct extraction, a signiRcant sensitivity difference
between direct and headspace sampling can occur
only for very volatile analytes. However, the choice of
sampling mode has a signiRcant impact on the extrac-
tion kinetics. When the Rbre is in the headspace, the
analytes are removed from the headspace Rrst, fol-
lowed by indirect extraction from the matrix. There-
fore, volatile analytes are extracted faster than
semivolatiles. Temperature has a signiRcant effect on
the kinetics of the process, since it determines the
vapour pressure of analytes. In general, the equilibra-
tion times for volatile compounds are shorter for
headspace SPME extraction than for direct extraction
under similar agitation conditions, for the following
reasons: (i) a substantial portion of the analytes is
present in the headspace before the extraction process
begins; (ii) there is typically a large interface between
sample matrix and headspace; and (iii) the diffusion
coefRcients in the gas phase are typically higher by
four orders of magnitude than in liquids. The concen-
tration of semivolatile compounds in the gaseous
phase at room temperature is small, consequently
headspace extraction rates for those compounds are
substantially lower. These rates can be improved by
using efRcient agitation or by increasing the extrac-
tion temperature.
In the third mode (SPME extraction with mem-
brane protection, Figure 2C), the Rbre is separated
from the sample by a selective membrane, which lets
the analytes through while blocking the interferences.
The main purpose for the use of the membrane bar-
rier is to protect the Rbre against adverse effects
caused by high molecular weight compounds when
very dirty samples are analysed. While extraction
from headspace serves the same purpose, membrane
protection allows the analysis of less volatile com-
pounds. The extraction process is substantially
slower than direct extraction because the analytes
II / EXTRACTION / Solid-Phase Microextraction 1419
have to diffuse through the membrane before they
can reach the coating. Use of thin membranes and
increase in extraction temperature, applied to analy-
sis of polyaromatic hydrocarbons (PAHs) in matrices
containing humic matter, result in shorter extraction
times.
Theoretical Aspects of Solid-phase
Microextraction Optimization and
Calibration
Thermodynamics
SPME is a multiphase equilibration process. Fre-
quently, the extraction system is complex, as in
a sample consisting of an aqueous phase with sus-
pended solid particles having various adsorption in-
teractions with analytes, plus a gaseous headspace. In
some cases speciRc factors have to be considered,
such as analyte losses by biodegradation or adsorp-
tion on the walls of the sampling vessel. In the dis-
cussion below we will only consider three phases: the
Rbre coating, the gas phase or headspace, and a ho-
mogeneous matrix such as pure water or air. During
extraction, analytes migrate between all three phases
until equilibrium is reached. The following discussion
is limited to partitioning equilibrium involving liquid
polymeric phases such as poly(dimethylsiloxane).
The method of analysis for solid sorbent coatings is
analogous for low analyte concentration, since the
total surface area available for adsorption is propor-
tional to the coating volume if we assume constant
porosity of the sorbent.
The mass of an analyte extracted by the polymeric
coating is related to the overall equilibrium of the
analyte in the three-phase system. Since the total mass
of an analyte should remain constant during the ex-
traction, we have:
C0Vs"C
f Vf#C
h Vh#C
s Vs [1]
where C0 is the initial concentration of the analyte in
the matrix: Cf , C
h and Cs are the equilibrium con-
centrations of the analyte in the coating, the head-
space and the matrix, respectively; Vf, Vh and Vs are
the volumes of the coating, the headspace and the
matrix, respectively. If we deRne the coating/gas dis-
tribution constant as Kfh"C f /C
h , and the gas/
sample matrix distribution constant as Khs"C h /Cs ,
the mass of the analyte absorbed by the coating,
n"C f Vf, can be expressed as:
KfhKhsVfC0Vs
n" [2]
KfhKhsVf#KhsVh#Vs
1420 II / EXTRACTION / Solid-Phase Microextraction
Also:
KH
Kfs" "KfhKhs"KfgKgs [3]
KF
since the Rbre/headspace distribution constant,
Kfh can be approximated by the Rbre/gas distribution
constant Kfg, and the headspace/sample distribution
constant, Khs, by the gas/sample distribution con-
stant, Kgs, if the effect of moisture in the gaseous
headspace can be neglected. Thus, eqn [2] can be
written as:
KfsVfC0Vs
n" [4]
KfsVf#KhsVh#Vs
The equation states, as expected from the equilib-
rium conditions, that the amount of analyte extracted
is independent of the location of the Rbre in the
system. It may be placed in the headspace or directly
in the sample as long as the volumes of the Rbre
coating, headspace and sample are kept constant.
There are three terms in the denominator of eqn [4]
which give measures of the analyte capacity of each of
the three phases: Rbre (KfsVf), headspace (KhsVh) and
the sample itself (Vs). If we assume that the vial
containing the sample is completely Rlled (no head-
space), the term KhsVh in the denominator, which is
related to the capacity (Ch Vh) of the headspace, can
be eliminated, resulting in:
KfsVfC0Vs
n" [5]
KfsVf#Vs
Equation [5] describes the mass absorbed by the
polymeric coating after equilibrium has been reached
in the system. In most determinations, Kfs is relatively
small compared to the phase ratio of sample matrix to
coating volume (Vf  Vs). In this situation the capa-
city of the sample is much larger compared to
capacity of the Rbre, resulting in a very simple
relationship:
n"KfsVfC0 [6]
The above equation emphasizes the Reld-sampling
capability of the SPME technique. It is not necessary
to sample a well-deRned volume of the matrix since
the amount of analyte extracted is independent of
Vs as long as KfsVf  Vs. The SPME device can be
placed directly in contact with the investigated system
to allow quantitation.
Prediction of distribution constants In many cases,
the distribution constants present in eqns [2]}[6]
which determine the sensitivity of SPME extraction
can be estimated from physicochemical data and
chromatographic parameters. For example, distribu-
tion constants between a Rbre coating and gaseous
matrix (e.g. air) can be estimated from isothermal GC
retention times on a column with a stationary phase
identical to the Rbre-coating material. This is possible
because the partitioning process in gas chromatogra-
phy is similar to the partitioning process in SPME,
and there is a well-deRned relationship between the
distribution constant and the retention time. The na-
ture of the gaseous phase does not affect the distribu-
tion constant, unless the components of the gas, such
as moisture, swell the polymer, thus changing its
properties. A most useful method for determining
coating-to-gas distribution constants uses the linear
temperature programmed retention index (LTPRI)
system, which relates retention times relative to the
retention times of n-alkanes. The logarithm of the
coating-to-air distribution constants of n-alkanes can
be expressed as a linear function of their LTPRI
values. For poly(dimethylsiloxane) (PDMS), this rela-
tionship is log Kfg"0.00415*LTPRI!0.188. Thus,
the LTPRI system permits interpolation of the
Kfg values from the plot of log Kfg versus retention
index. The LTPRI values for many compounds are
available in the literature, hence this method allows
estimation of Kfg values without experimentation. If
the LTPRI value for a compound is not available from
published sources, it can be determined from a GC
run using a GC column coated with the same material
as the Rbre.
Estimation of the coating/water distribution con-
stant can be performed using eqn [5]. The appropri-
ate coating/gas distribution constant can be found by
applying techniques discussed above, and the
gas/water distribution constant (Henry’s constant)
can be obtained from physicochemical tables or can
be estimated by the structural unit contribution
method.
Some correlations can be used to anticipate trends in
SPME coating/water distribution constants for
analytes. For example, a number of investigators have
reported correlation between the octanol/water distri-
bution constant, Kow, and Kfw. This is to be expected,
since Kow is a general measure of the afRnity of com-
pounds for the organic phase. It should be remem-
bered, however, that the trends are valid only for
compounds within homologous series, such as
aliphatic hydrocarbons, aromatic hydrocarbons or
phenols; they should not be used to make comparisons
between different classes of compounds, because of
different analyte activity coefRcients in the polymer.
Effect of extraction parameters Thermodynamic
theory predicts the effects of modifying certain
 II / EXTRACTION / Solid-Phase Microextraction 1421

extraction conditions on partitioning and indicates
parameters to be controlled for reproducibility.
The theory can be used to optimize the extraction
conditions with a minimum number of experiments
and to correct for variations in extraction conditions,
without the need to repeat calibration tests under
the new conditions. For example, SPME analysis of
outdoor air may be done at ambient temperatures
that can vary signiRcantly. A relationship that pre-
dicts the effect of temperature on the amount of
analyte extracted allows calibration without the need
for extensive experimentation. Extraction conditions
that affect Kfs include temperature, inorganic salt
concentration, pH and organic solvent content of the
water.
Kinetics
The kinetic theory is useful to optimize the extraction
conditions by identifying ‘bottlenecks’ in SPME and
indicating strategies to increase extraction speed. In
the discussion below we will limit our consideration
to direct extraction (Figure 3).
Perfect agitation Let us Rrst consider the case where
the liquid or gaseous sample is well agitated. In other
words, the sample phase moves rapidly with respect
to the Rbre, so that all the analytes present in the
sample have access to the Rbre coating. In this case,
the equilibration time, deRned as the time required to
extract 95% of the equilibrium amount (Figure 4) of
an analyte from the sample, corresponds to:
2(b!a)2
te"t95%" [7]
Df
Using this equation one can estimate the shortest
equilibration time possible for a practical system by
substituting appropriate data for the diffusion coefRc-
ient of an analyte in the coating (Df) and the Rbre-
coating thickness (b!a). For example, the equilibra-
tion time for the extraction of benzene from a highly
agitated aqueous solution with a 100
m PDMS Rlm
is expected to be about 20 s assuming diffusion coef-
Rcient of 10\5 cm2 s\1 in PDMS. Equilibration times
close to those predicted for agitated samples have

Figure 3 Graphic representation of the SPME/sample system configuration, with dimensions and parameters labelled as follows: a,
fibre coating inner radius; b, fibre coating outer radius; L, fibre coating length; d, vial inner radius; Cf, analyte concentration in the fibre
coating; Df, analyte diffusion coefficient in the fibre coating; Cs, analyte concentration in the sample; Ds, analyte diffusion coefficient in
the sample; Kfs, analyte distribution coefficient between fibre coating and sample; Kfs"Cf/Cs. (With permission from Louch et al. (1992)
Analytical Chemistry 64: 1187.)
1422 II / EXTRACTION / Solid-Phase Microextraction
Figure 4 Mass absorbed versus time for a well-agitated solu-
tion of infinite volume. (With permission from Louch et al. (1992)
Analytical Chemistry 64: 1187.)
been obtained experimentally for extraction of
analytes from air samples (because of high diffusion
coefRcients in gases) or when high sonication power
is used to facilitate mass transfer in aqueous samples.
However, in practice there is always a layer of unstir-
red water around the Rbre, although a high stirring
rate will reduce its thickness.
Practical agitation Independently of the level of agi-
tation, Suid contacting the Rbre surface is always
stationary, and as the distance from the surface in-
creases, the Suid movement gradually increases until
it corresponds to the bulk Sow in the sample. To
Figure 5 Boundary layer model configuration showing the different regions considered and the assumed concentration versus radius
profile for the case when the boundary layer determines the extraction rate.
model mass transport, the gradation in Suid motion
and convection of molecules in the space surrounding
the Rbre surface can be simpliRed by a zone of a de-
Rned thickness in which no convection occurs, and
perfect agitation in the bulk of the Suid everywhere
else. This static layer zone is called the Prandtl bound-
ary layer (Figure 5). Its thickness is determined by the
agitation conditions and the viscosity of the Suid.
The equilibration time can be estimated for practi-
cal cases from the equation below:
Kfs(b!a)
te"t95%"3 [8]
Ds
where (b!a) is the coating thickness on the Rbre,
Ds is the diffusion coefRcient of the analyte in the
sample Suid, Kfs is the distribution constant of the
analyte between the Rbre and the sample and  is
a boundary layer thickness. This equation can be used
to predict equilibration times when the extraction
rate is controlled by the diffusion in the boundary
layer. The extraction time calculated using eqn [8]
must be longer than the corresponding time predicted
by eqn [7].
Conclusion
SPME is gaining acceptance principally because of its
simplicity, speed and low cost of operation. The de-
tection limits are comparable to a total extraction
technique since all extracted analytes are introduced
to the analytical instrument in SPME versus only
a fraction for a total extraction techniques. Selectivity
 II / EXTRACTION / Solid-Phase Microextraction 1423

Figure 6 Reconstructed GC-MS chromatogram indicating short chain fatty acids in a sewage sample. Peak assignment: 1, acetic; 2,
propionic; 3, isobutyric; 4, butyric; 5, pivalic; 6, isovaleric; 7, valeric; 8, hexanoic acids. The peaks correspond to pyrenylmethyl esters of
these acids.

of the technique is controlled by chemical properties
of the coating and it is determined by the appropriate
distribution constants. Selecting the appropriate Rbre
allows discrimination against interferences and there-
fore a separate clean-up step is not necessary. In
addition, the coating can contain derivatization re-
agent, which can speciRcally bind target analytes,
resulting in high speciRcity and sensitivity of the pro-
cess. Figure 6 shows a chromatogram obtained after
selective headspace SPME extraction of low molecu-
lar weight carboxylic acids from a sewage sample by
poly(acrylate)-coated Rbre containing 1-pyrenyl-
diazomethane which selectively reacts with the target
analytes. New coatings and reagents will allow ex-
pansion of SPME applications to new areas such as
inorganic analysis and analysis of biomolecules.

Figure 7 Separation of purgeables A, B and C on a Vocol column. Conditions: 03}303C min\1 703; 2.1 atm, dedicated injector,
capacitor voltage 24 V, MS detector, mass range 45}250. Peak assignment: 1, chloromethane; 2, vinyl chloride; 3, bromomethane; 4,
chloroethane; 5, trichlorofluoromethane; 6, 1,1-dichloroethene; 7, dichloromethane; 8, 1,2-dichloroethene; 9, 1,1-dichloroethane; 10,
trichloromethane; 11, 1,1,1-trichloroethene; 12, tetrachloromethane; 13, benzene; 14, 1,2-dichloroethane; 15, trichloroethene; 16,
1,2-dichloropropane; 17, bromodichloromethane; 18, 2-chloroethyl vinyl ether; 19, cis-1,3-dichloropropene; 20, toluene; 21, trans-1,3-
dichloropropene; 22, 1,1,2-trichloroethane; 23, tetrachloroethylene; 24, dibromochloromethane; 25, chlorobenzene; 26, ethylbenzene;
27, tribromomethane; 28, 1,1,2,2-tetrachloroethane.
1424 II / EXTRACTION / Solvent Based Separation
In addition to solvent-free sample extraction,
SPME is also a solvent-free sample introduction tech-
nique which facilitates design of a simple, low volume
injection system. The net result is rapid desorption
and good chromatographic separation, especially
when Sash-heated injectors are used. Figure 7 illus-
trates 2.5 min extraction and separation of 28 Envir-
onmental Protection Agency volatile priority pollu-
tants, which is over an order of magnitude faster than
the standard purge and trap technique. This approach
is particularly useful in combination with online
SPME extraction. As eqn [6] indicates, it is possible
to integrate sampling with a sample preparation step.
This not only results in elimination of analyte losses
to container walls and degradation during the trans-
port, but also saves time and transport costs. This is
particularly true when online SPME extraction is
combined with Reld portable GCs.
Another interesting feature of SPME which is cur-
rently being explored includes speciation of analytes in
complex matrices. The small amount of extracting
phase does not disturb the equilibrium existing in the
natural system and therefore allows quantitation of
individual species or the determination of distribution
Solvent Based Separation
R. G., P. M. Harper and Martin Hostrup,
CAPEC, Technical University of Denmark, Lyngby,
Denmark
Copyright ^ 2000 Academic Press
Introduction
Separation involves removal of one or more of the
constituent parts from a mixture. A solvent is that
constituent of a solution that is liquid in the pure
state, is usually present in the larger amount, and has
dissolved the other constituent (a solute) of the solu-
tion. The solute may be a solid, a liquid or a gas. The
solvent may be a single compound or a mixture of
compounds. Solvent-based separation techniques
become necessary when separation or removal of
a solute(s) from a mixture become difRcult or in-
feasible by conventional separation techniques such
as distillation. If the addition of a solvent causes
a totally miscible liquid to split into two liquid phases
and produce the necessary property difference, the
solvent-based separation technique is commonly
known as liquid}liquid extraction. If the addition of
a solvent causes the coexisting vapour and liquid
phases to have different properties, the solvent-based
constants in a multiphase system. In addition, the Rbre
can be made very speciRc, so separation using
chromatographic systems may not be necessary.
Therefore development of coupling between SPME
with other analytical instrumentation, such as mass
spectrometry and inductively coupled plasma}mass
spectrometry will facilitate high sensitivity and a large
throughput.
See also: II/Extraction: Solid-Phase Extraction; Solvent
Based Separation. III/Environmental Applications: Solid-
Phase Microextraction; Solid-Phase Microextraction:
Overview.
Further Reading
Kolb B and Ettre LS (1997) Static Headspace Gas Chromato-
graphy. Theory and Practice. New York: Wiley-VCH.
Pawliszyn J (1997) Solid Phase Microextraction. Theory
and Practice. New York, NY: Wiley-VCH.
Schwarzenbach R, Gschwend P and Imboden D (1993)
Environmental Organic Chemistry. New York, NY:
John Wiley.
Young AD (1989) Boundary Layers. Oxford: BSP Profes-
sional Books.
separation technique is called extractive distillation.
Figure 1A and 1B highlight the change of the mixture
properties as a result of the addition of a
solvent. In Figure 1A, the difference between the
properties of the liquid and vapour for the binary
azeotropic mixture of ethanol}water with and with-
out the addition of solvents is highlighted. It is clear
from Figure 1A that addition of a solvent removes the
barrier of the azeotropic condition. Figure 1B high-
lights through a ternary diagram that addition of the
solvent causes the totally miscible binary liquid mix-
ture (components 1 and 2) to split into two liquid
phases, a solvent-rich phase and a solute-rich (1 or 2)
phase.
Examples of industrial processes employing sol-
vent-based separation techniques are numerous. Al-
most all chemical, petrochemical, biochemical and
pharmaceutical processes employ one or more sol-
vent-based separation techniques. In chemical and
petrochemical processes, solvents are used mainly to
separate components from liquid and/or gaseous
mixtures, while in biochemical and pharmaceutical
processes, solvents are typically employed for dissolv-
ing or removing solids. Use of a solvent to extract
aromatic compounds from a petroleum by-product
 II / EXTRACTION / Solvent Based Separation
1425

Figure 1 (A) VLE phase diagrams for ethanol}water (pressure 1 atm) with and without solvents (plotted on a solvent-free basis). (B) Ternary LLE diagram for
acetone}water}ethyl acetate. (C) Process flowsheet for separation of phenol from wastewater.
1426 II / EXTRACTION / Solvent Based Separation

Table 1 Classification of important solvent-based separation techniques

Separation Solute property Number and identity Separation barrier Separation Solvent function
technique of phase phenomena

Liquid}liquid Totally miscible Two liquid phases Total miscibility Property differences Addition of solvent
extraction solutes in liquid phases causes phase split
Extractive Solutes from Vapour and liquid Azeotropes or Property differences Addition of solvent
distillation azeotrope phases relative in vapour and breaks the
or are close volatilities liquid phases azeotrope but
boiling does not cause
liquid phase split
Azeotropic Solutes from Vapour and two Azeotropes or Property differences Addition of solvent
distillation azeotrope or are liquid phases relative in vapour and two breaks the
close boiling volatilities liquid phases azeotrope but also
causes liquid phase
split
Absorption Absorbed gases Vapour and liquid Solubility of gases Differences in Solvent must be able
in liquid phases solubility to dissolve the
solute (gas)
Stripping Entrained liquids Vapour and liquid Solubility of liquids Differences in Solvent must be able
in gases phases solubility to dissolve the
solute (liquid)
Leaching Solid particles Solid(s) and liquid Solubility of solids Differences in Solvent must be able
phase solubility to dissolve the
solute (solid)

or removal of a chemical species (undesirable by- Solvent Selection:

product or raw material) from a wastewater stream
through solvent-based separation are typical exam-
ples of industrial application. Figure 1C illustrates
the removal of phenol from water through solvent
based liquid}liquid extraction. An important feature
in this and most other vapour}liquid and/or
liquid}liquid solvent-based separation techniques is
that the solvent is recovered and recycled back to the
solvent-based separation unit.
A logical criterion for classiRcation of solvent-
based separation techniques is the number and identi-
ties of the coexisting phases and the function of the
solvent. Table 1 gives a list of some of the well-
known solvent-based separation techniques, classi-
Red in terms of the number and identities of the
coexisting phases and function of the solvent. It can
be noted from Table 1 that the selected solvent is
directly related to the separation task and the separ-
ation technique and indirectly related to factors such
as cost of operation, the efRciency of separation and
the environmental impact. Therefore, solvent selec-
tion plays an important role in solvent-based separ-
ation. While solvents and solvent-based separation
techniques have been known for a very long time, use
of efRcient search techniques, such as computer-aided
molecular design (CAMD) and computer-aided
database search, are fairly new. This article highlights
the computer-aided methods and tools related to sol-
vent selection.
Problem Formulation
Problem formulation is an important Rrst step in
solvent selection as it is necessary Rrst to deRne the
functions of the solvent before attempting to Rnd
suitable candidates. Each problem, characterized in
terms of solvent and solute properties, needs to ad-
dress a set of issues related to separation task, perfor-
mance, environmental impact and problem-speciRc
(special) considerations. The solvent selection prob-
lem is formulated in terms of a set of properties
(target properties) and their values (target values).
A two-step procedure, consisting of a problem identi-
Rcation step (identiRes the solvent functions and
issues that need to be addressed) and a criteria for
evaluation step (selects target properties and their
target) is recommended.
Properties
The properties of the selected solvent deRne, to
a large extent, the type of the solvent-based separ-
ation technique. Consider the binary azeotropic mix-
ture of ethanol}water and the solvents benzene or
ethylene glycol. If benzene is used as the solvent, the
resulting solvent-based separation process is called
azeotropic distillation because ethanol}water}ben-
zene forms a heterogeneous azeotropic system, as
shown in Figure 2A. If, on the other hand, ethylene
glycol is used as a solvent, the solvent-based

Figure 2 (A) Ternary VLLE diagram for ethanol}water}ben-
zene (solvent). (B) Ternary homogeneous VLE diagram for
ethanol}water}ethylene glycol (solvent). Key: 2, heterogeneous
liquid boiling surface; *, vapour line; 䉭, critical point; 䊐, azeo-
tropes. All temperatures in 3C.
separation technique is called extractive distillation
because ethanol}water}ethylene glycol forms a ho-
mogeneous azeotropic system, as shown in Figure 2B.
Table 2 gives a list of different types of solvent
properties that may be considered in the selec-
tion/design of a solvent. These properties are classi-
Red in terms of pure component, mixture and envir-
onmental. Table 3 expands on the nature of the en-
vironmental properties. While the pure component
and environmental properties are usually available
for a large number of chemical species, the mixture
properties usually need to be estimated through suit-
able property prediction methods.
Problem Identi\cation
Solvents are well known for their different applica-
tions and, therefore, functions. They may be em-
ployed as cleaning agents, as paint additives, as separ-
II / EXTRACTION / Solvent Based Separation 1427
ating agents and many more. Each application of the
solvent is related to different sets of desirable func-
tions and undesirable effects (or functions). The main
question that needs to be asked here is what functions
will the selected solvent perform? The answer de-
pends, to a large extent, on the properties of the
solute and/or the mixture to be separated. Properties,
for pure compounds and mixtures, provide a frame-
work for classifying the different solvent functions
and their undesirable effects in a systematic and struc-
tured way.
Criteria for Evaluation
Since in solvent selection problems, one is looking for
alternatives that match approximately the desirable
solvent functions but not the undesirable solvent ef-
fects, numerical values of properties can be used to
evaluate candidate solvents. Based on the identiRed
separation task, the question of which properties (tar-
get properties) should be considered in deRning the
solvent functions and what should be the property
values (target values) is addressed in this step. The
exact target values for the target properties are ob-
tained by trial and error. However, if a known sol-
vent is being substituted, then the target values are
obtained from the solvent that needs to be sub-
stituted. In Table 2, two types of criteria for evalu-
ation are shown } simple and general. As simple, the
minimum number of properties that may deRne the
desired solvent properties for each solvent based sep-
aration is highlighted, while as general, a compre-
hensive list of properties is highlighted.
Example
Consider the process from Figure 1C } the efSuent
water stream from an industrial process contains 7%
w/w of phenol, which needs to be removed through
liquid}liquid extraction. The desired solvent, when
added to the phenol}water system, must cause
a phase split such that the solvent-rich phase will
contain signiRcantly more phenol than water while
the water-rich phase will contain very little phenol or
solvent. It should be possible to separate easily the
solvent from phenol. That is, the solvent must not
form azeotrope, it must have a reasonable difference
in boiling point and vapour pressure from phenol,
and it must have a density lower than that of water in
order to have free convection Sow in the extraction
column. If the solvent has a high environmental im-
pact, the loss of the solvent through the water-rich
phase will have to be reduced. If the solvent is unable
to remove enough phenol, more solvent may need to
be used. It should pose a low risk of explosion (the
Sash point temperature should be as high as possible).
1428 II / EXTRACTION / Solvent Based Separation

Table 2 Solvent selection problem formulation with properties

Property Solvent design

L}L Extraction Extractive Azeotropic Solid separation Gas absorption

distillation distillation

Simple General Simple General Simple General Simple General Simple General

Pure
Solubility parameter * * * *
Surface tension * *
Viscosity *
Boiling point * * * * * *
Melting point * * * * * * * * * *
Density *
Vapour pressure * * * * * *
Heat of fusion *

Mixture
Selectivity * * * * *
Solvent loss. *
Solvent power * * * * *
Distribution coefficient *
Phase split * * * *
Azeotropes * * *
Mixture viscosity *
Henry’s law constant *
Environmental * * * * * * * * * *

To ensure a minimal loss of the solvent to the water have a reasonable difference in boiling point and
stream, the solvent should have very low miscibility vapour pressure from phenol, and must have a den-
in water and a high octanol}water partition coefRc- sity lower than that of water in order to have free
ient. It should be possible to separate the solvent convection Sow in the extraction column). For the
easily from phenol (must not form azeotrope, must process in Figure 1C, the target properties and their
target values are given in Table 4.
Table 3 Specific environmental concerns
Solvent Selection: Methods and Tools
Environmental Health Safety Environmental
Methods
property concern concern concern
Solutions of solvent selection problems formulated
Implicit above require a multistage approach (see Figure 3).
Toxicity * *
Biological persistence *
Chemical stability * Table 4 Problem formulation for separation of phenol from
Reactivity * * wastewater

Explicit Target property Target value

Biodegradability *
Vapour pressure * * * Partition coefficient (log P ) '1.5
Henry’s law * Solvent loss (0.0015
constant in water Liquid density at 298 K (0.95
log P * * Normal boiling point (450 K
Water solubility * Vapour pressure at 360 K '0.03 bar
Flash point * Flash temperature '300 K
Biological oxygen * Selectivity '8
demand Capacity '2
Vapour density * * Separation factor '80
Evaporation rate * * Other properties Must not form azeotrope
LD50 * * with phenol
Ozone depletion * Acceptable environmental
potential properties

Figure 3 Multilevel approach to solvent selection.
After problem formulation, a list of feasible solvent
alternatives is determined and ordered according to
a speciRed criterion. The best feasible candidates are
then analysed in terms of separation task, perfor-
mance, environmental impact and special consider-
ations in order to determine the most appropriate
solvent(s). If none are found, it is necessary to go back
to the problem formulation stage and relax some
target property values or go back to the stage for
determination of alternatives and use another search
space. Thus, solvent selection is also a design problem
requiring a trial-and-error solution approach.
List of solvent alternatives Determination of the list
of solvent alternatives is based on the ‘generate and
test’ paradigm. That is, Rrst generate a list of solvent
candidates and then analyse (test) the candidates to
determine those that match the speciRed target prop-
erty values. The methods available can be classiRed
into three types: database search, CAMD and hybrid.
The database search approach involves a search in
one or more databases for compounds that match the
speciRed target property values. For this approach, an
efRcient search engine (or computer-aided technique)
is needed. For solvent selection problems involving
only pure component target properties, efRcient search
engines based on so-called pattern matching are avail-
able. DifRculties are encountered, however, when mix-
II / EXTRACTION / Solvent Based Separation 1429
ture properties are also included in the target properties
and when the databases do not contain all the target
properties for all the compounds. In such cases, an
efRcient and comprehensive search is almost impossible
and reliable property estimation methods are needed.
In the CAMD technique, molecular structures of
chemically feasible compounds are generated, the
speciRed target properties for the generated molecules
are estimated and those that match the speciRed tar-
get property values are included in the list of alter-
natives. The CAMD technique is therefore a more
efRcient search technique that is able to overcome the
difRculties related to solvent selection problem for-
mulations involving pure component as well as mix-
ture target properties and incomplete databases.
CAMD techniques, however, depend on the accuracy
of the property estimation methods used for predic-
tion of target properties for the generated molecules.
The search space is not limited by the molecules
present in a database but by the number of molecular
structures that can be generated and by the applica-
tion range of the property estimation methods used.
Combining the search based on databases with
CAMD, a multilevel hybrid approach is obtained. In
this approach, in level 1, a database search is carried
out only with respect to the pure component target
properties. This gives an idea of the types of mole-
cules that are likely to be selected as solvents. Level 2
1430 II / EXTRACTION / Solvent Based Separation
uses this information as initial estimate and employs
CAMD to solve the solvent selection problem for the
pure component and mixture properties that it can
estimate with acceptable accuracy. At the end of level
2, a larger list of alternatives than level 1 is obtained.
In level 3, those molecules that can be found in the
database are identiRed and their target properties are
veriRed, resulting in an updated list of alternatives.
This list is now used for checking the remaining target
properties (such as environmental properties and
special properties that are found in special databases).
Screening out the molecules that do not satisfy the
target properties based on these databases produces
a further reRnement of the list of alternatives. Finally,
in level 4, selected molecules from level 3 are investi-
gated in terms of atomic structure, bond length, bond
angle, energies, etc., through links to molecular mod-
elling programs.
Final selection Since the list of alternatives contains
more than one solvent, all of which match the speci-
Red target property values, it is necessary to deter-
mine the most appropriate solvent from this list.
Therefore, it is necessary to deRne a selection cri-
terion, for example, an objective function ( F ) that is
either minimized or maximized. This objective func-
tion may be an explicit function of the target proper-
ties (see eqn [1]) or an implicit function of the target
properties (see eqn [2]):
F"(SP, SS) [1]
F"f(DS(SP, SS), T(SP, SS), P(SP, SS)) [2]
In the above equations, SP is solvent power, SS is
selectivity, DS is a vector of speciRed products, T is
a vector of operating temperatures and P is a vector of
operating pressures. Since the target properties of
eqn [1] are known for the solvents in a generated and
tested list of alternatives, use of eqn [1] simply means
ordering the molecules in ascending order and select-
ing the optimal for further analysis (for example,
pilot plant study). In this case, the solvent with the
maximum value of F is regarded as the optimal sol-
vent. In eqn [2], the evaluation of F needs other
calculations (such as process simulation) in order to
determine the values of DS, T and P corresponding to
an optimal F. Two solution approaches are com-
monly applied } an enumeration approach and a si-
multaneous solution approach. In the enumeration
approach, the optimal value for F in eqn [2] is deter-
mined for each solvent through process simula-
tion/optimization, generating a set of values for F, DS,
T and P corresponding to each solvent in the gener-
ated list of alternatives. The optimal solvent then
corresponds to the minimum (or maximum) F in the
generated set. In the simultaneous solution approach,
the solvent identity is an integer variable and adding
it as an optimization variable in the process optimiza-
tion problem gives a mixed integer nonlinear pro-
gramming (MINLP) problem formulation, which de-
termines the optimal solvent and the optimal F simul-
taneously.
Tools
From the above section, it is clear that the tools that
are needed for solution of the solvent selection prob-
lem are databases, search engines, property estima-
tion methods, process simulators and numerical
methods (such as a MINLP-solver). It should be noted
that all the tools might not be necessary for all solvent
selection problems. Also, different sets of tools are
needed depending on the chosen method of solution.
In this article, only the use of search engines with the
hybrid approach, which includes database search,
CAMD and property prediction, is highlighted.
Table 5 gives a list of various tools that may be used
in solving solvent selection problems.
Search engine The hybrid generate-and-test ap-
proach (search engine}CAMD algorithm) has four
levels. Each level has its own generate-and-test algo-
rithms. Higher levels use additional molecular struc-
tural information compared with lower levels. Levels
1}2 are group contribution based (thereby employing
macroscopic representation of the molecule), while
levels 3}4 are based on atomic (microscopic) rep-
resentation of the molecule. Switch from level 1}2 to
3}4 needs a conversion of macroscopic representa-
tion to microscopic representation.
Level 1 This level generates sets of building blocks
(fragments) by combining Rrst-order functional
groups. These sets are capable of forming at least one
feasible molecular structure. Simultaneous calcu-
lation of related properties (that are dependent only
on Rrst-order groups) and screening of the generated
structures is performed to control the problem size
and execution time. The algorithm here is based on
a modiRed set of rules. Building blocks are classiRed
according to type. Feasibility rules are based on the
number of groups from a speciRc class a compound
may contain. Valency rules are used to determine the
number of groups with one, two, three and four
connections that are to be used in molecule structure
generation. The main steps of the level 1 algorithm
are illustrated in Figure 4.
Level 2 This level generates molecular structures by
combining elements of the individual fragment sets
 II / EXTRACTION / Solvent Based Separation 1431

Table 5 List of tools for solvent selection problems

Tool Type Contact information

ProCAMD CAMD CAPEC

Synapse CAMD Molecular Knowledge Systems, Inc.
EFDB Electronic database (environmental fate) Syracuse Research Corporation
ChemBankNRTECS Electronic database (health, safety, SilverPlatter Information Inc.
physical properties, environmental data)
Dortmund Data Bank Electronic database (mixture and DDBST GmbH
physical properties)
PHYSPROP Electronic database (physical properties) Syracuse Research Corporation
SOLVDB Electronic database (solvents) Syracuse Research Corporation
NIST WebBook Online database (physical properties) NIST
CS Chemfinder Online database (physical properties, Cambridge Soft Inc.
links to other sources)
SMSWIN Phase behaviour calculations AstraZeneca
Process Design Studio Phase behaviour calculations CAPEC
ChemDraw 5.0 Ultra Property prediction Cambridge Soft Inc.
ACD/Labs Physico-Chemical Laboratory Property prediction Advanced Chemistry Development inc.
Cranium Property prediction Molecular Knowledge Systems, Inc.
ProPred 2.5 Property prediction CAPEC

from level 1 to form molecular structures. First- and
second-order groups are considered. The main fea-
ture of this algorithm is that it is pseudorecursive, all
allowed combinations are considered, and efRciency
is maintained by continuous removal of duplicate
structures. Also, the combination rules satisfy condi-
tions of chemical feasibility. Use of second-order
groups allows the estimation method to differentiate
between some isomers.
Level 3 In this level, the selected candidates from
level 2 are given an atomic representation. Note that
the atomic representation also deRnes the connect-
ivity of the molecules. Therefore, property prediction
methods based on connectivity indices can be
employed to predict properties that could not be
predicted earlier (due to unavailable group contri-
butions) or to verify previously estimated values.
Level 4 In this level, generation and testing enters
an interactive mode. For any selected candidate
from level 3, it is possible to use molecular mo-
delling programs such as MOPAC or Chem3D from
Cambridge Soft Corp. A three-dimensional graph (or

Figure 4 Hybrid CAMD search engine.

1432 II / EXTRACTION / Solvent Based Separation
molecular model) is created by applying a set of
standard or default bond lengths and angles for the
various types of connections. As a result the true
molecular model of a compound, which can be fur-
ther analysed in terms of conformers, stability, prop-
erties, etc., is obtained.
Application Examples
Problems
Solutions of solvent selection problems with the
database search approach and the hybrid approach
are illustrated. Tools listed in Table 5 have been used
for solution of these problems, which involve solvent-
based vapour}liquid, liquid}liquid and solid}liquid
separations. For the removal of morphine, all the
solution steps for solvent selection with the hybrid
approach are highlighted. For the other examples, only
the problem formulation in terms of target properties
and the Rnal results are presented. Also, for solution
with the database search approach, only pure compon-
ent target properties have been considered.
Database Search Approach
For the seperation of phenol from water by
liquid}liquid extraction, solution of the problem (as
deRned in Table 4) Rnds, among others, butyl acetate
and toulene as solvents that match the pure compon-
ent target properties.
For the puriRcation of ethanol from a binary mixture
of ethanol}water, solvents for extractive or azeotropic
distillation are sought. The pure component target
properties are: normal boiling point (Tb)(473 K;
melting point (Tm)'270 K; Sash point (FT)'320 K;
solubility paramater (
) between 15 and 20 MPa1/2
for azeotropic distillation or 28 and 35 MPa1/2 for
extractive distillation. Note that
of ethanol is
around 26 MPa1/2 and that of water is around
47.8 MPa1/2. A value of
far from water and closer to
ethanol will be selective to ethanol and will likely
cause a phase split. Benzene, toluene and cyclohexane
satisfy the target property values for azeotropic distil-
lation. Ethylene glycol satisRes the requirements for
extractive distillation. Figure 2A and 2B also conRrm
this result.
For the separation (removal) of phenol present as
a solid, a solvent is needed to dissolve it. The solvent
target properties may be deRned with Tm'270 K,
Tb(473 K and 23.5(
(25.5 MPa1/2. A search of
the database gives furfuryl alcohol, aniline, N,N-
dimethylformamide and furfural. The solvent function
of aniline related to dissolving solid phenol is validated
through the computed solid}liquid phase diagram for
the phenol}aniline mixture (see Figure 5).
Figure 5 Computed SLE phase diagram for aniline}phenol.
Hybrid Approach
The solvent selection problem for the removal of
phenol from wastewater has been solved with the
ProCAMD (see Table 5). The summarized results
from ProCAMD are shown in Figure 6. Compared
with butyl acetate and toluene, this solvent has been
found to have a higher F (eqn [1]) and is environ-
mentally acceptable.
IdentiVcation of a solvent for morphine
Problem formulation In the production of mor-
phine a solvent is needed for dissolving the solid.
Known solvents for morphine include cyclohexane,
tetrachloromethane, toluene and benzene. It is de-
sired to Rnd alternative solvents capable of dissolving
morphine. Furthermore, in order not to contaminate
the product with toxic substances, in case of solvent
inclusions after crystallization, the compound should
be nonaromatic and have a low toxicity.
Constraint selection Solubility is a mixture prop-
erty. To be able to predict solubility to some degree of
accuracy it is necessary to have access to a method for
calculation of activity coefRcients. For complex com-
pounds (such as morphine) very few group-
contribution-based property estimation methods are
able to describe the molecular structure (see
Figure 7). Therefore, the search for alternative sol-
vents is carried out using pure component pro-
perties as criteria for evaluation. It is well known
that two compounds having similar solubility
parameters (
) are highly likely to be miscible. The
search for solvents for morphine can therefore be
expressed as a search for compounds being liquid at
ambient temperatures and having a solubility para-
meter as close as possible to that of morphine
(
"26.3 MPa1/2).

Figure 6 Results for phenol}wastewater separation (screenshot from ProCAMD).
Design speciTcations
E Compound type acyclic alkanes, ethers, esters, al-
dehydes, ketones, alchohols;
E Tb'350 K; Tm (273 K;
E 22(
(30 (exclusion of the lowest ranking can-
didates);
E performance measure 26.3!
 should be as low
as possible.
Generation of alternatives The ProCAMD package
(see Table 5) was used, generating 348 candidates
fulRlling the requirements. After performing a struc-
ture search to identify substances with known CAS
registry numbers two candidates, 1,5-pentanediol
and acetol, were selected based on having a solubility
parameter close to that of morphine.
Figure 7 Molecular structure of morphine.
II / EXTRACTION / Solvent Based Separation 1433
Analysis and veriTcation To verify the predicted
values of the properties used as design speciRcations
a search in available databases was carried out and
the experimental values compared to the predicted.
Furthermore the RTECS database (see Table 5) was
consulted in order to investigate the health and envi-
ronment properties of the selected compounds. The
result of the investigations and the predicted proper-
ties are shown in Table 6.
Candidate selection From the data listed in Table 6
it is clear to see that among the known and generated
solvents 1,5-pentanediol is the most promising candi-
date and should be selected for further testing in an
experimental setting.
Future Developments
As current and future separation problems become
more difRcult due to complex molecular structure of
solutes, changes in environmental regulations and
demands for material and energy conservation, the
solvent selection problem is also becoming more difR-
cult. It is no longer feasible to attempt to solve the
solvent selection problem with a single database.
Computer-aided techniques provide the necessary
framework to solve the current and future solvent
selection problems. The current hybrid approaches,
1434 II / EXTRACTION / Steam Distillation

Table 6 List of solvents for separation of morphine

Solvent CAS-NO Predicted Experimental Compound

class
Tb (K) Tm (K)
(MPa1/2) Tb (K) Tm (K)
(MPa1/2) (RTECS)

Benzene 71-43-2 353.24 278.68 18.73 C,D,M,T,S

Toluene 108-88-3 383.78 178.18 18.32 C,M,T,S
CCl4 56-23-5 349.79 250.33 17.55 C,D,M,T,S
Cyclohexane 110-82-7 353.87 279.69 16.76 M,S
1,5-Pentanediol 111-29-5 491 253 27.0 512.15 257.15 26.45 S
Acetol 116-09-6 418 226 27.2 418.65 256.15 25.75 M

D, drug; S, primary irritant; T, reproductive-effector; M, mutagen; C, tumorigen.

however, need to integrate aspects of molecular mod-
elling and computational chemistry before acceptable
solutions to problems involving complex solutes
and tight environmental regulations can be obtained.
Finally, it should be noted that having a good so-
lvent means easier design/operation of the solvent-
based separation technique. Therefore, it is important
to formulate correctly the solvent selection problem
and to Rnd reliable results in the form of optimal
solvents.
Further Reading
Barton AFM (1985) CRC Handbook of Solubility Para-
meters and Other Cohesion Parameters. Boca Raton,
FL: CRC Press.
Fredenslund AA, Gmehling J and Rasmussen P (1972)
Vapour}Liquid Equilibrium Using UNIFAC. Amster-
dam: Elsevier.
Harper P, Gani R, Kolar P and Ishikawa T (1999) Com-
puter aided molecular design with combined molecular
modelling and group contribution. Fluid Phase Equilib-
ria 337: 158}160.
Horvath AL (1992) Molecular Design. Amsterdam:
Elsevier
Lo TC, Baird MHI and Hancon C (1987) Handbook of
Solvent Extraction. New York: John Wiley.
Marcus Y (1998) The Properties of Solvents. New York:
John Wiley.
Mavriounioutis M (ed.) (1998) Special issue on design of
chemical compounds. Computers and Chemical Engin-
eering Journal 22: 713.
Seader JD and Henley E (1998) Separation Process Prin-
ciples. New York: John Wiley.

Steam Distillation
L. Ramos, Free University, Amsterdam,
The Netherlands
Copyright ^ 2000 Academic Press
Sample preparation is nowadays the limiting step in
the trace analysis of organic pollutants in environ-
mental and biological samples. Looking forward to
the laboratory of the future, versatile and universal
sample enrichment techniques are required, which
can produce fast and valid data, with low costs in
terms of solvent consumption and operator involve-
ment. A selectivity higher than that of the classical
exhaustive extraction methods or the simultaneous
elimination of the interference material could be an
additional requirement, as it would reduce the
amount of solvents and adsorbents used by reducing
or eliminating the subsequent clean-up step. Possible
additional beneRts deriving from a low manual ma-
nipulation of the samples would be a reduction in the
risk of contamination and loss of the analytes, as well
as an easier automation of the process.
Steam distillation extraction}solvent extraction
(SDE) has been presented as such a universal sample
enrichment technique. SDE allows the simultaneous
extraction, clean-up and concentration of the target
compounds in a closed system, with short analysis times
(1}8 h) and by using small amounts of organic solvents
(a few mL). This paper reviews this assumption for the
case of the analysis of less volatile organic pollutants in
environmental samples. The SDE advantages and short-
comings for such an analysis have been discussed.
Introduction
The monitoring of toxic organic chemicals in envir-
onmental and biological samples is a major concern
in many different Relds. However, the large variety of
compounds of interest, the differences existing in

their environmental levels and physico-chemical
properties, and the complexity of the matrices typi-
cally investigated make the development of universal
analytical methods for such an analysis a very difR-
cult goal. This is especially true for the most toxic
organic pollutants as their high toxicity makes their
reliable detection and accurate quantiRcation at the
trace level more relevant.
Most of the procedures described in the literature
for the analysis of less volatile organic pollutants are
time-consuming, laborious and speciRc for the deter-
mination of an analyte (or family of compounds) in
a selected matrix. Examples of selective extraction of
the target compounds, allowing their determination
without any additional clean-up, can be found in the
literature. However, most of these procedures involve
sophisticated and expensive analytical techniques,
such as supercritical Suid extraction or gel per-
meation chromatography. On the other hand, the
efRciency of these methods have been recognized to
be highly matrix-dependent. Because of these unresol-
ved shortcomings, classical exhaustive extraction
techniques, i.e. liquid}liquid extraction, LLE,
solid}liquid extraction or Soxhlet extraction, are still
widely used in ofRcial methods and routine applica-
tions. Due to the low selectivity of these methods,
subsequent elimination of the co-extracted material is
recommended. Such a clean-up step is mandatory for
reliable trace level determination of lipophilic and
bio-accumulative pollutants in biological and com-
plex environmental samples.
Steam distillation-solvent extraction (SDE) has
been used mainly for the extraction and concentra-
tion of fragrance and Savour compounds. However,
a variety of SDE methods reporting sample prepara-
tion for the analysis of pollutants in environmental
samples can be found in the literature. Most of these
methods allow the simultaneous extraction, clean up
and concentration of the target compounds. The in-
vestigated compounds range from volatile polar and
non-polar pollutants to non-ionic surfactants. This
article reviews the suitability and the limitations of
SDE for the determination of less volatile trace or-
ganic pollutants, such as pesticides, polychlori-
nated biphenyls (PCBs), polychlorinated dibenzo-p-
dioxins and furans (PCDD/Fs), or surfactants, in en-
vironmental and biological matrices. The most rel-
evant variables affecting the efRciency of the SDE of
these compounds are discussed and the results of
some selected applications reviewed.
General Considerations
According to the theoretical model developed by
Rijks et al., in 1983, the efRciency of the SDE process
II / EXTRACTION / Steam Distillation 1435
increased with the extraction time and with the liquid
and vapour Sows. The process also depends on
analyte-speciRc parameters related to the activity co-
efRcient (calculated from the water solubility of the
analyte at 1003C) and the gas-liquid distribution co-
efRcient of the compound in water at the process
temperature (i.e., 1003C for water steam). Not unex-
pectedly, the recoveries increased with the afRnity of
the target compounds for the extracting solvent. This
theoretical model is applicable only under ideal con-
ditions, which are achieved when all volumes and
Sow rates remain constant and there is ideal mixing
and equilibrium at every stage. In spite of these lim-
itations, the model reSects the effect of several experi-
mental factors on the SDE process. In fact, the differ-
ent modiRcations carried out on the SDE devices
originally described by Likens and Nickerson in 1964
and by Flath and Forrey in 1977 reveal the inSuence
of several parameters on the recoveries of the target
compounds. The modiRcations were mainly focused
on increasing the size of the vapour chamber and/or
the condensing surface to allow a more complete
mixing of the solvent and steam vapours, as well as
on the miniaturization of the system. As a conse-
quence of the changes in design (Figure 1), the efR-
ciency of the extraction was increased, the analysis
time reduced and the Reld of SDE expanded through
the analysis of residue levels of less volatile pollutants
in environmental samples.
Due to the characteristics of the technique, the feasi-
bility of SDE for the analysis of less volatile com-
pounds depends on their (i) potential for forming azeo-
tropes with water and (ii) relative solubility in water
and in the extraction solvent. However, the SDE of the
target compounds from complex samples can be ex-
pected to occur only after destruction or degradation
of the main matrix components, which usually entrap
the analytes (see below). Therefore, as stated by Nash
in 1984, the applicability of the SDE technique to the
analysis of this kind of environmental matrices would
be limited by the resistance of the investigated com-
pounds to the selected degradation procedure. Alterna-
tively, in some cases, co-distillation solvents have been
used to improve the SDE efRciency by reducing the
surface tension of the water and by increasing the
extraction power (polarity) of the organic solvent.
Finally, rather different results have been published
about the suitability of adding anti-foam agents in
applications involving fatty samples (see Table 3).
Application of SDE to the Analysis
of Aqueous Samples
Water was one of the Rrst environmental samples
selected to evaluate the feasibility of the SDE
1436 II / EXTRACTION / Steam Distillation
Figure 1 A typical SDE modern design.
technique for the determination of less volatile or-
ganic contaminants levels. Table 1 summarizes rel-
evant data concerning some reported methods for the
analysis of this matrix.
Quantitative recoveries of spiked organoch-
lorinated pesticides, OCPs (globally, in the range
90}106 ppb), and PCBs (globally, in the range
70}104 ppb) in aqueous samples have been reported
using the SDE technique. The reported methods al-
lowed the simultaneous extraction and concentration
of the analytes in 1}1.3 h in a relatively small amount
of a non-polar solvent (1}15 mL). Usually, no addi-
tional treatment of the sample or the organic extract
was required. The SDE technique was favourably
compared with other widely used extraction proced-
ures, such as LLE or solid-phase extraction (SPE) by
Ramos et al. in 1995, e.g. similar recoveries have been
published for the analysis of PCBs in water at the ppb
(ng mL\1) level by using SDE, LLE or SPE. However,
the higher repeatability of the SDE procedure (rela-
tive standard deviations, RSD, lower than 10%) and
the small amount of organic solvent involved, as well
as the short sample preparation times, makes SDE
a valuable alternative technique for such an extrac-
tion, especially when a large number of analyses have
to be carried out.
Nevertheless, some limitations of SDE have also
been reported for less volatile pollutants in water
samples. Nash et al. in 1984 studied different para-
meters affecting the efRciency of the steam distillation
process. They concluded that this technique is prob-
ably limited to compounds with a vapour pressure of
about 1 kPa at 1003C. Their results also showed that
the performance of SDE depends on the concentra-
tion investigated and that recoveries tend to increase
with the spiking level.
Similar tendencies have been observed by Ramos
et al. in 1995 when using the SDE technique for the
extraction of water spiked with the 2,3,7,8-sub-
stituted-CDD/Fs at different levels of concentration
(0.25}2 ng mL\, 0.025}0.2 ng mL\ and 0.0025}
0.02 ng mL\). The recoveries obtained for the lower
and higher boiling point congeners (tetra- and octa-
CDD/Fs, respectively) are consistently lower than
those found for the rest of the investigated congeners:
respectively 40}76% and 73}137% at the highest
level of concentration investigated, 39}60% and
62}92% at the intermediate, and 37}55% and
25}72% at the lowest spiking level. These results also
show that the SDE recoveries for a given compound
decrease with the concentration level when using n-
pentane as the extraction solvent. The simple substi-
tution of n-pentane for a solvent more selective for
the PCDD/Fs (dichloromethane) increases recoveries
from 25}73% to 71}139% for tetra- to hepta-
CDD/Fs at the 0.0025}0.02 ng mL\ level. However,
no additional improvement is obtained for the
octa-CDD/F recoveries (38}56%). In spite of the
low recoveries obtained for OCDD/F, the pro-
posed SDE procedure compares favourably with re-
sults previously published by using LLE or SPE in
terms of repeatability, analysis time and solvent con-
sumption.
Good recoveries (in the range 84}100%) have
been reported by Meissner et al. for the analysis of
surfactants such as fatty alcohol sulfates and alkyl
polyglycosides in water (Table 1). SDE of the
fatty alcohols yielded by hydrolysis and subsequent
LLE of the original compounds is an attractive
technique for the effective clean-up and concentra-
tion of these complex mixtures of compounds at the
trace level. On the other hand, the application of
SDE to the extraction of fatty alcohol ethoxylates
with more than three ethoxy units in the molecule
cannot be accomplished due to their high solubility in
water.
 II / EXTRACTION / Steam Distillation 1437

Table 1 SDE methods for the analysis of less volatile organic pollutants in aqueous samples

Compound Spiking level Solvent (mL) Extraction Cc. factor a Post-treatment b Recovery RSD Ref.
(ng mL\1) time (h) (water : solvent, (%) (%)
v/v)

OCP 0.004}0.016 Isooctane/toluene 1 167 : 1 NRc 90}104 ? Hemmerling

(15) et al. (1991)
Arochlor 0.016 Isooctane/toluene 1 167 : 1 NR 98}100 ?
1016, 1242, (15)
1248, 1254
OCP 0.4}4.0 n-pentane (1) 1.3 50 : 1 NR 97}106 ? Godefroot
et al. (1982)
Arochlor 1260 10 n-pentane (1) 1.3 50 : 1 NR 81}104d ?
Toxic PCBs 0.01}1.0 n-pentane (2) 1 50 : 1 Concentration 70}115 (10 Ramos et al.
(1995)
PCDD/Fs 0.025}2.0 n-pentane (2) 1 50 : 1 Concentration 49}139 (10
PCDD/Fs 0.0025}0.02 Dichloromethane 1 50 : 1 Change of 49}139 (10
(2) solvent
Fatty alcohol 500 nM Ethyl acetate (2) 3e 100 : 1 Derivatization 87}100 5.6}7.0 Meissner
sulfates et al. (1999)
Alkyl poly- 2
M Ethyl acetate (2) 4e 25 : 2 Derivatization 84 ?
glycosides

a
Concentration factor.
b
Post-SDE treatment required.
c
NR, not required.
d
Recoveries for some selected peaks.
e
The SDE was conducted after hydrolysis with 4 M H2SO4 plus liquid}liquid extraction with diethyl ether of the hydrolysate and
concentration.

Application of SDE to the Analysis of However, the efRciency of the proposed SDE methods
Non-Fatty Environmental Samples for the extraction of endogenous pollutants from
Table 2 summarizes relevant data related to some
reported SDE methods for the analysis of less volatile
organic pollutants in non-fatty environmental sam-
ples. Most of the reported SDE applications referred
to the analysis of OCPs and toxic aromatic com-
pounds, e.g. PCBs, polychlorinated naphthalenes
(PCNs), or polynuclear aromatic hydrocarbons
(PNAs), in soils and sediments. Contrary to what
might be expected from the high complexity of these
samples, most of the methods did not include any
further pre-treatment of the matrix but blending with
the selected volume of water. Only a few procedures
involving drastic treatments (e.g. blending of the
sample with H2SO4 and K2Cr2O7) during the SDE to
guarantee the destruction of the soil or sediment
components in which the target compounds could be
entrapped, can be found in the literature.
The efRciency (or need) of such a drastic treatment
is difRcult to evaluate from the data published. In
general, high (quantitative) recoveries have been re-
ported for freshly spiked analytes (globally in the
range 78}102% for PCBs and OCPs at the
20}90
g g\ level) with all the procedures (Table 2).
weathered samples has been scarcely evaluated. In
these studies, Seidel et al. (1993) and Cooke et al.
(1980) found concentrations very close or below the
limit of detection have usually been reported for the
endogenous contaminants, but the lack of comparison
of the SDE results with those obtained by standard or
more exhaustive methods, i.e. Soxhlet extraction, do
not allow any discussion about the methods used.
Dunnivant et al. in 1988 reported recoveries
ranging from 47 to 99% for SDE of certiRed sedi-
ments with PCBs at the 2.34}24.6
g g\. However,
as quoted above, this SDE method involved a diges-
tion of the sample under drastic conditions.
In a closely related study, Nash et al. compared the
efRciency of steam distillation with subsequent or-
ganic solvent extraction to that of Soxhlet extraction
for the analysis of pesticides in soil, plant tissues and
air (polyurethane foam Rlters). Both procedures pro-
vided similar recoveries for the spiked samples (in the
ranges 80}90%, 80}90% and 90}100%, respective-
ly). However, the SDE levels determined for
weathered soils blended with water were 40}50%
lower than the concentrations found by the Soxhlet
procedure. The study also showed that the efRciency
Table 2 SDE methods for the analysis of less volatile organic pollutants in non-fatty environmental samples (symbols as in Table 1)
1438

Matrix Compound Spiking level Pre-treatment Solvent Extraction Post-treatment Recovery RSD Ref.
(g g\1) time (h) (%) (%)

Sediment Arochlor 1016 50 Blended with 2.5 L water Isooct/tola(15) 1 Elimination S 78 ? Veith et al. (1977)
Weathered PCBs, PCNs, } Blended with 0.8}1.1 L water n-Hexane (10) 2.2 Concentration } 1.1}23 Cooke et al. (1979)
sediment PNAs
Sediment Chlorinated 100}1000 Blended with 0.25 L water n-Hexane (10) 3 Elimination S 76}91 0.5}21 Onuska et al. (1985)
benzenes
Sediment Chlorinated 10}100 Blended with 0.25 L water n-Hexane (10) 3 Elimination S 71}88 1.4}33
benzenes
Sediment Chlorinated 1}10 Blended with 0.25 L water n-Hexane (10) 3 Elimination S 66}89 1.4}17
benzenes
Certif. Sediment PCBs 2.34}24.6 200 mL H2SO4#K2Cr2O7 n-Hexane (15)b 8 Alumina 47}99 0.3}5.4 Dunnivant et al. (1988)
Sediment PCBs 33.6}90.0 200 mL H2SO4#K2Cr2O7 n-Hexane (15)b 8 Alumina 102 4.7}9.3
Soil HCB 20.0 100 g soil#20 mL water# Petrolbenzine (?) 1 NR 100 2.7 Seidel et al. (1993)
II / EXTRACTION / Steam Distillation

10 mL ethanol#ultrasonic,
1 min
Weathered soil Endogenous } 100 g soil#20 mL water# Petrolbenzine (?) 1 NR } }
OCP 10 mL ethanol#ultrasonic,
1 min
Particulate PCDD 0.045}8.56c 150 g (sample#water)# n-Hexane (10) 3 Basic alumina 85}116d ? Townsend et al. (1989)
(homologues) HCl#sonication
Fruits, OCCs, OCPs 0.01}1.0 5}10 g sample blended n-Hexane (5) 1.5 NR 66}108 ? Hemmerling et al. (1991)
vegetables with 0.25 L water
Fruits, PCBs 0.1 5}10 g sample blended n-Hexane (5) 1.5 NR 68}89 ?
vegetables with 0.25 L water
Sediments Chiral PCBs } 5 g sample blended with n-Pentane (2) 1 NR } } Glausch et al. (1996)
4 g Cu#50 mL water
Sewage sludge Nonylphenol 100e 1 g sample blended with Cyclohexane (1}2) 3 NR (HPLC) (30f ? Lee et al. (1997)
polyethoxylates 0.1 L water Alumina (GC)
Toothpaste Fatty alcohol } 30 g sample#50 mL Ethyl acetate (2) 3 Derivatization } } Meissner et al. (1999)
sulfates 4M H2SO4#LLE
(diethyl ether, 10 mL)#
concentrat.
a
Isooctane/toluene.
b
Replace hexane layer after 1, 2, 4 and 8 h intervals.
c
Range for different homologues.
d
Mass balance calculations for the whole SDE system by comparison with the initial amount.
e
Total concentration for all NPnEO (n"1}17).
f
By comparison with SFE, except for NP1EO.

of the SDE depends on the soil type and, in agreement
with that mentioned for aqueous samples, on the
volatility of the selected compounds. The less volatile
the compound, the lower the recovery: SDE recove-
ries for DDT were only 21}60% of those found by
Soxhlet extraction.
Onuska and Terry observed a similar trend when
comparing the SDE and the Soxhlet efRciencies for
the extraction of spiked chlorinated benzenes from
a sediment. The concentrations found using SDE
were 14}36% lower than those using the Soxhlet
method, except for pentachlorobenzene and 1,3-dich-
lorobenzene, which were not determined by the latter
procedure. The authors also reported a decrease in
the SDE recoveries of the target compounds as the
investigated concentration level decreased (Table 2).
The recoveries of the studied chlorinated benzenes
decreased from 76}91% to 66}89% as the spiking
level decreases from 100}1000
g g\ to 1}10
g g\.
In a recent study, Meissner et al. used SDE for the
determination of fatty alcohol sulfates in cosmetics
(toothpaste) by combining this technique with a hy-
drolysis treatment. However, the application of SDE
to the analysis of nonylphenol polyethoxylates in sew-
age sludge by Lee et al. failed when compared with the
more efRcient supercritical Suid extraction technique.
In general, the published SDE methods for the
analysis of non-fatty environmental samples involve
longer extraction times (1}8 h) than those reported
for aqueous samples (1}1.3 h). In addition, and con-
trary to that proposed by the theoretical model of
Rijks et al., the recoveries of less volatile compounds
from non-fatty complex samples have been found to
be independent of the vapour Sow rates. However, it
is important to note that in this study by Seidel
ethanol was added to the water Sask to improve the
OCP recoveries, and that the possible effect of a co-
distillation solvent was not included in the theoretical
model.
Application of the SDE to the Analysis
of Fatty Biological Samples and Food
Due to the high lipophility of some of the most toxic
pollutants, such as OCPs, PCBs and PCDD/Fs, the
classical procedures for the analysis of these pollu-
tants in fatty samples were based on an exhaustive
extraction of the lipids from the matrix. Subsequent
removal of the co-extracted lipids has been widely
recognized as the main problem with these kinds of
methods, especially when analysing samples with
high fat contents such as dairy products. Because of
the characteristics of the SDE technique, the disrup-
tion of the strongly bound pollutant-matrix in these
II / EXTRACTION / Steam Distillation 1439
samples can be accomplished before SDE by degrada-
tion of the matrix components entrapping the target
compounds. Treatment with 1}2 M sulfuric acid fol-
lowed by ultrasonication in a bath and heating of the
sample during the SDE process has been found to be
one of the most efRcient procedures for breaking
down the matrix structure allowing steam distillation
of the analytes. Furthermore, this acid treatment al-
lowed a simultaneous clean up of the Rnal extract as
the matrix components form more polar products,
which can then be easily separated from the non-
polar analytes. According to the published results,
most samples submitted to this kind of treatment did
not require any additional clean up. Filek et al. report
good recoveries for the SDE of OCPs from dairy
products when using this type of acid pre-treatment:
in the range 83}126% for powdered milk and human
milk spiked at the 20.0}51.3 ng g\ level, and in the
range of 73}111% for a certiRed dairy product (OCP
levels ranging from 1.5 to 6.6 ng g\). However, the
SDE method failed when it was used for the extrac-
tion of the endogenous PCBs from dairy products
with different fat contents. According to the reference
method, the PCBs detected by Ramos et al., in 1998
ranged from 2 to 0.01 ng g\ in the investigated ma-
trices. Nevertheless, most of the PCB congeners were
found to be non-detectable with the SDE procedure
and, when found at quantiRable levels, the reported
concentrations were less than 26% of those deter-
mined by the reference method.
Rather similar results were reported by Seidel and
Lindner in 1993 for the analysis of the OCPs in dairy
products and human milk as none of the investigated
compounds were found to be at a quantiRable level.
However, no additional comparison with a reference
method was included in this study, in which 10 g of
sample was blended with water and ethanol. In this
case, the alcohol, added as a co-distillation solvent,
would also be able to disrupt the fat globule thereby
allowing the SDE of the analytes. An important short-
coming of this kind of approach is the formation of
large oil drops during the extraction, which increase
the diffusion layer and hinder the SDE process. Filek
et al. proposed blending of the sample with surfac-
tants has been proposed as a possible solution for the
case of fatty matrices without natural emulsiRers.
When no pre-treatment of the fatty sample was
carried out, a co-distillation (total, according to Yoon
et al., or partial, according to Ramos et al.) of the
lipids with the less volatile compounds occurred.
Then, a post-treatment for isolation of the target
compounds from the co-extracted matrix compo-
nents was required. Following the implication of
these results, it is rather surprising that neither pre-
nor post-treatment of the sample was included
 1440

Table 3 SDE methods for the analysis of less volatile organic pollutants in fatty environmental samples (symbols as in Table 1)

Matrix Compound Spiking level Pre-treatment Solvent Extraction Post-treatment Recovery RSD Ref.
(ng g\1) time (h) (%) (%)

Fish tissue PCBs 1700 Blended with 2.5 L water Isooct/tol.(15) 7}14 NR 82}85 ? Veith et al. (1977)
Muscle, liver, PCBs, PCNs, 5000 Blended with 0.1 L water n-Heptane (10) 2 Concentration 67}100 1.0}20 Dunnivant et al. (1988)
kidney DDTa
Muscle, liver, PCBs, PCNs, 1000 Blended with 0.1 L water n-Heptane (10) 2 Concentration 65}85 ?
kidney DDTa
Dairy products, Endogenous } 10 g sample#water#ethanol Petrobenzine (?) 1 NR } ? Seidel et al. (1993)
II / EXTRACTION / Steam Distillation

Human milk OCPs

Pumpkin seed OCPs 100 10 g sample#water#ethanol# Petrobenzine (?) 1 NR 65}89 6}12
surfactant
Certif. dairy OCPs 1.5}6.6 5 g sample#80 mL 2 M H2SO4# Petrobenzine (20) 1.5 Concentration 73}111 ? Filek et al. (1995)
product ultrasonic, 1 min#1 mL ethanol#
surfactant
Powdered milk, OCPs 20.0}51.3 5 g sample#80 mL 2 M H2SO4# Petrobenzine (20) 1.5 Concentration 83}126 ?
Human milk ultrasonic, 1 min#1 mL ethanol#
surfactant
Powdered milk Endogenous } 15 g sample#60 mL 1 M H2SO4# Dichlorometane (2) 60}90 SiO2-HSO4 (26 ? Ramos et al. (1998)
PCBs ultrasonic, 1 min
Dairy products PCBs 0.5 15 g sample#60 mL 1 M H2SO4# Dichlorometane (2) 60 SiO2-HSO4 (10 ?
ultrasonic, 1 min
Herbal essential OCPs 500}10 000 Blended with 0.05 L water ? ? LLE (hexane : 83}105 2.4}10 Rajendran et al. (1991)
oils ethyl ether)#
H2SO4
OPPs 500}10 000 Blended with 0.05 L water ? ? LLE (hexane : 72}116 0.5}10
ethyl ether)#
H2SO4

a
DDT#DDE#TDE.

in some of the Rrst reported applications of SDE for
the analysis of toxic aromatic compounds in biolo-
gical matrices. The investigated samples included Rsh
tissues, muscle, liver and kidney and, although satis-
factory recoveries (67}100%) were reported for the
spiked PCBs, PCNs and DDT (i.e. DDT#
DDE#TDE), it is important to note that the spiking
level in these experiments ranged from 1000}
5000 ng g\ (Table 3). Even at such a high level of
concentration, the authors reported an evident de-
pendence of SDE recoveries on the analyte concentra-
tion. In fact, Cooke et al. found that the PCB, PCN
and DDT recoveries from animal tissues decreased
from 67}100% to 65}85% when the spiking level
decreased from 5000 to 1000 ng g\. According to
this trend, it can be concluded that the very low levels
of the endogenous pollutants in environmental sam-
ples together with the typical complexity of the
matrix would be the main reasons for the disappoint-
ing results reported for some SDE applications in-
volving non-spiked fatty samples.
See also: II/Extraction: Analytical Extractions; Solid-
Phase Extraction; Solid-Phase Microextractions; Supercriti-
cal Fluid Extraction. Distillation: Extractive Distillation.
Further Reading
Cooke M, Nickless G, Povey A and Roberts DJ (1979)
Polychlorinated biphenyls, polychlorinated naphtha-
lenes and polynuclear aromatic hydrocarbons in Severn
estuary (UK) sediments. Science of Total Environment
13: 17}20.
Cooke M, Roberts DJ and Tillett ME (1980) Poly-
chlorinated naphthalenes, polychlorinated biphenyls
and DDT residues in British birds of Prey. Science of
Total Environment 15: 237}246.
Dunnivant FM and Elzerman AW (1988) Determination of
polychlorinated biphenyls in sediments, using sonication
extraction and capillary column gas chromatography-
electron capture detection with internal standard calib-
ration. Journal of the Association of the OfTcial Analyti-
cal Chemistry 71: 551}556.
Filek G, Bergamini M and Lindner W (1995) Steam distilla-
tion-solvent extraction, a selective sample enrichment
technique for the gas chromatographic-electron capture
detection of organochlorine compounds in milk powder
and other milk products. Journal of Chromatography
A 712: 355}364.
Glausch A, Blanch GP and Schurig V (1996) Enantioselec-
tive analysis of chiral polychlorinated biphenyls in sedi-
ments samples by multidimensional gas chromatogra-
phy-electron capture detection after steam distillation-
solvent extraction and sulfur removal. Journal of
Chromatography 723: 399}404.
Godefroot M, Stechele M, Sandra P and Verzele M (1982)
A new method for the quantitative analysis of organo-
II / EXTRACTION / Steam Distillation 1441
chlorine pesticides and polychlorinated biphenyls.
Journal of High Resolution Chromatography Commun-
ications 5: 75}79.
Hemmerling C, Risto C, Augustyniak B and Jenner K
(1991) Untersuchungen zur aufbereitung von
Lebensmittel- und Umwelproben fur die ruckstandsbes-
timmung von pestiziden und PCB mittels kontinuier-
licher wasserdampfdestillation. Die Nahrung 35:
711}719.
Lee HB, Peart TE, Bennie DT and Maguire RJ (1997)
Determination of nonylphenol polyethoxylates and their
carboxylic acid metabolites in sewage treatment plant
sludge by supercritical carbon dioxide extraction. Jour-
nal of Chromatography A 785: 385}394.
Meissner C and Engelhardt H (1999) Trace analysis of
surfactants derived from fatty alcohols. II. Hydrolysis
and enrichment techniques. Chromatographia 49:
12}16.
Nash RG (1984) Extraction of pesticides from environ-
mental samples by steam distillation. Journal of the
Association of the OfTcial Analytical Chemistry 67:
199}203.
Onuska FI and Terry KA (1985) Determination of chlorin-
ated benzenes in bottom sediments samples by WCOT
column gas chromatography. Analytical Chemistry 57:
801}805.
Rajendran N and Venugopalan VK (1991) Bioconcen-
tration of endosulfan in different body tissues of estuar-
ine organisms under sublethal exposure. Bulletin of
Environmental Contamination and Toxicology 46:
151}158.
Ramos L, Blanch GP, HernaH ndez L and GonzaH lez MJ
(1995) Recoveries of organochlorine compounds (PCBs,
PCDDs and PCDFs) in water using steam distillation-
solvent extraction at normal pressure. Journal of
Chromatography A 690: 243}249.
Ramos L, Tabera J, HernaH ndez L and GonzaH lez MJ (1998)
Selective extraction of polychlorinated biphenyls from
dairy products using steam distillation solvent extrac-
tion at normal pressure. Analytica Chimica Acta 376:
313}323.
Rijks J, Curvers J, Noy T and Cramers C (1983) Possibili-
ties and limitations of steam distillation-extraction as
a preconcentration technique for trace analysis of or-
ganics by capillary gas chromatography. Journal of
Chromatography 279: 395}397.
Seidel V and Lindner W (1993) Universal sample enrich-
ment technique for organochlorine pesticides in environ-
mental and biological samples using a redesigned simul-
taneous steam distillation-solvent extraction apparatus.
Analytical Chemistry 65: 3677}3683.
Townsend DI, Lamparski LL and Nestrick TJ (1989) La-
boratory simulation and potential mechanisms explain-
ing PCDD congener group ratio behaviour on partic-
ulates from combustion sources. Chemosphere 16:
1753}1757.
Veith GD and Kiwus LH (1997) An exhaustive steam-
distillation and solvent-extraction unit for pesticides and
industrial chemicals. Bulletin of Environmental Con-
tamination and Toxicology 17: 631}636.
1442 II / EXTRACTION / Supercritical Fluid Extraction
Supercritical Fluid Extraction
A. A. Clifford, University of Leeds, Leeds, UK
Copyright ^ 2000 Academic Press
Supercritical Fluids
It is now 170 years since Baron Charles Cagniard de
la Tour discovered that, above a certain temperature,
single substances do not condense or evaporate, but
exist only as Suids. In the following decades the
‘critical point’ was characterized, with its parameters:
the critical temperature and pressure. In recent years
Suids have been widely exploited at conditions above,
but not too far removed from, their critical temper-
atures and pressures. The term ‘supercritical Suids’
has been coined to describe these media. Their value
lies in the fact that they can have properties inter-
mediate between those we associate with gases and
liquids, and also that the properties can be controlled
by pressure as well as temperature. Consequently,
supercritical Suids can often provide optimum condi-
tions for both experiments and processes. Equally
important, especially as regulations become tougher,
is that supercritical Suids offer environmental advant-
ages. This is mainly because carbon dioxide and
water are available as solvents. The disadvantages of
supercritical Suids are that high pressures and some-
times temperatures are involved, and, in the case of
water, there are corrosion problems. As the techno-
logy to overcome them is available, these disadvan-
tages become cost and convenience factors to weigh
against potential advantages. Consequently, super-
critical Suids are being exploited in specialized areas.
Amongst these is supercritical Suid extraction (SFE),
on both an industrial and analytical scale.
Substances used as supercritical Suids include hy-
drocarbons, such as propane and ethene, water and
ammonia, Suorinated hydrocarbons and even xenon.
However, one compound, carbon dioxide, has so far
been the most widely used in extraction, because of
its convenient critical temperature, cheapness, chem-
ical stability, non-Sammability, stability in radioac-
tive applications and non-toxicity. Large amounts of
carbon dioxide released accidentally could constitute
a working hazard, given its tendency to blanket the
ground, but hazard detectors are available. It is an
environmentally friendly substitute for other organic
solvents. The carbon dioxide that is used is obtained
in large quantities as a by-product of fermentation,
combustion and ammonia synthesis and would be
released into the atmosphere sooner rather than later,
if it were not used as a supercritical Suid. Its polar
character as a solvent is intermediate between a truly
non-polar solvent such as hexane and weakly polar
solvents. Because the molecule is non-polar it is often
classiRed as a non-polar solvent, but it has some
limited afRnity with polar solutes because of its large
molecular quadrupole. It has a particular afRnity for
Suorinated compounds and is useful for working with
Suorinated metal complexes and Suoropolymers.
To increase the afRnity of carbon dioxide to a var-
iety of solutes, substances are added as ‘modiRers’ or
‘entrainers’. The characteristics they impart include
increased or decreased polarity, aromaticity, chiral-
ity, and the ability to further complex metal-organic
compounds. For example, methanol is added to in-
crease polarity, aliphatic hydrocarbons to decrease it,
toluene to impart aromaticity, [R]-2-butanol to add
chirality, and tributyl phosphate to enhance the sol-
vation of metal complexes. They are often added in
5% or 10% amounts by volume, but sometimes much
more, say 50%. They can have signiRcant effects
when added in small quantities and in these cases it
may be the effect on surface processes rather than
solvent character which is important. For example,
the modiRer may be effective in extraction by adsor-
bing onto surface sites, preventing the readsorption of
a compound being extracted.
Because supercritical Suid have properties inter-
mediate between those of gases and liquids to an
extent controlled by pressure, optimum conditions
can be sought for extraction. The medium can be
adjusted for compounds to be sufRciently soluble to
be removed, while at the same time the viscosity and
diffusion coefRcients can be high enough to bring
about relatively rapid mass transport. Table 1 shows
typical values for the density and viscosity of a gas,
supercritical Suid and liquid, taking carbon dioxide
as an example. Density is more than half that of the
liquid, giving rise to reasonable solubility. Moreover,
by controlling the solvent density SFE can, to some
extent, be made selective. In contrast, however, the
viscosity of a supercritical Suid is much closer to that
of a gas than that of a liquid. Thus pressure drop
through a supercritical extraction cell is less than for
the equivalent liquid process. Diffusion coefRcients,
also shown in Table 1 for naphthalene in carbon
dioxide, are higher in a supercritical Suid than in
a liquid. They are approximately inversely related to
the Suid density. The advantage shown in the table is
seen not to be so great and the main diffusional
advantage lies in the fact that typical supercritical
solvents have smaller molecules than typical liquid
solvents. The diffusion coefRcient for naphthalene in
 II / EXTRACTION / Supercritical Fluid Extraction
Table 1 The density,
, and viscosity, , of carbon dioxide and
the diffusion coefficient for naphthalene in carbon dioxide, D,
under gas, supercritical and liquid conditions

/kg m\3 /Pa s D/m 2 s\1
Gas, 313 K, 1 bar 2 16 5.1;10\6
Supercritical, 313 K, 100 bar 632 17 1.4;10\8
Liquid, 300 K, 500 bar 1029 133 8.7;10\9
a typical liquid would be closer to 1;10\9 m2 s\1.
Thus diffusion coefRcients in supercritical Suid ex-
periments and processes are typically an order of
magnitude higher than in a liquid medium. This has
the advantage of faster transport in the narrow pas-
sages typical in an extraction.
Application to Extraction
Because of the properties of a supercritical Suid, as
described above, SFE can be more rapid than liquid
extraction. Furthermore, the solvent is removed more
easily, and fractionation of the extract by reducing
the pressure in stages is feasible. SFE was Rrst ex-
ploited on a process scale and this application con-
tinues to develop. On an industrial scale the Rrst and
most famous example is the ‘natural’ decaffeination
of green coffee beans by the Hag process initiated in
Bremen. Hops are also extracted by SFE on a large
scale. Apart from these large-scale processes, more
than 30 high-value oils, Savours and essences are
extracted commercially in batch processes.
SFE is also used in chemical analysis to replace
liquid extraction for sample preparation for a wide
range of systems. SFE is now being used for the Total
Diet Study programme of the US Food and Drug
Administration. Usually, SFE is more rapid, less
laborious and involves solvents which are less hazard-
ous. Efforts still have to be made to make it more
quantitative, but in fairness to SFE, extraction is often
incomplete using a liquid. SFE is sometimes used on-
line with an analytical method such as gas
chromatography; it is most successful for some poly-
mer and plant extractions. Table 2 summarizes the
principle analytical applications of SFE.
Laboratory-Scale SFE
SFE is carried out on a laboratory scale for both
sample preparation and for initial studies on possible
industrial processes. A range of commercial equip-
ment is available to carry out experimental studies
1443
conveniently. A simple system is shown schematically
in Figure 1. It can be assembled in-house and shows
the principles involved. The Suid, typically carbon
dioxide, is supplied from a cylinder with a dip tube to
a pump, which can be a pump designed for liquid
chromatography capable of delivering up to 5 mL per
minute at a pressure of 400 bar and displaying the
pressure and the Sow rate. The pump head must be
accessible so that it can be cooled by circulating an
ethylene glycol and water mixture from a cooler, so
that the Suid substance is pumped as a liquid. An
alternative method of ensuring this is to use a Suid
supply with an overhead pressure of around 100 bar
of helium. In this case, cooling the pump head is not
necessary, but the Suid will contain a small percent-
age of helium. The pumped Suid substance then
passes into a controlled heater, which can be an oven
for gas chromatography. It Rrst passes through
a length (typically 0.3 m) of stainless steel tubing into
an extraction cell, rated for 400 bar at 1003C, which
is Rtted with a frit at the exit end (often both ends) to
keep the sample matrix to be extracted in place. The
exit tube is then connected to a restrictor to maintain
the pressure in the system. This can be of stainless
steel or, alternatively, a quartz capillary, in which
case the connector will have a graphitized ferrule. The
efSuent then passes through a collecting solvent to
trap the extracted compounds. Because of the cooling
effect as the Suid expands to atmospheric pressure, it
is usually necessary to heat the restrictor and the
simplest way of doing this is with a domestic hair
dryer. Evaporation of the collecting solvent may oc-
cur and it will be necessary to add solvent to the vial
during the extraction. This simple device, although
often satisfactory, can suffer from blocking of the
Table 2 Examples of the use of SFE in analytical sample prep-
aration
Matrix Examples of analytes extracted
Soils, sludges, water Agrochemicals, polychlorobiphenyls,
polycyclic aromatic hydrocarbons,
fuel hydrocarbons, phenols,
surfactants, metals
Food and animal tissue Veterinary drugs, pesticides,
anabolic steroids, mycotoxins, fats
Human milk and serum Drugs
Polymers, food packaging Low oligomers, polymer additives
Herbs, cosmetic products Flavours, fragrances
Plant tissue Alkaloids, various natural products,
triglycerides
Fly ash, engine emissions Polycyclic aromatic hydrocarbons,
dioxins
Sedimentary rocks Biomarker hydrocarbons
Fermentation broths Biologically active compounds
1444 II / EXTRACTION / Supercritical Fluid Extraction
Figure 1 Schematic diagram of a simple system for carrying out SFE on a laboratory scale.
restrictor and loss of extracted compounds because of
inefRcient trapping. Furthermore, it does not allow
independent control of the Sow rate and pressure.
More sophisticated commercial methods of pressure
control and trapping are available.
In a representative experiment, a 1 mL cell is
loaded with 0.5 g of the material to be extracted,
(previously dried and ground to particles of 0.1 mm
diameter). Carbon dioxide is pumped at a rate of
0.5 mL min\, measured as liquid at the pump. The
temperature is 503C and the pressure of 400 bar is
maintained by a restrictor of 25
m internal diameter
and 12 cm length. The efSuent is trapped in 3 mL of
dichloromethane, ready for analysis by gas chro-
matography after an internal standard had been ad-
ded. The extraction is carried out for 30 min. How-
ever, conditions for SFE vary widely and the details
for a particular application can be found in the many
reports now in the literature.
If a modiRer is required, a second liquid pump must
be added to the system and the output liquid fed into
a mixing chamber just before the shut-off valve in
Figure 1. ModiRers are usually added in relatively
small amounts, say 5% or 10% by volume. It is
possible to purchase cylinders of carbon dioxide al-
ready containing small amounts of common modi-
Rers, such as methanol or acetone. If a modiRer is
used, the trapping solvent is conveniently the same as
the modiRer, as modiRer will precipitate in the collec-
tion vial. Trapping is usually more efRcient if a modi-
Rer is used.
The experiment described above is described as
dynamic extraction, as the Suid is continuously Sow-
ing through the cell. Static extraction can also be
carried out in a similar system if a second shut-off
valve is inserted after the extraction cell. During an
experiment, the cell is pressurized with Suid and the
cell isolated so that contact between the matrix and
Suid can occur for a period of about 30 minutes.
A short dynamic stage is then carried out to remove
the Suid, containing the dissolved extract, from the
cell. For a static extraction, a modiRer may be added
as liquid to the cell before closing it.
SFE can readily be coupled to gas chromatography
by passing the restrictor through a septum into the
injection port of a chromatograph. This procedure
can be much more sensitive, as all the extracted ma-
terial is transferred to the chromatograph, whereas in
an off-line experiment, only a small fraction of the
collection solvent will be injected. Thus the procedure
 II / EXTRACTION / Supercritical Fluid Extraction
is applicable for example to the analysis of pesticides
at low levels. To carry out this procedure, the Rrst
section of the chromatographic column is cooled and
the carrier gas turned off. SFE is then carried out with
the carbon dioxide, or other Suid substance, passing
out through the column and the extracted materials
depositing at the inlet of the column. SFE is then
stopped and the carrier gas passed through the col-
umn to Sush out the carbon dioxide. The column is
then raised to the analysis temperature and
chromatography carried out.
Pilot and Process-Scale SFE
The basic process of extraction on a process scale is
analogous to that on a laboratory scale and is shown
schematically in Figure 2. The Suid substance, such
as carbon dioxide, is pumped as a liquid and therefore
is initially cooled to, say, 53C, which must allow for
some heating during pumping, and kept in a cooled
reservoir. A system for adding a proportion of liquid
modiRer, not shown, may be incorporated. The Suid
is then heated to the extraction temperature and
pumped into an extraction cell, which is maintained
at this temperature. The matrix to be extracted is
packed into the extraction cell in a mesh basket to
prevent it being carried out of the cell during extrac-
tion. Following extraction, the pressure is reduced to
precipitate the extract through a control valve. The
Sow rate of Suid is controlled by the rate of pumping
and the pressure in the extraction cell is controlled by
the setting of the control valve for a particular pump-
ing rate. Control systems may be used to control the
extraction conditions. Reduction of pressure causes
cooling of the Suid and so heat input is required, as
shown. The precipitated material is collected at the
base of the collection vessel, which has temperature
control and also pressure control from the control
valve on the Suid exit. A series of collection vessels at
successively lower pressures may be employed to trap
all the extract and separate it into fractions to some
extent. A trapping liquid, such as a vegetable oil, may
be used on a process scale to give a particular prod-
uct. Trapping onto a surface, such as active charcoal,
may also be used, particularly for volatile products,
followed by thermal desorption. On a process scale
the Suid leaving the collection vessel is likely to be
cooled for recycling.
Mechanisms and Kinetics of SFE
Although extraction is essentially a complex process
in which many factors, including procedural para-
meters, are involved, in a basic theoretical approach
1445
the process of extraction can be considered to involve
the three factors shown in the SFE triangle below.
The solute must, Rrstly, be sufRciently soluble in
the supercritical Suid to be removed by solution in the
Suid Sow. If this is not the case, it will be revealed by
interpretation of the kinetic recovery curve, as shown
below. If solubility is insufRcient, the situation may
be improved by adding a modiRer to the Suid, as
described earlier.
Secondly, the solute must be transported sufR-
ciently rapidly, by diffusion or otherwise, from the
interior of the matrix in which it is contained. The
diffusion process may be normal diffusion of the
solute, or it may involve diffusion in the Suid thor-
ough pores in the matrix. The time-scale for diffusion
will depend on the diffusion coefRcient and the shape
and dimensions of the matrix or matrix particles. Of
these the shortest dimension is of great importance, as
the times depend on the square of its value. Values for
this quantity of 1 mm or preferably less are usually
necessary.
Thirdly, the solute must be released by the matrix.
This last process may involve desorption from
a matrix site, passage through a cell wall, or escape
from a cage formed by polymer chains. It can be slow
and in some cases it appears that part of the substance
being extracted is locked into the structure of the
matrix. An example is the SFE of additives and lower
oligomers from polymers, which can give much lower
results than obtained by dissolving the polymer in
a solvent, or using liquid extraction at higher temper-
atures, which swells the polymer to a greater extent.
Thus SFE will not always give the total amount of
a compound in a sample, only the amount extractable
under particular SFE conditions. It may be that the
latter is of interest, for example for determination of
the migration of additives from polymers into food-
stuffs, but if the total amounts are required, SFE may
not be applicable. Preliminary experiments, and com-
parisons with other methods, are necessary. It can be
strongly temperature-dependent and thus higher tem-
peratures may improve the situation. The addition of
modiRers may often reduce the matrix effect; in fact
modiRers are often more important in this respect
1446 II / EXTRACTION / Supercritical Fluid Extraction
Figure 2 Schematic diagram of SFE on a process scale.
than in enhancing solubility. The mechanism is
thought to involve interactions with surfaces. It
should be emphasized here that the matrix effect also
occurs with liquid solvent extraction. The fact that
solvent strength can be varied in a supercritical Suid
means that the matrix effect is more obvious in this
medium and can be studied in more detail. The ad-
vantage is that conditions can often be found in SFE
where the matrix effect is minimized.
A related problem is the presence of water. Water is
not very soluble in many Suids, such as carbon diox-
ide, and it can ‘mask’ the substances to be recovered.
The rate of extraction may sometimes be equal to the
rate of water removal. It may be necessary to dry the
material to be extracted in air or by admixture with
Figure 3 Examples of schematic recovery curves, where recovery is controlled by (A) diffusion; (B) diffusion and matrix effects; and
(C) by solubility.
a drying agent, such as diatomaceous earth or anhyd-
rous magnesium sulfate. Reduction of the water con-
tent of plant material from, say, 80% (as measured by
mass loss at 1003C) down to 10% may be desirable,
provided valuable volatiles are not lost in the process.
However, water may assist extraction by acting as
a modiRer, as is believed to be the case for coffee
decaffeination.
ModiRers or entrainers added to the Suid, as dis-
cussed earlier, may be beneRcial to any of the above
factors. They may improve the solubility of the
compounds to be extracted and this was originally
thought to be their most important role. However
they often improve diffusion by absorption into
a polymer and swelling it, for example. ModiRers
 II / EXTRACTION / Supercritical Fluid Extraction
Figure 4 Extraction of lycopene from tomato paste at 1003C and 400 bar at two flow rates plotted against the volume of CO2 passed.
may improve the matrix factor by adsorbing on
surface sites. ModiRers, such as methanol, can reduce
the water problem by improving its solubility in the
Suid.
Figure 3 shows examples of the types of curves of
recovery versus time that can be obtained in SFE.
Curve (A) is a typical curve obtained when the pro-
cess is controlled by diffusion. When matrix effects
are signiRcant, the results may have the form of (B).
Curve (C) is an example of recovery behaviour when
the extracted compound is not very soluble in the
extracted Suid. It is thus highly desirable, when devel-
oping a procedure for a particular application, to
carry out kinetic experiments to obtain curves of
recovery versus time. The curves can then be used to
investigate the reaction mechanism, as well as deter-
mine a suitable extraction time. These developments
are detailed in some of the books listed in the bibli-
ography. Kinetic experiments are done by replacing
the collection vial periodically. As extraction is faster
initially, the time intervals are smaller near the begin-
ning of an extraction. A representative series of times
for changing over the collection vial is 2, 5, 10, 20,
30, 40 and 60 min. The total amount extracted can
also be compared with liquid extraction.
An example of an experimental recovery curve is
now given in Figure 4 for the extraction of lycopene
from 0.5 g of tomato paste, dried by mixing with
diatomaceous earth, at 1003C and 400 bar. Flow
rates of 2 mL min\ and 3 mL min\ per minute
(measured as liquid at the pump) were used, and the
results plotted not against time, but against the vol-
ume of CO2 passed (i.e. the time multiplied by the
Sow rate). The fact that the two curves approxim-
1447
ately coincide, indicates that the extraction is princi-
pally by the partition of lycopene between CO2 and
the tomato paste matrix, which in turn is related to
the solubility of lycopene in CO2.
Conclusions
Supercritical Suid extraction can be a clean alterna-
tive to liquid solvent extraction both for analytical
sample preparation and for production scale, because
environmentally friendly solvents such as carbon di-
oxide can be used instead of organic solvents. Some
applications are found to be more successful than
others. It requires more expensive equipment and
a greater commitment to process development than
liquid extraction. Nevertheless, it is being applied in
speciRc areas, such as for polymer additives in ana-
lytical chemistry and as a method for obtaining valu-
able compounds from plants on a process scale.
See Colour Plate 44.
See also: II/Chromatography: Supercritical Fluid:
Large-Scale Supercritical Fluid Chromatography. III/En-
vironmental Applications: Supercritical Fluid Extraction.
One-line Sample Preparation: Supercritical Fluid Ex-
traction. Polymers: Supercritical Fluid Extraction.
Further Reading
Bright FV and McNally MEP (1992) Supercritical Fluid
Technology. ACS Symposium Series 488. Washington
DC: American Chemical Society.
Clifford, T (1998) Fundamentals of Supercritical Fluids.
Oxford: Oxford University Press.
1448 II / EXTRACTION / Ultrasound Extractions
King JW and List GR (1996) Supercritical Fluid Techno-
logy in Oil and Lipid Chemistry. Champaign, Illinois:
American Oil Chemists’ Society.
King MB and Bott TR (1993) Extraction of Natural Prod-
ucts Using Near-Critical Solvents. Glasgow: Blackie.
Lee ML and Markides KE (1990) Analytical Supercritical
Fluid Chromatography and Extraction. Provo, Utah:
Chromatographic Conferences, Inc.
Lynch TP in Adlard ER (1995) Chromatography in the
Petroleum Industry. Journal of Chromatography Li-
brary Vol. 56. Amsterdam: Elsevier.
Ultrasound Extractions
C. Bendicho and I. Lavilla,
Universidad de Vigo, Facultad de Ciencias
(QuO& mica), Vigo, Spain
Copyright ^ 2000 Academic Press
Introduction
Sound is transmitted through a medium by inducing
vibrational motion of the molecules forming part of
it. Human hearing threshold is reached when the
frequency of sound is higher than 16}18 kHz. Ultra-
sound comprises the region of frequencies between
18 kHz and 100 MHz, the upper limit not being
sharply deRned (Figure 1). This broad region can still
be divided into two different regions: power ultra-
sound between 20 and 100 kHz and diagnostic ultra-
sound between 1 and 10 MHz. The above classiRca-
tion relies on the capability of energy transmission
into the medium at the lower frequencies, which
induces the cavitation phenomenon.
Relevant applications of ultrasonic energy include
its use in animal communications (e.g. bat navigation
and dog whistles), medicine for fetal imaging, under-
water range Rnding (SONAR) and nondestructive
testing for metal Saws. Recently, ultrasound has also
been considered a potential source for enhancement
of chemical reactivity. A large variety of chemical and
industrial processes rely on high intensity ultrasonica-
Figure 1 Sound frequencies (Hz, cycles per second). , hu-
man hearing 16 Hz}16 kHz; , power 20 kHz}100 kHz (clearing
plastic welding sonochemistry); X, high frequency 1 MHz}
10 MHz (medical diagnosis, chemical analysis). (From Mason TJ
(1991).)
McHugh MA and Krukonis VJ (1994) Supercritical
Fluid Extraction, 2nd edn. Boston: Butterworth-
Heinemann.
Page SH, Sumpter SR and Lee ML (1992) Fluid phase
equilibria in supercritical Suid chromatography with
CO2-based mixed mobile phases: a review. Journal of
Microcolumn Separations 4: 91.
Westwood SA (1993) Supercritical Fluid Extraction and its
Use in Chromatographic Sample Preparation. Glasgow:
Blackie.
tion, e.g., cleaning, drilling, soldering, acceleration of
chemical reactions, emulsiRcation, sterilization, Sota-
tion, homogenization, dissolution, deaggregation of
powder, disruption of biological cells, extraction,
crystallization, oxidation, etc. A further advantage of
the above-mentioned ultrasound-assisted processes is
the relative simplicity of both method development
and instrumentation.
A brief description of ultrasound fundamentals as
well as a discussion of its applications for solid}liquid
extraction is given below.
Fundamental Features of Ultrasound
Vibrations Induced by Ultrasound
Sound waves are usually represented as a series of
vertical lines, with intensity being related to separ-
ation between them, or as a sine wave where intensity
is related to the amplitude (Figure 2).
Ultrasonic irradiation of a liquid medium gives rise
to an acoustic pressure (Pa) which is added to the
hydrostatic pressure (Ph) which exists in the medium.
The acoustic pressure depends on time according to
the following expression:
Pa"PA sin 2ft
where f is the frequency of the wave ('16 kHz), t is
the time and PA is the maximum pressure amplitude
of the wave. At the point where the lines are close to
each other, pressure is higher than normal (i.e. com-
pression region), whereas at the point where the lines
are furthest apart, pressure is lower than normal (i.e.
rarefaction region).
The intensity of the wave can be deRned as the
energy transmitted per second per cm2 of Suid and
can be related to PA as follows:
I"P2A(2
c)\1

Figure 2 Sound motion in a liquid medium.
where c is the velocity of sound in the medium and

is the density of the medium.
Attenuation of Sound in a Liquid Medium
The intensity of the ultrasonic wave decreases with
increasing penetration into the medium. Molecular
vibration induced by the sound wave results in loss of
intensity of the wave, which is transformed into heat.
Heating occurs in the sites of compression and cool-
ing at the sites of rarefaction. Since the compress-
ibility of liquids is small, there is little heating
caused by ultrasound as waves pass through the
medium. The heating effect is caused by the degrada-
tion of acoustic energy due to absorption, following
the equation:
I"Io exp(!2d)
where I is the intensity at distance d from the ultra-
sound source and  is the absorption coefRcient.
The Phenomenon of Cavitation
The pressure wave caused by ultrasound transmitted
in a liquid medium will, in turn, cause an oscillation
of the molecules around their mean position. When
a large negative pressure (Pc) is applied to the liquid,
where Pc (rarefaction pressure)"Pa!Ph, the dis-
tance between molecules can overcome a critical dis-
tance R, under which the liquid breaks down so that
cavitation bubbles form. The R distance for water is
II / EXTRACTION / Ultrasound Extractions 1449
around 10\8 cm and the pressure involved is
10.1;10 kPa, where Pc"2/R,  is the surface
tension. The cavitation process can be observed at
much lower negative pressure (e.g. 10.1;104 kPa), as
a result of the presence of gas nuclei as dissolved gas,
suspended gas bubbles, or gas bubbles caused by heat
Suctuations within the liquid. The cavitation thre-
shold decreases with degassed liquids or as conse-
quence of the increase in hydrostatic pressure.
Cavitation can be divided into two classes: transi-
ent and stable. Stable cavities oscillate around some
equilibrium size (R0) over several rarefaction}com-
pression cycles. In contrast, transient cavities usually
exist over one acoustic cycle, increasing their size
during the cycle and collapsing into smaller bubbles.
The time required for a bubble to collapse is usually
shorter than the period of the acoustic wave, and
therefore Pm (i.e. pressure in the liquid at the moment
of transient collapse, Pm"Ph#Pa) can be considered
as constant during collapse. This time can be ex-
pressed as:
t"0.915Rm(
/Pm)1/2
where Rm is the radius of the cavity at the moment of
collapse.
Temperatures and pressures reached inside a cavi-
tation bubble containing nitrogen in water at ambient
temperature and pressure before collapsing are nearly
4200 K and 975 bar, respectively. The high temper-
ature existing inside cavitation bubbles accounts for
radical formation, whereas the shock wave caused by
bubble implosion may be responsible for the in-
creased chemical reactivity.
In]uence of Different Parameters on
the Cavitation Process
The different processes occurring during cavitation
(i.e. nucleation, bubble growth and collapse) can be
affected by parameters such as liquid medium, inten-
sity and hydrostatic pressure, which are among the
most important.
Thus, the formation of cavitation bubbles de-
creases on increasing ultrasonic frequency. This is due
to insufRcient time for the rarefaction cycle to allow
the growth of the bubble so that disruption of the
liquid can be produced.
As expected, cavitation is decreased in viscous me-
dia as a result of the increased negative pressure in the
rarefaction region needed for disruption of the liquid.
The number of nuclei for cavitation depends on
temperature. An increase of temperature from !10
to #503C causes an increase in sonochemical effects
as a result of the increased cavitation. Nevertheless,
1450 II / EXTRACTION / Ultrasound Extractions

when temperature exceeds 503C the decrease in sur-

face tension and increase in vapour pressure within
the cavity will result in a lower Pmax and, conse-
quently, sonochemical effects will diminish.
The increase in gas content within the liquid leads
to a lower cavitation threshold and intensity of the
shock wave released on the collapse of the bubble. It
has been observed that the use of monoatomic gases
(He, Ar, Ne) provides more effective cavitation than
diatomic gases (N2, O2, air).
External pressure also inSuences the cavitation
process. Thus, when the external pressure is increased
(Ph), a lower cavitation threshold and intensity of
cavitational collapse are observed. When Ph!
Pa'0, it means that the negative phase of the sound
will no longer exist, hence eliminating cavitation.
Finally, another factor that can inSuence cavitation
is intensity, which enhances cavitation. Figure 3 Cavitation effects in a homogeneous liquid.

Effect of Power Ultrasound on Heterogeneous Medium

Chemical Systems In this case, there are two types of cavitational
collapse that can affect the surface of solids
Homogeneous Medium (Figure 4): (1) cavitational collapse on the surface
Mechanical and chemical effects caused by cavitation of the solid due to the presence of surface defects,
fall into three different processes (Figure 3). First, the
cavitation bubble contains solvent vapour which is
subject to high temperatures and pressures on col-
lapsing. This promotes the formation of reactive spe-
cies, e.g. radicals. For example, when water is used as
solvent the following reactions take place:
H2OPH z # z OH

H z #OH z PH2O

H z #H z PH2

OH z #OH z PH2O2

H z #O2PHO2 z

H z #O2PHO2 z

H z #HO2 z PH2O2

HO2 z #HO2 z PH2O2#O2

H2O#OH z PH2O2#H z
Second, surface-active reagents can accumulate at the
interface between the bubble and the bulk liquid.
Finally, in the surrounding of the bubble, an intense Figure 4 Cavitational effects at a solid}liquid interface: (A)
shock wave will be produced causing enormous shear cavitational collapse on the surface of a solid particle; (B) cavita-
forces. tional collapse close to a surface of a solid particle.

entrapped gases or impurities; (2) cavitational col-
lapse close to a surface causing a microstreaming of
solvent to impinge on the surface (i.e. cleaning action
of ultrasound). It has been observed that ultrasonic
irradiation can cause particle rupture (i.e. disruption)
which results in a decrease in particle size and an
increase in surface area for reaction. Alternatively,
cavitational collapse in a medium containing two
immiscible liquids can cause the formation of an
emulsion.
Instrumentation
Among the several types of sonicator systems current-
ly available, mostly bath and probe-type sonicators
are used. Both systems are based on an electromag-
netic transducer (i.e. device capable of converting
mechanical or electrical energy into high frequency
sound) as a source of ultrasound power, commonly
operating at a Rxed frequency of 20 kHz.
Ultrasonic sources used now rely on the piezoelec-
tric effect discovered by Curie (1880). Ultrasonic pro-
cessors implement transducers which are based on the
changes in dimension of some materials on applica-
tion of an electrical potential across opposite faces.
When the potential is modulated at high frequency,
the material converts the electrical energy into mech-
anical energy (sound). A sufRciently high alternating
potential will result in the generation of ultrasound.
The Rrst ultrasonic transducer was reported by Gal-
ton in 1883, who tried to establish the threshold
frequency of human hearing.
Bath Systems
In these systems the transducer is usually placed be-
low a stainless steel tank, the base of which is the
source of ultrasound (Figure 5). Some tanks are also
provided with a thermostatically controlled heater.
Typically, the ultrasound power levels delivered by
Figure 5 Schematic diagram of an ultrasonic bath.
II / EXTRACTION / Ultrasound Extractions 1451
most commercial ultrasonic baths (e.g., 1}5 W cm\2)
are sufRcient for cleaning, degassing of solvents and
extraction of adsorbed metals and organic pollutants
from environmental samples, but are less effective for
extraction of analytes bound to the matrix. The
power should be great enough to cause cavitation
within the extraction vessel placed inside the bath;
this is not always achieved with commercial ultra-
sonic baths.
An important factor inSuencing extraction efRcien-
cy is the position of the extraction vessel inside the
bath. For a bath with a single transducer on the base,
the extraction vessel must be located just above the
transducer, since power delivery will be a maximum
at this position. In order to obtain reproducible re-
sults, the bath must be either thermostatted or pre-
heated at the equilibrium temperature (i.e. maximum
temperature measured in the liquid under continuous
running conditions) since most cleaning baths warm
up slowly during operation. An important drawback
of most cleaning baths is the lack of power adjust-
ment control.
Probe Systems
Probe-type sonicators are able to deliver to the ex-
traction medium up to 100-fold greater power than
that of an ultrasonic bath, so that a better perfor-
mance is expected. One main feature for the
successful application of ultrasonic probes for many
chemical processes is that the ultrasonic energy is not
transferred through the liquid medium to the extrac-
tion vessel but introduced directly into the system
(Figure 6). The ultrasonic probe consists of the fol-
lowing components:
E A generator which is the source of alternating elec-
trical frequency (typically 20 kHz). The generator
allows tuning to be carried out for optimum perfor-
mance.
Figure 6 Schematic diagram of a probe-type sonicator.
1452 II / EXTRACTION / Ultrasound Extractions
E Ability of the ultrasonic processor to be used in
pulsed mode operation which allows the medium
to cool between pulses of sonication.
E The upper horn element } a piece of titanium to
which the detachable horn is attached, forming
both the emitter or booster.
E A detachable horn, usually made of a titanium
alloy, which allows the vibration of the Rxed horn
to be transmitted to a chemical system. Tip erosion
can occur as a result of cavitation. Depending on
the volume of sample to be irradiated a range of
detachable horns can be used.
Despite the improved performance displayed by
probe-type sonicators for solid}liquid extraction
compared with cleaning baths, a series of problems
can arise with the use of these sonicators. Volatile
components can be lost due to the ‘degassing’ effect
of the ultrasound power. Ultrasound irradiation by
means of probes is accompanied by a large amount of
heat generated during operation, hence some cooling
of the sonication vessel is required.
Ultrasound-Assisted Extraction
Extraction techniques are widely accepted as a
prerequisite for analytical determination of both or-
ganic and inorganic analytes in a large variety of
samples.
As a part of an analytical process, sample prepara-
tion is considered to be an essential step so that the
entire process can be simpliRed. In this case, the
ability of many analytical systems to handle liquid
samples has brought about the development of separ-
ation methods which fulRl a main objective, i.e. to
obtain quantitative analyte leaching from the solid
matrix using a suitable solvent, with little or no
matrix release, so that matrix effects can be kept to
a minimum during the measurement steps. For speci-
ation applications, a last condition of a solid}
liquid extraction method must be the maintenance of
the species integrity during treatment.
Table 1 shows the most relevant methods for
treatment of solid samples based on analyte extrac-
tion. An important requirement of most techniques
shown is that solvents at high temperature (i.e. at
boiling point) or pressure must be used. In contrast,
operation with ultrasonic processors can be per-
formed at ambient temperature and normal pressure,
and mild chemical conditions can be used in most
cases.
Sonication is usually recommended for pretreat-
ment of solid environmental samples for the extrac-
tion of nonvolatile and semivolatile organic com-
pounds from solid, such as soils, sludges and wastes.
When comparing the different methods available for
analyte extraction from solid samples, sonication is
considered as an effective method since unsophisti-
cated instrumentation is required and solid}liquid
separations can usually be performed in a short time
using diluted reagents and low temperatures. To date,
most of applications of ultrasonic extraction have
been carried out for organic compounds, but the
usefulness of ultrasound for element extraction is still
to be explored. Some examples of solid}liquid extrac-
tion of some elements with the use of ultrasound are
shown in Table 2. It should be pointed out that for
many applications reported in this table, operation
conditions were intended to obtain a homogeneous
slurry so that a representative aliquot could be sam-
pled; speciRc optimization of the variables in-
Suencing ultrasound-assisted extraction processes
was not performed. SigniRcant variables inSuencing
the solid}liquid extraction process with a probe-type
sonicator are sonication time, vibrational amplitude
of the probe, acid concentration, particle size
and solid concentration in the liquid. In general,
the presence of an acidic liquid is an important pre-
requisite for quantitative extraction to be achieved;
nitric acid at low concentration (e.g. 3}5% v/v) is
usually chosen for extraction of elements from solid
samples.
Quantitative extraction can be achieved for some
analytes such as As, Cu, Pb, Cd, etc., from plant and
animal tissues. Nevertheless, incomplete extraction
has been observed from samples containing a typical
inorganic matrix (e.g. sediment). It is believed
that this Rnding is related to the ability of ultrasound
to penetrate the solid material. A further variable that
inSuences the solid}liquid extraction is the analyte}
matrix interaction. Thus, strongly bound analytes
should be more difRcult to extract, thereby re-
quiring more stringent extraction conditions. A
relationship between extractability and binding char-
acteristics of elements in the sample is yet to be
established.
The extraction efRciency obtained with ultrasound
could be increased by addition of glass beads which
promote particle disruption by focusing the energy
released by cavitation, and by physical crushing. Par-
ticle disruption could also be enhanced by increasing
hydrostatic pressure and viscosity. The use of
a bubbling gas during sonication gives rise to an
enhanced formation of H2O2 and hydroxyl radicals
(OH ) ) thus aiding analyte extraction from oxidizable
materials. In general, the use of probe-type sonicators
at the appropriate vibrational amplitude and sonica-
tion time is required so that extraction efRciency can
be improved for strongly-bound elements.
 II / EXTRACTION / Ultrasound Extractions 1453

Table 1 Extraction methods from solid samples.

Sample pretreatment method Principles of the technique

Accelerated solvent Sample is placed in a sealed container and heated to a temperature higher than its boiling
point, causing pressure in the vessel to rise.

Automated Soxhlet A combination of hot solvent leaching and Soxhlet extraction; sample in thimble is first
immersed in boiling solvent and then the thimble is raised for Soxhlet extraction with solvent
refluxing.

Forced-flow leaching Sample is placed in a flow-through tube, and solvent is pumped or pushed through high-
pressure nitrogen gas, while the tube is heated near the boiling point of solvent.

Gas phase After equilibrium, analytes partition themselves between a gas phase and the solid phase at
a constant ratio; with static headspace extraction, volatiles are sampled above the solid; with
dynamic headspace extraction, volatiles are sampled by continuously purging the headspace
above a sample with inert gas, trapping them on a solid medium, and then thermally
desorbing them into a gas chromatograph.

Homogenization Sample is placed in a blender, solvent is added, and sample is homogenized to a finely
divided state; solvent is removed for further work-up.

Pervaporation Volatile substances present in a heated donor phase placed inside a pervaporation module
evaporate through a porous membrane and the vapour condenses on the surface of a cool
acceptor stream on the other side of the membrane.

Solid}liquid extraction Sample is shaken together with the appropriate solvent in a container and the liquid separated
by filtration

Sonication Finely divided sample in a container is immersed in ultrasonic bath with solvent and subjected
to ultrasonic irradiation; an ultrasonic probe or cell disrupter can also be used.

Soxhlet extraction Sample is placed in a disposable, porous container (thimble); constantly refluxing solvent
flows through the thimble and leaches out analytes that are collected continuously.

Supercritical fluid Sample is placed in flow-through container and a supercritical fluid (e.g. CO2) is passed
through sample; after depressurization, extracted analyte is collected in solvent or trapped on
adsorbent and desorbed by rinsing with solvent.

Thermal A form of dynamic headspace analysis, but the sample is heated (controlled) to much higher
temperatures (as high as 3503C).

Contents based on Majors RE (1996) LC-GC International, 638 and Luque de Castro MD and Papaefstathiou I (1998) Trends in
Analytical Chemistry 17: 41.

Future Prospects
The use of ultrasound as a sample preparation
method for solid}liquid extraction is widespread in
many laboratories and can be regarded as fast and
effective. Extractions based on sonication have been
employed for the isolation of weakly-bound organic
compounds from solid samples such as soils, animal
tissue, plants, etc., and are comparable to methods
involving more intensive treatments (e.g., Soxhlet,
accelerated solvent, etc.). However, ultrasound ap-
plied to solid}liquid extraction of inorganic analytes
has rarely been attempted, perhaps owing to the inef-
Rcient sonochemical effects caused by most ultrasonic
baths, which are more extended than probe-type
sonicators. Ultrasound irradiation from high-inten-
sity ultrasonic processors opens the door to new per-
spectives, mainly concerning those analytes that are
strongly-bound to the matrix. Thus, extraction of
elements from solid samples is feasible under opti-
mized sonication conditions, hence avoiding the more
intensive treatments commonly employed for matrix
decomposition (i.e. dry or wet ashing procedures).
New possibilities of ultrasound lie in its use as selec-
tive extraction techniques for metal speciation in con-
junction with the appropriate leaching reagents.
Thus, ultrasound-accelerated sequential extraction
schemes for metal partitioning in environmental solid
samples (e.g. soil, sediment, sewage sludge) or selec-
tive extraction of physicochemical forms of elements
constitute new sample preparation strategies which
deserve further research.
1454 II / EXTRACTION / Ultrasound Extractions

Table 2 Percentage of metal extracted into the liquid phase of slurries prepared in an acidic diluent and subsequently homogenized
by sonication

Sample Element and percentage of extraction Sonication system Reference

Bovine liver Cd (111%) Bath 1

Bovine liver Mn (100%), Fe (72%) Probe 2
Cabbage leave Cd (89%), Pb (1%) Probe 3
Cabbage root Cd (86%), Pb (1.5%) Probe 3
Carbon Cr (14%) Probe 4
Carbon Cu (69%), Cr (2%) Probe 5
Lemon leaves Cd (67%), Cu (88%), Mn (98%) Bath 1
Orchard leaves Cd (100%), Cu (88%), Pb (98%) Bath 1
Oyster Cd (99%), Pb (98%) Bath 6
Prawns Se (88%) Bath 6
Rice flour Cd (100%) Bath 1
Sediment Cr (30%) Probe 4
Sediment Cu (60%), Cr (10%) Probe 5
Silica gel As (60%), Cr (65%), Ni (77%) Probe 7
Spinach Cu (98%), Cr (74%) Probe 5
Spinach Mn (100%), Zn (74%), Fe (36%), Cu (100%) Probe 2
Talc As (59%), Cr (61%), Ni (74%) Probe 7
Tomato leaves Mn (70%), Fe (70%), Cr (51%) Probe 2
Tomato leaves Mn (92%) Bath 1
Wheat flour Mn (97%), Fe (88%) Probe 2

1, Minami H et al. (1996). Spectrochimica Acta, Part B, 51: 211. 2, Miller-lhli NJ (1990) Fresenius Journal of Analytical Chemistry 337:
271; 3, Dobrowolski R and Mierzwa J (1993) Fresenius Journal of Analytical Chemistry 346: 1058. 4, Miller-lhli NJ (1994) Journal of
Analytical Atomic Spectrometry 9: 1129. 5, Miller-lhli NJ (1993) Fresenius Journal of Analytical Chemistry 345: 482.6, Mierzwa J et al.
(1997) Analytical Science 13: 189. 7, Mierzwa J and Dhindsa HS (1988) Atomic Spectroscopy 19: 6.

See also: III / Ultrasound-Assisted Metal Extractions.
Further Reading
Ashley K (1988) Ultrasonic extraction of heavy metals from
environmental and industrial hygiene samples for their
subsequent determination. Trends in Analytical Chem-
istry 17: 366.
BarceloH D (1993) Environmental Analysis. Techniques, Ap-
plications and Quality Assurance. Amsterdam: Elsevier.
Dean JR (1998) Extraction Methods for Environmental
Analysis. Chichester: Wiley.
Lorimer JP and Mason TJ (1987) Sonochemistry. Part I } The
physical aspects. Chemical Society Reviews 16: 239.
Luque de Castro MD and da Silva MP (1997) Strategies for
solid sample treatment. Trends in Analytical Chemistry
16: 16.
Majors RE (1996) The changing role of extraction in
preparation of solid samples. LC-GC International,
638.
Mason TJ (1991) Practical Sonochemistry. User’s Guide to
Applications in Chemistry and Chemical Engineering.
Chichester: Ellis Horwood.
Mason TJ and Lorimer JP (1988) Sonochemistry: Theory,
Applications and Uses of Ultrasound in Chemistry. Chi-
chester: Ellis Horwood.
Suslick KS (1988) Ultrasound: Its Chemical, Physical and
Biological Effects. Weinheim: VCH.
 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents
FLOTATION
Bubble+Particle Adherence:
Synergistic Effect of
Reagents
B. J. Bradshaw and C. T. O’Connor, University of
Cape Town, Rondebosch, South Africa
Copyright ^ 2000 Academic Press
Introduction
It is well known in the practice of Sotation that
mixtures of various collectors often behave with
greater effectiveness than would be expected
from their individual known characteristics. This
phenomenon is a classical example of synergism in
Sotation, in which the combined effect exceeds
the sum of the linearly weighted partial effects.
Such phenomena are not only consciously applied by
adding mixtures of reagents, especially collectors,
but may also occur inadvertently since many indus-
trial reagents are synthesized from less than absolute-
ly pure chemicals, resulting in the presence of small
amounts of different product molecules which
are often capable of having a positive synergistic
effect on the Sotation behaviour. Such synergism
can have a signiRcant effect not only on the
recovery but also on the selectivity of speciRc min-
erals in differential Sotation. The manner in
which reagents interact in order to achieve a synergis-
tic effect is a complex function of their chemical
nature as well as their chemisorptive or physisorptive
properties. The former will determine whether the
chemical composition of the reagent changes when
another compound is present through, for example,
a dimerization reaction. The latter will determine
how competitive or synergistic adsorption will inSu-
ence the ultimate Sotation behaviour. The analysis of
synergism between reagents in Sotation is compli-
cated by the fact that the roles and interactions of the
different classes of reagents are difRcult to
isolate due to the complexity of the Sotation process,
viz. the frother is added to stabilize the froth zone but
can also interact with the collector and affect the
performance of the collection zone.
This review Rrstly discusses those properties of
pure collectors, frothers, depressants and activators
which are pertinent to their potential synergistic
1455
behaviour. The interactions between collectors,
frothers and each other are then reviewed. The em-
phasis here is on sulRde minerals but similar ef-
fects have been extensively reported in the case of
oxide Sotation. Finally, an hypothesis is proposed to
explain the synergism observed when mixtures of
thiol collectors are used in the Sotation of pyrite. This
represents a typical sulRde mineral Sotation system
and will serve to highlight how the various sub-
processes of Sotation may be inSuenced in a synergis-
tic manner, thus inSuencing the ultimate Sotation
performance.
Functional Roles of Pure Reagents
Collectors
The predominant functional role of collectors is
to induce hydrophobicity by adsorption onto the
desired mineral and they are therefore concentrated
at the mineral}water interface. Collectors are hetero-
polar molecules containing a nonpolar hydrocarbon
chain, which renders the particle hydrophobic, and
a polar group that interacts with the mineral surface.
Collector molecules can be divided into three
classes: nonionic, which are largely insoluble and
used in the Sotation of coal and graphite; cationic,
which are typically amine salts and used in the Sota-
tion of silicates and sulRdes at alkaline conditions;
and anionic, which are used to Soat basic minerals
such as metal oxides and sulRdes. Fatty acids are used
for the Sotation of nonsulRde minerals such as apa-
tite, calcite, feldspar and hematite. Sulfonates and
sulfates are used for apatite as their frothing proper-
ties limit their usefulness for other systems. Sulfhydryl
or thiol collectors are used for the Sotation of sulRde
minerals and, of these, xanthates, Rrst patented in
1925, are still the most widely used.
The mechanism of mineral}collector bonding de-
pends on the collector type and the nature and charge
of the mineral surface and can occur via physisorp-
tion or chemical bonding. There are several modes of
chemical interaction of the collector with the mineral
surface. In the case of physisorption, the collector
does not interact with the mineral surface. The at-
tachment is due primarily to van der Waals forces
and the Gibbs free energy of adsorption is relatively
low. In the case of chemisorption, when the collector
interacts with the mineral surface without movement
of the metal ions from their lattice sites, this produces
1456 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents
monolayer coverage. When the surface chemical reac-
tions are associated with movement of metal ions
from their lattice sites, multilayers may form. If a re-
action occurs in the bulk solution between dissolved
ions and the collector, a hydrophobic surface will
only be established if there is bulk precipitation on
the mineral surface.
SulRde minerals are semiconductors and react elec-
trochemically with thiol collectors according to the
mixed potential model. This involves the cathodic
reduction of oxygen and the anodic oxidation of
collectors. The electrochemical potential of the sys-
tem and the thermodynamics of the respective reac-
tions determine the nature of the surface products.
Depending on the nature of the surface products
formed, the collector may however be physisorbed,
such as in the case of the neutral dithiolate, or
chemisorbed, as in the case of the metal thiolate.
Naturally, when mixtures of collectors are used,
a combination of these mechanisms and products
may occur, possibly resulting in an enhanced Sotation
performance.
Frothers
Frothers are added to create a stable dispersion of
bubbles in the pulp which will subsequently create
a reasonably stable froth and which will allow selec-
tive drainage from the froth of entrained gangue and
improve the Sotation selectivity. The frother also
affects the Sotation kinetics. They are nonionic
heteropolar molecules and, unlike collectors, are not
associated with particular categories of minerals. The
frothing ability of a compound is associated with
hydroxyl (}OH), ester (}COOR) and carbonyl (}CO)
chemical groups, and commercial frothers can be
divided into three main categories: alcohols, al-
koxyparafRns, polyglycols and polyglycol ethers.
The polar end of the frother molecule forms hydrogen
bonds with the water and no mineral}frother bonds
are formed. The nonpolar end is hydrophobic so that
the frother concentrates at the air}water interface and
is thus described as being surface-active. This af-
fects the surface tension, which indicates the dif-
ference between the surface activity of frothers and
causes a stable froth to form. In general, increased
surface activity results in increased Soatability and
froth stability.
Depressants
The role of depressants, which are either inorganic
salts, such as sodium silicate, sodium sulRte or or-
ganic compounds such as polysaccharides, dextrin
and starch derivatives, guar gums, carboxymethylcel
lulose and alginates, is to reduce the collection of
unwanted gangue which consists of typically tal-
caceous or other oxide minerals. This is done by
either enhancing the hydrophilic nature of the gangue
surface, by preventing the formation of hydrophobic
species which might adsorb on the gangue surface or
by preventing the coating of unwanted slimes on the
mineral surface. Mechanisms of depression also in-
clude the formation of large aggregates and the com-
plexation of the collector in solution.
Activators
Activators are speciRcally added to enhance Sotation
performance, usually by modifying the surface of the
particle in some way so as to make it more amenable
to interaction with the collector. They may however
have unexpected effects, for example, by com-
plexing with other ions in solution and rendering
particles less Soatable. Copper sulfate, for example, is
a well-known activator. Under certain circumstances,
in sulRde Sotation, the copper may ion-exchange with
surface ions, creating a readily Soatable particle but
in different pulp conditions may complex as a
hydroxy species and depress the particles. Such ef-
fects may be considered synergistic but fall outside
the scope of this article. Another commonly used
activator is sodium sulRde or bisulRde which is used
as a sulRdizing reagent for tarnished or oxidized ores.
Synergistic Interactions Between
Reagents
There has been a considerable amount of work done
on the effects of mixing reagents in Sotation.
Table 1 summarizes much of this literature with re-
spect to type of reagents mixed, minerals tested,
measurements made and the beneRts observed.
Collector+Collector Interactions
The use of mixtures of collectors has long been recog-
nized in plant practice and has been shown to en-
hance Sotation performance. These beneRts have
been reported for a wide range of collector mixtures
(anionic, cationic and nonionic) and include lower
dosage requirements, improved selectivity and rates
and extents of recovery and an increase in the recov-
ery of coarse particles. In many cases an optimum
ratio of constituent collectors was shown to exist.
Dithiophosphates are a class of thiol collectors that
are so widely used in mixtures that they are known as
promoters.
Using measurements obtained from experimental
techniques shown in Table 1, a number of mechan-
isms have been proposed by various authors to
 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents 1457

Table 1 The effects of mixing reagents in Flotation

Interactions Reagents a Mineral Techniques Benefit of mixture Reference

(ratios tested) b systems c

Collector : collector Ethyl X : amyl X Arsenopyrite (P) Batch flotation Higher rates of Plaskin et al. (1954)
Thiol}thiol (2 : 1, 1 : 2 mass) recovery with mixtures.
Optimum mixtures:
Ethyl X : amyl X : diethyl Arsenopyrite (P) ethyl X : amyl X
DTP (1 : 1 mass) (1 : 2) for arsenopyrite
and (1 : 1) for
galena

Ethyl X : butyl X : diethyl Galena (P) Radiographic More even collector Plaskin and
DTP (1 : 1 mass) adsorption coverage on Zaitseva (1960)
techniques mineral surface with
mixture

n-propyl DTC : n-hexyl Pyrite ore with quartz Batch flotation Increased recoveries Bradshaw and
DTC : cyclohexyl DTC : gangue (South Africa) for all mixtures. O’Connor (1994)
di propyl DTC (1.27% Sulfur) Optimum ratio:
(10 : 90; 50 : 50; 90 : 10) n-propyl DTC :
cyclohexyl DTC
(90 : 10)

Butyl X : butyl DTP Galena (P) Adsorption Preferential DTP Wakamatsu and
(50 : 50) Bubble pick-up adsorption from Numata (1979)
mixture with no
increased
mass picked up by
bubble

Isopropyl DTC : iso Chalcopyrite ore Batch flotation Better results with Falvey (1969)
propyl X (1 : 2 mass) (Canada) (1.1% Cu) DTC : X mixture
than with pure DTC

Di-isobutyl DTP : iso Platinum group metal Batch flotation Recovery improved Mingione (1984)
butyl X (30 : 70; 50 : 50; (PGM) ore from 73.2% for
70 : 30 mass) pure X to 80% with
70 : 30 mixture

Di-isobutyl DTP : SMBT Auriferous pyrite ore Recovery improved

(50 : 50 mass) (0.38 g/t Au, 1% Sulfur) from 73.8% for
pure SMBT to 79.9%
with mixture

Di-isobutyl DTP : SMBT Sphalerite ore Recovery improved

(50 : 50; mass) (1.5% Zn) from 90% for pure
SMBT to 95% with
mixture

Isobutyl X : cyano Chalcopyrite/pyrite Batch flotation Chalcopyrite recovery Jiwu et al. (1984)
diethyl DTC with quartz gangue increased from
(12 : 44 mass) (China) 92.4% to 92.8% with
12 : 44 mixture

DTP : MTP (types Mixed copper sulfide Batch flotation Optimum recovery at Mitrofanov et al.
unspecified) ore 75 : 25 due to (1985)
(75 : 25; 50 : 50; 25 : 75) combination of
collector properties
1458 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents

Table 1 Continued

Interactions Reagents a Mineral Techniques Benefit of mixture Reference

(ratios tested) b systems c

Ethyl X : di-ethyl DTC Hazelwoodite (SP) Adsorption Optimum ratio: 33 : 66 Critchley and Riaz
(80 : 20; 66 : 33; 50 : 50; Surface tension for lower surface (1991)
33 : 66; 20 : 80) Microflotation tension, increased
microflotation
recovery and extent
of adsorption

SMBT : amyl X Gold and arsenopyrite Batch flotation Gold and arsenopyrite Van Lierde and
(70 : 25 mass) ore (France) recovery increased Lesoille (1991)
with use of mixture

Collector : collector Isopropyl X : dicresyl Mixed copper sulfide/ Batch flotation Enhanced rate and Adkins and
Thiol}thiol DTP (95 : 5) oxide ore (2.9% Cu) recovery with mixture. Pearse (1992)
Recovery from
80}83% Cu

n-butyl X : cyclohexyl Pyrite ore with quartz Batch flotation Recovery increased Bradshaw and
DTC (95 : 5; 90 : 10; gangue (South Africa) for all mixtures. O’Connor (1997)
85 : 15; 50 : 50) (0.83% S) Highest recovery for
50 : 50 mixture

n-butyl X : cyclohexyl Pyrite (P) Bubble loading Increased bubble

DTC (90 : 10) Thermochemical loading and heat of
adsorption with mixture

Thiol}anionic Ethyl X : sodium oleate Pyrite (polished section) Surface tension Largest contact angle Valdiviezo and
(10 : 90; 20 : 80; 40 : 60; Gold (polished section) Contact angle corresponded to Oliveira (1993)
60 : 40; 80 : 20) low surface tension
with 3 : 1 mixture

Thiol}anionic Ethyl X : amino acid Chalcocite (P), Microflotation Higher recoveries Hanson et al.
polymers glycine (1 : 1) galena (P), pyrite (P) obtained for all (1988)
sulfides with mixture

Butyl X : hydrolysed Mixed sulfide ore with Batch flotation 90 : 10 mixture Orel et al. (1986)
gold increased gold
polyacrylamide (90 : 10) recovery 3% above that
obtained with pure X

Thiol}cationic Ethyl X : ammonium Pyrite (P), quartz (P) Surface tension Lowest surface tension Buckenham and
bromide for 1 : 1 mixture Schulman
(05 : 1; 1 : 1; 2 : 1; 4 : 1) Microflotation Increased recovery (1963)
with all mixtures

Collector : frother Ethyl X : alkyl alcohols Chalcocite (P) Frothability Enhanced frothability Leja and
with X added to Schulman (1954)
alcohols

Ethyl X :
-terpinol Chalcocite (P) Microflotation Increased recovery Lekki and

with increasing Laskowski (1971)
dosage of frother
with xanthate.
Frothability Only froths in 3 phase
 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents 1459

Table 1 Continued

Interactions Reagents a Mineral Techniques Benefit of mixture Reference

(ratios tested) b systems c

Ethyl X :
-terpinol Chalcocite ore Batch flotation Increased recovery Lekki and Laskowski
(1 : 1) due to joint (1975)
frother}collector
interactions

Butyl X : 41G Galena Contact angle Contact angle on Harris (1982)

(polished section) mineral increased with
addition of frother to X
Xanthogen formate : Copper sulfide ore Batch flotation Collector dosage Crozier and Klimpel
MIBC (Chile) Plant practice reduced 40% to (1989)
achieve same recovery,
which reduced cost and
selectivity

Collector : frother Ethyl X : alkyl alcohols No mineral Surface tension Reduced film Manev and Pugh
Range of molar Film thickness thickness and surface (1993)
concentrations tension with increasing
addition of X

Frother : frother MIBC, pine oil, cresylic Various copper sulfide Plant practice Survey of 66 plants Crozier and Klimpel
acid, PPG ores showed 37% used (1989)
mixtures of frothers

a
Reagents tested as components of the mixture are separated by a colon. Where more than two reagents are in the list, all the reagents
listed have been tested at all the ratios specified in brackets.
b
Ratios are mole ratios unless otherwise specified as mass ratios (mass).
c
In cases where the origin or grade of the ore is not included in Table 1, this information was not available in the original reference.
X, Xanthate class of reagents; DTC, dithiocarbamate class of reagents; DTP, dithiophosphate class of reagents; MTP, monothiophos-
phate class of reagents; SMBT, sodium mecaptobenzonthiazole; PPG, polypropylene glycol; 41G, a proprietary frother containing
triethoxybutane manufactured by NCP; MIBC, methyl isobutyl carbinol; (P), pure mineral sample with no gangue component; (SP),
synthetically prepared pure mineral sample.

explain the fact that the mixtures give a Sotation
performance greater than that expected from the con-
tributions of each individual component. These pro-
posals are based on effects related to adsorption
of the collectors on the surface of the particle, interac-
tions between the reagents, either in the bulk or at the
surface, or changing froth characteristics.
When using mixtures of collectors it has often been
observed that there is a greater surface coverage of
the adsorbed collectors on the mineral than would
have been expected from their weighted averages.
This could either enhance the overall hydrophobicity
of the mineral surface or result in an adsorbed surface
layer of collector molecules more suitable for
frother}collector interactions. The increased mineral
hydrophobicity could result from the formation of
a more evenly distributed surface species. The change
in hydrophobicity can be measured by, for example,
changes in contact angle, bubble loading and ulti-
mately the recovery in batch Sotation tests. It has also
been proposed that, for certain systems, when a mix-
ture of collectors is exposed sequentially to a surface
which, by deRnition, must have a heterogeneous
distribution of energetically different sites, the
weaker collector will adsorb preferentially on the
strong sites and the strongly adsorbing collector, ad-
ded subsequently, will adsorb on the weaker sites. In
this way as many sites as possible are utilized for
adsorption, thus enhancing the hydrophobicity.
Single collector addition may only result in adsorp-
tion on strongly adsorbing sites, forming nonuniform
coverage and thus a less than optimal adsorption
capacity. It is possible that such an effect will not
be observed if the collectors are pre-mixed before
addition, thus emphasizing the fact that synergism
may depend on the sequence of addition as much as
on the presence of a mixture.
The grade of the concentrate is largely a function of
the depressant used, which affects the froth zone
characteristics. The presence of hydrophobic solids in
the froth phase will destabilize the froth, causing
bubble coalescence in the froth which results in im-
proved drainage and consequently increased selectiv-
ity and grades. The presence of hydrophilic or only
1460 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents

Figure 1 The effect of mixtures of collectors on batch flotation performance of a low-grade pyrite ore at pH 4. Values measured
(squares) were compared with those predicted from the linearly additive mole ratio contribution of potassium n -butyl xanthate (PNBX)
and dithiocarbamate class of reagents (DTC; triangles) for (A) % sulfur recovery obtained for 25% grade; (B) % sulfur grade obtained
for 80% recovery; and (C) water recovery obtained after 7 min (g).

slightly hydrophobic minerals can stabilize the froth physisorbed and chemisorbed surface products can
zone and thereby decrease the grade achieved. The also affect the froth structure and inSuence the
use of a combination of collectors resulting in both Rnal grade achieved. It is also often observed that
 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents
enhanced performance is achieved when a strong
collector with no frothing properties is used with
a weaker collector with frothing properties. The for-
mer increases coarse particle recovery and the latter
increases Rne particle recovery. This is however not
a true synergistic effect since the combined ef-
fect is the sum of the individual effects.
Collector+Frother Interactions
Before the collision of a mineral particle and an air
bubble, adsorbed layers of reagents are present at
both interfaces. At the time of collision, there are
interactions between these layers which are af-
fected by the nature and charge of the respective
molecules. Any associated molecules are anchored to
the mineral group by the polar groups of the collec-
tor. The strength of this Rlm determines the tenacity
of attachment of the mineral to the bubble and the
ultimate success of the Sotation process. When the
molecular associations between frother and collector
are suitably balanced the appropriate mechanical
properties of the Rlm at the interface are created,
resulting in good recoveries and grades. If the collector
or frother dosages are too high, the molecules would
be too densely packed and penetration and successful
attachment would not take place. This supports the
well-known phenomenon that too high a dosage of
reagents can result in reduced recoveries. In this case
synergistic interactions between the frother and collec-
tor that improved Sotation performance at the lower
dosages are no longer possible at the higher dosage.
Frother molecules can accumulate at the mineral
surface, without enhancing its hydrophobicity and, at
Figure 2 The mass loading per bubble for bubbles of average diameter of 1.2 mm of pyrite with equimolar amounts of potassium
n -butyl xanthate (PNBX), dithiocarbamate class of reagents (DTC) and the 90 : 10 mixture of collectors added at pH 4.
1461
the time of collision with a bubble, re-orientate quick-
ly, facilitating mineral}bubble attachment. this pro-
duces a stable three-phase froth and strong tenacity of
mineral}bubble attachment. An alternative explana-
tion is that at the mineral}water interface the alkyl
chains of frother and collector molecules are held
together by van der Waals forces. Frothers are able to
hydrogen-bond with the oxygen atom in the collector
molecule. These associations are only formed when
a mineral is present. The frother’s ability to interact
with the collector is thus more signiRcant than its
surface activity, which is required to produce a stable
froth zone. This also explains why detergents are not
suitable frothers. It has moreover often been shown
that the collector can affect frothing properties
and that the frother can affect mineral hydro-
phobicity.
The surface activity and thus frothability of the
frother is very sensitive to the presence of small
amounts of other substances, such as impurities or
collector molecules. The chemical nature of certain
combinations of frothers and collectors may result in
interactions occurring at the point of collision of the
pure components. The properties of frothers can
sometimes be additive, with the mixing of stronger
and weaker frothers to form medium-strength
frothers.
Synergistic Interactions +
A Case Study
Synergistic enhancement of Sotation performance has
been observed in batch Sotation tests with a low
1462 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents
Figure 3 The difference in heat flux measured when equimolar amounts of potassium n -butyl xanthate (PNBX), dithiocarbamate
class of reagents (DTC) and 90 : 10 mixture of collectors are added to pyrite at pH 4.
grade pyrite ore using thiol collectors at pH 4.
The collectors tested were potassium n-butyl
xanthate (PNBX) and an alkyl dithiocarbamate
collector. Performance was analysed using grade-
recovery data as well as water and mass recoveries
and the rate of sulfur recovery. The froth surface was
analysed using digital image analysis. In all experi-
ments the total molar concentration of collector was
constant.
Figure 1 shows the batch Sotation results as repre-
sented by sulfur grade at 80% recovery, the sulfur
recovery at 25% grade and the water recovery, all as
a function of mole ratio of components. It is clear that
the grades and recoveries are greater than would be
expected from a merely linearly additive effect
and are synergistically enhanced. Obviously pure col-
lectors may not show linearity with respect to dosages
but in the present case the dosages were in the range
where these differences were minimal. The
change in water recovery, however, was linearly
proportional to the molar contribution of the compo-
nents and clearly the synergistic effect was only
inSuencing the behaviour of the solid particles.
Digital image analysis of the froth showed that, when
the mixture of collectors was used, the froth was
more mobile and the froth surface bubble size was
larger. This may be due to the frother}collector inter-
actions, decreasing froth stability, increasing drain-
age of entrained material and increasing the grades
obtained.
In order to elucidate the mechanisms of synergism,
the extent of bubble loading and the heats of adsorp-
tion were measured for the respective collectors and
collector mixtures using pure pyrite at pH 4,
(Figure 1). Figure 2 shows that, for bubbles of aver-
age diameter of 1.2 mm, increased bubble loading
resulted from the use of a mixture of collectors and
Figure 3 shows that when a mixture of collectors was
used there was a stronger adsorption than in the case
of the pure xanthate, where multilayer adsorption of
dixanthogen is indicated, and in the case of dithiocar-
bamate where pseudo-monolayer adsorption is in-
dicated. In this example, the synergistic effect
observed is attributed to increased mineral hydropho-
bicity, which is thought to be due to the weakly
adsorbing dixanthogen adsorbing in multilayers
around the strongly adsorbing dithiocarbamate,
which acts as a sort of anchor on the surface of the
mineral particle. The ultimate result is an increase in
bubble loading, an improvement in froth character-
istics and a greater grade and recovery.
Further Reading
Bradshaw DJ and O’Connor CT (1997) The synergism of
mixtures of thiol collectors in the Sotation of low grade
pyrite ores. In: Hoberg H and von Blottnitz H (eds)
Proceedings of the XX International Minerals Process-
ing Congress, Aachen, vol. 3, pp. 343}354. Germany:
GMBH Publishers.
Fuerstenau DW (1995) Where are we in Sotation chemistry
after 70 years of research? In: Proceedings of the XIX
International Minerals Processing Congress, San Fran-
cisco, SME, vol. 3, pp. 3}18.
Harris PJ (1982) Frothing phenomena and froths.
In: King RP (ed.) Principles of Flotation, pp. 237}250.
Johannesburg: South African Institute of Min.
Metall.
 II / FLOTATION / Bubble+Particle Adherence: Synergistic Effect of Reagents
Klimpel RR and Hansen RD (1988) Frothers. In:
Somasundaran P and Moudgil BM (eds) Reagents in
Mineral Technology, pp. 387}409. New York: Marcel
Dekker.
Laskowski JS (1993) Frothers and Sotation froth.
Mineral Processing and Extraction Metall. Review 12:
61}89.
Leja J (1989) Interactions of surfactants. Mineral Process-
ing and Extraction Metall. Review 5: 1}22.
Leja J and Schulman JH (1954) Flotation theory: molecular
interactions between frothers and collectors at
solid}liquid interfaces. Transactions of the A.I.M.E.
199: 221}228.
Lekki J and Laskowski JS (1971) On the dynamic ef-
fect of frother}collector joint action in Sotation. Trans-
actions of Institute of Min. Metall. 80: C174}C180.
Mingione PA (1984) Use of dialkyl and diaryl dithiophos-
phate promoters as mineral Sotation agents. In: Jones
MJ and Oblatt R (eds) Reagents in the Minerals Indus-
try, pp. 19}24. London: Inst. Min. Metall.
List of References from Table 1
Adkins SJ and Pearse MJ (1992) The inSuence of
collector chemistry on the kinetics and selectivity in
base metal sulRde Sotation. Mineral Engineering 5:
295}310.
Bradshaw DJ and O’Connor CT (1994) The Sotation of
pyrite using mixtures of dithiocarbamates and other
collectors. Mineral Engineering 7: 681}690.
Bradshaw DJ and O’Connor CT (1997) The synergism of
mixtures of thiol collectors in the Sotation of low grade
pyrite ores. In: Hoberg H and von Blottnitz H (eds)
Proceedings of the XX International Minerals Process-
ing Congress, Aachen, vol. 3, pp. 343}354. Germany:
GMDB Publishers.
Buckenham MH and Schulman JH (1963) Molecular asso-
ciation in Sotation. Transactions of the Institute of Min-
eral and Metallurgy 7: C1}C6.
Critchley JK and Riaz M (1991) Study of synergism be-
tween xanthate and dithiocarbamate collectors in Sota-
tion of heazlewoodite. Transactions of the Institute of
Mineral and Metallurgy 100: C55}C57.
Crozier RD and Klimpel R (1989) Frothers: plant practice.
Mineral Processing and Extractive Metallurgy Review 5:
257}279.
Falvey JJ (1969) Dialkyl Dithiocarbamate as Froth Flota-
tion Collectors. US patent no. 3 464 551.
Hanson JS, Barbaro M, Fuerstenau DW, Marabini A and
Barbucci R (1988) Interaction of glycine and a glycine-
based polymer with xanthate in relation to the Sotation
of sulRde minerals. International Journal of Mineral
Processing 23: 123}135.
Harris PJ (1982) Frothing phenomena and froths. In:
King RP (ed.) Principles of Flotation, pp. 237}250.
Johannesburg: South African Institute of Mining and
Metallurgy.
1463
Jiwu M, Longling Y and Kuoxiong S (1984) Novel frother
collectors for Sotation of sulRde minerals}CEED. In:
Jones MJ and Oblatt R (eds) Reagents in the Minerals
Industry, pp. 287}290. London: Institute of Mining and
Metallurgy.
Leja J and Schulman JH (1954) Flotation theory: molecular
interactions between frothers and collectors at
solid}liquid interfaces. Transactions of the American
Institute of Mining, Metallurgical and Petroleum Engin-
eers 199: 221}228.
Leja J (1989) Interactions of surfactants. Mineral Process-
ing and Extraction Metallurgy Review 5: 1}22.
Lekki J and Laskowski JS (1971) On the dynamic ef-
fect of frother}collector joint action in Sotation. Trans-
actions of the Institute of Mining and Metallurgy 80:
C174}C180.
Lekki J and Laskowski JS (1975) A new concept of frothing
in Sotation systems and general classiRcation of Sotation
frothers. In: Proceedings of the XI International Min-
erals Processing Congress, Universita de Calgari, Cal-
gari, pp. 427}448.
Manev E and Pugh RJ (1993) Frother/collector interactions
in thin froth Rlms and Sotation. Colloids and Surfaces
70: 289}295.
Mingione PA (1984) Use of dialkyl and diaryl dithiophos-
phate promoters as mineral Sotation agents. In: Jones
MJ and Oblatt R (eds) Reagents in the Minerals Indus-
try, pp. 19}24. London: Institute of Mining and Metal-
lurgy.
Mitrofanov SI, Kuz’kin AS and Filimonov VN (1985)
Theoretical and practical aspects of using combina-
tions of collectors and frothing agents for sulphide
Sotation. 15e Congres International de Mineralurgie,
vol. 2, pp. 65}73. St. Ettienne: Societe de l’Industrie
Minerale et du Bureau de Recherches Geologiques et
Mineres.
Orel MA, Chibisov VM and Lapatukhin IV (1986) Use of
mixtures of butyl xanthate and hydrolyzed ployacrylam-
ide when Soating gold-containing ore. Soviet Journal of
Non-ferrous Metals 27: 97}98.
Plaskin IN and Zaitseva SP (1960) Effect of the
combined action of certain collectors on their distribu-
tion between galena particles in a Sotation pulp. (Min-
tek translation no. 1295, June 1988.) Naachnye Soob-
shcheniya Institut Gonnogo dela Imeni AA Skochin-
skogo, Akademiya Nauk SSSR, Moskva, Report no. 6,
pp. 15}20.
Plaskin IN, Glembotskii VA and Okolovich AM (1954)
Investigations of the possible intensiRcation of the Sota-
tion process using combinations of collectors. (Mintek
translation Feb. 1989.) Naachnye Soobshcheniya
Institut Gonnogo dela Imeni AA Skochinskogo,
Akademiya Nauk SSSR, Report no. 1, pp. 213}224.
Valdiviezo E and Oliveria JF (1993) Synergism in aqueous
solutions of surfactant mixtures and its effect on the
hyrophobicity of mineral surfaces. Mineral Engineering
6: 655}661.
1464 II / FLOTATION / Bubble+Particle Capture
Van Lierde A and Lesoille M (1991) Compared ef-
fectiveness of xanthate and mercaptobenzothiazole as
gold and arsenopyrite collectors. In: Proceedings of the
XVII International Minerals Processing Congress, Dres-
den, vol. IV, pp. 111}119.
Bubble^Particle Capture
J. Ralston, University of South Australia,
The Levels, Adelaide, Australia
Copyright ^ 2000 Academic Press
Introduction
Bubble}particle capture is the heart of froth Sotation.
For efRcient capture to occur between a bubble
and a hydrophobic particle, they must Trst undergo
a sufRciently close encounter, a process that is
controlled by the hydrodynamics governing their ap-
proach in the aqueous environment in which they are
normally immersed. Should they approach quite
closely, within the range of attractive surface forces,
the intervening liquid Rlm between the bubble and
particle will drain, leading to a critical thickness at
which rupture occurs. This is then followed by move-
ment of the three-phase-line contact line (the bound-
ary between the solid particle surface, receding liquid
phase and advancing gas phase) until a stable wetting
perimeter is established. This sequence of drainage,
rupture and contact line movement constitutes the
second process of attachment. A stable
particle}bubble union is thus formed. The particle
may only be dislodged from this state if it is supplied
with sufRcient kinetic energy to equal or exceed
the detachment energy, i.e. a third process of detach-
ment can occur.
The capture (or collection) efRciency E of a
bubble and a particle may be deRned as:
E"EC;EA;ES [1]
where EC is the collision efRciency, EA is the at-
tachment efRciency and ES is the stability efR-
ciency of the bubble}particle aggregate. This dissection
of capture efRciency into three parts was originally
proposed by Derjaguin and Dukhin (1960}61) and
focuses attention on the three zones of bubble}particle
capture where, in order, hydrodynamic interactions,
surface forces and forces controlling bubble}particle
aggregate stability are dominant.
Wakamatsu T and Numata Y (1979) Fundamental study on
the Sotation of minerals using two kinds of collectors.
In: Somasundaran P (ed.) Fine Particle Processing,
American Institute of Mining, Metallurgical and Petro-
leum Engineers, New York, pp. 787}801.
This article describes each of the substeps in the
bubble}particle capture process. The individual pro-
cesses and efRciencies are focused upon, since they
provide the key to understanding the substeps. Our
knowledge of the various efRciencies has been
enhanced by six important publications, referred to in
Table 1, each of which signalled major advances in our
understanding and catalysed further research in this
interdisciplinary Reld of colloid and Sotation science.
Processes and Substeps
Process 1: Collision Ef\ciency
For a batchwise Sotation process, the Sotation recov-
ery (the mass of particles recovered in a given time t)
R is given by:


3GhECEAES
R"1!exp!t "1!exp(!tk)
2dbV
[2]
where G is the volumetric gas Sow rate of a swarm of
bubbles of diameter db passing through a particle
suspension of volume V and depth h, and:
3GE EAESh
k" ! [3]
2dbV
The Sotation rate constant k is directly analogous to
that obtained in chemical reaction kinetics. Its value
will be partly determined by the substep(s) in
bubble}particle collision, attachment and detach-
ment processes, as well as by physical variables such
as G. (For a constant G and constant bubble size
distribution, db will be an appropriate average.)
Equation [2] has been shown to apply, for
example, to a system of monodisperse polystyrene
latex particles Soating under batchwise conditions.
A plot of ln (1!R) versus t yields the rate constant k.
For systems that are polydisperse in particle size
and/or in which particles of different hydropho-
bicities are present, the recovery then becomes the
sum of a series of exponential terms and the plot of
 II / FLOTATION / Bubble+Particle Capture 1465

Table 1 Key papers in understanding fundamental flotation substeps (details of references are given in Further Reading)

Date Area of research

1948 A fundamental paper by Sutherland on the kinetics of the flotation process appeared in Australia. This paper
invoked induction time, described particle size effects in flotation, and catalysed other similar approaches.
While it was preceded by other efforts, this paper was the first comprehensive effort to describe recovery, size
and time data in a fundamental manner.

1960}61 In Moscow, Derjaguin and Dukhin produced a key paper on the theory of flotation of small and medium-sized
particles. Hydrodynamics, surface forces and diffusiophoresis were all used in this theory. This seminal work
resulted in an acceleration of fundamental flotation research worldwide.

1972 Blake and Kitchener, working together in London, published some very careful measurements of the thickness
of aqueous films on hydrophobic quartz surfaces. Film thicknesses, measured as a function of salt concentra-
tion, were shown to depend on the electrical double layer force. Film instability occurred on hydrophobic
surfaces at film thicknesses less than about 60 nm. This value, which was smaller than the range of the
electrical double layer force, represented the combined effects of hydrophobic force, surface heterogeneities
and external disturbances. Blake and Kitchener’s film thickness studies hinted at the length dependence of
hydrophobic forces, information which was subsequently obtained by surface force experiments after 1982.

1976 Scheludko and colleagues in Bulgaria considered how particles might become attached to a liquid surface and
developed the capillary theory of flotation.

1977 Anfruns and Kitchener published the first measurements of the absolute rate of capture of small particles in
flotation. This was the first critical test of collision theory under conditions where the bubble and particle surface
chemistry was characterized and controlled.

1983 Schulze published a key textbook on the physicochemical substeps that are important in flotation, drawing on
a wide range of hydrodynamic, surface chemical and engineering information. Originally published in German,
once translated into English the book captured an international audience.

c. 1980}present There has been a strong interest in developing reliable collision models (Dai et al., 1998). The surface force
apparatus and, recently, the atomic force microscope colloid probe technique, have provided very useful
insight into electrical double layer, van der Waals and hydrophobic forces (Israelachvili, 1985; Fielden
et al., 1996). Thin film drainage has been investigated between a rigid and a deformable interface (Miklavcic
et al., 1995). Attachment efficiencies have been measured (Hewitt et al., 1995). Reliable methods for
measuring contact angles on particles have been developed (Diggins et al., 1990). Major theoretical and
experimental advances in describing dynamic contact angles on well-defined surfaces have been made (Blake,
1993).

ln (1!R) versus t will show curvature, reSecting the
different contributions to the recovery from the
various particles present in the mixture.
In the metallurgical literature, R versus t data are
frequently analysed by assuming that the pulp con-
sists of ‘fast’ and ‘slow’ Soating components, allow-
ing the respective rate constants (kf and ks) and
fractions (ff and fs) to be determined. Although this is
a gross simpliRcation of the real multicomponent
situation, much valuable information may be gleaned
from such an analysis. In fact the latter is frequently
used to examine the Sotation behaviour of particles
of a speciRc size range in Sotation circuits, where the
behaviour of an individual Sotation cell or bank of
cells may be approximated to a batchwise process.
Derjaguin and Dukhin were the Rrst to distinguish
three zones of approach of a bubble and a particle
on the basis of the different kinds of force in
each zone (Figure 1). This model is a very useful
one and helps to identify the various contributions
to capture efRciency. However, it should not
be taken to mean that there are well-deRned bound-
aries between each zone; rather they grade into
one another, the importance of the various
contributing effects in each zone being more ac-
curately identiRed as further information becomes
available.
Zone 1 is a region far from the bubble surface
where hydrodynamic forces are dominant, control-
ling EC in eqn [1]. Hydrodynamic drag forces act to
sweep the particle around the bubble, viscous forces
tend to retard this relative motion between the two,
while particle inertial and gravitational forces move
the particle towards the bubble.
1466 II / FLOTATION / Bubble+Particle Capture
Figure 1 Hydrodynamic (1), diffusiophoretic (2) and surface
force (3) zones of interaction between a bubble and a particle.
(Reproduced with permission from Derjaguin BV and Dukhin SS
(1960}61). Theory of flotation of small and medium-size particles.
Transactions of the Institute of Mining and Metallurgy 70:
221}246, Figure 1).
Broadly speaking, all models of collision efR-
ciency predict that EC decreases with particle size at
constant bubble size down to a particle diameter of
about 0.5
m. Then, Brownian diffusion prob-
ably takes over as the predominant capture mech-
anism (although this has not been proven), the
collision efRciency increasing with decreasing size
as the tiny particles (virtually ‘solute molecules’)
move towards the bubble surface. In 1948 Sutherland
made the Rrst signiRcant contribution to the treat-
ment of collision efRciency. His hydrodynamic
treatment of the process of particle and bubble ap-
proach in zone 1 was carried out without any consid-
eration of particle inertia, bubble deformation or Rlm
thinning, deRciencies that were in part recognized by
Sutherland and Wark in 1955.
The Sutherland theory, based on potential
theory or streamline Sow, shows that the concentra-
tion, C, of mineral Soated at a time t is related to its
initial concentration, C0, by the recovery, R as:
  
C0!C 3
R RpVt NB
R" " 1!exp!t

C0 cosh2 (3Vt /2Rb)
[4]
where R and Rp are the bubble and particle radii,


Vt is the bubble}particle relative velocity,  is the
induction time, NB is the number of bubbles per unit
volume, and is the fraction of particles retained in
the froth following bubble}particle attachment.
The reader should note the relationship between
eqns [2], [3] and [4], which are the basis for a Rrst-
order model, largely based on pulp microprocesses.
Despite the deRciencies of the Sutherland model, his
‘Rrst approximation theory’ yields results that are in
fair agreement with experimental determinations of
particle trajectories, touching angles and collision
efRciencies, obtained from model experiments
performed in a vertical Sow tube with individual
particles and a single bubble. For more detailed treat-
ments of the hydrodynamic aspects of bubble}
particle collision the reader is referred to the extensive
literature available.
The inability of collision theories to describe ad-
equately the collection process between bubbles and
smooth and angular particles was vividly demon-
strated by Anfruns and Kitchener in 1977. Their
experiments, the Rrst measurements of absolute rate
of capture, gave results in only fair agreement with
collision theory, assuming every collision resulted in
capture of their very hydrophobic particles.
Process 2: Attachment Ef\ciency
Derjaguin and Dukhin identiRed zone 2 in Figure 1
as that region where diffusion effects are
important. A strong electric Reld exists in this zone,
since the liquid Sow around the moving bubble gives
rise to a tangential stream at its surface that destroys
the equilibrium distribution of adsorbed ions there.
Where surfactant is present it is continually swept
from the upper to the lower surface of the bubble.
Transport of ionic surfactant to the moving bubble
surface therefore takes place, leading to the establish-
ment of a concentration gradient. A strong electric
Reld of order 3000 V cm\1 is established when
the cation and anion diffusion coefRcients
differ, as they generally do. Hence charged par-
ticles entering zone 2 will experience an elec-
trophoretic force in precisely the same way as in an
electrophoresis cell and will be either attracted to-
wards, or repelled from, the bubble surface. The term
‘diffusiophoresis’ was coined for this phenom-
enon, i.e. the ‘diffusiophoretic force’ therefore
acts on the particle as an additional force.
To date, however, evidence conRrming the pres-
ence or absence of diffusiophoresis in Sotation is
equivocal and sparse. Apart from noting its possible
contribution to capture efRciency, it is not pur-
sued further here.
In zone 3, surface forces predominate once the thin
Rlm between the bubble and the particle is reduced

much below a few hundred nanometres. These forces
can accelerate, retard or even prevent the thinning of
the liquid Rlm between the particle and the bubble.
From a thermodynamic point of view, the free energy
of a liquid Rlm differs from the bulk phase from
which it is formed. This excess free energy was origin-
ally called the ‘wedging apart’ or ‘disjoining’ pressure
by Derjaguin and represents the difference be-
tween the pressure within the Rlm, pf, and that in the
bulk liquid adjacent to the solid surface, pl. Note that
for a bubble pushed against a Sat solid surface, im-
mersed in water, pb, the pressure within the bubble, is
equal to pf. Derjaguin and his school, as well as
Scheludko, performed experimental measurements of
disjoining pressures, providing both the Rrst real veri-
Rcation of the DLVO theory of surface forces (named
after Derjaguin, Landau, Verwey and Overbeek), as
well as the Rrst accurate experimental estimates of the
Hamaker constant. The disjoining pressure (
) de-
pends on the Rlm thickness, h, and:

(h)"pf!pl [5]
For mechanical equilibrium in a stable Rlm
(h)'0
and d/dh(0.
If the liquid Rlm is stable at all thicknesses the
liquid is said to wet the solid completely and the solid
is said to be hydrophilic. This occurs, for example,
when an air bubble approaches a clean silica surface
immersed in water } in this instance the Hamaker
constant is negative and the corresponding van der
Waals force is repulsive for the silica}water}air triple
layer. For an unstable Rlm the thin Rlm must drain,
then rupture, and the resulting three-phase line of
contact (tplc, vapour}water}solid) must expand to
form a wetting perimeter before the particle can ad-
here to the bubble. Each of these events will have
a characteristic time associated with it, the sum of
which must be less than the contact time between the
bubble and the particle if bubble}particle capture is
to occur. The contact time is generally of the order of
10\2 s or less. The induction time,  (see eqn [4]) is
normally taken as the time required for bubble}par-
ticle adhesion to occur, once the two are brought into
contact, i.e. it is the sum of the thin Rlm drainage and
tplc spreading times (tfilm#ttplc) and is synonymous
with the attachment time. Rupture is a very fast
process and is not a signiRcant contributor to .
When a bubble is pressed against a solid surface,
through water, the intervening Rlm is generally not
plane parallel. Rather the edge of the Rlm thins quick-
ly and a small, thicker dimple is trapped in the centre,
because the bubble is deformable. This is essentially
a kinetic phenomenon, caused by the outSow being
greatest at the very edge of the Rlm in the initial stages
II / FLOTATION / Bubble+Particle Capture 1467
of drainage. The existence of this dimple has been
detected experimentally. Hydrodynamic theories
attempting to describe the proRle and evolution of the
dimple have been proposed but with very limited
success in describing experimental data. Surface
deformation of bubble surfaces can also occur under
the inSuence of electrostatic interactions (and
possibly other surface forces as well) aside from any
kinetic effects.
An unstable Rlm arises when there is a net attract-
ive force between the particle and the bubble. This
normally occurs when there is an attractive hydro-
phobic force involved, since the van der Waals and
electrostatic forces are repulsive, except in rare cir-
cumstances. The measurement of this hydrophobic
force, its length dependence and theoretical origins are
subjects of intense research effort. In recent times
it has become possible to measure the hydrophobic
force, in a conRguration relevant to the Sotation pro-
cess, by attaching a small particle to the cantilever in
an atomic force microscope (Figure 2). The particle is
then pressed against a captive bubble and the
force}separation distance proRle determined. In this
fashion, the various surface forces may be explored.
Experimental evidence relating to Rlm drainage
in systems where soluble surfactants are present is
rather equivocal. Adsorption and desorption pro-
cesses coupled with possible molecular reorientation
make any theoretical interpretation difRcult. Un-
fortunately these are the very systems that are of
primary interest to mineral processing. Furthermore
additional complications ensue when one considers
a particle approaching a bubble in Sotation. The
nature of the bubble surface (i.e. whether it is mobile
or immobile) will inSuence the thinning of the thin
Rlm between bubble and particle. This makes any
solution of the Navier}Stokes equation for Rlm
drainage difRcult, particularly in the case of the
Figure 2 Experimental arrangement for the measurement of
forces between a particle and a bubble using the atomic force
microscope colloid probe technique [from Fielden, Hayes and
Ralston (1996), Langmuir, 12, 3721}3727, with permission].
1468 II / FLOTATION / Bubble+Particle Capture
angular particles that are normally present in Sota-
tion. It is worth recalling at this point the observa-
tions that smooth spheres Soat more slowly than
angular particles under otherwise identical condi-
tions, presumably because the asperities on the angu-
lar particles lead to increased Rlm drainage rates
and/or rupture.
The kinetics of movement of the tplc are of central
importance in many processes, apart from Sotation.
During the movement of the tplc a dynamic contact
angle is established. Irrespective of whether the ‘sur-
face chemical’, ‘hydrodynamic’ or mixed ‘surface
chemical/hydrodynamic’ approaches are used, there
is as yet no general theory that adequately describes
tplc kinetics on all surfaces. One cannot generally
calculate ab initio what the spreading velocity of the
tplc will be when an air bubble spreads over a mineral
surface immersed in water in the presence of a surfac-
tant. Part of the problem at least is due to the fact that
poorly characterized experimental systems have been
used where any generalization has been obscured by
the same time-dependent adsorption/desorption/
molecular reorientation processes that complicate
thin Rlm drainage rate studies. Physical and chemical
surface heterogeneities on the particle surface also
strongly inSuence the tplc kinetics.
At present only the crudest estimates of tfilm and
ttplc can be made. Hence various experimental
methods for determining  are frequently resorted to.
A potentially valuable approach to the calculation of
induction times, based on bubble deformation and
restoration, has been developed.
These experimental methods for determining in-
duction times are generally based on either pressing
a bubble against a smooth mineral surface or against
a bed of particles. The disadvantages of all current
methods for determining  include: (1) insufR-
cient understanding of the process of bubble defor-
mation and energy dissipation during bubble}particle
collision; (2) insufRcient information concerning
the behaviour of the attractive hydrophobic forces
during the bubble}particle interaction (e.g. how the
thin Rlm of liquid evolves during the time a particle
slides or rolls around a bubble; it may well be incor-
rect to assume that bubble}particle interaction ceases
when the particle passes the bubble equator); (3) the
absence of data on tfilm, e.g. inSuence of surfactant
type and concentration on thin-Rlm drainage mecha-
nisms and rate; and (4) the absence of data on ttplc as
a function of hydrophobicity, physical and chemical
surface heterogeneities and surfactant type.
The most appropriate method for determining in-
duction times is probably through direct observation
of bubble}particle interactions in a Sotation cell un-
der well-deRned conditions. The necessary theory can
then be developed. For the present the Sutherland and
similar approaches (eqn [4]) serve as useful approxi-
mations in determining  from experimental Sotation
data of the type normally generated.
Kinetic effects certainly have a strong inSuence
on bubble}particle collision and attachment ef-
Rciencies. The latter provides the key to selective separ-
ations in Sotation. Once attachment has occurred, the
interplay between particle size and contact angle in the
environment of the Sotation cell becomes of paramount
importance and is the next subject of our discussion.
Process 3: Stability Ef\ciency and Detachment
Flotation limits for coarse particles The essential
problem in understanding bubble}particle aggregate
stability is to determine whether or not the adhesive
force, acting on the tplc, is large enough to prevent
the destruction of the aggregate under the dynamic
conditions that exist in Sotation. It is important to
understand the physics of the problem before moving
on to a mathematical description. Let us consider
a smooth spherical particle located at the Suid inter-
face. Once the equilibrium wetting perimeter has
been established following spreading of the tplc, the
static buoyancy of this volume of the particle will act
against the gravitational force (Figure 3). The hydro-
static pressure of the liquid column of height Z0 acts
against the capillary pressure. The ‘other detaching
forces’ require further discussion. Since they arise
from the particle motion relative to the bubble, velo-
city-dependent drag forces will oppose the detach-
ment of the particle from the bubble. An analysis of
these forces is extremely complex and has not been
reported to date. Therefore any force balance will
necessarily be quasistatic and approximate.
The net adhesive force, Fad, is equal to the sum of
the attachment forces, Fa, minus the detachment
forces, Fd, i.e.:
Fad"Fa!Fd [6]
Figure 3 Location of a smooth spherical particle at a fluid inter-
face. (From Schulze (1983) Physicochemical Elementary Pro-
cesses in Flotation. Amsterdam: Elsevier.)
 II / FLOTATION / Bubble+Particle Capture 1469

The particle will not remain attached to the bubble E Hydrostatic pressure, Fh, of the liquid column of
if Fad is negative but will be present in the liquid height Z0 on the contact area:
phase.
The mathematical description of the various forces Fh"! r20l gZ0"!
R2p(sin2 )l gZ0 [9]
that dictate the equilibrium position of particles at
liquid}vapour or liquid}liquid interfaces has fol- E Capillary pressure, Fp, in the gas bubble which acts
lowed an evolutionary trail. Analogous processes of on the contact area
r20:
interest, for example, include pigment ‘Sushing’,
where a solid particle is induced to transfer from one Fp"P
r20
A
liquid phase to another by appropriate surface modi-
Rcation with surfactants, and the stabilization of which for a spherical bubble is given approxim-
emulsion droplets by solid particles. ately as:
The actual problem of the balance of forces operat-


ing on a particle at a liquid}air interface has been 2
Fp+
R2p sin2  !2Rbl g [10]
considered by Sutherland and Wark, who considered Rb
the case of a gas bubble attached to a plane solid
surface of inRnite extent and used this as a model for E Gravitational force, Fg:
bubble}particle adhesion in Sotation. Since this work
there have been other very notable contributions. It 4
was Princen who proposed the Rrst extensive and Fg"
R3 p g [11]
3 
generalized treatment of the forces acting on a par-
ticle at Suid interfaces. This theory was developed where p is the particle density.
further by Schulze in 1977 and expanded in 1983. E Extra detaching forces, Fd, which are denoted ap-
Consider the case of a spherical particle at a proximately and generally as the particle mass
liquid}air interface. We assume that the system is in multiplied by a generalized acceleration bm in the
a quasistatic state and that the contact angle corres- Sotation cell:
ponds to that for a static system. The dynamic contact
angle can depart signiRcantly from the static value, 4
depending in part on the velocity of the tplc. If the Fd+
R3 pbm [12]
3 
particle oscillates around its equilibrium position, the
tplc would be expected to move to some extent. It is worth remarking that it is bubble}particle
Hence a full analysis would need to account for the aggregates that are actually accelerated in the Sota-
velocity-dependent drag forces mentioned above and tion device, thus p is in fact an approximation
link these to contact angle dynamics. Since this is an ("p! ).
intractable problem at present, a simpler approach is
necessary. At equilibrium, the sum of these forces,  F, must
Let us suppose that a spherical particle of radius equal zero.
Rp is attached to a bubble of radius Rb where Rb is The energy of detachment, Edet, corresponds to the
much greater than Rp, as shown in Figure 3. By un- work done in forcing a particle to move from its
derstanding the forces that operate on the particle, it equilibrium position, heq() at the liquid}vapour in-
is possible to calculate the energy of detachment. The terface to some critical point, hcrit(), where detach-
forces acting upon the particle are as follows: ment occurs and the particle moves into the liquid
phase. The sum of the various forces, F, is related to
E Capillary force, Fc, acting in the vertical direction Edet by:
along the tplc:


S
hcrit( )

Fc"2r0 sin 0 "!2
Rp sin [sin(#)] Edet" F dh() [13]

S
heq( )
[7]
Equation [13] may be solved by introducing the
where  is the liquid}vapour surface tension. various forces and carrying out a numerical integra-
E Static buoyancy of the fractional volume of the tion. The detachment process takes place when the
immersed particle, Fb: kinetic energy of the particle equals Edet. The kinetic
energy of the particle is given by 23
R3ppV2t, where Vt is

the relative (turbulent) velocity of the particle,
Fb" R3p l g[(1!cos )2(2#cos )] [8]
3 acquired due to stresses on the bubble}particle
1470 II / FLOTATION / Bubble+Particle Capture

aggregate in the turbulent Reld of the Sotation cell, as where F is the Helmholtz free energy, L is the contact
the aggregate collides with other bubbles or aggre- line, T is the temperature, V is the volume and W is
gates or due to other modes of excitation. Vt is deter- the thermodynamic work. The Young}DupreH equa-
mined experimentally as the velocity of gas bubbles in tion becomes:
the Sotation cell and p is the density of the particle.
The maximum Soatable particle diameter based on 
S/V!S/L"L/V cos $ [17]
the kinetic theory, Dmax,K, is given as: r

  
3 S 2
hcrit( )
The line tension is important for small contact radii
Dmax,K"2
R3p g
2
pV2t 3 and can oppose or reinforce L/V cos . It counteracts
S
heq( )
the formation of the tplc in Scheludko’s theory which


2p 3h neglects thin Rlm drainage and other hydrodynamic
; 1! !cos3 # sin2 
 2R effects. Experimental data for hydrophobic,
 angular quartz particles between about 10 and 35
m


3 in average diameter follow a general trend that is
! 2 2 sin  sin(#)
a Rp predicted by eqn [15], although quantitative agree-
ment is poor. If a pseudo-line tension, embracing

 
2
!
(Rp sin )2 !2R  g dh
Rb
this in turn enables Dmin in eqn [15] to be re-
Equation [14] may be solved by numerical integra-
tion or by plotting each of the kinetic and detachment
energies as a function of Rp at constant  and p and
speciRed Vt.  refers to the density of the Suid and  is
the surface tension at the liquid}vapour interface.
This equation has been shown to describe adequately
both the detachment of a sphere from a liquid}vapour
interface and the behaviour of hydrophobic angular
quartz particles between approximately 30 and
120
m in diameter under Sotation conditions.
Flotation limits for Vne particles The only theoret-
ical study to date dealing with the limit of Soatability
of Rne particles was published by Scheludko and
co-workers in 1976. The limit is the critical work of
expansion required to initiate a primary hole or three-
phase contact line during bubble}particle approach
} a requirement that is met by the kinetic energy of
the particles. The matching of these two quantities
enables a minimum particle diameter, Dmin,K, for So-
tation to be obtained:
1/3
surface heterogeneities, replaces  in eqn [15], then
expressed in terms of a critical bubble radius
[14] below which attachment does not occur. Recon-
ciliation between theory and experiment is then
achieved although the concept of pseudo-line
tension needs to be placed on a Rrmer experimental
foundation.
The Future
In terms of our fundamental understanding, there is
no entirely adequate collision model that can cor-
rectly account for particle size and inertial effects in
the presence and absence of soluble surfactants. Thin
Rlm drainage is poorly understood when one of the
interfaces is both physically and chemically hetero-
geneous, and the other is deformable. The nature of
the hydrophobic interaction between a particle and
a bubble requires both experimental and theoretical
veriRcation. There is no reliable model at present to
describe the movement of a three-phase contact line
over a physically and chemically heterogeneous sur-
face. Thus major research challenges exist that, if they
are to be successfully overcome, must embrace

 
32 1/3
systems where surfactants are both present and
Dmin,K"2 [15]
V  1!cos 
2
t
absent.
From a separation technology point of view, froth
where  is the line tension, opposing expansion of the Sotation will continue to be one of the principal
tplc. Molecules that are present in a line have a free means by which ores are successfully beneRciated for
energy that is different from those at a surface } in many years to come. Increasingly the technique is also
fact there is an excess linear free energy and a linear being used in the deinking of paper, soil remediation,
tension in an analogous fashion to that of excess plastics recycling and heavy metal ion decontamina-
surface free energy and surface tension. tion, to name but a few examples. Both research and
In fact, practice are expected to accelerate strongly over the


F next decades as new techniques and theoretical ap-
" [16]
L T,V,W proaches are used.

Further Reading
Anfruns JF and Kitchener JA (1977) Rate of capture of
small particles in Sotation. Transactions of the Institu-
tion of Mining and Metallurgy, Section C 86: C9}C15.
Blake TD and Kitchener JA (1972) Stability of aqueous
Rlms on hydrophobic methylated silica. Journal of the
Chemical Society, Faraday Transactions I 68:
1435}1442.
Blake TD (1993) Dynamic contact angles and wetting
kinetics. In: Berg JC (ed.), ch. 5. Wettability.
New York: Marcel Dekker.
Collins GL and Jameson GJ (1976) Experiments
on the Sotation of Rne particles: the inSuence of particle
size and charge. Chemical Engineering Science 31:
985}991.
Crawford R and Ralston J (1988) The inSuence of particle
size and contact angle in mineral Sotation. International
Journal of Minerals Processing 23: 1}24.
Dai Z, Dukhin SS, Fornasiero D and Ralston J (1998) The
inertial hydrodynamic interaction of particles and rising
bubbles with mobile surfaces. Journal of Colloid and
Interface Science 197: 275}292.
Derjaguin BV and Dukhin SS (1960}61) Theory of
Sotation of small and medium-size particles. Transac-
tions of the Institute of Mining and Metallurgy 70:
221}246.
Diggins D, Fokkink LGJ and Ralston J (1990) The wetting
of angular quartz particles. Colloids and Surfaces 44:
299}313.
Drelich J and Miller JD (1992) The effect of surface
heterogeneity on pseudo-line tension and the Sota-
tion limit of Rne particles. Colloids and Surfaces 69:
35}43.
Fielden ML, Hayes RA and Ralston J (1996) Surface
and capillary forces affecting air bubble}particle
interactions in aqueous electrolyte. Langmuir 12:
3721}3727.
Column Cells
I. M. Flint, Canadian Process Technologies Inc.,
Vancouver, BC, Canada
M. A. Burstein, NPACI, Edcenter on Computational
Science and Engineering, SDSU, San Diego, CA, USA
Copyright ^ 2000 Academic Press
Introduction
History
The Rrst pneumatic Sotation cell, which used air
sparging through a porous bottom and horizontal
slurry Sow, was patented in 1914 by Callow. The Rrst
countercurrent column Sotation device was designed
and tested by Town and Flynn in 1919. Cross-current
II / FLOTATION / Column Cells 1471
Hewitt D, Fornasiero D, Ralston J and Fisher LR (1993)
Aqueous Rlm drainage at the quartz}water interface.
Journal of the Chemical Society, Faraday Transactions
89: 817}822.
Hewitt D, Fornasiero D and Ralston J (1995) Bubble par-
ticle attachment. Journal of the Chemical Society, Fara-
day Transactions 91: 1997}2001.
Israelachvili JH (1991) Intermolecular and Surface Forces,
2nd edn. London: Academic Press.
Laskowski JS and Ralston J (1992) Developments in Min-
eral Processing. Colloid Chemistry in Mineral Process-
ing. Amsterdam: Elsevier.
Lynch AJ, Johnson NW, Manlapig EV and Thorne CG
(1981) Mineral and Coal Flotation Circuits: Their Simu-
lation and Control. Amsterdam: Elsevier.
Miklavcic SJ, Horn RG and Bachmann (1995) Colloidal
interaction between a rigid solid and a Suid drop. Jour-
nal of Physical Chemistry 99: 16357}16364.
Ralston J (1992) The inSuence of particle size and contact
angle in Sotation. In: Colloid Chemistry in Mineral
Processing, ch. 6. Amsterdam: Elsevier.
Scheludko A, Toshev BV and Bojadjiev DT (1976) Attach-
ment of particle to a liquid surface (capillary theory of
Sotation). Journal of the Chemical Society, Faraday
Transactions 72: 2815}2828.
Schulze HJ (1983) Physico-chemical Elementary Processes
in Flotation: An Analysis from the Point of View of
Colloid Science Including Process Engineering Consid-
erations. Amsterdam: Elsevier.
Sutherland KL (1948) Kinetics of the Sotation process.
Journal of Physical Chemistry 52: 394}425.
Sutherland KL and Wark IW (1955) Principles of Flotation.
Melbourne: Australasian Institute of Mining and Metal-
lurgy.
Ye Y and Miller JD (1989) The signiRcance of bubble}par-
ticle contact time during collision in the analysis of
Sotation phenomena. International Journal of Mineral
Processing 25: 199}219.
pneumatic Sotation machines were widely used in
industry in the 1920s and 1930s, but were later
replaced by the impeller-type Sotation devices in
mineral-processing plants. Dissolved-air Sotation
became the main type of Sotation for water treatment
applications. These substitutions were the result of
the absence of effective and reliable air spargers for
Rne bubble generation and the lack of automatic
control systems on the early columns. During this
period, both the poor Sotation selectivity and entrain-
ment of slimes characteristic of impeller-type cells
were offset by the use of complex Sow sheets using
large numbers of cleaner stages and recycle
lines. Column Sotation devices were reintroduced
1472 II / FLOTATION / Column Cells
for mineral processing in Canada by Boutin and
Wheeler in 1967, at which time washwater was added
to the froth to eliminate entrainment of hydrophilic
materials to the Soat product. By the late 1980s col-
umn Sotation had became a proven industrial techno-
logy in the mineral industry. These separators are
routinely used on their own or in conjunction with
other types of devices within separation circuits.
This technology is currently being applied to liquid}
liquid separations (oil}water, organic solvent}liquid),
solid}liquid, or solid}solid separations in many
industries.
Comparative Strengths and Weaknesses
Column cells are Sotation devices that also act as
three-phase settlers where particles move downwards
in a hindered settling environment. Within the vessel
there is a distribution of particle residence times depen-
dent on settling velocity that may impact on the Sota-
tion of large particles. Impeller devices do not suf-
fer from this effect to the same degree but do
require higher energy input to suspend larger particles.
The low turbulence in columns means particles
usually have low momentum, which in turn may
reduce the probability of collection by passing
bubbles. As a result, Rne particle recovery may be
hindered when compared to the capabilities of impel-
ler-type designs.
The mechanism of particle}bubble collision in col-
umns is different from intensive mixing devices
such as impeller cells. Under the low intensity mixing
caused only by a rising bubble swarm, particle drift
from the liquid streamlines is caused mainly by grav-
ity and inertial forces and also by interception, while
in mechanical cells, according to many researchers,
bubble}particle collision occurs at their relative
movement within a turbulent vortex or at adjacent
vortices. Also, as velocities of both bubble and par-
ticle during the attachment are slower under the qui-
escent conditions in a column, the contact time is
generally higher. Therefore, probabilities of both
collision and adhesion (components of attachment
probability) are different to those in mechanical
Sotation processes.
The lower velocity gradient and less intensive shear
forces in the vicinity of rising bubbles under low
turbulent conditions in a column lead to reduced
detachment probability. The latter is most important
for improvement of recovery for coarse, heavy or
weakly hydrophobic particles.
A column can support a deep froth bed and may
use washwater to maintain a downward Sow of
water in all parts of the vessel. This essentially elimin-
ates the entrainment of hydrophilic particles in the
Soat product when the vessel is used for solid}solid
separation. This property, along with the absence of
stray Sows of feed material to the Soat product by
turbulence, means that column devices are normally
superior to impeller-type machines for the selective
separation of Rne particles.
In immiscible liquid separation duties, columns do
not emulsify the material like impeller devices.
The bubbles used in a column are usually generated
within the size range that maximizes interfacial sur-
face Sux and collection intensity through the vessel.
Dissolved air systems nucleate micrometer-sized
bubbles on particles which require very low down-
ward liquid velocities in large volume vessels to separ-
ate the bubble and water. Also, dissolved air systems
cannot provide air hold-up higher than approxim-
ately 4}6%, due to limited gas solubility and lower
Sooding limits caused by the microbubbles. In mech-
anical cells, bubbles are usually generated by shear
action of the impeller; thus, bubble size is dependent
on both air Sow rate and impeller rotation speed. As
such, bubble size cannot be controlled independently
of cell turbulence.
The height-to-diameter ratio of a column is signiR-
cantly higher than the impeller-type machines. As
a result, control and consistency of Sow are more
critical. The column requires much less Soor space to
operate.
Control Systems
Control systems are designed to maintain separation
in a changing environment by maintaining operating
variables at their optimum values for process perfor-
mance. The conRguration used depends on the varia-
bility of the vessel feed, the ability of the operating
and instrumentation staff, the availability of de-
tectors and other parts, capital costs and the goals of
the project. The most basic system only controls the
interface level, between the aqueous suspension and
froth phases, while complex systems can integrate
expert systems or other forms of artiRcial intelligence
into a full-grade/recovery adjustment strategy.
All columns perform best when Sows are constant,
therefore operation should be as close to steady-state
conditions as possible. Good control systems limit
damage due to variations by maintaining constant
Sows in earlier stages, establishing a recycle within
the column system, or compensating by changing
conditions within the vessel.
Level
The goal of a level control system is to maintain
a constant aqueous suspension depth despite changes

Figure 1 Example of level control loop. LT, level transmitter; LIC, level indicator and control; LY, level D/A signal conversion; LCV,
level control device; MAG, magnetic flow detector; FT, flow transmitter; FI, flow indicator; PID, proportional-integral-derivative; PLC,
programmable logic controller.
in feed Sow, Soatable material concentration or air
rates. An example of this control is found in Figure 1.
In water}oil separation, a periodic level rise may be
organized to dump an accumulated organic pad. The
simplest method of controlling level is to adjust the
discharge height of the underSow using a ‘gooseneck’
or alternative form of gravity control. If this is not
possible, then the level must be detected and that
signal used to control either a variable-speed pump or
control valve through a controller device. Detection
devices include Soats; pressure, capacitance, conduc-
tance, and ultrasonic transducers, or combinations of
these devices. The set point for the level is determined
from the desired froth depth. Generally, the higher
the level, the greater is the recovery of the Soating
component and the lower its content in the overSow
II / FLOTATION / Column Cells 1473
(froth product). In more complicated systems, the
level control may be used with froth or oil pad depth
data to control overSow grade, with Sow-monitoring
devices for predictive control based on incoming feed,
or multiple monitors to compensate for variations in
air rate or feed composition.
Air
The purpose of the air loop is to control a volumetric
Sow of air through the column or to maintain a three-
phase density within the vessel. In basic control
systems air rate is controlled manually based on
a monitored air Sow rate. In slightly more advanced
systems, the Sow is controlled through an automatic
valve to compensate for pressure changes (Figure 2).
1474 II / FLOTATION / Column Cells
Figure 2 Example of air control loop. PI, pressure indicator; FIT, flow indicator and transmitter; FIC, flow indicator and controller; FY,
flow D/A signal conversion; FCV, flow control device.
Air rate may be linked to predictive- or recovery-
based systems.
Bias
Bias is deRned as a downward Sow of liquid through
the froth zone. Positive or downward bias is usually
used when two suspended substances must be separ-
ated from each other. If multiple separation stages are
in operation, it is usually used on the last stage. The
downward Sow of water through the froth is control-
led by varying the water rate added to the froth zone.
This Sow may be monitored by temperature, con-
ductance or by Sow differences (water added to
froth minus overSow water, or amount of liquid in
underSow minus its amount in feed). The actual bias
needed depends on the distribution of the water
through the froth and the hydrophilic particle sizes.
Bias may be estimated using the difference in
slurry Sows (Figure 3) or, more accurately, by Rrst
calculating the liquid volumetric Sows using Sow and
density meters.
Advanced Controls
It is possible } although not common } to control
a column to separate according to a grade}recovery
response curve. As grade increases or decreases in the
feed, level, air rate and bias may be adjusted to
achieve the most economical performance. This type
of control requires a good predictive model based on
theoretical knowledge, past experience and test work
that uses information from upstream processes to
adjust column parameters in anticipation of changes
(feed-forward control). Predictive systems provide
feed-forward control and can incorporate either
knowledge base or models (statistical or determinis-
tic) into the control loop. Excessive complexity of
models or control strategy does not improve the re-
sults as the uncertainty in parameters grows. Such
a system also requires extensive online detection
equipment such as density, Sow and pressure meters.
When these controls are implemented they are either
model-based systems or some form of artiRcial intelli-
gence (knowledge base, neural network systems
based on fuzzy logic principles).
Operating Parameters
Process-operating variables are those inputs to the
separator that may change with time and can be used
to control the production quantity and quality. These
include column control variables such as gas rate,
washwater rate and froth or oil pad levels. There are
also variables that may be controlled but are usually
not even monitored, like bubble size distribution, and
variables that depend on other parts of the operation
such as volumetric feed rate, and feed solids charac-
teristics: concentration, liberation and particle size
distribution.
Gas (Air)
Gas (air) rate is an effective parameter to control
separation since the probability of particle collision

Figure 3 Example of differential feed}underflow bias control loop. LCV, level control device; LY, level D/A signal conversion; LIC,
level indicator and control; FT, flow transmitter; MAG, magnetic flow detector.
with bubbles is dependent on the number of bubbles
and their size distribution. The maximum particle
surface Sux removed depends on bubble surface area
Sux. As surface area Sux increases, so does the prob-
ability of material}bubble aggregation (collection)
within a speciRc range. This range is bounded by the
increased mixing intensity as Sooding limits are ap-
proached and the increase in bubble size that is usu-
ally associated with an increase in gas Sow. The total
removal capacity, known as carrying capacity, can
also be controlled by the gas rate since it is propor-
tional to the speciRc bubble surface area. The carry-
ing capacity is determined as the maximum amount
of material which can be transported into froth in
unit time from a unit cross-sectional area of a col-
umn. It varies depending on particle size (for solid
separation) and density of the Soating substance. The
carrying capacity can be estimated from the balance
II / FLOTATION / Column Cells 1475
of the available bubble surface area and particle sur-
face Sux. The normal range of superRcial air velocity
is 1.0}2.5 cm s\1. In buoyant material separations,
high gas rates may reduce the three-phase density of
the aqueous suspension within the column to a den-
sity lower than that of the product. This will cause an
unstable pad that will sink if not quickly removed
from the system.
Volumetric Feed Rate
The volumetric feed rate determines the vessel reten-
tion time and strongly inSuences vessel mixing. An
increase in superRcial suspension velocity results in
lower gas limits as Sooding will occur at lower gas
rates and increases the size of microbubbles which
become entrained by downward Sow to the under-
Sow. However, higher slurry velocities also decrease
1476 II / FLOTATION / Column Cells
the negative inSuence of mixing on grade and recov-
ery (higher Peclet number) and lessen the retention
time difference between Rne and coarse particles
due to the settling. Typically, superRcial feed velocity
is 0.5}1.3 cm s\1.
Feed Solids
An increase in the percentage of solids contained in
the feed increases the residence time of those solids in
the case of constant-column throughput of solids.
The maximum solids load is determined by the vis-
cosity of the system and may be only 0.25}2%
(weight/weight) for paper de-inking applications to
almost 70% for calcite/silica separation.
Washwater
Mineral separation columns can provide a positive
bias which causes displacement of the feed liquid
phase with washwater in the overSow. This substitu-
tion virtually eliminates entrained Rnes from the over-
Sow product. Washwater distribution on to or into
froth and its Sow rate should be individually tuned
for each application depending on feed and concen-
trate size distributions, froth stability, height and
mobility, and on process objectives. Excessive wash-
water supply causes froth disruptions, loss of recov-
ery and dilution of products. Typically, superRcial
washwater Sow rate does not exceed 0.15 cm s\1,
although optimal rates depend on washwater distri-
bution design and froth rheology. Washwater is not
normally used in mineral roughing or scavenging
operations, oil}water separations, or systems where
entrainment is not a factor.
Froth Depth (Solid Separation)
The froth level maintained within the column is
highly variable depending on the application. Some
vessels may be operated with no froth, such as
oil}water separators, or mineral columns operating
on very large particles. In other cases, like molyb-
denite Sotation, a froth as deep as 1.5 m may be run
to ensure minimal entrainment and high selectivity. In
general, a deep froth gives more opportunities for
grade/recovery control and compensates for poor
washwater distribution. Froth depth in mineral
(solid}solid separation) column Sotation typically
varies from 15 to 300 cm. The gas hold-up in froth
gradually increases upwards due to froth sineresis
and drainage along plateau canals. The entrained Rne
particles return back to the lower (collection) section
of the column by net downward liquid Sow in the
froth (in the case of positive bias). Experimental data
conRrm that, in some cases, upgrading of the product
occurs mainly in the froth zone, and not at the collec-
tion stage.
It is important to note that an increase in froth
depth decreases the volume of the remainder of the
column which may be detrimental to overall perfor-
mance.
Organic Pad (Liquid Separation)
In an oil separation vessel a hydrocarbon pad may be
maintained at the top of the column. A deep pad
minimizes water entrainment into the overSow but
may increase the stripping of light hydrocarbons.
When high air rates are used and the organics pad is
not removed, droplets of organic phase may form and
drop through the aerated zone of the column. Air
rates must be lowered or the organics pad continu-
ously removed as a froth to prevent sinking of the
Soated organics.
Bubble Size
Some types of spargers allow the change of bubble
size distribution at nearly constant overall air rate.
Both break-up and coalescence of bubbles occur after
formation by the sparging devices which results in an
equilibrium size distribution above a certain distance
from the spargers. The average and deviation of this
distribution depend on the surface tension at the
air}water interface and turbulence in the cell. Gener-
ally, smaller bubbles provide higher collection inten-
sity and carrying capacity, but loaded microbubbles
may sink or be entrained in the downward slurry.
Also, maximum gas rate (at column Sooding point or
transition to a churn-turbulent regime) is reduced
with decreasing bubble size, meaning that there is
a speciRc bubble size that gives the maximum upward
rising surface area Sux. The point of column Sooding
can be estimated (in the assumption of cross-section
Sow uniformity and narrow bubble size distribution)
from the drift Sux model. In many cases a combina-
tion of smaller bubbles that provide the separation
and coarser transport bubbles that coalesce with the
smaller bubbles results in optimal Sotation rates.
Column Circuits
Column cells can be used to perform many functions.
These include separation within a grinding circuit
(unit cell), as an initial (primary or rougher) or scav-
enging separator whose purpose is maximum recov-
ery of material, or as a Rnal separator (cleaner or
recleaner) used to produce a pure product. They can
also be used to process bleed streams from other
processes. There are many examples of column
usage, including base metal and industrial mineral

separation, iron ore puriRcation, coal cleaning, sol-
vent extraction and oil}water separation, paint recov-
ery and newspaper de-inking. In addition, columns
can be used to remove hydrophobic substances, or
materials dissolvable in hydrophobic liquids, from
water or soils. Examples are DDT, polycyclic aromatic
hydrocarbons (PAHs) or other dangerous chemicals,
oil production from tar sands, or the puriRcation or
removal of algae or bacteria from cultures. All of these
separations fall into three categories: solid}solid,
solid}liquid and liquid}liquid separations.
Solid^Solid Separations
In order to get a good separation, the solids present
must be liberated: that is, not physically or chemically
attached, be suspended in a liquid medium and the
Sotation kinetics of the materials must be differ-
ent. One or more stages of separation may be needed,
depending on the kinetics and chemistry of the separ-
ation. To achieve sharper separation when dif-
ference in Sotation rate of components is not high
and/or material is not completely liberated, compli-
cated Sowsheets including multiple recycle lines and
regrinding are used. Regrinding operations for mid-
dlings are used to avoid over-grinding of the bulk of
material as it would cause reduction in Sotation rate
and selectivity for Rne particles. For Rnely dis-
seminated ores, entrainment is a substantial factor
Figure 4 Hydrophilic product, solid}solid four-stage separation circuit. Example of iron ore.
II / FLOTATION / Column Cells 1477
reducing sharpness of separation. Entrainment is
a process of particle transfer to froth without their
attachment on to bubble surfaces. This phenomenon
can be explained by movement of small particles in
the wake behind the rising bubble or within the static
layer of liquid surrounding it. In machines with inten-
sive mixing (impeller cells) the entrainment can also
be caused by local upward slurry Sows. These Sows
are not present in columns therefore reducing overall
entrainment intensity and improving separation
efRciency. A classical Sotation Sowsheet includes
several cleaning stages generally linked by recycle of
the cleaner tailings to previous stages. When more
than one material is Soatable and separation depends
only on degrees of hydrophobicity (molybdenite-
chalcopyrite), four to six stages may be required. If
insufRcient recovery is achieved in the primary
vessel (rougher Sotation), scavenger cells may be
used. In general, all stages do have a common separ-
ation goal. For example, silica (impurity) is Soated
away from hematite in a four stage iron ore circuit in
Figure 4. This circuit, or variations of it, is common
when the valuable product is hydrophilic or an under-
Sow product of the column. The example gives four
stages of separation; however, in many cases fewer
stages are required.
The circuit for a hydrophobic product is shown in
Figure 5. The second cleaner stage of this circuit is
generally not needed unless the separation is between
1478 II / FLOTATION / Column Cells
Figure 5 Hydrophobic product, solid}solid four-stage separation circuit. Example of copper or plastics float.
hydrophobic materials with similar Sotation rates. As
an example, this conRguration or variations of it can
be used in phosphate, copper, zinc and plastics separ-
ations, or for soil remediation.
Solid^Liquid Separations
In many circumstances a solid is present in a liquid
stream that must be removed. Flotation is often a vi-
able precursor stage, used to increase the percentage
of solids, prior to Rltration. This type of system can be
used to Soat coal and associated PAHs from run-
off water and upgrade the presentage of solids
from p.p.m. levels up to 10}25%. Figure 6 gives an
example of such a circuit where PAHs from coking
coal are Soated from a contaminated site run-off
Figure 6 Example of solid}solid separation: PAH from run-off water. Input of approximately 500 p.p.m. solids; filter feed of
approximately 24% solids.
water without removing the naturally occurring sand
and silt.
Flotation can also be considered as an alternative
to settling of naturally hydrophobic materials in
wastewater treatment. This type of separation may
also be used to remove bacteria or algae from water,
or many solid substances from reaction vessels.
Liquid^Liquid Separations
Immiscible liquids of any kind can be separated from
water by Sotation. The bubbles act to increase the
kinetics of the naturally Soating droplets such as
diesel, crude oil, kerosene or the organics used in
solvent}liquid extraction processes. Some examples
are hydrocarbon separation from water on oil

Figure 7 Treatment of oil platform process water; generalized circuit.
production platforms prior to Rnal release of water,
site run-off remediation and organics separation
in hydrometallurgy. Columns are capable of remov-
ing freely Soating hydrocarbons but usually not emul-
siRed or dissolved hydrocarbons. In order to remove
emulsiRed forms of hydrocarbons, a pre-aeration unit
must be installed.
Oil production application Large amounts of water
are involved in the extraction and production of oil.
Column cells are used in the water treatment stage of
production prior to release of the water back into the
environment. In a typical circuit, as shown in
Figure 7, water from the process is Rrst passed
through a cyclone or corrugated plate separator then
to a column. The hydrocarbon concentrate from both
of these vessels is returned for processing.
Site run-off remediation Sites that contain hydro-
carbon contamination such as reRneries and distribu-
II / FLOTATION / Column Cells 1479
tion depots often have run-off waters that con-
tain entrained hydrocarbons. These can be treated
effectively with Sotation technology using a cir-
cuit containing a column either working on its own or
in conjunction with a settling tank. If emulsiRed
hydrocarbons are present, a pre-aeration unit may
be required on the column in order to achieve con-
tamination level under 15 p.p.m.
Organic+aqueous separation in solvent}liquid
extraction circuits The solvent liquid extraction cir-
cuits employed in most hydrometallurgical processes
require the removal of essentially all of an organic
solvent from an aqueous medium in more than one
stream. Initial separation is usually done in settlers.
Columns with or without a pre-aeration unit can be
used as secondary separation devices prior to Rlter-
ing. The advantage of columns over many other
devices is their ability to compensate for wide Suctu-
ations in both aqueous and organic Sows.
1480 II / FLOTATION / Cyclones for Oil / Water Separations
Conclusion
Column Sotation has become the standard proven
industrial Sotation technique rather than an experi-
mental method during the last decade. Nevertheless,
its use in mineral-processing plants is mainly re-
stricted at present to cleaning operations. The future
of Sotation equipment development lies in the combi-
nation of the advantages of impeller and column
Sotation and in the use of pneumatic machines in
roughers.
As a greater share of Sotation operations are
used for unconventional areas such as environmental
applications (water treatment, soil remediation,
etc.) and ultraRne and colloid particle separation,
special machines will be developed combining
Sotation attachment at intensive aeration and mixing
conditions and three-phase separation in a quiescent
environment. This leads to the concept of pre-aer-
ation in a reactor (a unit for attachment of recovering
phase on to gas bubbles) and de-aeration in a separ-
ator (a unit for separation of loaded bubbles from the
bulk of three-phase suspension). Additional coarser
bubbles can be added in a separator as a carrier to
enhance the removal of loaded microbubbles by co-
alescence.
This concept and other types of new combined
Sotation machines will provide for more effec-
Column Flotation Cells
See II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
Cyclones for Oil/Water Separations
M. T. Thew, University of Bradford,
Bradford, UK
Copyright ^ 2000 Academic Press
Synopsis
Though the solid}liquid hydrocyclone has been estab-
lished for most of the 20th century, satisfactory
liquid}liquid separation performance did not arrive
until the 1980s. The offshore oil industry had
a need for compact, robust and reliable equipment for
removing Rnely divided contaminant oil from water.
This need was satisRed by a signiRcantly differ-
tive and efRcient separation for a wide range of
applications.
See also: I/Flotation. II/Flotation: Froth Processes and
the Design of Column Flotation Cells.
Further Reading
Agar GE, Huls BJ and Hyma DB (eds) (1991) Column ’91.
Proceedings of an International Conference on Column
Flotation, June 2}6, 1991. Sudbury, Ontario, Canada:
Canadian Institute of Mining and Metallurgy and Petro-
leum.
Boutin P and Wheeler DA (1967) Column Sotation devel-
opment. Canadian Mining Journal 88: 94.
Finch JA and Dobby GS (1990) Column Flotation. New
York: Pergamon.
Gomez CO and Finch JA (eds) (1996) Column ’96. Proceed-
ings of the International Symposium on Column Flotation,
August 26}28, 1996. Montreal, Quebec, Canada: Cana-
dian Institute of Mining and Metallurgy and Petroleum.
Pal R and Masliyah J (1991) Process dynamics and control
of a pilot Sotation column. Canadian Metallurgical
Quarterly 30: 87}94.
Rubinstein JB (1995) Column Flotation, Processes, Designs
and Practices. Basel, Switzerland: Gordon and Breach.
Yingling JC (1993) Parameter and conRguration optimiza-
tion of Sotation circuits, part I: a review of prior work.
International Journal of Mineral Processing 38: 21}40.
ent type of hydrocyclone, which of course had no
moving parts.
After explaining this need more fully and compar-
ing it with solid}liquid cyclonic separation in mineral
processing, the advantages that the hydrocyclone
conferred over types of equipment installed earlier to
meet the duty are given.
Separation performance assessment criteria are
listed prior to discussing performance in terms of feed
constitution, operator control and the energy re-
quired, i.e. the product of pressure drop and Sowrate.
The environment for petroleum production sets
some constraints for materials and this includes the
problem of particulate erosion. Typical materials
 II / FLOTATION / Cyclones for Oil / Water Separations 1481

used are mentioned. Relative cost data for types of oil
separation plant, both capital and recurrent, is out-
lined, though sources are sparse. Finally, some
pointers to further development are described, as the
oil industry looks to equipment installed on the sea
bed or even at the bottom of the wellbore.
Introduction to Liquid^Liquid
Hydrocyclones
This article covers the application of hydrocyclones
to remove or concentrate dispersed oil from water.
Two main classes of operation relating to their use
with water-continuous liquid exist. Firstly as re-
movers of oil contaminant from water (clean-up
units) and secondly as a method of de-watering crude
from wet oil Relds (concentrator units). It excludes
usage in relation to oil spills at sea, though feasibility
studies have been technically successful. It also ex-
cludes applications where dispersed water (brine) is
found in oil, though articles on this application } as
yet only on the fringe of commercialization } may be
found in the Hydrocyclone Conferences listed in the
bibliography. Operation with oil-continuous liquid
is difRcult since interfacial effects are larger
and break-up more likely as the brine droplets are less
viscous than the continuous liquid.
Table 1 lists the key stages in arriving at the pres-
ent near-universal usage of hydrocyclones for re-
moval of oil contaminants from water in the oil
industry offshore and more recently onshore.
Concentrators, sometimes called ‘dehydration hydro-
cyclones’, are used in very wet oilRelds as the initial
stage to reduce the water (brine) content from, say,
95% to 50% or less. Under some conditions the
overSow stream inverts to become oil continuous.
The two decades since de-oiling hydrocyclones
were Rrst seen to be capable of reaching legal stan-
dards of cleanliness offshore, e.g. 40 mg kg\1
maximum free oil on the UK Continental Shelf, are
brief when compared with the much longer and wide-
spread employment of solid}liquid hydrocyclones as
in mineral processing or china clay production. Some
salient comparisons between the two categories and
the consequences of the clean-up or concentrator
duty are set out in Table 2.
Flow\eld and Geometry
The dispersion of Rne oil droplets and their low dif-
ferential density necessitate a high radial acceleration
Reld. Since the oil is buoyant it will migrate towards
the vortex core. To understand separation perfor-
mance it is helpful to stress consequences of these two
points. The oil is much more sensitive to the Sow
pattern than solid particles so this leads to the require-
ment for a low turbulence, reasonably linear vortex
core and low peak shear SowReld to avoid droplet
break-up leading to lower oil removal efRciencies.
The overSow stream should be a small fraction of
the feed, since the oil content in clean-up applications
is typically below 1%. An approximate volumetric
balance gives the interrelationship between the para-
meters of a de-contaminating hydrocyclone. If the
volumetric feed concentration of oil Cf&1% and the
overSow and underSow concentrations Co ,Cu are as-
sumed to be 50% and 0 respectively, then the under-
Sow rate Qu will be 98% of the feed rate Qf. For
a wet oil concentrator however assuming Cf&10%
Qu will be &80% of Qf. The overSow (reject) stream
Sow rate clearly is Qo"Qf!Qu.
Colman and Thew at the University of Southamp-
ton in the late 1970s found that a cone angle as small
as 1}1 123 (total angle) produced the necessary stable
vortex, with a small relatively fast moving core mov-
ing to the overSow. This reverse axial Sow penetrated
a long distance downstream, so that in the early work
a cylindrical tailpipe was put on the end of the cone.
They also found that an enlarged entry section,
as illustrated on Figure 1, reduced the pressure
drop and reduced peak shear; the left-hand side of
Figure 1 shows the approximate proportions of the

Table 1 Advances in the use of oil}water hydrocyclones

1950s and 1960s Sporadic work on liquid}liquid hydrocyclones, especially in relation to use in the atomic energy industry
1965 Bradley produces his classic text on ‘Hydrocyclones’ clarifying the problems of liquid}liquid separation
1974 Kimber and Thew achieve 90% oil separation at Southampton University, UK
1978}80 The Southampton Group achieve '99% separation with crude oil (Colman and Thew, 1980 Hydrocyclone
conference)
1983}84 First field trials offshore
1985 First large offshore installation (15 m3 min\1)
Late 1980s Installations worldwide
1990s Virtually only method of oil}water separation. Number of manufacturers grows, prices fall. Higher mechanical
packing density in pressure vessels
Early 1990s Use in concentrator mode begins
Late 1990s ‘Downhole’ trials begin
1482 II / FLOTATION / Cyclones for Oil / Water Separations

Table 2 Comparisons between solid}liquid and oil}water hydrocyclones

Factor Solid}liquid Oil}water

Differential density Often water}quartz, 1650 kg m\3 50}300 kg m\3

Split ratio (QU/QF) Fixed; set by outlet orifice size Controllable by external valves
Outlets Usually one or both open to atmosphere Closed system
Axial pressure gradient near the centre line Usually very small (air core) Substantial, overflow at lower pressure
than the underflow
Pressure drop 1}5 bar 5}20 bar
Inlet pressure 1}5 barg 10}102 barg
Shear May cause some size reduction due to Likely to cause droplet break-up
particle-particle interaction
Particle size 1
m}10 mm Usually droplets &1
m}100
m
Concentration (by volume) Varies greatly; can be slurry at underflow Typical oil contaminant concentration in
feed to first stage (0.1%, oil
contaminant concentration in feed to
second clean-up stage 5}10%, well-
head oil concentration in feed to wet oil
concentrator 5}50%
Orientation Fixed, ‘g’ important No limitation, lateral acceleration up to
&0.1 ‘g’ on floaters unimportant

Southampton bi-cone design, which had twin tangen-
tial inlets to produce a linear core. The right-hand
side illustrates a typical later development with
a single 1803 wrap-round involute inlet and a curved
wall. Note the absence of a projecting vortex Rnder,
since there is no loss of oil in any short circuit Sow in
the end wall boundary layer.
Both the clean-up and concentrator units use the
same wall proRle but the latter has a larger overSow
(or reject) port.
Later systematic work by Young (1994) came up
with similar proportions.
Stronger swirl increases the acceleration Reld, but
also raises the pressure drop and too much swirl
seems to increase vortex instability and shear. Using
a swirl number S (S"DRXi(2Ai)\1, where Ai is the
total inlet area measured at a point where the Sow is
normal to the radius from the hydrocyclone axis to
the centroid of the area, Xi , and DR is a reference
diameter of the hydrocyclone), the range of values for
de-oilers is typically 7}11 or more usually 8}10. The
reference value DR for the bi-cone design is at the
junction of the two cone angles (see Figure 1), and for
a curved wall at, say, the point where the tangent is at
about 103 to the cyclone axis.
The axial pressure gradient near the centre line is
not zero: see section on pressure drop later.
De-oiling hydrocyclones have shown a range of
sizes but there are no large units as for solid}liquid
separators, since large droplets do not exist and thus
no requirement for units with a relatively small accel-
eration Reld. Size is a compromise, taking into ac-
count the factors shown in Table 3. Early installa-
tions showed DR values rising from about 30 mm to
70 mm, but more recently this has tended to drop
back to about 15}30 mm.
Installation
Unlike solid}liquid units, clean-up or concentrator
hydrocyclones have hitherto almost always been in-
stalled in pressure vessels, as shown schematically in
Figure 2. This allows easy fabrication of individual
units (sometimes called liners) from relatively thin
walled material and reduces the number of connec-
tions to be made. Several assemblies complete with
instrumentation and controls are built into a skid.
Some larger units have serious vibration at very high
Sows so that a mid-length damper is built in.
The overSow or reject stream, being a small pro-
portion of the feed, is readily incorporated into
a manifold which may double up as a mounting plate.
Though series operation is possible and has been
laboratory proven, in the Reld the use of a single stage
with many units in parallel has hitherto been world-
wide. Even though the turn-down ratio (comparison
of maximum and minimum usable Sowrates) for an
individual unit is limited } see later discussion of the
inSuence of feed Sowrate on separation } the installa-
tion of units in banks which can be valved out in turn,
allows overall turn-down ratio of about four for two
banks, eight for three banks and so on.
Increased cleaned water throughput of an existing
installation can be achieved by the addition of more
hydrocyclone units.
Apart from modularity and improved separation,
clean-up hydrocyclones have other advantages over
the equipment formerly used such as induced gas
 II / FLOTATION / Cyclones for Oil / Water Separations 1483

Figure 1 Hydrocyclone proportions. (left) Southampton University bi-cone design; (right) Typical later industrial development.

Sotation (IGF) plant or gravity separators with plate
packs. The installed size and weight of hydrocyclones
has only been found to be about 10% of the older
plant. Since hydrocyclones have negligible free sur-
face effects they are insensitive to orientation and
the motion of Soating equipment. Unlike IGF they
have no large requirement for chemicals and their
cleaned underSow water does not require subsequent
processing apart from occasional Socculant addition
to reduce oil content below the legal limit.
As the unit residence time is only 1}2 s the hydro-
cyclones cannot cope with a slug of oil. Any zones of
high shear upstream from these de-oiling hydrocyc-
lones will reduce the size of oil droplets, thus worsen-
ing the separation efRciency. New installations
can largely alleviate this problem by re-design to
re-position Sow-control valves to come after a hydro-
cyclone rather than upstream from it.
Performance
Though variables will often interact, for convenience
their effects are discussed separately.
Separation Ef\ciencies
To achieve the purity required of the underSow
stream from a clean-up unit, its output rate is usually
1484 II / FLOTATION / Cyclones for Oil / Water Separations

Table 3 Factors influencing the size of de-oiling hydrocyclones

Factor Influence

Flowfield Reynolds Number, ufDI/
Pressure drop
Separation of smaller drops*
Avoidance of excessive shear
Cost for a given flow rate
Packing densityH
In early development its perceived influence on separation suggested larger units. Now
ignored
Larger units have higher pressure drop, for the same peak radial acceleration field. Even
though the reservoir pressure is often high enough to remove the need for pumps this factor
tends to limit size
As the hydrocyclone size reduces to a DR of approximately 20 mm improved separation
efficiency of droplets smaller than about 15
m is achieved
For DR values below about 10 mm experimental observation suggests that high shear causes
droplet break-up and consequent loss of separation efficiency
Favours few, larger units if cost for unit alone is considered
Units are normally installed in a pressure vessel. Higher overall throughputs can be achieved
for a fixed volume pressure vessel fitted with many smaller rather than fewer larger units.
This favours reduced cost for complete assembly.

*Generally dominant factors in current practice.

restricted to about 98% of the feed stream rate. This of the valuable product and its purity in the product
can result in the overSow reject stream having a sig- stream . For the clean-up units the fractional
niRcant water content; fortunately the oil droplets in recovery 0 is however deRned by the fraction of oil
this stream coalesce readily causing the oil and water in the feed which appears in the reject (overSow)
phases to separate easily. The oil may then be stream, and u for this operation is deRned in terms of
pumped into an oil storage vessel or delivery line. the oil (waste) content of the product (underSow)
The inability of hydrocyclones to achieve perfect stream:
phase separation in a single step has previously been
noted by Bradley (1965). Qo;Co
For an oil-water hydrocyclone operating in the 0"
Qf;Cf
concentrator mode, the yield of oil in the overSow
product stream is the principal criterion. and
DeRnitions for purity and yield follow below.
The performance of any separation unit is Cu
u"1!
usually deRned in terms of the fractional recovery  Cf

Figure 2 Pressure vessel installation of oil}water hydrocyclones. (One unit shown.)

 II / FLOTATION / Cyclones for Oil / Water Separations
if the split ratio R f and its complement F are deRned
by Rf"Qu /Qf and F"Qo /Qf .
Then a relationship between the fractional recov-
ery parameter 0 and the underSow purity u follows:
Qu;Cu
0"1!}
Qf;Cf
thus 0"1!Rf;(1!u).
One bonus for the de-oiling duty is that the produc-
ed water, i.e. brine, is often warm. Higher temper-
ature is beneRcial as the water viscosity is reduced
and the differential density increased. Possible
problems arising solely from the elevated temper-
ature, with reduced interfacial tension and easier dis-
tortion of oil drops as they are less viscous, seem
unimportant. Thus values of u of 0.99 or even 0.999
have been obtained, even though the differential
density may only be 100 kg m\3 with a mean drop
size 20}30
m.
Generalized prediction of u and 0 is very difR-
cult. Apart from the practical difRculty of predic-
ting the feed drop size distribution, droplet break-up
is inSuenced by interfacial tension. In the complex
liquid mixtures of crude oil and produced water,
interfacial tension is inSuenced by surface-active
agents whose presence and effects are difRcult
to determine.
Correlation of experimental results could ideally
use the Stokes Number (St"2Qfd2/9D3R
,
where "differential density, d"character-
istic droplet diameter,
"viscosity of the continuous
Suid). Strictly, this only applies to dilute dispersions
with Stokesian Sow with a droplet at a Reynolds
Number below unity. The major problem which
tends to restrict usage to laboratory investigations is
in the determination of d. This is usually taken as the
d50 diameter of a droplet that has an equal chance of
reporting to underSow or overSow, but in most ap-
plications it is impractical to measure it.
In]uence of Feed Characteristics
Flow rate At low Sow rates the tangential velocity is
too low to generate an adequate inward radial accel-
eration. This is reSected in the Reld results of Mel-
drum et al. in 1987 which have been converted from
tabular results to the plot in Figure 3. These show
sharply decreasing values of u for a 60 mm DR hydro-
cyclone on the Hutton Reld and for a 35 mm
DR hydrocyclone on the Murchison Reld for feed
Sows below &100 L min\1 and 60 L min\1, respec-
tively.
Figure 3 also shows rapid falls in u values for the
35 mm Murchison unit for feed Sows above
1485
Figure 3 Field results: separation efficiency (u ) vs flow rate
(Q f ). Calculated from Meldrum et al. data.
&200 L min\1 and for the 60 mm Hutton unit
above &425 L min\1, which reSects a split ratio
falling below the critical value. This is discussed fully
later but essentially as Sow rate rises the outlet pres-
sures fall, and as the overSow outlet is at the lower
pressure, a stage is reached where the overSow stream
is so diminished that oil can only leave via the under-
Sow.
The turn-down ratios illustrated by Figure 3 are
3 for the Murchison unit operating on a feed concen-
tration Cf (0.1% and 4.5 for the Hutton unit, but
other Relds with a larger driving pressure available
have achieved values up to about 7. Pumped installa-
tions or oilRelds with lower reservoir pressures might
fall to about 2, for a single unit.
A plateau region separation efRciency is usual.
Higher Sow rates raise the acceleration Reld but,resi-
dence time falls, turbulent re-mixing rises and if inter-
facial tension is low droplet break-up may become
signiRcant. Within the plateau region, rapid transi-
ents in Sow rate should not reduce separation ef-
Rciency appreciably.
Oil}water ratio Over a wide range of oil concentra-
tions clean-up and concentrator hydrocyclones are
considered to act as a Sow divider, i.e. u remains
constant and Cu will rise in proportion to Cf if droplet
size remains constant, however Reld results show this
to give a pessimistic view as droplet sizes rise with oil
content. This means that Cu may even remain con-
stant as Cf rises.
As Cf rises the split ratio may need to be raised also.
For clean-up the value of F is often maintained at
2}3%, but a rational control strategy is discussed
later. In concentrator operations when Cf rises to
1486 II / FLOTATION / Cyclones for Oil / Water Separations
Table 4 Transient oil separation performance
df u % drop (u steady state } u transient)

m Steady-state 5 s injection 2 s injection
50 0.93 0 1
17 0.53 4 6
df"mean feed drop size, BP Forties oil in cold tap water,
Cf steady state &500 mg kg\1.
5}10% the ratio F/Cf may be conservatively main-
tained at 2. This means the overSow stream is 50%
water; for it to be, say, 70% oil (favourable result)
F/Cf falls to about 1.4. A moderate transient increase
in oil content, provided the F/Cf ratio remains satis-
factory, has very little effect on u. Table 4 shows
results for a short-lived oil pulse, obtained with on-
line oil content measurements compared with steady-
state results.
A larger increase in Cf, but still within the accept-
able range, may show a rise in u if droplet size has
also risen. Any further rise in Cf will exceed the
capacity of the overSow and u will fall substantially
when F/Cf falls below about 1.2.
Droplet size Like all separation devices based on
differential density, in a hydrocyclone, reduction
in droplet size will give poorer results. An acceptable
minimum value for df under favourable conditions
may be as low as 5}10
m (elevated temperature,
'150 kg m\2, adequate interfacial tension), how-
ever the usual acceptable minimum of the feed drop-
lets is 15}20
m.
Droplets of 20
m are less vulnerable to low values
of interfacial tension in promoting break-up than
larger ones. However, the presence of surfactants that
drastically lower interfacial tension will almost cer-
tainly reduce the effectiveness of hydrocyclone
separators, as shown by Colman, Thew and Corney
at the First Hydrocyclone Conference (Cambridge,
UK) in 1980.
In laboratory work, grade efRciency curves df vs
u have been obtained as for solid}liquid hydrocyc-
lones, but because of the considerable difRculties
in obtaining representative samples and in measure-
ment such curves are seldom available in the Reld.
When sampling, both isokinetic conditions and avoid-
ance of droplet break-up are necessary and gas bubbles
may complicate interpretation. A suitable technique
was discussed by Colman, Thew and Corney.
Free gas Until fairly recently most clean-up hydro-
cyclones were installed downstream from three-phase
separators. This means that the free gas content of the
hydrocyclone feed was a fairly minor constituent. As
a result any gas core was relatively small in diameter.
Because of the signiRcant axial pressure gradient at
the centre line this gas leaves with the overSow (oil-
rich) stream.
Provided the gas content of the feed is reasonably
invariant with time, laboratory tests have demon-
strated that oil separation is little affected up to
a threshold of 20}30% by volume free gas. This
Rgure relates to conditions at the hydrocyclone entry.
Field experience has generally conRrmed this satisfac-
tory picture except when the gas Sow exhibits signiR-
cant slugging. An entering gas slug does not have the
angular momentum to maintain rotation of liquid
and with the breakdown of inward radial acceler-
ation, separation performance falls sharply. Amongst
the thousands of units in service such a loss of perfor-
mance is uncommon.
One consequence of appreciable gas leaving via the
overSow is that it reduces the area for liquid to an
annulus in the overSow exit port. This changes the
relationship between the pressure drop and Sow rate
for the overSow liquid thus adversely affecting
the control.
With the pressure drop in the hydrocyclone, some
evolution of dissolved gas would be expected. This
does occur but is too slow to be appreciable within
the hydrocyclone and is manifest downstream from
it, being most noticeable in the overSow stream. The
evolving gas has been used to achieve post-cyclone
separation of some more very Rne drops in a suitable
vessel possibly because the evolving gas bubbles nu-
cleate on the oil droplets. In terms of Henry’s Law the
mass of gas coming out of solution in the overSow,
will be proportional to Qo;H (pf!po) where pf is
the upstream feed pressure, po is the downstream
overSow pressure and H is Henry’s constant. How-
ever, the volume evolved will also depend inversely
on the absolute pressure and in any case Henry’s Law
gives a maximum value as it relates to equilibrium
conditions.
Solid particles Crude oil}brine mixtures commonly
contain small amounts of reservoir solids. This prob-
lem is growing as more Relds have larger produced
water contents and it is made worse by the trend to
produce from unconsolidated formations. The
amount may range from a few hundred mg L\1 to
about 10 g L\1 in worst cases. The solids may be
water-wetted or oil-wetted, both usually report to the
underSow unless the oil-wetted solids are very Rne, in
which case they may tend to be neutrally buoyant. In
practice the overSow stream very seldom contains
solids. OverSow blockages are rare and when they do
occur they are usually associated with debris left in
the system at installation or after maintenance.
 II / FLOTATION / Cyclones for Oil / Water Separations
Erosion, if it occurs, is usually restricted to the inlet
region where velocities are higher. The use of harder
materials, for example Stellite, in the inlet region has
allowed long periods of satisfactory operation.
A number of installations have run continuously for
Rve years or more.
A development of the last 2}3 years is the arrival of
de-sanding hydrocyclones installed ahead of the de-
oiling units. These have been used prior to the choke
with a containment vessel to withstand the very high
pressure. Both relatively large units in appropriate
steels and smaller ceramic units are entering service.
In off-shore Relds the water in the oil}water
mixtures may be quite corrosive, particularly if it is
sour (containing H2S). This has necessitated the use of
alloy steels in the fabrication of the clean-up hydro-
cyclones. However, particularly for low cost, low
Sow land-based installations, cheaper materials may
be satisfactory. Polyurethane units, possibly in a car-
bon steel casing or even bare are on the threshold of
commercial usage.
Pressure Drop (the Cost Implications for
Separation)
As in solid}liquid hydrocyclones, pressure drop
p&Qm f where 2.1(m(2.2. In dimensionless
terms deRning:
p
Euler Number NEu"
 u2f

ufDi
Reynolds Number NRe"
v control layout is shown on Figure 4. The effec-
where uf"Qf/Ai and the inlet port diameter"Di
(for multiple or noncircular inlets Di is the diameter
of the circular port with equal Ai).
Leads to the dimensionless relationship
NEu&NnRe
where
0.1(n(0.2
The use of uf in terms of Qf and Ai will be reSected by
changes in the swirl number S.
Values of  and v are based on arithmetic averages
of the mixed liquids. Uncertainty in the value of the
kinematic viscosity v is not serious due to the small
value of n. Typical Euler Number values are in the
range 10}20 and do not seem to be affected by
variations in Cf.
The de-oiling hydrocyclones have a substantial
pressure gradient along the hydrocyclone axis not
present in solid}liquid hydrocyclones with an air
core. Thus pressure drop between the feed and the
1487
overSow pfo is greater than the pressure drop be-
tween the feed and the underSow pfu, the relation-
ship between the two being variable and set by ex-
ternal valves. The ratio between the two is important
for control and optimization and this pressure drop
ratio (PDR"pfo/pfu) varies with split ratio, Rf and
also its complement F. pfo has two principal compo-
nents } one due to the radial pressure gradient and the
other arising from the velocity through the overSow
port. The radial pressure gradient and pfu are both
proportional to the u2f and S, since uf"Qf/Ai. For
a Rxed geometry the PDR"B1#B2;F2 where
B1 and B2 are constants, where B1"f1(S) and B2"f2
1
(S, A\0 ). The oil concentrated overSow rate is regu-
lated by using a Rxed PDR value as set-point.
Pumped installations In younger oilRelds the driv-
ing pressure stems from reservoir pressure. In older
Relds where pumping is necessary, this represents
a cost. To reduce droplet break-up a low shear posit-
ive displacement pump should be used, however cen-
trifugal pumps have proved satisfactory provided
their speed is not too high and the duty is not too far
from their best efRciency point. Fields with elec-
trical submersible pumps have utilized de-oiling hy-
drocyclones satisfactorily.
Operator Control
For water clean-up units the PDR set-point is usually
in the range 2}3 but for concentrators, with their larger
overSows, values below 1.5 may be set. A simpliRed
tiveness of the control is restricted to some extent by the
relative sizes of the underSow and overSow exit ports.
The PDR is a weak function of the feed Reynolds
Number (NRe"uf Di/v), see Figure 5. This depend-
ence is usually ignored.
Critical split ratio (FCRIT) At low oil contents it is
desirable to reduce F thereby reducing the amount of
the overSow stream. But below a critical value
FCRIT the axial pressure gradient near the centre line is
inadequate to sustain the reverse Sow. Though oil
still gathers in the vortex core it is unable to reach the
overSow outlet, becomes remixed near the inlet and
leaves via the underSow so u falls sharply. The ef-
fect is illustrated on Figure 6. FCRIT rises with an
increase in the diameter Do of the overSow port and is
therefore higher for concentrator units. Though
extremely small overSow outlets will permit very
small FCRIT values, they are impractical as the smallest
upset is too much for them to handle.
Table 5 summarizes the factors to be considered in
control of the split ratio via the easily measured
1488 II / FLOTATION / Cyclones for Oil / Water Separations
Figure 4 Simplified control layout.
variable, PDR. Hydrocyclone manufacturers will ad-
vise on the size of the overSow outlet. It may require
enlargement during the life of an oilReld.
Cost Comparisons
Cost data are sparse but some information produced
by BP about Rve years ago has been recast on Table 6
in terms of ratios. The Sow rate used for deriving the
information was applicable to a Reld producing about
16 000 m3 per day of oil.
Figure 5 Pressure drop ratio (PDR) vs inlet Reynolds Number.
Southampton University laboratory results and field results de-
duced from the data of Meldrum et al.
Table 6 shows that both the capital and running
costs increase in the sequence hydrocyclone-IGF-Rl-
ter/coalescer-centrifuge. Hydrocyclones also produce
the most compact plant and are the least sensitive to
orientation or lateral acceleration as encountered in
installations on a Soating base.
Future Developments
In 1979 an oil industry task force recommended plate
pack gravity separators or IGF for produced water
clean-up. Six years later the Rrst large hydrocyclone
installation } about 15 m3 min\1 } was operating
successfully in the North Sea, so the points below
relate only to the immediate future, perhaps prior to
2003.
E Improving oil separation (in hydrocyclones) means
removing smaller drops. There is probably limited
scope for further optimizing geometry, though the
claims for computational Suid mechanics (CFD) in
rapid optimization are likely to become valid in
a year or two. (Adequate representation of turbu-
lence in conRned swirling Sow has proved difR-
cult.)
E For new systems and even for some retro-Rts, there
is often room for worthwhile improvement by
reducing shear upstream of the hydrocyclones or
re-locating them. The resulting larger drops ease
the separation task.
 II / FLOTATION / Cyclones for Oil / Water Separations 1489

Figure 6 Separation efficiency ( u ) vs complement split ratio (F ). The collapse in separation consequent on too low an overflow is
shown.

E If the split ratio needs frequent adjustment, a vari-
able area overSow (two values would be adequate)
simply controlled would be very useful. Laboratory
tests have shown that this is readily feasible but
moving parts are still not readily accepted, as they
are associated with reduced reliability.
E As the cost of dealing with produced water
mounts, economic/technical studies have strongly
suggested hydrocyclone installation at the bottom
Table 5 Factors affecting PDR
of the wellbore, ‘downhole’. If the geology is suit-
able the water could be re-injected in adjacent
strata. Preliminary trials using electric submersible
pumps (ESP) (some concepts suggest two) have
shown and are showing considerable promise. Ac-
cess after installation is costly so there are prob-
lems in control and reliability, though ESP
speed variation is proven. Reduced standards of
separation may still be acceptable. A halfway stage
to use pre-choke hydrocyclones on the sea bed is
also being investigated.

Factor Correction Comment

to PDR
Table 6 Relative cost for clean-up plant (16 000 m3 d\1)
F(FCRIT IncreaseH Oil is lost in underflow, uP0
F/Cf(2 IncreaseH Some oil lost to underflow Plant type Capital Running Oil separation
(conservative criterion) cost cost performanceR
Excessive free gas IncreaseH Effective overflow area for
in underflow liquids is reduced. Condi- HydrocyclonesH 1.0 1.0 3
tion may be difficult to Induced gas flotation 1.1 1.1 4
detect Filter/coalescer 1.3 1.6 2
F too high Reduce p fo excessive Centrifuge 4.6 4.0 1
F too high Reduce Too much water in overflow
stream HCapital cost is 6}7; annual running cost based on data from
several North Sea fields with installations &10 years old.
HOpening the valve in the overflow raises the flowrates hence R Ranking order for oil content in cleaned flow with identical oil
increasing PDR. content and mean drop size in the feed; 1 is best.
1490 II / FLOTATION / Dissolved Air
E De-sanding hydrocyclones are now being installed
upstream from the choke in some Relds. There
could be an energy saving if de-sanding and de-
oiling could be performed in a single unit, but
simultaneous optimization of both functions is un-
likely. A successful laboratory research project has
been reported in France, but initial Reld trials in
West Africa were disappointing.
E Heavy oils, i.e. those with a higher density and
viscosity, appear unpromising for cyclonic separ-
ation processes. Nevertheless success has been re-
ported for commercial de-oiling units used in the
concentrator mode in trials in western Canada.
E Feasibility studies and preliminary Reld trials are in
progress on integrated de-watering plus de-oiling
cyclonic separation and/or de-sanding plus de-oil-
ing. The attraction is an ultra-compact plant suited
particularly to Soating installations. Though de-
watering units have not yet met with widespread
success, the impending arrival of compact, robust
electrocoalescers to raise water droplet size prior to
separator entry, could transform the situation.
E With success in dealing with petroleum it is surpris-
ing that applications to edible oils, which are about
10 times more valuable, have yet to materialize.
Laboratory trials have been very satisfactory. Not
only is lost oil a revenue drain but it generates
a potential environmental hazard.
Further Reading
Note: The Rve conferences on hydrocyclones all contain
several papers on oil}water hydrocyclones
Dissolved Air
D. Shekhawat and P. Srivastava,
Michigan State University, East Lansing, MI, USA
Copyright ^ 2000 Academic Press
Introduction
Dissolved air Sotation (DAF) is a solid}liquid separ-
ation process for the removal of Rne suspended
material from an aqueous suspension. The basic prin-
ciple underlying DAF is Henry’s law, which gives
the solubility of air in water. According to Henry’s
law, the solubility of air in water is directly propor-
tional to its partial pressure. A supersaturated solu-
tion of water is produced using high pressure in
a saturator. The bubbles are generated by the pressure
release of this water stream. These bubbles attach to
First Hydrocyclone Conference, Cambridge (UK) (1980)
Priestley G and Stephens HS (eds). CranReld: BHRA.
Second Hydrocyclone Conference, Bath (UK) (1984) Watts
GA and Pickford R (eds). CranReld: BHRA.
Third Hydrocyclone Conference, Oxford (UK) (1987)
Wood P (ed.). CranReld: Elsevier-BHRA.
Fourth Hydrocyclone Conference, Southampton (UK)
(1992) Svarovsky L and Thew MT (eds). Dordrecht:
Kluwer.
Hydrocyclones 96 Conference, Cambridge (UK) (1996)
Claxton D, Svarovsky L and Thew MT (eds). London:
MEP.
Vortex Separation: Fifth International Conference on Cyc-
lone Technologies, Warwick (UK) (2000) Svarov-
sky L and Thew MT. Organised and published by
BHR Group, CranReld.
Meldrum N (1987) Hydrocyclones: a solution to produced
water treatment. Proceedings of the 19th Annual
Offshore Technology Conference, Houston, Texas,
USA.
Paige R and Ferguson M (1993) Water injection:
practical experience and future potential (A BP
Study). Conference on Offshore Water and Envir-
onmental Management. Business Seminars Interna-
tional, London.
Smyth IC and Fay B (1998) Further developments of the
HydrosepTM System for downhole oil/water separation.
Conference on Downhole Production and Subsea Pro-
cessing, Aberdeen. Organised by BHR Group, London:
MEP.
Svarovsky L (1984) Hydrocyclones. London: Holt,
Rinehart and Winston.
Young GAB, Wakley WD et al. (1994) Oil-water separ-
ation using hydrocyclones: an experimental search for
optimum dimensions. Journal of Petroleum Science 11:
37}50.
suspended material present in the aqueous stream,
causing them to Soat to the surface, where they are
collected as Soc.
DAF can be carried out by vacuum or pressurized
methods. In the vacuum Sotation method the water
to be treated is saturated with air at atmospheric
pressure. The bubbles are produced by applying a
vacuum to the Sotation tank, releasing the air as Rne
bubbles. The vacuum Sotation process has several
disadvantages. These are (a) the amount of air avail-
able for Sotation is limited by the vacuum achievable,
(b) it is a batch process, and (c) it requires special
equipment to produce and to maintain high vacuum.
These disadvantages limit the application of vacuum
Sotation and it is only used in wastewater sludge
thickening.

The pressure Sotation process is the most widely
used DAF technique. High pressure water is saturated
with air. This pressurized water forms small bubbles
when injected into water at atmospheric pressure.
Three types of pressurization processes can be used
in DAF: full Sow, partial Sow and recycle Sow
pressurization. The entire inlet stream is pressurized
in full Sow pressure DAF. It is commonly used
when the wastewater contains large amounts of sus-
pended solids and the pressurization process does
not affect the treatment efRciency of the sys-
tem. Partial Sow pressurization is used where
the wastewater contains moderate to low concentra-
tions of suspended solids. In the recycle Sow pressur-
ization system, 10}25% of the clariRed efSuent
is recycled through a pressure vessel to the
Sotation tank. The Socculation process in not
disturbed in the recycle Sow system because of in-
tense mixing and pressurization as clear water is
pumped. A recycle Sow system is cost-efRcient
because it pressurizes only part of the water, thus
requiring less compressor power. Recycle Sow pres-
sure Sotation is the best-suited system for most DAF
applications.
DAF is an effective alternative to sedimenta-
tion. The advantages and disadvantage of DAF rela-
tive to sedimentation are as follows:
Advantages
1. ClariRcation rates are higher in DAF, resulting in
smaller Socculation tank volumes.
2. More concentrated sludge solids are produced in
DAF than from sedimentation.
3. DAF uses lower amounts of coagulants and Soccu-
lent aids.
4. Oxygenation effects in DAF reduce odour
problems.
5. DAF provides better removal of low density par-
ticles and algae, which can plug Rlters.
Figure 1 Schematic diagram of the dissolved air flotation process for water treatment.
II / FLOTATION / Dissolved Air 1491
Disadvantage
1. DAF processes are more costly to operate and
maintain than sedimentation processes.
Process Description
A schematic diagram of a DAF process for wastewater
treatment is shown in Figure 1. Its essential elements
are a Socculation tank, a Sotation tank, an air com-
pressor, an air saturator, a recycling pump and a hy-
drosweep system. The wastewater is pumped to the
Socculation tank after being treated with coagu-
lant/Socculent agents such as aluminium sulfate.
A portion of the clariRed efSuent is recycled for
pressurization. Compressed air is introduced into the
discharge stream of the recycle pump, and the water is
saturated with air at high pressure. The pressurized
water stream is introduced to the Sotation tank
through nozzles, where Rne bubbles (20}100
m) in
diameter are formed. The bubbles attach themselves to
suspended solid particles, causing the agglomerates to
Soat to the surface of the tank. The Soat can be mechan-
ically skimmed from the surface, and the clariRed water
is taken from the bottom of the Sotation tank.
Principles of Dissolved Air Flotation
DAF facilities are composed of the following four
principal steps:
1. coagulation and Socculation prior to Sotation
2. bubble generation
3. bubble}Soc collision and attachment in the mixing
zone
4. rising of the bubble}Soc aggregates in a Sotation
tank
Coagulation and Flocculation Prior to Flotation
Coagulation and Socculation are often considered as
a pretreatment step in DAF processes. Favourable
1492 II / FLOTATION / Dissolved Air
conditions for bubble attachment to particles requires
coagulation conditions that reduce particle charge
and produce hydrophobic particles. Coagulant dos-
ages and pH conditions that satisfy these criteria
depend on the coagulant type and raw water charac-
teristics, including particle concentration, hardness,
and concentration and type of natural organic matter
(NOM). Unlike in sedimentation, large Soc particles
are not needed in DAF. Flocculation tanks are de-
signed to produce strong Socs with particle size distri-
butions of 10}30
m and short Socculation times, in
the range of 10}15 min.
Bubble Generation
Small air bubbles, 100
m or less, are formed by
injection of supersaturated pressurized recycle water
into a Sotation tank using specially designed nozzles.
The process of bubble formation involves two steps:
nucleation and growth. During the Rrst step the large
pressure difference across the nozzle produces
bubble nuclei spontaneously. Air bubbles grow at
a Rxed number of nucleation centres due to air trans-
ferred from the water. As the excess air is transferred
from the dissolved to the gas phase, the bubbles grow
in size. Additional bubble growth may occur as the
bubbles rise due to a decrease in hydrostatic pressure
or coalescence.
Measurements of bubble sizes for DAF systems
indicate that bubbles maintain a steady-state size
range of 10}100
m. A reasonable estimate of aver-
age bubble diameter is 40
m. The steady-state size
depends on the saturator pressure and the injection
Sow rate. The injection Sow must provide a rapid
pressure drop and be sufRcient to prevent back-
Sow and bubble growth on pipe surfaces in the vicin-
ity of the injection system. To ensure small bubbles,
pressure differences (saturator gauge pressures) of
400}600 kPa are recommended.
Bubble Floc Collision and Attachment in
the Mixing Zone
There are three possible mechanisms for forming ag-
gregates of bubbles and particles:
1. entrapment of preformed bubbles in large Soc
structures (Soc size much larger than bubble size
scale)
2. growth of bubbles whose nuclei formed on par-
ticles or within Socs
3. particle collision with adhesion to preformed
bubbles
For DAF processes, the third mechanism is the most
important.
Classically, the contact angle between the adsorbed
bubble and particle has been used to characterize the
extent of bubble}Soc adhesion. Here the contact
angle must be Rnite and large enough that the energy
of adhesion of water to the solid particle is less than
the energy of cohesion of water. A larger contact
angle indicates both hydrophobicity and good ad-
hesion. The magnitude of the contact angle, however,
depends on the size of the bubbles and particles.
A different view of particle}bubble attachment
of colloidal particles by small bubbles is that a Rnite
contact angle need not form. In this heterocoagula-
tion model, the stability of charged particles and
bubbles is described. Attachment requires reduction
in electrical charge interactions and attraction by
London}van der Waals forces as particles are trans-
ported to bubble surfaces.
Both the contact angle and the heterocoagulation
models predict experimentally observed trends that
two conditions are necessary for favourable Sotation:
charge neutralization of the particles and production
of hydrophobic particles. Bubble attachment to par-
ticles requires hydrophobic particle surfaces or hy-
drophobic regions on the particles. For many par-
ticles, hydrophobicity can be increased by reducing
the negative charge. Other particles, such as freshly
precipitated or amorphous Al(OH)3, have polar
surface groups that make them hydrophilic. This
hydrophilic effect may be reduced by charge
neutralization, but aluminium hydroxide particles
have a polymolecular coating of water which hinders
bubble adhesion.
Rising of the Bubble+Floc Aggregates in
a Flotation Tank
Following bubble attachment and reduction in
particle density, particle}bubble agglomerates rise
to the surface of the Sotation tank in the separation
or clariRcation zone. The rise velocity of the
particle}bubble agglomerate may be calculated using
Stokes law.
DAF Modelling
The design and operation of DAF facilities has largely
been based on experience and results from pilot-plant
studies. In recent years, a conceptual model of DAF
has been developed, based on a single collector colli-
sion theory in laminar Sow conditions (SCC model).
A kinetic model has also been presented, based on the
population balance model of bubbles and Socs in
a turbulent Sow condition (PBT model).
For modelling purposes, DAF processes can be
divided into two zones: reaction zone (regions where
 II / FLOTATION / Dissolved Air 1493

the saturated recycle Sow is introduced) and separ-
ation zone.
Reaction Zone Modelling
For the reaction zone efRciency (dNfl/dt), deRned
as the reduction of number of primary particle Socs
with time, Soc and bubble size (dfl and db) and con-
centration are deRned as relevant process parameters:
dNfl/dt"!(3/2)(
pb T)(bvb Nfl)/db
where
pb"particle bubble attachment efRciency;
T"total single collector efRciency; db"bubble
diameter; vb"bubble rise velocity; Nfl"Soc number
concentration; b"bubble volume concentration.
Particle (Soc)}bubble interception is considered to
be the most relevant kinetic mechanism for DAF
efRciency, depending on the Soc and bubble size
(dn and db) and incorporated in T. This term also
considers Soc}bubble collision mechanisms related to
Brownian diffusion, settling and drag.
A summary of these parameters, their dependence
on the system variables and desirable operational
conditions is given in Table 1.
Separation Zone Modelling
Assuming laminar Sow conditions, the efRciency
of the separation zone, vflb (deRned as Soc}bubble
agglomerate rising velocity), is deRned by the
Soc}bubble agglomerate size (dflb) and density (flb)
(eqn [2]). These are dependent on the Soc size}density
ratio and concentration, and the size and number of
bubbles comprising the Soc}bubble agglomerate:
vflb"gdflb(pw!pflb)/18  [2]
Eqn [3] represents a necessary prerequisite for ef-
Rcient DAF, vos being the DAF overSow rate and
[1]
m being the fraction of DAF tank dead space:
vflb'vos /(1!m) [3]
Applications of Dissolved Air Flotation
DAF is best applied to remove materials that normal-
ly settle slowly, persist in remaining in suspension or
have a tendency to Soat. Prior to the 1960s it was
mainly utilized in the area of mining and metallurgi-
cal industries. Now, DAF Rnds numerous applica-
tions, e.g. mineral processing, water puriRcation,
wastewater treatment, waste sludge thickening,
wastewater reclamation, recycled paper de-inking,
and many more. It is widely used for drinking water
puriRcation in many Scandinavian countries, South
Africa, the Netherlands, the UK and others. In drink-
ing water clariRcation, DAF has been applied in
combination with Socculation for the removal of

Table 1 Summary of conceptual reaction zone model parameters

Parameter Dependence Comments

Pretreatment parameters

pb (particle}bubble attachment efficiency)
Np (particle number concentration)
Reaction zone}flotation tank
T (total single collector efficiency)
d b (bubble diameter)
1. Particle}bubble charge interactions 1. Favourable flotation; requires
reduction in particle charge and
hydrophobic particles
2. Hydrophilic nature of particles 2. Increase
pb to 1: optimum coagulation
and charge neutralization
1. Raw water quality 1. Concentration and size of particles;
concentration of NOM
2. Coagulant type and conditions 2. Coagulants may add particles
3. Flocculation time 3. Flocculation may reduce Np and
increase dp
1. Particle}bubble collisions from diffusion 1. Increase T : produce floc size of
and interception tens of
m
2. Minimum T for dp of &1
m 2. Short flocculation times
Controlled by pressure difference across 1. Desire microbubbles: range
nozzle and injection flow 10}100
m, median 40
m;
smaller bubbles: better performance
2. T varies as d b\2; rate of collection of
particles varies as d \b
1

b (bubble volume concentration 1. Saturator pressure 1. Increasing b increases Np: more

bubbles for collection of particles
2. Recycle ratio 2. Increase b: more bubble volume for
reducing floc density

Reproduced with permission from Edzwald (1995).

1494 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions
algae and humic substances. The Rrst water treatment
plant based on the DAF process was established in
South Africa in 1969. Since then it has received
worldwide attention for research and development on
all aspects of DAF.
The Rrst DAF plant in the USA was set up at the
Millwood water treatment plant in Westchester
county (35 miles north of New York city) in August
1993. Now, several other plants based on DAF are
operating or are under study in the USA. It is postu-
lated that DAF is an emerging technology in the USA
that will become more important because of existing
and proposed regulations that require Rltration of
surface waters and increased removal of protozoa
cysts such as Cryptosporidium and Giardia. Large
scale pilot-plant trials of water treatment have been
carried out in the UK for removal of Cryptosporidium
using DAF. Well-operated chemical coagulation-
based treatment using DAF should be capable of
achieving 99% removal of Cryptosporidium
oocysts.
DAF is also used in the forest industry, food-
stuff industry, meat-processing industry, seafood
industry, potato processing, pulp and paper industry,
petroleum industry, poultry industry, producing re-
Rned sugar from raw juices, separation of grease, oil,
Rbres and other low density solids, chemical process-
ing plants, storm water cleaning, and other similar
industries.
Future Trends
There is great potential for DAF. Its use has been
limited due to lack of knowledge of the process by
users, designers and other regulatory agencies. The
Electrochemistry: Contaminant Ions and Sul\de Mineral
Interactions
J. T. Smit and J. Gnoinski,
Anglo American Research Laboratories (Pty) Ltd.,
Johannesburg, South Africa
R. F. Sandenbergh, University of Pretoria,
Pretoria, South Africa
Copyright ^ 2000 Academic Press
Introduction
Mineral separation by Sotation is based on the selec-
tive levitation and separation of mineral particles by
gas bubbles. This is carried out by the selective con-
version of the surfaces of the minerals to be Soated
design and operation of DAF methods are currently
tested on empirical data and data from costly and
time-consuming pilot-plant models. More informa-
tion is needed on the performance, designs and costs
of the DAF process.
See also: I/Flotation.
Further Reading
Derjaguin BV, Dukhin SS and Rulyov NN (1984) Kinetic
theory of Sotation of small particles. In: Matijevic E and
Good RJ (eds) Surface and Colloid Science, vol. 13, New
York: Plenum Press, pp. 71}113.
Edzwald JK (1995) Principles and application of
dissolved air Sotation. Water Science and Technology
31: 1}23.
Edzwald JK, Malley JP and Yu C (1991) A conceptual
model for dissolved air Sotation in water treatment.
Water Supply 9: 141}150.
Fukushi K, Tambo N and Matsui Y (1995) A kinetic
model for dissolved air Sotation in water and waste-
water treatment. Water Science and Technology 31:
37}47.
Hall H, Pressdee J, Gregory R and Murray K (1995)
Cryptosporidium removal during water treatment using
dissolved air Sotation. Water Science and Technology
31: 125}136.
Ives KJ and Bernhardt HJ (eds) (1995) Flotation processes
in water and sludge treatment. Water Science and Tech-
nology 31.
Kitchener JA and Gochin RJ (1981) The mechanism of
dissolved air Sotation for potable water: basic analysis
and proposal. Water Research 15: 585}590.
Takahashi T, Miyahara T and Mochizuki H (1979) Funda-
mental study of bubble formation in dissolved air pres-
sure Sotation. Journal of Chemical Engineering, Japan
12: 275}280.
from their typical hydrophilic nature to hydrophobic,
to which the gas bubbles may attach to effect the
levitation. This conversion is usually achieved by the
selective attachment of collectors to the surface of
the mineral or by natural processes; an example
of the latter is the formation of elemental sulfur or
a metal-deRcient sulRde layer on sulRdes.
Typical collector agents are organic substances con-
sisting of an ionic, i.e. hydrophilic, end that attaches to
the mineral and a nonionic hydrocarbon end that
creates the hydrophobicity of the mineral surface.
A widely used collector in selective sulRde Sotation is
the xanthate ion (O-alkyldithiocarbonate, ROCS2).
 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions
Most of the sulRde minerals are electronic semicon-
ductors or electrically conductive. This implies that
the reactions required for the creation of a hydro-
phobic surface on sulRdes may be electrochemical.
For example, the xanthate ion may be oxidized at an
anodic area of a local cell on the mineral surface to
form hydrophobic dixanthogen:
2 P(ROCS2)2#e\
2ROCS\ [1]
The corresponding reduction reaction is the catalytic
reduction of oxygen. Apart from this type of electro-
chemical reaction, electrochemical interaction be-
tween dissimilar sulRde minerals of different rest
potentials occurs when there is electrical contact
between them in a sufRciently conductive electro-
lyte, and this is of a galvanic nature. For contact of
this nature to occur the various sulRdes must be either
present in composite particles, i.e. middlings, or be
brought into such frequent collision contact in the
Sotation pulp, that signiRcant electrical charge
transfer can take place. In the laboratory these
phenomena can be studied by electrically connected
mineral electrodes, or stirred/Suidized beds made
up of sulRde minerals only. Although these
laboratory methods indicated signiRcant electro-
chemical interaction, studies conducted at the
authors’ laboratories on simulated mineral feed of
realistic plant composition showed no signiRcant
charge transfer. This was possibly due to the com-
plexity of the Sotation pulp chemistry, which makes
it difRcult to distinguish between the contribution
of chemical and electrochemical processes to overall
plant performance.
Theory and Principles of
Electrochemical Interaction between
Mineral Species
Reactions on mineral surfaces in which there is
a change of oxidation state for the species involved
are generally electrochemical in nature. This adds to
the complexity of multi-mineral systems, in the sense
that, apart from interactions through a common
aqueous phase, by for instance dissolution}precipita-
tion reactions, galvanic interactions through electri-
cal contact between minerals must be considered also.
Galvanic interactions between minerals will cause the
more inert mineral to act predominantly as cathode,
and reduction of dissolved oxygen would typically
occur on its surface. This will stimulate anodic
counter-reactions such as the oxidation of xanthate
to dixanthogen or metal xanthates and that of metal
sulRdes to metal-deRcient sulRdes or elemental sulfur,
1495
all of which will promote hydrophobicity on the
surface of the more reactive mineral.
Similarly, contact of a less reactive sulRde mineral
particle with more reactive steel, generated in abund-
ance by industrial grinding operations, in the form of
loose particles or layers smeared onto the mineral
surfaces, may depress the mixed potential of the
galvanic couple to such an extent that the oxidation
reactions necessary for the hydrophobization of
the sulRde surface will be slowed down. In extreme
cases, reactions may become thermodynamically
impossible.
In the following sections the role of electrochemical
interactions in the separation of complex sulRde ores
will be further explored. Particular attention will be
paid to the use of pulp potential as a monitoring and
control tool and the role of electrochemical reactions
in the development of hydrophobicity.
Mixed Potential Theory
The interaction between collector reagent and min-
eral takes place at the mineral}solution interface. For
ease of reference, and in view of the vast body of
investigative work done on it, our discussion will
focus on the use of xanthates as collector agents. In
the case of sulRde minerals, which are generally
semiconductors, the interactions with xanthate col-
lectors involve charge transfer across the electrical
double layer at the solid}liquid interfaces. Woods et
al. suggested three ways by which the xanthate ion
could confer hydrophobicity to a mineral surface.
Firstly the anodic reaction leads to dixanthogen
formation (see eqn [1]). Secondly they distinguish
between dixanthogen produced by the anodic reac-
tion of the xanthate ion and the xanthate ion adsorp-
tion at a lower potential which is held by electrostatic
attraction:
C2H5OCS2P(C2H5OCS2)ads#e} [2]
Finally there is chemisorption for which the anodic
oxidation of the xanthate ion on lead sulRde is:
PbS#2ROCS2PPb(ROCS2)2#S0#2e\ [3]
The corresponding cathodic reaction typically re-
quires the reduction of oxygen in industrial Sotation
systems. If it assumed that the process is Faradaic in
nature, i.e. no charge accumulation can occur, the
oxidation and reduction reactions will be coupled by
the Sow of charge. The rate of the electrochemical
reactions can now be determined by considering the
driving force available for the process and the kinetics
of the individual processes. If the respective electrical
1496 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions

and ionic resistance of the mineral and solution is Table 1 Rest potentials at various dissolved oxygen contents
low, the system will with time reach a common po- (Reproduced from Cheng and Iwasaki (1992) with permission
Copyright Gordan and Breach Publishers.)
tential called the ‘mixed potential’ at which the indi-
vidual reactions will take place at steady state. This Mineral Rest potential (V vs SHE) Range reported
will typically be the case for solutions with a high salt in 6.25;10\4M KEX at 0}7 ppm O2
loading and with anodic and cathodic areas in close solution
proximity. The situation may be further complicated Mild steel !0.515 to !0.255
by the involvement of more than two half cell reactions Sphalerite !0.15
and also by cathodic and anodic areas of varying sizes Stibnite !0.125
as is typically the case in galvanic interactions. Realgar !0.12
Orpiment !0.10
The mixed potential theory has been used to
Antimonite !0.09
account for the collectorless Sotation of sulRde min- Covellite #0.05
erals such as chalcopyrite, by considering the contri- Bornite #0.06
butions of both surface oxidation and oxygen reduc- Chalcocite #0.06
tion reactions to the common potential. For example, Chalcopyrite #0.14 0.115}0.355
Galena #0.14 0.142}0.172
Trahar has shown that surface oxidation of sulRde
Molybdenite #0.16
minerals results in the formation of hydrophobic sul- Pyrrhotite #0.21 0.055}0.290
fur layers and thus enhances Sotation. It has been Pyrite #0.22 0.389}0.445
suggested that for sulRde minerals, surface oxidation Arsenopyrite #0.22 0.277}0.303
involves the progressive removal of metal atoms,
leaving a hydrophobic, metal-deRcient sulRde layer
with a crystal lattice only marginally altered from the potential of a mineral surface, is the potential asso-
original structure. More recent studies by Buckley ciated with a Rnite reaction rate in a speciRc solution
and Woods, using X-ray photoelectron spectroscopy, environment (see Table 1). This is illustrated by Fig-
have conRrmed that sulfur species are indeed formed ure 2 from the work of Gardner and Woods, which
on the mineral surfaces. The concepts are sum-
marized in Figure 1.
In the presence of xanthate collector conditions for
the formation of dixanthogen have been shown by
Allison et al. to occur when the rest potential of the
mineral is greater than the reversible potential of the
xanthate/dixanthogen couple ER(0.13V at pH 7.0),
which for these minerals is the active collector species
in xanthate-based Sotation, except for galena, where
the metal xanthate was indicated, as discussed by
Cheng and lwasaki (see Further Reading). The rest

Figure 2 Pyrite electrode at 253C in 0.05 M Na2B4O7 solution

Figure 1 Generalized depiction of galvanic interaction between
electrically connected particles.
(pH 9.2) containing 1000 ppm of three potassium alkylxanthates.
(A) Cyclic voltammograms at 4 mV s\1; (B) Contact angles mea-
sured after holding the electrode at each potential for 30 s. The
vertical lines are the Er values for the xanthates. (Reproduced
with permission from Gardner and Woods (1977) Copyright
CSIRO Publishing.)
 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions
Figure 3 Galena electrode at 253C in 0.05 M Na2B4O7 solution
(pH 9.2) containing 1000 ppm of three potassium alkylxanthates.
(A) Cyclic voltammograms at 4 mV s\1; (B) contact angles mea-
sured after holding the electrode at each potential for 30 s. The
vertical lines are the Er values for the xanthates. (Reproduced
with permission from Gardner and Woods (1977) Copyright
CSIRO Publishing.)
clearly indicates that of pyrite hydrophobicity only
develops at potentials more noble than the reversible
potential for xanthate/dixanthogen reaction. For
galena the response shown in Figure 3 is somewhat
different, with a signiRcant current Sow occur-
ring below ER due to the contribution of the
chemisorbed reaction. It is also interesting to observe
a zero contact angle at !0.2V.
Galvanic Interaction between
Sul\de Minerals in a Pulp
As indicated earlier, galvanic interactions arise be-
tween two or more dissimilar minerals, and/or metals
that are in electrical contact with each other and with
an electrolyte. Electrochemical reactions at the min-
1497
eral surfaces result in coupled current and ion Sows.
The cathodic reaction is generally the reduction of
oxygen to hydroxide, while the oxidation reaction
involves the oxidation of the sulRde mineral. The
current Sow depends on the surface area and con-
ductivity of the mineral as well as the chemical com-
position of the electrolyte.
Minerals can only be separated by Sotation if they
are physically separate, i.e. liberated from each other.
Short periods of galvanic contact between sulRde
particles are unlikely to result in the development of
the longer-term hydrophobicity that would be re-
quired for Sotation. Polarization studies by Gardner
and Woods on lead sulRdes have indicated that the
formation of hydrophobic substances, in this case lead
xanthate, is reversible and thus unlikely to endure long
enough for bubble contact to be established. In the
context of the selective Sotation of sulRdes it would be
the middlings, where the different sulRdes would
still be in physical contact, that would be the most
inSuenced by galvanic interactions.
In the case of middlings it is possible that the
Soatability may even be better than that of pure
minerals, due to the greater spatial separation and
electric potential differentiation of the anodic
and cathodic sites on such composite particles com-
pared to single mineral particles. The possibility for
spatial separation will increase with increasing con-
ductivity of the solution and will be more important
in solutions of high salinity.
As an indication of the galvanic interactions that
may develop between different sulRdes, a list of
rest potential values has been reproduced in Table 1.
The rest potential values mentioned were determined
at near neutral pH values and will generally decrease
with increasing pH. Because of this effect, many
sulRdes may be depressed by an increase in pulp pH,
as their potentials move further away from the dixan-
thogen/xanthate equilibrium potential.
This disregards any chemical changes that may
occur on the mineral surfaces due to a rise in alkalin-
ity. High pH conditions typically develop at the cath-
odic sites, which favour the precipitation of metal
hydroxides and would encroach on the anodic reac-
tion site if the spatial separation of the sites is not
large.
Consider the Sotation of a middlings particle con-
taining chalcopyrite and pyrite. In the absence of
a xanthate collector, pyrite acts as a cathode of the
local pyrite-chalcopyrite cell. Oxidation of the chal-
copyrite surface is the predominant reaction balanced
by the corresponding reduction reaction on the pyrite
surface.
Buckley and Woods demonstrated that the col-
lectorless Soatability of chalcopyrite and pyrite
1498 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions

middlings particles increases with the amount of
quartz added. This was attributed to the adsorption
of hydrophilic iron hydroxides from the sulRde min-
eral surfaces on the quartz surface.
The uptake of xanthate ion strongly depends on the
rest potential of the sulRde mineral. For sulRde min-
erals with rest potentials above #0.13V, xanthate
ions are oxidized at the mineral surface to dixantho-
gen, which imparts hydrophobicity to the mineral
surface. For bornite and chalcocite, whose rest poten-
tial was below that of the xanthate/dixanthogen re-
versible couple, metal xanthate was identiRed. For
sphalerite and stibnite, the reaction products could
not be positively identiRed. Rao, Moon and Leja also
indicated that contact between various sulRdes and
iron will result in the depression of the potential to
such an extent that the oxidation of xanthate to
dixanthogen will no longer be possible. This is in-
dicated in Figure 4.
During electrochemical interaction between sulRde
species, ionic charge transfer takes place through the
Sotation liquor, while electronic charge transfer takes
place through the solid interface; solid phase conduct-
ivity, as well as water conductivity is thus important.
As an example, it is the experience on the Phalaborwa
igneous complex that plant water conductivities
range generally between 180 mS (fresh industrial
water) and extremes of ca. 500 ms, with a middle
range of 200}300 mS.
For separate mineral particles, the solid phase
charge transfer would rely on particle collision, in
which the gangue particles have a shielding inSuence.
This reduces the galvanic interactions to a point
where electrochemical interactions between fully lib-
erated minerals are unimportant in Sotation plant
practice, unless plant waters are highly conductive,
and both pulp densities and sulRde mineral concen-
trations are high enough.
Reaction Products Affecting
Flotation Performance
General
The reaction products of galvanic interaction may
inSuence the Sotation efRciency of composite
minerals by direct depression or activation of min-
erals, or by affecting Sotation froth character-
istics. These reaction products, as will be shown, are
not unique to galvanic processes, but their rate of
formation may be enhanced by such interactions. The
spatial separation of the anodic and cathodic reac-
tions in galvanic interactions favours the kinetics in
the sense that the reaction products formed at the

Figure 4 Mixed potentials of sulfide minerals alone and in contact with metallic iron as a function of xanthate concentration.
Nonoxidizing conditions, argon purging } 400 cc min\1 natural pH, 253C. (Reproduced with permission from Rao, Moon and Leja
(1976) Flotation, A.M. Gaudin Memorial Vol. Copyright American Institute of Mining Metallurgical and Petroleum Engineers.)
 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions 1499

anodic and cathodic sites do not directly interact to Activating Species

deposit potentially reaction stiSing product on the
The release of activating species may be accelerated
anodic site.
by galvanic interaction. Activation of sulRde species,
The reaction products of galvanic interaction be-
raising their Soatability above that which is achiev-
tween mineral species can be distinguished on the
able naturally, can occur due to an enhancement of
basis of their location. Firstly, the reaction product
the hydrophobicity of the mineral, or due to the
may take the form of a surface modiRcation of the
insertion or attachment of ions which are more react-
mineral, e.g. a metal-deRcient sulRde layer, supported
ive towards collector reagents than the host species.
and to a greater or lesser degree stabilized, by the
In the latter case, the main, but not only, ion to
underlying, unaltered phase. Secondly, the product
consider is copper, which may attach itself to a par-
may be chemically distinct from, and physically at-
ticle as follows, by direct replacement:
tached to, the original mineral particle. Examples
of this are elemental sulfur and ferrous hydroxide
coatings. Finally, the reaction product may detach 2Cu(OH)2#2MS#H2O#2e\
and remove itself from the original mineral, like P2M(OH)2#Cu2S#HS\#OH\ [7]
sulfate anions, copper cations, or ferric hydroxide
particulates.
The reaction path can be favoured by the presence
The anodic reaction of sulRdes is presently thought
of a suitable cathodic particulate mineral, like pyrite,
to lead to the formation of metal-deRcient, sulfur-rich
reacting along the lines of the following equation:
surface species, by releasing an active metal ion which
may form a metal hydroxide (M(OH)2):
MS#2H2OPM(OH)2#S0#2H##2e\ [8]
MS#xH2O#0.5xO2PM1 S#xM(OH)2 [4]
\x The effect of galvanic coupling in a conducting
particle will cause an electron Sow, which will facili-
The formation of metal-deRcient sulRde at the sur-
tate the simultaneous cathodic reduction of the MS
face will tend to activate the surface and cause the
surface by reaction with Cu(OH)2, and the anodic
metal hydroxy species to detach. However in the case
oxidation of MS both reactions producing hydropho-
of Fe2#, species remain largely attached, leading to
bic surface coverings of Cu2S and S0, respectively.
a blanketing effect that tends to hinder particle}
Another example is the activation of pyrite by lead
bubble attachment. At suitable pH values, the release
ions. In support of the assertion that true electro-
of reactive cations may lead to the unwanted activa-
chemical interaction is limited to composite particles,
tion of sulRde minerals, notoriously by copper ions.
Zhang et al. found, importantly, that interaction be-
As a general precaution against this reaction path,
tween composite particles containing pyrite and
dissolved oxygen levels can be lowered. However,
sphalerite was negligible in the absence of metal ions
a lowering of pulp oxidative potential tends to lead to
in solution. In their presence, sphalerite was found to
a general depression of Sotation. An alternative is the
successfully compete with pyrite for the activating
elimination of metal-deRcient sulRde species, or ele-
ions, and through their action to compete more suc-
mental sulfur, by reaction with aqueous sulfur diox-
cessfully for xanthate, thus depressing the Soatability
ide:
of pyrite. Competition for activating species thus
seems to be an important factor in the interaction
M1 S#SO23\PMS#S2O23\ [5]
\x between minerals in Sotation.
This is a possible mechanism for galena depression Depressing Species
with sulfur dioxide in Sotation, in addition to other
mechanisms postulated, i.e. a lowering of copper ion The formation of hydrated, oxidized surface species
activity in solution, xanthate decomposition and like iron hydroxide and basic sulfates, increases par-
a lowering of the oxidative potential below that ticle hydrophilicity, and will thus depress Sotation.
necessary for xanthate oxidation to dixanthogen. High thiosulfate levels, which may arise when milling
A reaction path for the cathodic reaction of chal- under relatively nonreducing conditions, e.g. fully
copyrite, at neutral pH values and in oxygen-starved autogenous grinding, or laboratory grinding in porce-
pulps, was also proposed by Li and Iwasaki: lain mills, may lead to the precipitation of insoluble
thio-salts, which may also depress Sotation. The gen-
2CuFeS2#3H2O#2e\ eration of soluble sulRde species, especially notable in
stagnant, anaerobic water reservoirs with bacterial
PCu2S#2Fe2##3HS\#3OH\ [6] action, may lead to mineral depression, since the
1500 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions
sulRde and xanthate ions compete for the same sur-
face sites.
Froth Characteristics Affecting Species
The effect of froth structure on Sotation is usu-
ally related to its stability. Stable froths have small
bubbles and a high entrained water content. The
solids in the entrained water are approximately at
their concentrations in the pulp; consequently their
overSowing concentrates will be of low grade. The
presence or absence of Rne, colloidal particles has
a profound effect on Sotation froth structure and
drainage, and thus overall Sotation performance. The
most notable example of this is the deleterious ef-
fect that copious quantities of (naturally Soatable)
talc particles have on sulRde Sotation, with adverse
results in grade, recovery, and rates of recovery. The
same froth modifying effects can be noticed
when dealing with colloidal precipitates. In practice,
the noticeable effects are largely limited to iron
hydroxides, due to their abundance in natural sys-
tems. Thus, a change in froth structure may be noted
when hydrated ferrous hydroxide particles are oxi-
dized to (less hydrated) ferric oxide. The latter species
allows a more desirable, less slimy, froth structure. As
another example, it has been strongly suggested that
the common practice of copper sulfate addition
in Sotation plants, apart from activating effects,
also has strong froth structure modifying effects.
Application to Plant Practice
Collectorless Flotation
For collectorless Sotation, the formation of a metal-
deRcient sulRde layer on the particle surface must
generally be targeted. Formation of such a layer may
be accelerated and spatially accentuated by galvanic
interaction. When considering the Sotation of sulRde
minerals in the absence of collector reagents, in the
context of electrochemical mineral interaction, three
factors can induce Soatability. First of all, Soatability
can be natural, i.e. due to the crystal structure and
chemical bonding of a mineral. Examples of such
minerals are molybdenite, stibnite, and the arsenic
sulRdes realgar and orpiment. Secondly, collectorless
Sotation can be self-induced, i.e. under the right pulp
oxidative potentials, surface products will form
which induce hydrophobicity. Examples of this are
pyrrhotite and chalcopyrite. Rao and Finch found
that pyrite/sphalerite selectivity could be enhanced by
Rrst recovering the pyrite which is naturally Soatable,
due to a chemically formed sulfur layer, in the ab-
sence of a collector. A third cause of Soatability is
mineral size. As minerals decrease in size, their
recovery into Sotation froths increases due to entrain-
ment, rather than selective attachment to froth bub-
bles. A good example of this is galena Sotation from
complex ores. At one mine site, carrying out sequen-
tial copper}lead}zinc Sotation, about three-quarters
of lead recovery into the copper concentrate was
found to be made up of galena particles smaller than
10
m. Since this effect is physical rather than
chemical, it can only be signiRcantly affected by
a change in physical parameters, e.g. froth lamellae
thickness and particle size.
In the collectorless Sotation of pyrite}chalco-
pyrite}quartz mixtures, Johnson found a dependence
of Sotation behaviour on the pyrite/chalcopyrite sur-
face area ratio, which would be consistent with elec-
trochemical interaction. In this work it was however
shown as well that copper solubilization was not
enhanced in the presence of pyrite, but rather re-
duced; this points to copper deposition on pyrite } in
other words, to activation rather than direct electro-
chemical interaction. Interestingly, interaction was
reduced in the presence of quartz, due to adsorption
of metal ions onto the quartz surfaces. In this respect,
adsorption studies on other gangue minerals showed
that such scavenging of potentially activating ions
from solution may be substantial.
Pulp Oxidative Potential Control
Hayes and Ralston showed that the control of pulp
oxidative potentials allows Sotation selectivity, and is
therefore a worthwhile approach in the Sotation of
complex sulRde ores, in addition to pH strategies.
Direct electrochemical interactions between phys-
ically separated sulRde minerals, in which one af-
fects the other’s Sotation behaviour directly through
an anode}cathode relationship, have so far not been
convincingly demonstrated on plant scale.
Galvanic and electrochemical interaction between
sulRde minerals, and general chemical reactivity, is to
a large degree dependent on the presence of oxygen in
solution. The control of oxygen levels is thus gener-
ally the objective and result of pulp oxidative poten-
tial control. Since industrial Sotation relies heavily on
the use of ambient air, it has been proposed to regu-
late the oxygen concentrations entering Sotation by
admixture of nitrogen, or by partial re-circulation of
process air released from the froth surface. The for-
mer approach is expensive whereas the latter depends
on cells speciRcally designed to collect and re-circu-
late air leaving the top surface of the froth. Cylin-
drical cells seem to be most effective in this
respect. The gas composition of bubbles generated by
pressure differentials in Sotation cells is depen-
dent on a suite of factors, including the magnitude of
 II / FLOTATION / Electrochemistry: Contaminant Ions and Sul\de Mineral Interactions
the pressure drop, dissolved substances, and nature of
the nucleating surfaces. BeneRts in Sotation results
were shown when regulating certain reagent additions
on the basis of pulp oxidative potential, e.g. sulfuric
acid, rather than pH. However, beneRts might well be
mostly due to froth structure improvements.
Trahar has shown that interactions between sulRde
minerals were much decreased if they are ground
separately, and only combined in the Sotation cell. In
this case, no mineral interaction between chal-
copyrite, galena and sphalerite could be statistically
proven. This demonstrates that galvanic interaction
between sulRde minerals is only practically noticeable
when mechanical contact exists. Grano et al. have
also found that Sotation selectivity between sulRde
minerals is most sensitive to milling and precon-
ditioning parameters, more so than to oxidative po-
tentials during Sotation itself, and mostly due to the
presence of mild steel particles originating from
equipment wear. For this reason, amongst others, re-
search into comminution techniques which maximize
mineral separation, whilst minimizing smearing, over-
grinding, and steel consumption, must be a priority in
the minerals industry. The effect of mild steel
particles generated during ore comminution is mainly
due to oxygen consumption, corrosion inhibitors being
essentially ineffective. Full oxidation of these par-
ticles during conditioning removes their deleterious
effect. Even for real ores containing signiRcant
quantities of more than one sulRde mineral, reason-
able correlations exist between the behaviour of min-
erals in the ore and single minerals, as found
by Grano et al. This indicates the limited extent
of electrochemical mineral interactions in general
practice.
Conclusion
True electrochemical interactions between sulRde
minerals on industrial size plants are thought not to
be of practical signiRcance, except when physical
contact between dissimilar sulRdes exists (middlings),
and/or at high pulp densities, high sulRde concentra-
tions in the Sotation feed, and high water conductivi-
ties. More importantly, sulRde minerals are found to
interact through competitive adsorption of activating
ions, the reduction of oxygen levels in the Sotation
pulp, and froth modifying activity of mineral oxida-
tion products. Middlings particles, composed of two
or more sulRdes, do however experience electro-
chemical interaction, the result of which appears to
be an enhancement of Soatability, leading to a reduc-
tion in concentrate grades. The solution to such
a problem is however more to be sought in comminu-
tion technology than electrochemical intercession.
1501
Oxidative potential control offers advantages in
industrial Sotation separations, but its effect
does not appear to be an interference with electro-
chemical mineral interactions.
Further Reading
Allison SA, Goold LA, Nicol MJ and Granville A (1972)
Metallurgical Transactions 3: 2613}2618.
Buckley A and Woods R (1981) Investigation of the surface
oxidation of sulRde minerals via ESCA and electro-
chemical techniques. Interfacial phenomena in mineral
processing, Yarar B and Spottiswood DJ (eds) Engineer-
ing Foundation 3}17.
Cheng X and Iwasaki I (1992) Pulp potential and its im-
plications to sulRde Sotation. Mineral Processing and
Extractive Metallurgy Review 11: 187}210.
Gardner JR and Woods R (1977) An electrochemical in-
vestigation of contact angle and of Sotation in the pres-
ence of alkylxanthates. II. Galena and pyrite surfaces.
Australian Journal of Chemistry 30: 981}991.
Grano S, Ralston J and Smart RStC (1990) InSuence of
electrochemical environment on the Sotation behaviour
of Mt. Isa copper and lead-zinc ore. International Jour-
nal of Mineral Processing 30: 69}97.
Hayes RA and Ralston J (1988) The collectorless So-
tation and separation of sulRde minerals by Eh control.
International Journal of Mineral Processing 23:
55}84.
Li X and Iwasaki I (1992) The effect of cathodic
polarisation on the Soatability of chalcopyrite in the
absence of oxygen. Minerals and Metallurgical Process-
ing 9: 1}6.
Plaksin IN and Shafeev RSh (1963) InSuence of surface
properties of sulphide minerals on adsorption of Sota-
tion reagents. Bulletin of the Institute of Minerals and
Metallurgy 680: 715}722.
Rao SR and Finch JA (1987) Electrochemical studies on the
Sotation of sulphide minerals with special reference to
pyrite}sphalerite } II. Flotation studies. Canadian Me-
tallurgical Quarterly 26(3): 173}175.
Rao SR, Moon KS and Leja J (1976) Effect of grinding
media on the surface reactions and Sotation of heavy
metal sulphides. Flotation, A.M. Gaudin Memorial Vol.
American Institute for Minerals Metals and Petroleum
Engineering, pp. 509}527.
Trahar WJ (1984) The inSuence of pulp potential in sul-
phide Sotation. Principles of Mineral Flotation, The
Wark Symposium. Australasian Institute of Mining and
Metallurgy. Jones MH and Woodcock JT (eds),
Parkville. Victoria, Australia (40): 117}135.
Woods R, Young CA and Yoon RH (1990) Ethyl xanthate
chemisorption isotherms and Eh-pH diagrams for the
copper/water/xanthate and chalcocite/water/xanthate
systems. International Journal of Mineral Processing 30:
17}33.
Zhang Q, Xu Z, Bozkurt V and Finch JA (1997) Pyrite
Sotation in the presence of metal ions and sphalerite.
International Journal of Mineral Processing 52: 187}201.
1502 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles

Flotation Cell Design: Application of Fundamental Principles

B. K. Gorain, J. P. Franzidis and E. V. Manlapig, bottom of the cell tank to the discharge or tailings
Julius Kruttschnitt Mineral Research Centre, box (Figure 1).
Indooroopilly, Queensland, Australia
Hydrodynamic Zones
Copyright ^ 2000 Academic Press
A mechanical Sotation cell necessitates generation
of three distinct hydrodynamic zones for effec-
Introduction tive Sotation. The region close to the impeller
encompasses the turbulent region necessary for solids
The froth Sotation process is commonly employed suspension, dispersion of gas into bubbles and
for the selective separation of a mineral species bubble}particle interaction for collection of
from a liquid}solid suspension of both valuable minerals on the surface of the bubbles. Above the
and unwanted gangue mineral particles. The valuable turbulent region lies the quiescent zone where the
mineral species (which needs to be separated) is
rendered hydrophobic by controlling its surface
chemistry to provide the potential conditions for
the attachment of the particles to air bubbles. The
bubbles and particles are made to interact with each
other inside a Sotation machine. The Sotation ma-
chine, depending on its operating conditions, pro-
vides an environment for the bubble}particle attach-
ment and permits levitation of bubble}particle ag-
gregates to the froth. The manner in which bubbles
and particles interact with each other depends on the
cell operating conditions and the type of Sotation
machine used.
Flotation machines, in general, may be categorized
into four different classes: (i) mechanical or con-
ventional cells; (ii) energy-intensive pneumatic cells;
(iii) column cells; and (iv) froth separators. Of these,
mechanical Sotation cells have dominated the min-
eral industry since the early days of Sotation and
account for a signiRcant amount of minerals pro-
cessed. The aim of this article is to describe the opera-
tion and design of mechanical Sotation cells.

Cell Operation
A mechanical Sotation cell essentially consists of
a vessel or a tank Rtted with an impeller or rotor. The
impeller agitates the slurry to keep particles in sus-
pension, disperses air into Rne bubbles and provides
an environment in the cell tank for interaction of
bubbles and hydrophobic particles and their sub-
sequent attachment and therefore separation of valu-
able mineral particles from the undesired gangue
mineral particles. The bubble}particle aggregates
move up in the cell by buoyancy and are removed Figure 1 Schematic diagram of a mechanical flotation cell. 1,
Discharge box; 2, concentrate launders; 3, feed box; 4, cell lip; 5,
from the cell lip into an inclined drainage box called
bearing shaft; 6, drive pulley with guard; 7, three-phase induction
a launder (Figure 1). The launder product is com- motor; 8, air inlet pipe with control valve; 9, concentrate launder
monly known as concentrate. The particles that do discharge point; 10, impeller shaft; 11, tailings discharge point,
not attach to the bubbles are discharged out from the 12, base support for the cell tank.
 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles
Figure 2 Hydrodynamic zones in a mechanical flotation cell.
bubble}particle aggregates rise up in a relatively less
turbulent region. This region also helps in reducing
the number of gangue minerals which may have been
entrained mechanically or entrapped between bub-
bles for upgrading of valuable minerals. The region
above the quiescent zone is the froth zone that serves
as an additional cleaning step and improves the grade
of the concentrate product. The three hydrodynamic
zones in mechanical Sotation cells are depicted in
Figure 2.
The conSicting functional requirements in dif-
ferent zones in a mechanical Sotation cell are a chal-
lenge in terms of the cell design and a Rne balance of
hydrodynamic conditions is necessary for the opti-
mum recovery of valuable minerals in a cell.
Gas Dispersion
One of the most important hydrodynamic conditions
in a mechanical Sotation cell is dispersion of gas into
Rne bubbles. The bubble generation mechanism in
Figure 3 Schematic diagram of formation of bubbles in mech-
anical cells (after Grainger Allen, 1970; courtesy of Transactions
of the Institute of Metallurgy, UK).
1503
a mechanical cell is a two-stage process. Firstly, air
cavities are formed at the trailing edge of the impeller
blades, which is the low pressure region. Thereafter,
bubbles form by shedding of vortices from the tail of
the cavity, as shown in Figure 3.
The dispersion of air into bubbles can be character-
ized by three properties: bubble size, gas hold-up and
superRcial gas velocity. Mean bubble size in industrial
mechanical cells varies, in general, from 0.5 to 2 mm;
gas hold-up varies from 5 to 15% and superRcial gas
velocity varies from 0.6 to 1.5 cm s\1, depending on
cell operating conditions (impeller speeds and air
rates) and the cell duty in plant operation } roughers,
scavenging, cleaners, etc. Recent studies have shown
that bubble size, gas hold-up and superRcial gas velo-
city cannot describe the gas dispersion in a mechan-
ical cell adequately when taken individually; but
when taken together the gas dispersion properties
determine the bubble surface area Sux Sb in the cell,
which has been shown to characterize gas dispersion
very well. Typical Sb values in industrial cells vary
from 30 to 60 s\1. The concept of Sb has been found
to be useful in metallurgical scale-up and cell optim-
ization, design and selection.
Mode of Air Entry
In mechanical cells there are two modes by which air
is introduced into the cells; one is the forced air entry
mode carried out using a blower and the other is the
self-induced air entry mode, in which air is sucked
into the cell by vortexing. The two cell designs can be
distinguished by the difference in vertical loca-
tion of the impeller in the cell. In the forced air-type
machine, the impeller is located close to the cell
bottom with a deeper impeller submergence, and an
external air blower is used to supply air under pres-
1504 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles
sure through the hollow shaft to the impeller region.
Self-induced air machines utilize a standpipe which
shrouds the drive shaft which is solid. The impeller is
located almost in the midpoint of the cell which
draws air through the space between the standpipe
and the solid shaft.
Flow Patterns
The impeller in a mechanical Sotation cell, during
rotation, generates a vortex at the bottom of its
blades drawing slurry from its lower section and
discharging out from its upper section of the blades.
Air is introduced through the impeller shaft or in the
spacing between the shaft and a standpipe depending
on whether the cell is forced air type or self-induced
type, as described above. The dispersed air bubbles
come in contact with the slurry close to the impeller
discharge point. The aerated slurry Sow then leaves
the impeller mechanism for the surrounding tank
volume. The impeller, therefore, acts as a pump
drawing in slurry from below and expelling the
aerated slurry to the cell volume. A typical Sow pat-
tern of a mechanical cell is shown in Figure 4.
Cell Design
The essential components of a mechanical Sotation
cell are described below.
Figure 4 Typical flow patterns in a mechanical flotation cell (courtesy of Outokumpu Mintec Oy, Finland).
Cell Tank
The proRle of a cell tank is rectangular with truncated
corners, U-shaped, conical or cylindrical, depending
on cell type and size. Typically, mechanical cells are
designed with a rectangular tank bottom for cells
with volume up to 3 m3 and a U-shaped bottom for
cells with volume up to about 38}45 m3. Cells larger
than 38}45 m3 are typically cylindrical with either
a conical or a Sat bottom. Figure 5 shows a schematic
of different tank designs.
In a typical plant, the mechanical cell tanks are
arranged in a series called a bank. The number of cells
in a bank varies depending on cell size, application
and plant circuit conRguration. The tailings from the
Rrst cell move on as the feed to the second cell and so
on and the tailings from the last cell form the Rnal
tailings of the bank. The concentrates from dif-
ferent cells are combined in different ways de-
pending on the requirements of the circuit. For
example, in a cleaner bank, the concentrates from the
Rrst two cells may be combined to form the Rnal
concentrate products, whereas the concentrate prod-
uct from the rest of the cells may be combined and
recirculated to the feed of the cleaner bank.
Feed Box and Discharge Box
Each bank in a Sotation circuit (which could also be
an individual cell) is usually Rtted with a feed box
 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles
Figure 5 Typical tank designs in mechanical flotation cells.
with a rectangular opening at the bottom of the
box to allow entry of slurry into the cell bank
for Sotation. The feed box is rectangular or half-
cylindrical in shape depending on cell type and size.
A tailings box or discharge box is also Rtted at the
end of the bank (or on the opposite side of feed box
in an individual cell) to allow discharge of tailings.
The discharge box is also rectangular or half-cylin-
drical in shape. Figure 6 shows a typical arrangement
of a feed and a discharge box in mechanical Sotation
cells. For some cell types and sizes, a dart valve or
overSow weir are Rtted in the discharge box to con-
trol pulp level in the cell tank. For other designs,
a discharge box is not used and a pinch valve is Rtted
to the tailings outlet pipeline instead for pulp level
control.
Cell Launders
Launders in Sotation cells are located outside the
overSow lip to collect and transport the froth or
concentrate product out of the cell tank. Launders are
typically located on the top of the cell tank, as shown
in Figure 1. Launders are designed with a slope of
about 10}153 for smooth transportation of froth
without blockage in the launders.
The design of launders varies with cell size and
type. The launders are located on opposite sides ad-
jacent to the feed and discharge boxes in rectangular
cell tank designs, as shown in Figure 1. Launders on
three sides are also common in rectangular cells ar-
Figure 6 Schematic of a mechanical cell showing feed box and discharge box and concentrate launders.
1505
ranged in series. Large cylindrical cells have concen-
tric launders which can be either internal or external
or both, depending on the capacity of launder neces-
sary for froth removal.
Impellers or Rotors
The impeller or agitator, also referred to as the rotor,
is considered to be the heart of a mechanical Sotation
cell as it provides the energy to perform the following
functions necessary for the Sotation process:
1. Suspension of solids in the cell tank.
2. Dispersion of air into bubbles.
3. Creation of microturbulence for effective bub-
ble}particle collision.
4. Suction of air into the cell in self-induced type
cells.
The design of an impeller varies with cell type.
Most impeller designs have a Sat circular disc with
different shapes of blades or Rngers Rtted to the
disc concentrically to the lower section of the disc.
The shape of the blades or Rngers varies from cylin-
drical to tapered (half-spherical). The half-spherical
impeller design is more popular in the new design of
cells and details of this design will be discussed later
in this article. The top section of the disc connects
to a drive shaft which in turn connects to the
pulley/gear-motor drive assembly. The impeller is
located in the centre of the cell cross-section with its
1506 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles
Figure 7 Shapes of different impellers and stators. (A) Bate-
man; (B) Dorr-Oliver; (C) Outokumpu; (D) Wemco (courtesy of
Bateman Process Equipment, Dorr-Oliver, Outokumpu Mintec Oy
and Baker Process, respectively).
submergence varying with cell type and mode of
air entry. Figure 7 shows the shapes of different
commercially available impellers.
Stators or Diffusers
A stator or diffuser is an important component
of a mechanical Sotation cell, which surrounds the
impeller and acts as an internal bafSe useful in
reducing pulp vortex in the cell. The tangential Sow
of the agitated slurry (due to rotation of the impeller)
is transformed to a radial direction for effective
dispersion of gas and solids in the cell tank. This
reduction in the vortex Sow helps in maintaining
a stable pulp}froth interface, essential for Sotation.
A stator consists of a number of blades arranged in
a concentric circle with gaps between the blades to
facilitate movement of slurry in the cell tank. A stator
is usually mounted on the bottom of the cell tank
surrounding the impeller concentrically from its bot-
tom. In some cell designs the stator is Rtted to the
standpipe such that the stator shrouds the impeller
from the top and hangs with an open space at the
bottom: this is commonly known as an overhung
stator.
The impellers and diffusers are moulded and
coated with rubber or polyurethane for abrasion
resistance.
Impeller Drive Assembly
The impeller connected to the shaft (hollow or solid)
is driven by a three-phase induction motor with the
help of V-belts, pulleys or gear box. A typical drive
arrangement in a mechanical cell is shown in Fig-
ure 1. The size of the drive and the motor pulleys
determine the speed at which an impeller operates.
Cell Types and their Designs
Most of the industrial mechanical Sotation cells in the
early days (before the 1970s) were of the cell-to-cell
type (tanks of different cells connected in a row)
for small plants and multistage cleaner Soats where
the pumping action of the impellers permitted the
transfer of intermediate Sows without external
pumps. With the emergence of large Sotation cells,
since the early 1980s, dictated by economic consider-
ations, open Sow cells (with slurry Sowing openly
through a series of cells in a bank) have become
prominent.
In the 1980s many mechanical cell designs were
prevalent around the world. The major ones are:
1. Agitair cells from Galigher company, USA.
2. Aker machines from Aker Trondelag, Norway.
3. BCS cells from Minemet Industrie, France.
4. Booth cells from Booth Company, USA.
5. Denver cells from Denver Equipment Limited,
Joy Industrial Company, USA.
 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles 1507

6. Krupp cells from Krupp Polysius AG, West Ger-

many.
7. Maxwell cells from Technequip Ltd, Canada.
8. Mechanobre cells from Machineoexpert V/O,
USSR.
9. OK cells from Outokumpu Oy, Finland.
10. Sala cells from Sala International AB, Sweden.
11. Wedag cells from KHD Humboldt Wedag from
West Germany.
12. Wemco cells, Wemco division, Envirotech, USA.

Only a handful of these cell manufacturers have

survived the competitive global market by improving
their products or by mergers or by diversiRcation.
Manufacturers of Wemco and Outokumpu cells,
through research and development, have consistently
updated their technology to remain competitive. The
recent Tankcells (designated as OK-TC) and Smart-
cells from the manufacturers of Outokumpu and
Wemco cells, respectively, are an example. Some new
designs, such as the Bateman BQ and Svedala RCS Figure 8 A schematic diagram of the Bateman flotation cell
cells, have emerged in the mid 1990s. The companies (courtesy of Bateman Process Equipment, South Africa).
which manufactured Denver and Sala cells have been
procured by Svedala and their cells are marketed by
Svedala’s Pumps and Process division. The Agitair from 5 m3 (BQR 50) to 100 m3 (BQR 1000). The tank
cells are now marketed by Baker Process (previously dimensions of different cells of varying sizes are
known as EIMCO Process Equipment Company). given in Table 1. The unit cell design Bateman cells
KHD Humboldt Wedag have stopped manufacturing are called HiFlo2+ and HiClean2+ machines.
mechanical cells and now market a newly developed The Bateman mechanism consists of a hemispher-
pneumatic cell known as PneuSoat. ical-shaped impeller which is connected to a solid
Presently there are Rve major manufacturers of drive shaft. The impeller is designed with no disc on
mechanical Sotation cells. Details of the design fea- the top and the impeller blades have both the top and
tures of different cells are described in the sec- bottom opened. The drive shaft is shrouded with
tions below. a stand pipe. The Bateman mechanism utilizes the
forced air entry mode and air is supplied into the
mechanism through the gap between the standpipe
Bateman Cell and the shaft. The mechanism utilizes an overhung-
The Bateman Sotation mechanism was developed in type stator (or diffuser) connected to the bottom
1993 and is presently marketed by the Bateman Pro- of the standpipe, which is a horizontal hood with
cess Equipment Limited. The BQR series of Bateman bafSe plates projecting downwards (Figure 8).
cells have a round tank design with cell sizes varying
Dorr-Oliver Cells
Table 1 Bateman cell tank dimensions for different cell sizes The Door-Oliver cell is marketed by Dorr-Oliver,
(courtesy of Bateman Process Equipment, South Africa) a global corporation and member of the Krauss-
Model Volume Height Depth Installed motor Maffei Group.
(m 3) (m) (m) (kW) The Dorr-Oliver Company Limited manufacturers
Sotation cells in a wide range of sizes. Cells with
BQR 50 5 2 2 NA
a volume of 0.03 m3 (DO 1) to 2.8 m3 (DO-100) have
BQR 100 10 2.5 2.5 45
BQR 200 20 3.2 3 55 a Sat-bottom tank design. Cells with volumes from 4.2
BQR 300 30 3.6 3.4 75 to 44 m3 come with a U-shaped tank bottom. Cells
BQR 400 40 4 3.75 75 with volumes from 50 to 150 m3 are available with
BQR 500 55 4.2 4.34 115 a round tank with a conical bottom. Details of tank
BQR 750 75 5.2 4.5 132
dimensions for the Dorr-Oliver cells are given in
BQR 1000 100 5.5 4.95 132
Table 2.
1508 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles

Table 2 Dorr-Oliver cell tank dimensions for different cell sizes (taken from Dorr-Oliver flotation cell brochure; courtesy of Dorr-Oliver,
Australia)

DO conventional cells

Model Volume (m 3) Length (m) Width (m) Height (m) Installed motor (kW)

DO 1.0 0.03 0.3 0.3 0.33 0.55

DO 3.5 0.1 0.45 0.45 0.5 0.55
DO 10 0.3 0.65 0.65 0.66 1.1
DO 25 0.7 0.9 0.9 0.86 2.2
DO 50 1.4 1.2 1.2 0.97 4.0
DO 100 2.8 1.52 1.52 1.22 5.5
DO 150 4.2 1.83 1.83 1.53 7.5
DO 300C 8.5 2.29 2.29 1.88 7.5
DO 300 8.5 2.29 2.29 1.88 11.0
DO 500C 14 2.69 2.69 2.46 15
DO 600 17 2.95 2.69 2.46 22
DO 1000 28 3.35 3.35 2.89 30
DO 1350 38 3.81 3.58 3.22 37
DO 1550 44 3.96 3.96 3.22 45

Tank design

Model Volume (m 3) Height (m) Diameter (m) Installed motor (kW)

DO 1750 50 3.86 4.32 56

DO 3500 100 5.49 4.65 93
DO 5300 150 6.71 4.72 131

The Door-Oliver mechanism consists of a hemi-
spherical-shaped impeller Rtted to a hollow shaft. The
mechanism utilizes the forced air entry mode in which
air is introduced to the impeller through the hollow
shaft. The stators for the Dorr-Oliver cells are gener-
ally mounted on the bottom but the large cells mecha-
nisms are designed with an overhung stator.
Figure 9 shows a schematic diagram of a large
Dorr-Oliver cell with a tank design.

Outokumpu Cells
Outokumpu Mintec, a Finnish company which
belongs to the Outokumpu Group, operates inter-
nationally and has been the manufacturer of the
Outokumpu Sotation cell for the last 30 years.
Outokumpu produces different Sotation ma-
chines which can be catgorized as:

1. OK conventional Sotation machines: for rougher,

scavenger and cleaner Sotation.
2. OK-TC (TankCell) Sotation machines: for
rougher and scavenger Sotation.
3. SK Sotation machines: for Skim-Air Flash Sota-
tion in the grinding circuit.
4. HG Sotation machines: for cleaner Sotation.

The OK conventional Sotation cells are available in

Figure 9 Schematic diagram of a large Dorr-Oliver cell (cour-
tesy of Dorr-Oliver, Sydney, Australia).
volumes up to 38 m3. Conventional cells have a rectan-
gular tank design for cell volumes up to 3 m3; above
3 m3 and up to 38 m3 the cells have a U-shaped tank.
TankCell designs are available from a volume of 5 m3
to a volume of 160 m3 and are essentially a cylindrical
 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles 1509

horizontal disc on the top which is attached to a num-

ber of narrow vertical slots tapered downwards. The
impeller has separate slots for air and slurry move-
ment. The mechanism has a forced air type entry
mode in which air is brought into the impeller
through a hollow shaft.
The stator in the OK mechanism is mounted on
the bottom of the tank. There are two stator designs
used in an Outokumpu cell: one is known as the
multi-mix or conventional stator and the other is
known as free-Sow. The multi-mix stator is
typically used for Rne particle Sotation, whereas the
free Sow stator is typically used for coarse particle
Sotation.

Figure 10 A schematic diagram of Outokumpu TankCell (cour-
tesy of Outokumpu Mintec Oy, Finland).
cell with a Sat bottom (Figure 10). The tank dimen-
sions of different cells are given in Table 3.
The OK impeller mechanism is designed with
a hemispherical-shaped impeller consisting of a
Svedala Cells
The former manufacturers of Denver Sotation cells
(Denver Equipment, USA) and Sala cells (Sala
International in Sweden) have merged together to
form the Svedala Pumps and Process Division, which
is part of the worldwide Svedala Industri group. Both
Denver and Sala cells are available through Svedala
companies located worldwide.
The Svedala Sotation cells include mechanical So-
tation cells in the AS range (previously known as Sala
cells), in sizes from 0.03 to 16 m3; the DR range

Table 3 Outokumpu cell tank dimensions for different cell sizes (taken from Outokumpu flotation cell brochure; courtesy of
Outokumpu Mintec Oy, Finland

OK conventional cells

Model Volume (m 3) Length (m) Width (m) Height (m) Motor installed (kW)

OK-0.5-R 0.5 NA NA 0.84 2.75}3.75

OK-1.5-R 1.5 NA NA 1.08 5.5}7.5
OK-3-R 3 1.52 1.52 1.21 7.5}11
OK-8-U 8 2.29 2.29 1.88 15}22
OK-16-U 16 2.95 2.69 2.46 30}45
OK-38-U 38 3.49 3.59 3.23 55}75

Model Volume (m 3) Height (m) Diameter (m) Motor installed (kW)

Tank Cells
OK-5-TC 5 2.45 2.2 7.5
OK-10-TC 10 2.85 2.7 15
OK-20-TC 20 3.45 3.3 37
OK-30-TC 30 3.9 3.9 45
OK-40-TC 40 4.3 4.1 45
OK-50-TC 50 4.6 4.6 75
OK-70-TC 70 5 5 90
OK-100-TC 100 5.3 5.6 110
OK-130-TC 130 5.4 6.3 132
Extra Hard Duty
OK-100-TC-XHD 100 4.6 6.3 90
OK-130-TC-XHD 130 4.8 6.7 110
OK-160-TC-XHD 160 5.1 7.1 132
1510 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles

Table 4 Svedala cell tank dimensions for different cell sizes bottom. The blades are connected to a circular hori-
(taken from Svedala flotation brochure; courtesy of Svedala, UK) zontal disc located just above the centre of the blades.
Model Volume Height Diameter Installed motor The mechanism is designed with an overhung stator
(m 3) (m) (m) (kW) with vertical vanes projecting downwards connected
to the mechanism standpipe. Depending on cell
RCS 5 5 1.9 2 15
application, the DV mechanism can be modiRed in two
RCS 10 10 2.4 2.6 22
RCS 15 15 2.5 3 30 different ways to suit the application. The design
RCS 20 20 3 3.25 37 of the mechanism which allows entry of air through
RCS 30 30 3.4 3.7 45 a hollow drive shaft is known as the DVH mechanism
RCS 40 40 3.8 4.1 55 (deep vane and hollow shaft), whereas the design
RCS 50 50 4.1 4.5 75
which allows entry of air through a concentric stand-
RCS 70 70 4.6 5 90
RCS 100 100 5.2 5.6 110 pipe is known as the DVS mechanism (deep vane and
RCS 130 130 5.6 6.1 132 solid shaft).
RCS 160 160 6.1 6.5 160 Figure 11 shows a schematic of the Svedala RCS
RCS 200 200 6.5 7 200 Sotation cell, showing the DV mechanism and the cell
tank design.

(previously known as Denver cells), in sizes from 0.09

to 42.5 m3; and cell-to cell machines in sizes from
0.08 to 14.2 m3.
In 1995 Svedala developed a new design of Sota-
tion cell known as the RCS Flotation machine which
comes in sizes from 5 to 200 m3. The tank dimensions
of different cells sizes are shown in Table 4.
The RCS Flotation machine utilizes a new DV
(deep vane) mechanism. The DV mechanism consists
of vertical rectangular blades or vanes tapered at the
Wemco Cells
Wemco Sotation cells are manufactured by Baker
Process, which also makes Agitair cells and pyramid
column cells.
There are two major Wemco designs, the Wemco
1#1 design and new SmartCell design (Figure 12).
The 1#1 design comes in cell sizes from 0.57 to
85 m3. The SmartCell design comes in sizes from 8.5
to 160 m3. The Wemco 1#1 cell utilizes the self-
induced air entry mode and consists of a rotor,

Figure 11 A schematic diagram of Svedala RCS cell (courtesy Figure 12 A schematic diagram of Wemco SmartCell (courtesy
of Svedala, UK). of Baker Process, USA).
 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles 1511

Table 5 Wemco cell tank dimensions for different cell sizes (taken from Wemco flotation cell brochure; courtesy of Baker Process,
USA)

Wemco 1#1

Model Volume (m3) Length (m) Width (m) Height (m) Installed motor (kW)

44 0.57 1.12 1.12 0.57 3.75

56 1.1 1.42 1.42 1.1 5.5
66 1.7 1.52 1.68 1.7 7.5
66D 2.8 1.52 1.68 2.8 11
84 4.2 1.6 2.13 4.2 22
120 8.5 2.29 3.05 8.5 30
144 14.2 2.74 3.66 14.2 45/55
164 28.3 3.02 4.17 28.3
190 42.5
225 85

Wemco SmartCells

Model Volume (m3) Height (m) Diameter (m) Installed motor (kW)

300 8.5 1.6 2.59 30

500 14 2.44 2.84 30
750 21 2.57 3.45 40
1500 42.5 2.82 4.32 75
2500 71 3.66 5.31 100
4500 127 4.65 6.2 200
5650 160 4.88 6.83 1.14

Wemco Agitairs

Volume (m3) Length (m) Width (m) Height (m) Installed motor (kW)

1.13 1.22 1.22 0.76

1.7 1.52 1.52 0.76
2.83 1.52 1.52 1.19
4.25 1.6 2.13 1.35
8.5 2.29 3.05 1.35 18.5
14.15 2.74 3.66 1.6 30
28.3 3.05 4.17 2.36 45

disperser, standpipe and a hood. The larger cells are
designed with a false bottom and draught tube. The
SmartCell Sotation machine utilizes the Wemco
1#1 aeration mechanism which is reconRgured and
embedded with an expert control system. The dimen-
sions of different Wemco cells are shown in
Table 5.
The air and pulp circulation in the Wemco cells
are determined by the rotor size, speed and
submergence in the pulp. Liquid circulation and air
transfer are a function of rotor speed, size and sub-
mergence.
Present and Future Trends
Traditionally, Sotation machine design closely fol-
lows the trend of comminution machines in mineral-
processing plants. Due to economic considerations in
the processing of low grade ores, the present com-
minution machines such as crushers, semi-autogous,
autogenous and ball mills are designed for very high
capacities. The Cadia Hill Mine in New South Wales,
Australia, which treats a copper-gold ore at the rate
of 2100 tonnes per hour, utilizes a 12 m diameter
SAG mill (with a 20 MW motor) and two 6.5;11 m
ball mills (each with a 8.75 MW motor). To be com-
patible with the comminution circuit, large capacity
150 m3 Sotation cells are used in the rougher circuit.
At present, cells are large as 300 m3 are being
designed by various manufacturers. Installation of
large cells has many advantages:
1. reduction in capital costs;
2. reduced size of plants;
3. reduced power consumption;
4. reduced maintenance;
1512 II / FLOTATION / Flotation Cell Design: Application of Fundamental Principles
5. easy control;
6. reduced reagent consumption.
However, with increase in cell size, the problem of
machine design and metallurgical scale-up becomes
more acute. The scale-up features that may have been
tolerated on smaller cells are not applicable to larger
cells. The simple similitude considerations used in
terms of dimensionless numbers (power number,
Froude number, air Sow number, Reynolds number)
are not sufRcient to design large machines. The
development and evaluation costs rapidly increase
with cell size, which calls for a more rational and
fundamental basis in cell design. Extensive research at
the Julius Kruttschnitt Mineral Research Centre in
Brisbane has shown that bubble surface area Sux or
Sb is an important criterion for metallurgical scale-up,
which will gain more prominence in the future and
will be considered as a parameter in conjunction with
other important dimensionless numbers used in ma-
chine design and scale-up.
An increase in cell sizes also requires more ef-
fective froth transportation due to the increase in
travel time of bubble}particle aggregates which re-
sults in high drop-back and low froth recovery. To
address the problem of froth transportation and stab-
ility in large cells, new design features such as internal
launders, double launders, high capacity launders,
booster cones, froth crowders, cross-launders and
beehive launders are emerging. More work will be
carried out by cell manufacturers and researchers to
understand froth transportation and froth recovery.
The effect of the interactions of different
launder designs, froth crowders and cell-operating
parameters such as impeller speed, air rate and froth
depth will be the subject of further investigation
for better cell design and optimization of cell
operation.
The design differences of various cells mar-
keted by different manufacturers are in fact dif-
ferences in impeller/stator mechanisms and air input
systems (either self-induced or forced air type through
a standpipe with a solid shaft or through a hollow
shaft). However, the design of tanks is similar for
different cell types, and resembles the cylindrical
design of the old Maxwell cells. The launder and
froth crowding devices in different designs are
tailor-made to suit different applications.
The large new Sotation cells are equipped with
integrated control systems. The recent trend of instal-
lation of a few large cells in a circuit will see more
control instrumentation like air Sow control, variable
speed drive for speed control, as well as online
measurement equipment for monitoring bubble size,
superRcial gas velocity, gas hold-up and bubble sur-
face area Sux, which will be used for better cell
performance optimization. Froth vision equipment
will also gain prominence for better control of froth
in the large Sotation cells.
The development of Sotation cells will continue
as more and more Rne particle processing will be
necessary in future. The large Sotation machines
will have to be designed to generate very small
bubbles and a high degree of microturbulence for
effective bubble}particle collision to remain
competitive against other novel technologies like high
intensity pneumatic cells. Entrainment will be a ma-
jor issue in concentrators, which will need reRnement
of froth-washing technologies in mechanical Sotation
cells.
Acknowledgements
The authors would like to thank the manufacturers of
Bateman, Dorr-Oliver, Outokumpu, Svedala and
Wemco cells for providing sale catalogues, pictures
and other information regarding their Sotation
cells.
Further Reading
Bezuidenhout G (1995) The Bateman Sotation machine.
XIX International Mineral Processing Congress 3:
231}236.
Degner VR (1988) Flotation machine design. In: Klimpel
RR and Luckie PT (eds) Proceedings of Industrial Prac-
tice of Fine Coal Processing, SME/AIME, ch. 16, pp.
135}146. Somerset, CA.
Grainger Allen TJN (1970) Bubble generation in froth
Sotation machines. Transactions of the Institute of Min-
ing and Metallurgy 79: C15}C22.
Harris CC (1986) Flotation machine design, scale-up and
performance: database. In: Advances in Mineral Pro-
ceedings, ch. 37, pp. 618}638. SME/AIME.
Nitti T and Tarvainen M (1982) Experiences with large
Outokumpu Sotation machines. In: XIV International
Mineral Processing Congress, pp. VI 7.1}7.12. Toronto,
Canada.
Schubert H (1985) On some aspects of the hydrodynamics
of Sotation process. In: Forssberg KSE (ed.) Flotation
of Sulphide Minerals, pp. 337}355. Amsterdam:
Elsevier.
Smith EL, Prevett MJ and Lawrence GA (1982) An
improved mechanism for large Sotation cells. In: XV
International Mineral Processing Congress, pp. VI
9.1}9.19.
Young P (1982) Flotation machines. Ming. Mag. 146:
35}59.

Foam Fractionation
G. Narsimhan, Purdue University,
West Lafayette, IN, USA
Copyright ^ 2000 Academic Press
Introduction
Foam concentration/fractionation is a separation
technique in which surface-active solutes are either
concentrated from a dilute solution or separated from
a mixture by preferential adsorption at a gas}liquid
interface created by sparging an inert gas through the
solution. These gas bubbles entrain the surfactant
solution and form a stable foam with a large
gas}liquid interfacial area. As the foam moves
through the column, the surfactant solution tends to
drain due to gravity and capillary forces. This results
in a decrease in the amount of liquid in the foam. The
reduction in the entrained liquid is Rrst associated
with the bubbles forming the closest spherical pack-
ing, after which they will deform to a dodecahedral
shape and then possibly coalesce. Consequently, there
is an increase in the gas}liquid interfacial area per
unit volume of the liquid. The surfactant tends to
adsorb preferentially at the gas}liquid interface. At
the top of the column, the foam is sent to a foam
breaker where the foam is broken either mechanically
or chemically. This results in either enrichment or
concentration of more surface-active protein because
of the recovery of adsorbed protein from the
gas}liquid interface into the bulk entrained liquid.
In the case of a dilute solution of a single protein,
Table 1 Foam fractionation of proteins
Proteins separated Experimental set-up
Choline esterase Batch
Pepsin, rennin Batch
Sodium cholate Batch
Apple proteins Semi-batch
Bovine serum albumin Batch
Bovine serum albumin Continuous
Potato proteins Batch with recycle
Catalase, amylase Batch
Streptokinase Batch
Lysozyme, human serum albumin Batch
Acid phosphatase Batch
Urease, catalase Batch
Bovine serum albumin-DNA, lysozyme-DNA Batch
Placental proteins Continuous
II / FLOTATION / Foam Fractionation 1513
the extent of enrichment would depend upon the
relative amount of adsorbed protein compared to that
in the bulk entrained liquid. In the case of a mixture
of proteins in solution, the separation of a protein
from the mixture would depend upon the extent of
preferential adsorption of that protein at the
gas}liquid interface. Since the adsorption isotherm
usually leads to a much higher proportion of adsor-
bed protein at very low bulk concentrations, foam
concentration is very effective for extremely di-
lute solutions.
Because of the presence of hydrophilic and hydro-
phobic functional groups, proteins are surface active.
Therefore, foam-based separations are viable for con-
centration/separation of protein solutions. Foam-
based separation has been applied to various proteins
and enzymes. Experimental investigation has been
summarized in Table 1. This review highlights the
theoretical aspects of prediction of enrichment and
separation of proteins and enzymes in a foam frac-
tionation column.
Different Modes of Operation
of a Foam Column
Figure 1 depicts the different modes of operation
of a foam fractionation column. The simplest mode is
the production of a protein-rich concentrate phase
from a dilute aqueous protein solution. This can be
operated as semi-batch mode (Figure 1A), in which
a pool of protein solution is maintained at the bottom
Reference
Schultz, 1937; Bader et al., 1944
Andrews and Schultz, 1945
Bader et al., 1944
Davis et al., 1949
Schnepf and Gaden, 1959
Gehle and Schugerl, 1984
Ahmad, 1975a,b
Brown et al., 1990
Uraizee and Narsimhan, 1996
Weijenberg et al., 1978
Charm et al., 1966
Holmstrom, 1968
Lalchev and Exerowa, 1981
London and Hudson, 1953
London et al., 1954
Lalchev et al., 1982
Sarkar et al., 1987
1514 II / FLOTATION / Foam Fractionation
Figure 1 Different modes of operation of a foam fractionation
column.
of a column and is sparged with an inert gas which
forms the foam. The foam is continuously removed at
the top of the column, sent to a foam breaker and the
top product collected. Since the most surface-active
protein is preferentially removed from the solution,
the solution would progressively get depleted in that
protein as time progresses. As a result, the pool would
get enriched in other components in the case of mix-
tures. In continuous operation, a feed stream of pro-
tein solution is introduced into the pool and the
bottom product withdrawn (Figure 1B). Sparging of
gas bubbles mixes the liquid pool well enough so that
the bottom product is at the same composition as the
liquid pool. In addition, the continuous foam column
can also be operated in stripping, enriching or com-
bined modes. In the stripping mode, the object is to
remove, almost completely, protein from a dilute
solution. In this mode, the feed is introduced into the
foam and trickles down countercurrently through
the rising foam (Figure 1C). The protein concentra-
tion in the liquid below the feed-point falls with foam
depth, due to it being adsorbed on to the rising bubble
surface. There is a net upSow of solution through the
foam maintained by entrained up-Sowing liquid from
the pool. If the foam column is deep enough the
protein adsorbed on the bubble surface
F, will be in
equilibrium with the feed liquid concentration cF, and
the pool liquid concentration will be very low. Conse-
quently, the bottom product is stripped of more pro-
tein than that in the simple mode of operation. In the
enriching mode (Figure 1D), the feed stream is intro-
duced into the liquid pool and part of the top product
that is obtained by collapsing the foam is reSuxed
into the column. Protein-rich reSux Sows down
countercurrently through the foam resulting in fur-
ther enrichment of protein in the top product. In the
combined mode (Figure 1E), the feed is introduced
into the foam and the external reSux is used. Part of
the column above the feed acts as an enricher,
whereas the bottom part of the column acts as a
stripper.
It is reasonable to assume that the residence time of
the bubbles through the liquid pool is sufRciently
large for protein adsorption to reach close to equilib-
rium so that the surface concentration of protein at
the gas}liquid interface can be assumed to be close to
the equilibrium value. Also, if bubble coalescence in
the foam bed is negligible, the concentration of pro-
tein in the interstitial liquid can be expected to be the
same as that in the liquid pool. For simple mode of
operation of the foam column with a continuous feed
stream consisting of a dilute protein solution, the top
product concentration cD, is related to the pool con-
centration cB via:
6G
B
cD"cB# [1]
dD
where G is the gas Sow rate, D is the top product Sow
rate, d is the bubble size and
B is the equilibrium
surface concentration of protein at the gas}liquid
interface corresponding to the pool concentration. In
the above equation, the Rrst term on the right-hand
side is the contribution to the protein concentration
from the bulk interstitial liquid before the foam is
collapsed and the second term is the contribution
from the adsorbed protein at the gas}liquid interface
which is recovered into the bulk upon collapse of the
foam. A mass balance around the column now gives
the following equations for the top product concen-
tration cD and the bottom product concentration
cB respectively (Lemlich, 1968):
6G
B B
cD"cF# [2]
d F(F!B)
and:
6G
B
cB"cF! [3]
Fd
where cF is the feed concentration, cB is the pool
concentration, F is the feed Sow rate, and B is the
bottoms Sow rate. In the case of binary mixture of
 II / FLOTATION / Foam Fractionation 1515

two proteins, the separation efRciency S, deRned simple mode of operation. The separation efR-
as the ratio of the two enrichments, is given by: ciency for a binary mixture is given by:

6G
2(cB, 2) B 6.59G
2(cF, 2) 1
1# 1#
cD, 2 cF, 1 dcF, 2 F(F!B) cD, 2 cF, 1 dcF, 2 (F!B)
S" " [4] S" " [8]
cF, 2 cD, 1 6G
1(cB, 1) B cF, 2 cD, 1 6.59G
1(cF, 1) 1
1# 1#
dcF, 1 F(F!B) dcF, 1 (F!B)

where the subscripts 1 and 2 refer to components

1 and 2 and
i (cB, i) is the equilibrium surface concen-
tration of component i corresponding to the bulk
concentration cB, i. It can easily be seen that the sepa-
ration ratio is greater than unity if component 2 is
more surface active than 1. Also, in the above equa-
tion factor 6 arises because the area per unit volume
of spherical bubbles of diameter d is 6/d. If the
dodecahedral shape of the bubbles in the foam is to be
accounted for, factor 6 is to be replaced with 6.59.
For a Langmuir adsorption isotherm, the surface
concentration of proteins is related to the bulk con-
centration via:
Ki ci

i" , i"1, 2 [5]
1# Ki ai ci
i
where Ki is the equilibrium constant, ci is the bulk
concentration and ai is the area occupied by a protein
molecule.
In the stripping mode, the feed stream is introduced
into the foam (Figure 1C). For a long stripping col-
umn, the protein concentration of downSowing inter-
stitial liquid will approach that of entrained liquid in
the foam. The two concentrations will approach each
other at the feed level. Therefore, the protein concen-
tration of the interstitial liquid at the top can be taken
to be the feed concentration. Therefore, mass balance
around the foam column yields (Lemlich, 1968):
Analysis of Foam Column
for the Prediction of Liquid Hold-up,
Enrichment and Separation Factor
Various phenomena that take place in a foam column
are shown schematically in Figure 2. Bubbles are
formed by the sparger into the liquid pool. Proteins
adsorb on to the bubbles during their formation and
their passage through the liquid pool. The rate of
adsorption of protein depends on the rate of dif-
fusion of protein molecules to the gas}liquid interface
as well as on the adsorption activation energy at the
bubble surface. The extent of the surface coverage at
the gas}liquid interface is dependent on the time of
formation of the bubbles and its residence time in the
liquid pool (Uraizee and Narsimhan, 1995). The
foam bed consists of an assemblage of gas bubbles
separated by thin liquid Rlms, creating a large
gas}liquid interfacial area. The size distribution of the
bubbles depends on the type of sparger employed for
bubble formation. A sintered disc with Rne pores

6.59G
F
cD"cF# [6]
d(F!B)

and:

6.59G
F
cB"cF! [7]
Bd

where
F is the equilibrium surface concentration of

the protein at the gas}liquid interface corresponding
to the feed concentration. Since
F5
B, B4F, com-
parison of eqns [2] and [3] with eqns [6] and [7]
indicates that the stripping mode yields a leaner bot- Figure 2 Schematic of various phenomena that take place in
tom product and richer top product compared to the a foam column.
1516 II / FLOTATION / Foam Fractionation
usually results in a wide distribution of bubble sizes
whereas either capillaries or oriRces of uniform sizes
lead to more or less uniform bubble sizes. Since the
volume fraction of liquid in a foam is usually very
small, the gas bubbles are distorted and are usually
approximated by a dodecahedron (Narsimhan and
Ruckenstein, 1986). A typical gas bubble is shown in
Figure 3A. The neighbouring gas bubbles are as-
sumed to be separated by planar Rlms of the continu-
ous liquid phase. Where three bubbles touch, their
Rlms drain laterally into a Plateau border. This is
a channel whose length is the length of a side of the
touching dodecahedral bubbles, and whose walls
have a sharp concave curvature of radius Rp (Fig-
ure 3B). This lateral Sow is caused by a pressure drop
P between the liquid pressure in the Rlm, which is
essentially the air pressure in the bubble, and the
pressure of the liquid in the Plateau border. If  is the
surface tension of the bubble}liquid interface, then:
 foam, N is the number of bubbles per unit volume of
P" [9]
Rp
The liquid in the Plateau border drains under grav-
ity. Consequently, the liquid hold-up decreases with
foam height. The lateral Sow out of the thin Rlms
separating the gas bubbles will cause them to thin
further, possibly causing them to rupture because of
instability resulting from the growth of thermal and
mechanical perturbations thus leading to bubble co-
alescence. Coalescence leads to internal reSux of the
liquid from the ruptured Rlms into the Plateau bor-
ders and a decrease in the interfacial area because of
an increase in the bubble size. The former tends to
enhance separation (enrichment) whereas the latter is
detrimental. The former effect is usually pre-
dominant, so that coalescence leads to higher separ-
ation (enrichment). Only when coalescence is excess-
ive, collapse of the foam bed occurs. When there is
a broad distribution of bubble sizes, diffusion of
gas from smaller to larger bubbles may occur because
of the difference in the capillary pressure (being
Figure 3 Schematic of a bubble in a foam column.
inversely proportional to bubble size) thus leading to
the growth of larger bubbles at the expense of smaller
ones.
In order to predict the liquid hold-up as a function
of foam height, one needs to solve the balance equa-
tions for drainage of liquid from thin Rlms into the
Plateau borders. The equations describing the rate of
change, with vertical position, of the volumetric
hold-up of the liquid in the Rlms, caused by their
drainage into the Plateau borders and bubble coales-
cence is given by (Uraizee and Narsimhan, 1995):
d N
! (nf Af xf)!Nnf Af V! nf Af xf "0
dz 2
[10]
where xf is the Rlm thickness, nf is the number of Rlms
per bubble, Af is the area of the Rlm,  is the number
of bubbles entrained per unit cross-section of the
the foam, and V is the velocity of drainage of the Rlm
and  is the coalescence frequency.  and N can
be related to the superRcial gas velocity G, liquid
hold-up , and the bubble volume v through:
G 1!
" , N" [11]
v v
As before, the equation describing the rate of change,
with vertical position, of volumetric liquid hold-up in
the Plateau borders, caused by Sow from the Rlms
into the Plateau borders and bubble coalescence, and
gravity drainage is given by (Uraizee and Narsimhan,
1995):


d d 4
! (npapl)# NnpapuR
dz dz 15
N
#Nnf Af V# nf Af xf "0 [12]
2
 II / FLOTATION / Foam Fractionation 1517

where np is the number of Plateau borders per bubble, column is much smaller than the entrainment of the
ap is the area of cross-section of Plateau border, R is liquid at the foam}liquid interface. Hence, the mater-
the radius of the bubble, l is the length of the Plateau ial balance around the foam yields:
border, and u is the velocity of gravity drainage of
Plateau borders. Similarly, the protein balance in the G0 4
foam can be written as: " N0npap0uR0 [17]
1!0 15


d d 4
! (np ap lcp, i)# NnpuRcp, i #Nnf Af Vf cf, i
dz dz 15
N N
# nf Af xf cf, i# nf Af
i"0, i"1, 2
2 2
[13]
where cp, i and cf, i are the protein concentrations in the
Plateau border and Rlm respectively. In the absence of
coalescence, they would be equal. However, coales-
cence would enrich the liquid in the Plateau border
because of reSux of adsorbed protein from the rup-
tured thin Rlms. In the above equation,
i is related to
the bulk concentration ci via the Langmuir adsorption
isotherm given by eqn [5]. In eqns [12] and [13],
V and u are the velocities of drainage of Rlms and
Plateau borders, respectively. For an immobile
gas}liquid interface, the velocity of drainage of Rlms
into the Plateau borders can be evaluated from the
Reynolds equation:
The inlet bubble size R0 depends on the type of
sparger and the superRcial gas velocity G. The above
two equations can be solved for xf 0 and ap0. Also, the
protein concentration in Rlms and Plateau borders at
the foam}liquid interface can be taken as equal to the
pool concentration, i.e.:
cf 0"cp0"cB [18]
The pool concentration should satisfy the overall pro-
tein balance given by:
FcF"BcB#TcT [19]
where F, B and T refer to feed, bottom and top
product Sow rates expressed per unit area of cross-
section of the foam column. The overall mass balance
can be written as:
F"B#T [20]

2 Px3f Eqns [10] to [13] can be solved with the initial condi-
V" [14]
3 R2f tions [16] to [20] to give the proRles of xf, ap and cp, i.
The liquid hold-up  at any foam height can then be
where Rf is the radius of the Rlm,  is the viscosity, calculated via:
and P is the pressure drop under which the Rlm
drains. The velocity of drainage of the Plateau bor- "Nnf Af xf#Nnpapl [21]
ders for immobile gas}liquid interface is given by:
The enrichment ei for each component is given by
gap (Uraizee and Narsimhan, 1995):
u" [15]
20(3
(Nnf Af xf cf, i#Nnp ap lcp, i#Nnf Af
i)T
ei"
where  is the density of the liquid. cF, iT
Eqns [10], [12] and [13] are initial value problems [22]
which have to be solved with proper initial conditions
at the foam}liquid interface to evaluate xf and ap and
cp, i as a function of foam height. where (2)T refers to the evaluation of the quantity
The liquid hold-up at the foam}liquid interface within the parenthesis at the top of the column. The
(z"0) can be set to the void fraction of spheres separation factor S is then given by:
(Uraizee and Narsimhan, 1995):
e2
S" [23]
0"Nnf Af xf 0#Nnpap0l"0.26 [16] e1

As the liquid hold-up at the top of the column is much The above analysis assumes adsorption equilibrium
smaller than 0.26, the Sow rate at the top of the for the surface concentration of proteins at the
1518 II / FLOTATION / Foam Fractionation

air}water interface. Uraizee and Narsimhan (1995)

have modiRed this analysis to account for the kinetics
of adsorption of proteins on to the gas bubbles during
their travel through the liquid pool before the
formation of foam and demonstrated the effects
of different parameters including the kinetics of
adsorption and pool height on enrichment and recov-
ery of proteins.

Effect of Operating Conditions

on Enrichment and Separation
The operating conditions that can be varied in a foam
column are the superRcial gas velocity G, the bubble
size R, the column height L, feed Sow rate F, the feed
concentration cf and the mode of operation. In addi-
tion, the separation will also be inSuenced by the
viscosity of the feed and the extent of bubble coales-
cence in the foam column.
Protein enrichment depends on the total amount of
protein selectively adsorbed at the gas}liquid inter-
face as well as on the liquid hold-up in the foam.
Smaller liquid hold-ups result in a larger interfacial
area per unit volume of the liquid and therefore in
larger enrichment. At higher superRcial gas velocities,
more liquid is entrained by the gas bubbles from the
liquid pool leading to higher liquid hold-ups in the
foam column and consequently to smaller enrich-
ment. As the bubble size increases, a larger propor-
tion of the liquid that is entrained by the foam is
distributed in the Rlm, resulting in a faster drainage
rate. On the other hand, an increase in the bubble size
results in a decrease in the interfacial area per unit
volume. Because of the above two opposing ef-
fects, there exists an optimum bubble size at which
enrichment may be maximum (Narsimhan and
Ruckenstein, 1986) for one component protein solu-
tion as shown in Figure 4. In addition, this maximum
is found to be more pronounced at smaller superRcial
gas velocities. Narsimhan and Ruckenstein (1986)
have developed a population balance model to ac-
count for the bubble size distribution in the descrip-
tion of drainage and coalescence in a foam bed. Their
model was able to predict the change in the bubble
size distribution as a result of coalescence. The results
indicated collapse of the foam bed for broader inlet
bubble size distribution with a coefRcient of vari-
ation above a critical value. In the case of a mixture
of proteins, however, the separation efRciency
would depend on the preferential adsorption of one
protein over the other components as can be seen
from eqns [22] and [23]. As expected, the separation
efRciency is higher for the protein which adsorbs
the most at the gas}liquid interface with a higher
value of
. As a result, the separation efRciency
Figure 4 Effect of the inlet bubble size on the enrichment for
"10\2 P, s"10\4 sP, and c0"10\7 gmol mL\1. (Repro-
duced with permission from Narsimhan and Ruckenstein,
1986a.)
would be higher for larger values of Langmuir ad-
sorption parameter Ki as can be seen from eqn [5]. An
increase in the viscosity of the feed would result both
in a larger amount of liquid entrained by the foam as
well as slower liquid drainage leading to larger liquid
hold-up. Also, an increase in the viscosity of the feed
would tend to stabilize the foam resulting in lower
bubble coalescence. Both these effects will result
in lower protein enrichment. Bubble coalescence in
a foam column leads to: (i) an increase in the protein
concentration due to internal reSux with subsequent
increase in the surface concentration; (ii) a decrease in
the liquid hold-up because of increased liquid drain-
age rates as a result of larger bubble sizes; and (iii)
a decrease in the total surface area because of larger
bubble sizes. The Rrst effect results in more pro-
tein adsorption per unit area at the gas}liquid inter-
face. The second effect leads to higher surface
area per unit volume of the liquid. The third ef-
fect leads to a decrease in the total amount of protein
adsorbed at the interface. Consequently, the Rrst two
effects lead to an increase in the enrichment and
separation whereas the second and third effects
lead to lower recovery. The second effect may be
predominant since coalescence was found to result in
an increase in protein enrichment as well as recovery
(Uraizee and Narsimhan, 1995). The separation ef-
Rciency, as one would expect, depends on the relative
 II / FLOTATION / Foam Fractionation 1519

surface activities of proteins in a binary mixture.

For larger values of Langmuir isotherm constant
Ki (more surface active), the separation efRciency
increases. In fact, the calculations show that the
separation efRciency increases linearly with the
ratio K2/K1 (Uraizee and Narsimhan, 1997). How-
ever, the separation efRciency was found to de-
crease rapidly with the feed concentration of the
protein (Uraizee and Narsimhan, 1997).
Brown et al. (1990) measured enrichment and re-
covery in a continuous foam concentration column
for bovine serum albumin (BSA). In their experi-
ments, foam was generated by sparging nitrogen gas
through a glass frit. As a result, the foam consisted of
nonuniform size distribution of bubbles. They com-
pared the experimental data with predictions based
on a model similar to the one described above but
neglecting drop coalescence. Their experimental data
showed a decrease in the protein enrichment with
superRcial gas velocity. The model predictions agreed
fairly well for the highest feed concentration of
0.1 wt% as shown in Figure 5. The experimental
enrichments were found to be larger than the model
predictions (even for the largest bubble size in the
foam) with the deviation being larger at lower feed
concentrations. This was believed to be due to the Figure 6 Comparison of experimental results with model pre-
dictions for BSA; feed concentration 0.1 wt%, bubble diameter
fact that drop coalescence in the foam column be-
1.9;10\3 m, gas velocity 2.6;10\3 m s\1, foam height
came increasingly important at lower feed concentra- 1.3;10\1 m, F"2;10\5 m s\1, pH 4.8, ionic strength 0.1 M.
tions as conRrmed by experimental measurements of (䢇) Experimental data. (䉭) Model predictions accounting for
bubble size with the height of the column. kinetics of adsorption as well as coalescence. (N) Model
Uraizee and Narsimhan (1996) also observed a de- predictions accounting only for kinetics of adsorption. ( 䉬) Model
prediction accounting only for coalescence assuming equilibrium
crease in enrichment with gas velocity for foam
surface concentration is shown in the inset. (Reproduced with
concentration of BSA in their continuous foam con- permission from Uraizee and Narsimhan (1996).)

centration experiments in which the foam was gener-

Figure 5 Effect of superficial gas velocity on protein enrichment
for cF"0.1 wt%. F"0.02 cm s\1, I"0.1 M, pH"7, z"5 cm.
The curves refer to model predictions for different bubble sizes.
(Reproduced with permission from Brown et al., 1990.)
ated by sparging nitrogen through a capillary bundle
thus resulting in a foam of uniform bubble sizes. In
their experiments, the residence time of the bubbles in
the liquid pool was varied by varying the pool height.
Interestingly, protein enrichment was found to in-
crease with pool height at sufRciently high pool
heights, thus indicating the importance of kinetics of
adsorption of protein on to the gas}liquid interface
on enrichment. At low pool heights, however, they
observed an increase in protein enrichment with a de-
crease in pool height due to excessive bubble coales-
cence in the foam. Their model, which accounted
both for the kinetics of protein adsorption as well as
coalescence, was able to explain the increase in pro-
tein enrichment due to bubble coalescence at small
pool heights and an increase in enrichment with pool
heights at larger pool heights. A comparison of the
experimental data with their model predictions is
shown in Figure 6.
1520 II / FLOTATION / Foam Fractionation
Ahmed (1975) observed an increase in the sep-
aration efRciency of albumin with the superRcial
gas velocity with the value reaching a plateau at
sufRciently high gas velocities. Schnepf and
Gaden (1959) and Ahmad (1975) reported a max-
imum protein enrichment at the isoelectric point of
the protein which can be explained by the maximum
protein adsorption at the interface due to the absence
of electrostatic energy barrier for adsorption. How-
ever, this maximum was found to be considerably less
pronounced at higher protein concentrations. Protein
enrichment was also inSuenced by the change in the
bubble size at different pH (Brown et al., 1990).
Separation efRciency of albumin was found to
decrease dramatically as the foam height increased
from 3 to 17 cm (Ahmed, 1975). Even though enrich-
ment increased with foam height because of internal
reSux resulting from increased drop coalescence, the
top product Sow rate was also found to decrease
dramatically due to faster drainage. As a result, the
protein separation was less at higher foam heights.
Ahmed (1975) also found that the introduction of the
feed stream into the foam instead of liquid pool
increased the separation efRciency because the
foam column was operated in the combined mode
with an enricher and stripper.
In conclusion, the main attractive features of foam
fractionation are its low capital and operating costs.
Therefore, it can be employed as a Rrst step for
preconcentration/separation before more expensive
separation methods can be used. More work is
needed to establish the applicability of foam frac-
tionation as a viable separation method for mixtures
of proteins and to develop new processes based on
this technique. Few experimental data are available
on the adsorption isotherm and kinetics on to
gas}liquid interface for mixtures of proteins. More
importantly, it is necessary to probe denaturation (if
any) of proteins and enzymes when subjected to
foaming.
See also: II /Flotation: Bubble-Particle Capture; Froth
Processes and the Design of Column Flotation Cells;
Historical Development.
Further Reading
Ahmed SI (1975) Laws of foam formation and foam frac-
tionation. 1. The effect of different operating
parameters on the foam fractionation of albumin from
a solution containing organic and inorganic materials.
Separation Science 10: 673.
Ahmed SI (1975) Laws of foam formation and foam frac-
tionation. 2. The inSuence of different association
conditions on surfactants, glycerides, sugar and salts on
the foam fractionation of albumins. Separation Science
10: 689.
Bader R, Schultz F and Stacey M (1944) A crystalline serum
mucoprotein with high choline esterase activity. Nature
154: 183.
Bader R and Schultz F (1946) Fractionation by adsorption
and crystallization on foam. Part II. Experiments
with bile salts. Transactions of the Faraday Society 42:
571.
Brown LK, Narsimhan G and Wankat PC (1990) Foam
fractionation of globular proteins. Biotechnology and
Bioengineering 36: 947.
Charm SE, Morningstar J, Matteo C and Paltiel B (1966)
The separation and puriRcation of enzymes through
foaming. Analytical Biochemistry 15: 498.
Davis SG, Fellers CR and Esselen WB (1949) Application of
foam fractionation procedures to the isolation of fruit
juices. Food Technology 3: 198.
Gehle RD and Schugerl K (1984) Protein recovery by
continuous fractionation. Applied Microbiology Bio-
technology 20: 133.
Holmstrom B (1968) Foam concentration of streptokinase
from crude culture Rltrates. Biotechnology and Bioen-
gineering 10: 551.
Lalchev Z, Dimitrova L, Txvetkova P and Exerowa
D (1982) Foam separation of DNA and proteins from
solutions. Biotechnology and Bioengineering 24: 2253.
Lalchev A and Exerowa D (1981) Concentration of pro-
teins by foaming. Biotechnology and Bioengineering 23:
669.
Lemlich R (1968) Principles of foam fractionation. In: Perry
ES (ed.) Progress in Separation and PuriTcation, vol. 1,
pp. 1}56. New York: Interscience.
London M, Cohen M and Hudson P (1954) Some general
characteristics of enzyme foam fractionation. Bio-
chimica Biophysica Acta 13: 111.
London M and Hudson P (1953) Studies on the puriRcation
of acid prostatic phosphatase. Archives of Biochimica
Biophysica Acta 46: 141.
Narsimhan G and Ruckenstein E (1986) Hydro-
dynamics, enrichment and collapse in foams. Langmuir
2: 230.
Narsimhan G and Ruckenstein E (1986) Effect of
bubble size distribution on the enrichment and collapse
in foams. Langmuir 2: 494.
Sarkar P, Bhattacharya P, Mukherjee RN and Muckerjee
M (1987) Isolation and puriRcation of protease from
human placenta by foam fractionation. Biotechnology
and Bioengineering 29: 934.
Schnepf RW and Gaden EL (1959) Foam fractionation of
proteins: Concentration of aqueous solutions of BSA.
Journal of Biochemical, Microbiological and Technolo-
gical Engineering 1: 1.
Schultz F (1937) Adsorption on foams. Nature 139:
629.
Uraizee F and Narsimhan G (1995) A model for
continuous foam concentration of proteins: Effects
of kinetics of adsorption of proteins and coalescence
of foam. Separation Science and Technology 36(6):
847.
 II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
Uraizee F and Narsimhan G (1996) Effects of kinetics of
adsorption and coalescence on continuous foam concen-
tration of proteins: Comparison of experimental results
with model predictions. Biotechnology and Bioengineer-
ing 51: 384.
Froth Processes and the Design of Column Flotation Cells
I. M. Flint, Canadian Process Technologies Inc.,
Vancouver, BC, Canada
M. A. Burstein, NPACI Edcenter on Computational
Science and Engineering, San Diego State University,
San Diego, CA, USA
Copyright ^ 2000 Academic Press
Introduction
The function of a Sotation column is selectively to
separate certain suspended solid particles or liquid
droplets based on their surface properties. Bubbles rise
and particles (drops) settle within the vessel, and colli-
sions are highly dependent on gravitational momentum.
The vessel is a multiphase contacting/heterocoagulation
device where the dispersed phase to be removed at-
taches to the bubbles and accumulates at the top of the
column in the form of froth. The latter overSows to
launders. In this quiescent system, transport, dispersion
and mixing of materials are induced by the motion of
gas bubbles in the continuous liquid medium.
For the purpose of designing columns, immiscible
liquid droplets are considered as acting as solid
spheres of an appropriate size and density: thus,
a ‘particle’ may represent either a solid or a liquid.
Almost all Sotation columns are operated in the
countercurrent regime where slurry moves down-
wards against a continuous rising bubble swarm. This
type of Sow increases efRciency (selectivity) of
separation as the distance between discharge ports for
overSow and underSow is large. In some cases, for
example for the Sotation of very coarse particles,
co-current columns can be considered in order to
increase particle residence time and reduce loaded
bubble rise time. Unless otherwise stated, all of this
article is related to countercurrent columns.
Initial Design Data
The feed transport Suid must be characterized in
terms of liquid Sow rate and chemical composition.
Component solids or immiscible liquid Sow rate,
material composition and size distribution must
also be known. In all cases, mean values, standard
1521
Weijenberg DC, Mulder JJ, Drinkenburg AAH and
Stemerding S (1978) The recovery of protein from po-
tato juice waste water by foam separation. Journal of
Engineering and Chemical Processing, Design and
Development 17: 209.
deviations and design maxima and minima are
required.
Test work must be done, or approximations made,
to determine the Sotation characteristics of the mater-
ial to be separated, including rate constants and max-
imum recovery for all material and particle (droplet)
size fractions. Process targets must be well under-
stood, including the desired quality of products and
recovery. Data error must be minimized since it dir-
ectly impacts on the accuracy of the design scale-up.
Site-speciRc information is also required for Rnal
designs. This includes limitations in dimensions due
to plant layout, civil engineering speciRcations, in-
cluding such items as wind loading, earthquake con-
siderations, supporting platforms and others.
General Dimensions
Typically, columns range in height from 6 to 15 m.
This height is dictated by the dimensions of the dif-
ferent zones within the column but is most inSuenced
by the collection zone height.
Column cross-sections are usually round or rectan-
gular. Cylindrical columns do not have special Sow
conditions at the corners. They, therefore, usually
have a more uniform air and feed distribution, less
tank weight due to the self-supporting nature of the
structure and less wall area per unit operating vol-
ume. Rectangular columns use floor space more
efRciently and are easier to bafSe. The cross-sectional
area is usually constant throughout the vessel and is
determined by carrying capacity and residence time
considerations in the collection and froth zones. Typi-
cal industrial cell cross-sectional areas range from less
than 1 m2 to more than 12 m2.
Column Zones
The Sotation column, as generally built, is composed
of a number of distinct zones. Under the spargers
there is a dead volume (underSow zone) which is only
used to remove slurry from the vessel. The volume
between the spargers and the feed port is called the
collection zone. The volume between the feed port
1522 II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
Figure 1 Column hydrodynamic zone.
and the froth interface is called the cleaning (recollec-
tion) zone and above the interface is the froth zone.
The froth zone may be further divided into the wash-
ing zone, if it is under the wash water distributor, and
the free drainage zone, if it is above. These zones are
illustrated in Figure 1.
Under]ow Zone
The physical dimensions of this zone should be mini-
mized since its roles are to ensure that there are no
small bubbles entrained by the downward Sow to the
underSow stream, that sloughing of the solids par-
ticles does not occur, and that the outSow from the
base of the column does not create unwanted Sow
patterns within the active zones of the vessel. Bubble
entrainment to the underSow is obvious, as in this
case frothing occurs in the tailing sump or in the next
open tank downstream of the column.
A zone underneath the spargers does not contribute
to the Soatation collection or separation. Design con-
siderations of the zone include base cone angle and
placement, underSow exit port conRguration and
height to spargers.
Base Cone Design
In most solids separation applications, the base
can be designed Sat. The solids will eventually
form a false bottom at the angle of repose under the
speciRc Sow conditions. Depth of the base cone
should be selected considering angle of repose of
particles. If sloughing of solids is considered a prob-
lem, then the column can be designed with either
a real or false bottom at an angle greater than the
angle of repose.
Under]ow Port
Generally, the output port is designed to pull from the
cross-sectional centre at the base of the vessel in order
to minimize both the Sow differences within the
collection zone and large scale slurry circulation pat-
terns.
The spargers are placed at a level such that the
furthest descent of bubbles is above the highest ex-
pected solids settling point.
Collection Zone
The collection zone is characterized by a stream of
individual bubbles rising against a descending liquid
or slurry. This is the zone in which the bubble}par-
ticle attachment occurs. The capacity of the column is
dictated in part by the intensity of bubble}particle
collision (number of collisions per unit time), the
probability of attachment, and the bubble surface
area Sux through the vessel (removal ability) in this
zone.
When sizing a column, certain assumptions are
made. These include that the column operates with
dispersed bubbles that rise as a swarm without slugs.
It is also assumed that the Sow of bubbles, liquid and
solids is uniform across the column, and that there are
no large scale vortices.
There are four main collection zone design criteria
which determine the vessel dimensions: Soatable
particle residence time, mixing characteristics,
maximum gas rate and bubble loading. The resulting
volume can usually be achieved with a range of
height-to-diameter options. The Rnal dimensions
are also dictated by layout and economic consider-
ations. It should be emphasized that this zone must
be designed in parallel with the sparging system
and froth zone as each of these parts inSuences the
others. The placement of the column within the
 II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
operating circuit will also impact on the Rnal design
and operation.
Solids Settling
Particles settle within the column system since there is
no mechanical agitation to suspend them. As such,
each particle will have a speciRc hindered settling
velocity dependent on size, density of the particle and
the effective density and viscosity of the suspen-
sion with modiRcations due to bubble-induced mix-
ing. Mixing enhances particle suspension, so small
and/or light particles do not have their own trajectory
and follow liquid Sow more than in two-phase sys-
tems. The settling velocity generally has little inSu-
ence on residence time for particles smaller than
about 20
m, but becomes an important design con-
dition for larger particles.
Particle Residence Time Distribution
(RTD)
Material residence time depends upon the inherent
mineral settling velocity under the conditions within
the column and the superRcial velocity of the slurry.
The total collection zone height divided by the sum-
mation of the hindered settling and slurry velocities
gives a total average residence time for each particle
size and density. More precisely, the particle resi-
dence time is a stochastic parameter inSuenced by the
turbulent mixing and potential macrocirculation pat-
terns within the column.
Elementary Processes
Flotation depends on the elementary processes of
collision and attachment. In columns the probability
of collision between a particle and a bubble remains
virtually constant within the collection zone. There is
a higher probability of attachment at the bottom of
a countercurrent column since bubble surface cover-
age by particles is low for newly formed bubbles. This
maximizes the recovery of the small proportion of
particles targeted for Sotation that are still present in
the lower parts of the zone.
The relative movement of slurry and rising bubbles
is the main source of mixing energy in columns. This
results in a low intensity of the turbulence (low energy
dissipation and large internal scale of turbulence) and,
therefore, low relative bubble}particle velocities and
accelerations. Bubble}particle collision efRciency
is due to gravitational and inertial particle drift from
the liquid streamlines around the rising bubble and
due to the interception. The probability of particle
detachment from bubbles is limited since the velocity
gradient around the bubble is minimized.
1523
Rate Constant
The Soating ability of a material is generally referred
to as a rate constant, similar to chemical processes,
and is assumed (for simplicity) to be of Rrst-order
kinetics for each mineral component and size frac-
tion. The value of this term is dependent on a com-
plex function involving the collision/attachment and
detachment, as well as processes occurring in other
zones of the Sotation column (see below). These data
are generally determined through test work. As mech-
anism and intensities of subprocesses (collision, at-
tachment, detachment, entrainment) in column and
impeller Sotation can be substantially different,
the lab and pilot tests for column design and scale-up
should also be made in columns. First-order Sotation
rates for the components can be determined from the
recoveries in a continuous lab Sotation column, or by
simulation of kinetic tests by recycling column tails
back to the feed line.
Taking into consideration separate Rrst-order kin-
etic models for individual subprocesses and taking
into account free bubble surface reduction in the
upward direction would lead to complicated nonlin-
ear kinetic equations. These are important in under-
standing the physics of the process, but cannot be
used for scale-up or control, due to unavoidably high
error in determining their coefRcients from ex-
perimental data.
Carrying Capacity
The removal capability of the bubbles is called the
carrying capacity and is the general term which char-
acterizes the maximum amount of solids carried by
the air bubbles (either in reference to the maximum
capacity of the column, or to the maximum capacity
per air volume). This refers to the fact that only
a speciRc amount of particles can be attached to and
removed by a certain bubble area. Thus, the max-
imum Soatable solids removal or the surface area of
attached particles is related to bubble surface area
Sux.
Typically, the distance between spargers and the
slurry}froth interface is between 6 and 12 m. This
leads to a substantial mass of particles attached to
a bubble (bubble loading).
As bubbles become loaded by collected particles,
the contact time between particle and bubble reduces
due to the shortening of particle trajectory along the
free bubble surface. This means that the rate of collec-
tion slows as loading increases, especially when the
lower section of a bubble is virtually covered by
attached particles. Detachment probability is also
much higher for particles attached to the upper hemi-
sphere of a bubble.
1524 II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
Smaller bubbles can carry more solids than larger
bubbles, assuming an equal gas rate and that the
loaded bubbles have sufRcient buoyancy to move
against slurry Sow. A smaller weight of Rne particles
can be carried at a constant gas rate and bubble size
distribution than that of coarse particles.
Carrying capacity limitations should be taken into
account when estimating height-to-diameter ratio for
columns working at high overSow (froth) yield. Typi-
cal carrying capacity per unit column cross-sectional
area for base metal minerals Sotation is 2.5 t/(m2 h)
and for coal Sotation 1.5 t/(m2 h).
Gas Rate and Bubble Size
Column cells are operated in the bubbly region where
bubbles rise in a swarm. The Sow regime in the
column may change to the ‘churn-turbulent’ condi-
tion when coalescence is caused by gas entering a re-
gion faster than it can leave as small bubbles. As
smaller bubbles have lower swarm rising velocity, the
Sooding occurs at a lower gas rate for Rne bubble
dispersions. Thus, there is a link between maximum
gas rate at the Sooding point and bubble size. Also,
Sooding is enhanced by countercurrent slurry Sow;
the higher its superRcial velocity, the lower the gas
rate at the Sooding point. At bubble size ranges used
for column Sotation, Sooding occurs typically at
a superRcial air velocity of 2.5}3 cm s\1. More pre-
cise values can be calculated from the drift Sux
model.
It is also possible for uniform countercurrent
froth Sow to occur in the column even at lower
superRcial air and slurry rates when the bubbles are
very stable (gas hold-up at both bubble and froth Sow
regimes can also be estimated based on the drift Sux
model).
Mixing
Columns are commonly sized with a dispersion
method which uses the Peclet number, a dimension-
less criterion, to characterize mixing. It is assumed
that an axial dispersion model adequately reSects
Sow structure in the collection zone. It is also possible
to use a tanks-in-series Sow model. The Peclet num-
ber reSects the ratio between the downward path of
particle and the average length of its stochastic drift
due to mixing (diffusion). It is equal to UL/D,
where U is the mean velocity of the phase of interest
(for particles it is the sum of downward liquid velo-
city and a hindered settling velocity), L is the charac-
teristic length scale for the apparatus (collection zone
height of the column), and D is the turbulent disper-
sion (diffusion) coefRcient. The latter can be
determined by a tracer technique or by using one of
several approximation formulae. D ranges from 0,
for perfectly mixed systems, to inRnity, for plug
Sow. The following variables have an effect
on the Peclet number: bubble size and number
of bubbles (which are dependent on gas rate and
surface tension), slurry rate, particle settling
velocity, collection zone height and diameter. At
a constant collection zone volume, a taller column is
better from a Sow structure perspective as less mixing
is induced. Peclet number can be estimated using one
of the experimental relationships, or from particle
residence time distribution (RTD) similar to that in
chemical reactors or separation equipment. RTD can
be directly measured using a tracer method. Disper-
sion of the RTD can be used to calculate turbulent
diffusion D and other column Sow structure
criteria.
The absence of an agitator limits the formation of
large scale Sow loops unless the column is operated in
a high air rate, churn-turbulent Sow or the feed distri-
bution of either air or slurry is not even. Low mixing
intensity and lack of circulation contours cause par-
ticle residence times to be highly dependent on the
particle settling velocity. Reduced mixing leads to
lowering of local upward Sow intensities which min-
imizes particle entrainment to the froth. Thus, at
a constant collection zone volume (slurry retention
time), its increased height leads to lower mixing
intensity and improved (due to this) metallurgical
results up to the point when restrictions in carrying
capacity limits concentrate (Soat product) yield.
Also, higher superRcial slurry velocity reduces
negative inSuence of mixing and slime entrainment
intensity.
Careful design and positioning of any bafSes
(horizontal or vertical), the feed system, and any
internal piping that may be needed minimize local
turbulence. The feed pipe must be located high
enough in the column to maximize the collection zone
length but also low enough to limit turbulence at the
slurry}froth interface.
Entrainment
Fine and/or light hydrophilic particles may pass up-
wards through the collection zone by entrainment.
There are two forms of entrainment. In the Rrst,
a portion of feed water containing suspended Rne
particles passes into the froth. This type of entrain-
ment can be minimized by maintaining a net down-
ward Sow of water through the upper column zones.
The second form of entrainment is the capture of
particles in the eddies behind a rising bubble. These
particles are also rejected in the froth zone operating
with positive bias.
 II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
Baf]ing
Columns may be vertically bafSed in order, both
to reduce mixing and to lend additional structural
support. An important condition to achieve effec-
tive operation with a bafSed column is an even feed
and air distribution between the compartments, other-
wise detrimental circulation patterns may form between
the bafSed sections. This overall circulation can
make a bafSed column less effective in terms of
Sow structure than a column without bafSes. Nor-
mally, bafSes are installed above and below the
feed distributor in a column, leaving space around feed
pipe(s) and air spargers open to allow even distribution
of the slurry and air bubbles, respectively.
Horizontal bafSes (plates) are not typically
used, though tests have conRrmed their ability to
improve Sotation of coarse particles due to less short-
circuiting in the wall part of the bafSed column.
Physical Dimensions
The total volume of the collection zone is determined
by residence time considerations, having also ac-
counted for mixing and hindered settling of coarse
particles, to achieve target recoveries. A formula
based on an axial dispersion model and Rrst-order
Sotation kinetics is typically used. A minimum dia-
meter is then calculated to allow sufRcient bubble
surface area to Soat the required solids. The diameter
and height must be larger than these minimum num-
bers and any combination can be used as long as the
volume remains above its minimum. The volume
should provide for the necessary retention time with
a correction for mixing, but should not exceed it
substantially. This is critical in the case of selective
Sotation when both components are Soatable and
have different but nonzero rate constants.
The selection of the vessel dimensions is an iterative
process since a change in many of the variables will
change the overall mixing in the vessel.
Access
Periodic maintenance is required, and access to the
inside of the column may be needed. Therefore, ac-
cess manholes and appropriate clearances must be
maintained within the vessel.
Cleaning Zone
The purpose of this zone is to buffer the froth
zone from the turbulence of the feed port. It is located
above the feed port and below the interface with the
froth. It is characterized by rising bubbles that may be
highly loaded with solids rising from the collection
zone and falling solids that have been entrained in the
1525
bubbles’ wake, or have been rejected in the froth zone
by loss of bubble surface area. If a sufRcient
amount of wash water is used, this zone may have
a net downward Sow of slurry. Only a limited number
of previously uncollected particles occurs in this zone
due to the turbulent mixing or entrainment. Since
collection of these particles can also occur in the collec-
tion zone, the height of the cleaning zone should be
minimized but must be sufRcient to allow damp-
ing of the feed turbulence below the froth interface.
In some circumstances the cleaning zone is the
overSow point of the Soat product. This occurs when
there is no froth zone either because a froth cannot be
maintained in a solids Soat, or because a liquid}liquid
separation is being performed. In the latter case an
organic pad may be present.
Froth Zone
This zone is usually present in solid}solid or
solid}liquid separations.
The froth zone in a column cell is characterized by
a rising bed containing a matrix of bubbles, which are
loaded with hydrophobic material, water lamellae
between bubbles and Plateau}Gibbs canals. En-
trained hydrophilic material may be found initially
either in lamellae or in canals. Film (lamella) thinning
and bubble coalescence in froth (syneresis) and drain-
age in Plateau}Gibbs canals are the main mechanisms
of gas hold-up increase and concentrate upgrading
with height in the froth. This is caused by reduction of
the air}liquid interface area and subsequent particle
detachment. Tracer tests indicate that, in some cases,
more upgrading is observed within the froth than
between slurry and lower froth layers.
Quiescent conditions in columns create a stable
froth that allows the addition of wash water. This
water displaces the liquid phase of the feed slurry,
with entrained associated Rne particles, from the
froth lamellae and Plateau}Gibbs canals and allows
the production of an essentially entrainment-free
overSow. In some cases, addition of small amounts of
water into the froth also improves the stability and
rheological properties of the countercurrent froth.
A presence of highly hydrophobic, angular par-
ticles large enough to bridge the lamella between
bubbles, without a population of smaller hydrophilic
particles, causes froth destabilization. In this case the
froth zone design is critical. In extreme cases a froth
bed may not be possible.
Channelling
Uneven distribution or excessive addition of wash
water can cause formation of channels in the froth
1526 II / FLOTATION / Froth Processes and the Design of Column Flotation Cells
and possible froth collapse. Care must be taken in
the design of the distributor to ensure even cross-
sectional wash water Sows.
Froth Zone Dimensions
Although the froth zone usually has the same cross-
sectional area as the collection zone, it may be necked
to promote crowding which increases the upward
velocity in the froth. This may be done when small
amounts of froth are generated, reagent conditions
dictate that the froth will not be stable, or the size
distribution of solid particles in the froth promotes
coalescence of bubbles. It is more common to pre-
serve the overall area and apply internal bafSing
and launders. Internal bafSes may be added to
support the froth, or to contain or localize froth
collapse.
Internal Launders
Syneresis and coalescence occur within the froth
zone. Thus, relative to a localized section of froth,
bubble surface area is lost with time as that section
travels from the slurry}froth interface to the overSow
points. Furthermore, analysis of particle RTD in froth
indicates that horizontal transport to the froth
launder is very slow. For larger diameter columns,
dead zones can form in the central part of the vessel.
Columns normally do not have froth skimmers or
paddles. Therefore, fast froth removal is critically
important for operation and is often achieved by
a series of internal froth launders.
Organic Pad
Liquid}liquid column applications may be operated
with an organic pad on the top of the vessel. Organics
Soated in the collection zone gather at the surface of
the vessel. These may overSow a weir continuously if
the organics concentration is sufRcient or if low
concentrations of organics in the overSow stream are
acceptable. Otherwise, the pad accumulates and is
dumped on a regular basis. If some or all of the
organic compounds present in the system are volatile,
a pad may not be suitable or dumping must be fre-
quent to prevent excessive stripping.
Air-Sparging Systems
The purpose of the sparging system is to distribute
evenly the appropriate-sized bubbles near the bottom
of the column. The sparging system is critical and
must be designed taking into consideration many
elements, including bubble size distribution, max-
imum air rate, bubble coalescence and induced mix-
ing; uniformity of air hold-up across the vessel, min-
imization of scale formation, resistance to wear, re-
quired air pressure and maintenance considerations.
There are many types of spargers used in column
cells. Pneumatic (porous media or perforated) spar-
gers form bubbles at small oriRces. Pneumohydraulic
spargers break up an air stream into bubbles by
a water jet as an air-water mixture is distributed into
the column. The air jet spargers form bubbles through
the high velocity injection of air into the column.
There are also a class of spargers termed external
spargers that aerate the feed slurry, or portion of the
underSow, and use the column as a de-aeration or
bubble separation vessel rather than for particle col-
lection. Combination of external spargers for slurry
pre-aeration with microbubbles and/or dissolved air
with internal spargers to facilitate microbubble buoy-
ancy by adding larger bubbles is optimal for a wide
range of applications (see below). In recent years, the
general trend among major column suppliers is to use
air jet and external types of spargers, although speci-
Rc circumstances dictate the use of other types.
Care must be taken when designing the bubble
distribution system to ensure that an even Sow of
bubbles is generated. Poor air distribution can cause
large scale Sow patterns in the column that are detri-
mental to performance. Macrocirculation zones can
also be caused by a misalignment of the column either
by bows along the length or by offsets from the
vertical.
Pre-Aeration
Columns, by nature, have low turbulent momentum
between the bubbles and particles, meaning that
smaller particles have slow Sotation kinetics in these
vessels. The column is, however, a good separator of
bubbles from the feed slurry, especially if wash water
is added. This feature virtually eliminates hydrophilic
entrainment. In order to improve the collection of Rne
particles, a pre-aeration system or intense Sotation
device can be used. These devices act by creating
a turbulent zone, where the inertial momentum of
both bubbles and particles is high (due to high inten-
sity turbulence and velocity gradients) enabling high-
er recovery of the smallest Soatable particles by
microbubbles. If the pressure in the pre-aeration de-
vice is substantial, a portion of air is dissolved and
then released in a column; normally, nucleation of air
bubbles occurs at a solid surface, thus a collision stage
of Sotation process is eliminated for the cavitation
bubbles.
Pre-aeration devices then feed a modiRed column
which acts as a recollection device for the larger
particles and a bubble coalescence/separation system.

Civil Engineering and Material
of Construction
The Rnal column design must be site-speciRc. There
may be height and/or area considerations due to re-
strictions of space, and weight and loading consider-
ations due to foundation requirements. In addition,
some environmental considerations such as wind
load, earthquake zone and rainfall intensity will af-
fect steel thickness, foundations and attachments,
braces and access platforms. As columns are normally
much taller than mechanical Sotation cells, they are
often located outside, and these factors can play an
important role in column design. There are also pro-
cess considerations like per cent solids, wear factors,
chemical composition of the slurry (pH, etc.) and
particle size distribution which affect the phys-
ical structure, pipe sizing and materials of construc-
tion. In special cases, these units may be designed as
pressurized vessels or as enclosed systems.
For example, many oil}water separation columns
are pressurized or some installations use circulating
inert gases to minimize oxidation. When columns are
installed for oil}water separation duties, mainly on
offshore platforms, a circulating hydrocarbon gas
(propane) is often used instead of air.
Conclusions
Despite its simple design, the scale-up and modelling
of column Sotation is a complex problem. It includes
analysis of three-phase three-dimensional Sow in col-
lection and cleaning zones and in the washed thick
froth layer. In the last few years, a technique for
column design has been developed. Its adequacy has
been conRrmed by many columns installed world-
wide for a wide range of mineral and other applica-
tions.
Historical Development
Z. Xu, University of Alberta, Edmonton, Alberta,
Canada
Copyright ^ 2000 Academic Press
Flotation is a versatile, surface wettability-based sep-
aration process, usually taking place in an aqueous
medium. In Sotation, a water-repellent (hydropho-
bic) target to be separated is attached to a carrier
lighter than the medium in which separation occurs.
The target varies from Rne particulates (solid or
II / FLOTATION / Historical Development 1527
Special attention should be paid to the carrying
capacity of air bubbles and to secondary upgrading in
the froth.
Design of air-sparging systems, feed distributors
and also froth discharge systems is critically impor-
tant for successful column operation.
Unconventional design and use of pre-aeration sys-
tems are the main trends in Sotation column develop-
ment at present.
See also: II/Flotation: Column Cells.
Further Reading
Clift R, Grace JR and Weber ME (1987) Bubbles Drops
and Particles. New York: Academic Press.
Dobby GS and Finch JA (1986) Flotation column scale-up
and modelling. CIM Bulletin 79: 89}96.
Finch JA and Dobby GS (1990) Column Flotation. Oxford:
Pergamon.
Levenspiel O (1972) Chemical Reaction Engineering. 2nd
edn. New York: Wiley.
Lynch AJ, Johnson NW, Manlapig EV and Thorne CG
(1981) Mineral and Coal Flotation Circuits, Their Simu-
lation and Control. New York: Elsevier.
Masliyah JH (1979) Hindered settling in a multi-species
particle system. Chemical Engineering Science 34:
1166}1168.
Ross VE and van Deventer JSJ (1988) Mass transport in
Sotation column froths. Column Flotation ’88. Proceed-
ings of the International Symposium, Phoenix, Arizona.
Littleton, Colorado: Society of Mining Engineers.
Rubinstein JB (1995) Column Flotation, Processes, Designs
and Practices. Basel: Gordon and Breach.
Schuhmann R. (1942) Flotation kinetics 1: methods for
steady state study of Sotation process. Journal of Phys-
ical Chemistry 46: 891}902.
Zhou ZA, Xu Z and Finch JA (1994) On the role of
cavitation in particle collection during Sotation } a criti-
cal review. Mineral Engineering 7: 1073}1084.
liquid) to ions and molecules, while the most com-
monly used carriers are air bubbles due to their ready
availability, easy handling and very low cost. Com-
pared to other light Suids (e.g. parafRn oil), air
has the highest hydrophobicity, and its low density
facilitates mass transfer of bubble-target aggregates
from the bulk medium to the interface where froth
forms and is collected/removed. Flotation was prac-
tised around a century ago, mainly for mineral separ-
ation applications. It is difRcult, if not impossible,
to pin down who should be given credit for the
1528 II / FLOTATION / Historical Development

Table 1 Key stages in flotation process development These ingredients form one of the three foundations
on which a Sotation system is built. This is shown in
Year Concept introduced Inventors Figure 1 as Sotation chemistry. Although neither the
1860 Oil as a carrier Haynes principles involved in Sotation nor these basic ingre-
1901 Gas as buoyant medium Potter/Froment dients have been changed since, the technology along
1902 Ultraflotation/carrier flotation Cattermole with the fundamental understanding of the processes
1905 First generation flotation have evolved greatly. Developments in each of these
machines Hoover
three foundations are summarized in this article with
1908 Frother (organic compounds) Higgins/Sulman et al.
1912 Activator (CuSO4 for sphalerite) Bradford emphasis on recent advances.
1913 Depressants (SO2 for activated
sphalerite) Bradford
1925 Modern flotation collectors Flotation Chemistry
(xanthate for sulfides) Keller
The search for new Sotation reagents for various
mineral separation systems has been one of the major
aims in Sotation chemistry development. Although
xanthate, Rrst used more than 70 years ago, remains
development of the Sotation process in the various the principal collector for sulRde mineral Sotation,
key stages. Nevertheless, Table 1 provides a general long chain surfactant has been introduced as the col-
picture of how Sotation has evolved since its Rrst lector in oxide, silicate and sparingly soluble salt
applications in mineral processing. mineral Sotation systems. The early trial-and-error
As shown in this table, up until the 1930s all the approach in screening and searching for a new Sota-
ingredients required for selective Sotation had been tion collector has evolved into today’s scientiRc de-
proposed, including: (i) a collector to render target sign. Using quantum chemistry and electron density
particles water-repellent by its adsorption; (ii) calculations, the structures of highly selective collec-
a frother to stabilize bubbles and promote foaming; tors have been proposed. A surfactant, with oxygen
(iii) an activator to induce or enhance collector ad- and nitrogen as the binding elements in its functional
sorption on target particles; and (iv) a depressant to group (e.g. hydroxyoximes), was found to be
destroy collector adsorption on unwanted particles; a powerful and more selective collector for oxide
along with bubble generation in a Sotation machine. minerals, while those with sulfur and nitrogen as the
binding elements (e.g. thionocarbamate) is particu-
larly selective for sulRde minerals. A common fea-
ture of these new collectors is their electron donor
character, forming a Rve- to six-member closed ring
structure with surface metallic elements. Many Rve-
membered heterocyclic compounds (e.g. oxazole- or
thiazole-based collectors) have recently been found to
be of special selectivity in base metal ore Sotation.
A general correlation between Sotation performance
and collector chain structure (e.g. short versus long,
single versus double, straight versus branched, single
bond versus double bond, parafRnic versus aro-
matic, etc.) has also been established and a detailed
list of newly developed collectors was compiled by
Nagaraj in 1988. The use of a collector mixture has
Figure 1 A flotation system shown as a tetrahedron with chem- shown improved collecting power and selectivity, and
ical, mechanical and physical aspects as the foundation through warrants further development.
which process dynamics is modelled and controlled. The triangu-
The invention of a water-soluble frother by Tveter
lar base plane emphasizes the interrelated nature between the
three foundation elements. The chemical aspect involves control (a polypropylene glycol ether, known as Dowfroth)
of chemistry at air} and solid}aqueous interfaces by collectors, was considered to be one of the major advances in
depressants, dispersants, activators, bacteria and frothers; the frother development. Following the advent of various
mechanical aspect concerns energy dissipation for bubble gen- types of synthetic, propylene-based frothers, the ef-
eration, particle dispersion, surface cleaning, hydrodynamic for-
fort in frother development has been directed to es-
ces and bubble}particle contact; the physical aspect deals with
the wetting phenomena and the nature of interactions between tablishing a correlation between frother structures,
bubble}particle, bubble}bubble and particle}particle pairs in frothing characteristics and their effect on recov-
aqueous solutions involved in a flotation system. ery and selectivity. To this end, increasing branching

in frother molecules has been identiRed as increasing
Sotation selectivity, often at the cost of reduced re-
covery. The use of mixed frothers in a Sotation sys-
tem to generate air bubbles with a wide range of sizes,
each suitable for particles of a given size range, has
also drawn considerable attention. An increased over-
all recovery has been demonstrated by using a mix-
ture of 1 : 1 polyglycol : methyl isobutyl carbinol
(MIBC) as compared to a single frother at the same
total concentration level. The synergistic effect
of a collector and a frother on bubble-particle collec-
tion has also been recognized, although the practical
application has not gained its fair share of attention.
The stabilization of air bubbles by simple inorganic
electrolyte should not be overlooked. Developed in
the early 1930s for natural hydrophobic coal Sotation
without using a frother, salt Sotation provides a dif-
ferent avenue for recovering natural hydrophobic
minerals, as the surface active frother tends to adsorb
on natural hydrophobic minerals with unfavourable
orientations for Sotation, consuming added chem-
icals and reducing their Soatability. In 1995 Weissen-
born and Pugh conRrmed that the hydration shell
around the added inorganic (ionic) species or frother’s
polar groups is responsible for froth stabilization.
Development in the activator appears rather lim-
ited, although most of the positively charged metal
hydroxy species have been found suitable for activa-
tion in silicate Sotation. In sulRde Sotation, copper
sulfate remains the only activator extensively used
today. In contrast, development in depressants has
taken on a different pace. Shortly after the intro-
duction of sodium dichromate (for PbS) and SO2 (for
ZnS) in 1913, sodium cyanide (1922) and alkali sul-
Rtes (1923) appeared to be the popular depressants to
use and remain the major depressants in modern
sulRde Sotation plants. Meanwhile, sodium silicate
(1928) and macromolecular starch (1931) have be-
come important depressants/dispersants in oxide, sili-
cate and sulRde Sotation systems. In addition to
nonionic dextrins, cationic polysaccharides and an-
ionic carboxymethyl cellulose have been found to be
effective depressants because of their multi-an-
choring nature with mineral surfaces. Recent ef-
forts have been directed to the search for polyamines,
which are effective in iron sulRde depression,
driven by the environmental pressure of reducing SO2
emission from smelters. Combined with SO2, di-
ethylenetriamine (DETA) has been found effec-
tively to depress the pyrrhotite in pentlandite Sota-
tion, although the depression mechanism remains to
be identiRed.
The control of the mineral surface property by
biotreatment is an emerging area and represents
a special branch in Sotation reagent development.
II / FLOTATION / Historical Development 1529
This approach is of special importance for desulfuri-
zation in coal Sotation, selective depression in base
metal sulRde Sotation and hydrophobization of non-
sulRde minerals. The success of biotreatment in these
systems lies in the extremely high selectivity of bac-
teria, such as Thiobacillus(T-)ferrooxidans, towards
the oxidation of pyrite, without any adverse ef-
fect on the Sotability of coal, resulting in a high
desulfurization efRciency in coal Sotation. Also
reported is an improved Soatability of sphalerite by
pretreatment of T-ferrooxidans in an acidic medium.
However, a high dose of T-ferrooxidans has been
found to be detrimental to sphalerite and galena Sota-
tion. Although sulfate-reducing bacteria have a min-
imal effect on the Soatability of molybdenite and
galena, they have been found to depress the Soatabil-
ity of chalocopyrite and sphalerite, resulting in highly
selective Sotation. Brunet et al. (1998) reported that
the combination of T-ferrooxidans, T-thiooxidans
and Leptospirillum accelerated pyrite oxidation. The
high selectivity of a bioprocess warrants the rapid
growth of biotreatment in mineral Sotation.
Accompanying the development of various Sota-
tion reagents is the recognition of surface reac-
tions/adsorption and the understanding of collec-
tor/mineral interactions in selective Sotation. The
theory of sulRde Sotation with xanthate family collec-
tors has advanced from simple surface chemical reac-
tions to a generalized electrochemical}chemical
process. Recognizing the electrochemical nature of
collector adsorption on sulRde surfaces was a quan-
tum leap in sulRde Sotation chemistry. The applica-
tion of the mixed potential theory to a sulRde Sota-
tion system provides a scientiRc explanation for a
required oxygen level to induce the Soatability,
and accounts for the role of pulp electrochemical
potential (Eh) in sulRde Sotation for a given
collector chemistry. An important consequence of
electrochemical involvement in sulRde Sotation is the
development of self-induced (also known as ‘collec-
torless’) Sotation by either controlled oxidation or
sulRdization of pre-oxidized sulRde minerals. The use
of cyclic voltammetry allows a direct correlation be-
tween collector adsorption (determined by charge
integration), under a given applied electrode poten-
tial, and contact angle, which in turn determines the
Soatability of sulRde minerals. An important out-
come from electrochemical studies is a new mecha-
nism for differential Sotation of complex sulRdes
by pulp potential control. However, the controversy
regarding the collector reaction product on sulRde
minerals is yet to be resolved. To this end, modern
spectroscopic methods are useful. Surface reactions
have been studied extensively using various surface
analytical techniques, including: (i) Fourier transform
1530 II / FLOTATION / Historical Development

infrared spectroscopy (FTIR in both in situ and
ex situ modes); (ii) Raman spectroscopy; (iii)
Auger and X-ray photoelectron spectroscopy (AES
and XPS); (iv) Suorescence spectroscopy; (v)
electron spin resonance spectroetry; (vi) laser ioniz-
ation mass spectrometry (LIMS); and (vii) time-of-
Rght}secondary ion mass spectrometry (TOF-SIMS).
For example, the monolayer formation of polysulRde
as a sulfur oxidation product on PbS has been con-
Rrmed from synchrotron XPS characterization. How-
ever, whether or not polysulRde is responsible for
collectorless Sotation remains to be established. Fol-
lowing the pioneer work on in situ spectroelec-
trochemical characterization of sulRde Sotation
chemistry by Leppinen et al. in 1988, the develop-
ment of a spectroelectrochemical cell (Figure 2), com-
bined with polarized FTIR spectroscopy, sets up an
entirely new direction for sulRde Sotation chemistry
research. Using polarized infrared radiation, the ori-
entation of the adsorbed molecular species can be
derived, as shown in Figure 3. However, further re-
search efforts are required to quantify the mo-
lecular orientation and to derive its practical implica-
tions in sulRde Sotation practice.
In oxide and silicate mineral Sotation, the interac-
tion (i.e. adsorption of the collector, mostly surfac-
tant) has been generally considered to be electrostatic
rather than chemical in nature. An electrostatic inter-
action model has proven satisfactory when applied to
silica and alumina Sotation with ionic collectors of
opposite charges from the surfaces. Progress has been
made in predicting the point of zero surface charge,
based on the minimum solubility theory and the sign
of surface charge from the hydration energy of lattice
ions. A more quantitative description of surface
charge distribution has been made possible following
the development of the surface triple-layer model in
combination with the surface site-binding theory.
Early adsorption studies have revealed the formation
of surfactant hemimicelles on mineral surfaces at
a bulk surfactant concentration of approximately
one-hundredth of its critical micelle concentration
(cmc). The formation of ionomolecular complexes
has been found to enhance the Soatability of oxides.

Figure 2 Schematic diagram of an in situ spectroelectrochemical cell suitable for studying sulfide flotation chemistry. Pushing
a movable sulfide mineral working electrode against the CaF2 window with a screw type of mechanics ensures not only elimination of
bulk water films to increase the sensitivity of infrared spectroscopy, but also reproducible positioning of the electrode (after each
electrode polarization) for quantitative analysis. The use of polarized infrared radiation in external reflectance mode allows identification
of molecular orientation.

Figure 3 In situ infrared spectra obtained with a copper elec-
trode polarized under electrode potentials of 150 (dotted lines)
and 350 (continuous lines) mV/SHE (standard hydrogen elec-
trode) in 2;10\3 mol L\1 potassium ethylxanthate solutions. By
comparing the spectra obtained with s- and p-polarized infrared
beams, a near perpendicular orientation of adsorbed copper xan-
thate on copper electrode (inset) was derived to account for the
absence of the band at 1050 cm\1, associated with COC molecu-
lar vibrations, with the s-polarized infrared beam. In contrast,
a random orientation of dixanthogen, formed under a higher
applied electrode potential, was ascertained by a similar spectral
feature of characteristic dixanthogen bands obtained with both
polarization modes.
This is consistent with recent observations on en-
hanced hydrophobicity of mica surfaces in a mixed
cationic amine and neutral alcohol surfactant solu-
tion. The increased overall surfactant adsorption den-
sity at the solid}liquid interface is accounted for by
screening elecrostatic repulsion between adjacent sur-
factant head groups.
Recently, a detailed study using a well-deRned
basal plane of crystalline sapphire in a surface forces
apparatus showed that the formation of monolayer
hemimicelles requires a near-cmc surfactant concen-
tration in the surface region, while its bulk concentra-
tion has to be well below the cmc. Based on the
well-known Stern}Grahame equation, this condition
II / FLOTATION / Historical Development 1531
cannot be satisRed in the absence of any attractive
driving force of electrostatic and/or chemical nature
required to preconcentrate the surfactant in the sur-
face region to the cmc level. As a result, the lack of
hydrophobic monolayer formation and hence ef-
fective Sotation is anticipated.
Clearly, effective oxide Sotation requires cre-
ation of a chemical environment that maximizes the
surface concentration of the surfactant at as low
a bulk concentration as possible. Changing suspen-
sion pH to control surface charge density in oxide
Sotation serves as an excellent example. Under cer-
tain circumstances, activation by hydrolysed metal
ions, which provide the linkage between an anionic
collector and a negatively charged mineral, is neces-
sary to induce Soatability. It should be noted that the
selectivity of separation in oxide Sotation is relatively
poor if the electrostatic force is the only driving force
for collector adsorption. This is particularly true in
Rne particle Sotation, as heterocoagulation between
different minerals often induces a secondary lock-
ing which destroys the selectivity. To this end, search-
ing for collectors which chemically anchor on to
targets remains the focus in oxide Sotation systems.
For sparingly soluble mineral Sotation, solution
chemistry calculation has been proven to be one of
the most valuable tools in searching for separation
windows. Since bulk solution chemistry controls the
Sotation response, a bulk precipitation followed by
surface deposition, with a switch-on type of adsorp-
tion characteristics, has been considered to be the
most favourable mechanism in Sotation of soluble-
type minerals, where the monolayer adsorption is
hardly recognizable.
A recent trend in laboratory studies of sulRde Sota-
tion chemistry is to use a mixed mineral system. With
this approach, an enhanced xanthate adsorption on
anodic minerals by galvanic contact of dissimilar
minerals was revealed. Also derived from this type of
research is the depression of pyrite by copper sulfate
addition in a sphalerite/pyrite mixed mineral system,
shown in Figure 4, as opposed to pyrite activation in
a single mineral system. A similar approach has been
used in oxide and salt-type Sotation systems.
In summary, our understanding of the interaction
mechanism of collectors with minerals in a Sotation
system has evolved signiRcantly following the devel-
opment of modern instrumentation. Future advances
in the fundamental understanding of Sotation sys-
tems are anticipated with the introduction of the
atomic force microscope in mineral Sotation re-
search. A combination of electrochemistry, in situ
spectroscopy and surface imaging at a molecular level
will enable us to pinpoint the mechanism and roles of
collector}mineral interactions in Sotation.
1532 II / FLOTATION / Historical Development
Figure 4 Flotation recovery of pyrite in the presence of
10\5 mol L\1 iso-propylxanthate alone (squares) or with cupric
ions (triangles), sphalerite (inverted triangles) or both (circles).
The flotation of pyrite was depressed by a combination of cupric
ions and sphalerite, although cupric ions alone activated pyrite
flotation, illustrating the importance of studying flotation chemistry
with mixed mineral systems in the context of the separation
practice.
Flotation Mechanics
Bubbles are an indispensable component of froth So-
tation. Bubble generation in Sotation machines forms
the second foundation of a mineral Sotation system.
It is interesting to note that the development of Sota-
tion machines follows the evolution of bubble genera-
tion methods, although a Sotation machine has to
fulRl three basic functions: (i) generation of sufR-
cient amount of bubbles with suitable sizes (c.
0.5}2 mm); (ii) dispersion of solid materials; and (iii)
effective collision between particles and bubbles,
in addition to providing a quiescent zone for froth
formation. Table 2 summarizes the major steps in the
early stages of Sotation machine development. Up
until 1911, all available bubble generation methods
had been practised in Sotation machines. Driven by
Table 2 Early developments in bubble generation and flotation
machines
Year Methods of bubble generation Flotation acromy
1904 Electrolysis Electroflotation
1904 Pressure reduction Vacuum flotation
1905 Air dispersion by agitation Mechanical cell
Pressurization/pressure release Dissolved air flotation
1911 Air dispersion by spargers Pneumatic cell
an increased demand for metals and a need for cost
reduction, the development of Sotation machines in
the subsequent 50 years was directed to the design of
large volume cells, with the beneRts of Sexible process
control and reduced capital cost, plant space, speciRc
power and maintenance. Taking DO-3500 (Denver)
as an example, cells of volume as large as 100 m3 are
now in operation. Radical changes incorporated in
this new super-large cell include the use of a pump-
type rotor, an overhung vane-type stator, a round
tank of conical bottom and radial discharge of froth
as in Sotation columns. The most recognized develop-
ment in Sotation machines today is, however, the
commercialization of Sotation columns, followed by
various innovative designs of aeration systems to gen-
erate microbubbles in response to slow Sotation ki-
netics of Rne particles.
The success of conventional Sotation columns in-
itiated a surge in the development of novel Sotation
devices, some of which are summarized in Table 3.
Detailed analysis of these new devices shows a com-
mon feature } the generation of Rne bubbles online
with high energy dissipation (e.g. high turbulence in
Suid). Major advances in these devices are illustrated,
for example, in a fast Sotation column, shown in
Figure 5. The in-line generation of Rne bubbles (feed
aeration) by either a static mixer or a simple Venturi
tube ensures a high bubble}particle collision efR-
ciency. Partial recycling of tailings to the feed allows
fugitive valuables to be captured, while a deep froth
with wash-water addition cleans up entrained un-
wanted gangues. Essentially, a Sotation column in
this conRguration is equivalent to a Sotation circuit:
(i) a rougher in the middle; (ii) a cleaner on the top;
and (iii) a scavenger at the bottom.
The improvement of froth quality, by deep froth
and froth washing in the Sotation column, brought
the recent development of Outokumpu’s HG tank,
which features an adjustable booster cone to control
froth quality. A hidden feature in these newly de-
signed Sotation devices is the role of hydrodynamic
cavitation. The importance of hydrodynamic cavita-
tion in Sotation is the complete elimination of the
bubble}particle collision step, resulting in a 100%
increase in Sotation rate constant, seen in a case study
using the set-up shown in Figure 6. The preferential
nucleation of bubbles on hydrophobic particles is
anticipated to contribute to improvement in Sotation
selectivity. With hydrodynamic cavitation, strong
mechanical agitation, which is otherwise required to
provide the kinetic energy necessary to overcome
energy barriers for bubble}particle attachment, can
be minimized. As a result, a more quiescent environ-
ment is created for enhanced froth-pulp disintegra-
tion. Along the same line of thinking, the use of
 II / FLOTATION / Historical Development 1533

Table 3 Examples of new flotation devices developed in the minerals industry since the 1980s

Device Features Inventors

Air-sparged hydrocyclone Centrifugal force/porous cylinder aerator Miller

Pneumatic cell Slot aerator under pressure Bahr
Packed column Unlimited froth height Yang
Microcell Static mixer/tailing recycle Yoon
Jameson cell Self-aspiration: plunging slurry jet Jameson
Contact cell Feed aeration Amelunxen
Ken Flote Conditioning with dissolved air Parekh et al.
USBM rapid cell Static mixer Jordan et al.
Rapid flotation column Feed aeration/tailing recycle Xu et al.
Next generation Hydrodynamic cavitation Zhou et al.

ultrasonication or vibroacoustic modulation, to facil-
itate gas nucleation and bubble}particle attachment,
has been tested. A gas nucleation mechanism, most
likely by hydrodynamic cavitation with innovative
engineering of the cavitation tube or ultrasonic
modulation, is anticipated to be the main feature of
the next generation of Sotation devices. A reac-
tor}separator design as seen in Figure 5 is desirable to
optimize individually bubble}particle contact and
bubble}pulp separation.
Flotation Physics
It is evident that, for effective Sotation, the thin
liquid Rlms between an air bubble and a target have
to be thinned and ruptured (Rlm stability), while
association between different species (hetero-co-
agulation) needs to be avoided to separate one species
from another. The physics of the Sotation system (i.e.
surface forces between various phases) controls such
phenomena as thin Rlm stability and coagulation. As
a macroscopic process, Sotation is often analysed in
terms of micro subprocesses and the knowledge
about them would serve as an encyclopedia of
colloidal science. Fundamental studies in Sotation
physics have evolved, to today’s role of hydrophobic
forces in Sotation, from pioneer work by Wark
(capillary forces), Sutherland (contact angles),
Derjaguin (interparticle forces), Derjaguin and Dukin
(elementary stages of Sotation), Klassen (role of
hydration shells in Sotation), Scheludko (thin Rlm
stability) and Schultz (hydrodynamic forces).
The capillary phenomena conRrmed the existence
of relatively short range molecular forces, manifested
in observed surface tension. The Soatability of min-
erals has been frequently correlated to contact angles
of solid against water (or, more precisely, contact
angle hysteresis in Sotation practice, which involves

Figure 5 Schematic of a modern flotation column featuring feed line aeration (insert) and partial recycle of tailings. The main thrust of
a column in this configuration is its equivalence to a virtual flotation circuit with the capability of generating fine bubbles for fine particle
flotation.
1534 II / FLOTATION / Historical Development

Figure 6 Schematics of a flotation system with reactor}separator design exploiting hydrodynamic cavitation in fine particle flotation.
A 10-fold increase in removal efficiency of fine oil contaminants in the soil-washing process has been demonstrated in a laboratory-
scale test. Using the setup at INCO’s Matt separation site (Sudbury, Canada), a doubling of the pentlandite flotation rate at
a comparable concentrate grade was obtained. The main attraction of this configuration is that it can be readily implemented in the
existing flotation circuit by simply bridging the conditioning tank with mechanical cells using a cavitation-type reactor.

consideration of both adhesion and detachment). The
contact angle phenomena manifest the competition
for water molecules by solids and by water itself,
assuming that the interaction between air and other
phases (solid or water) is negligible. The work of
cohesion of water greater than the work of
water}solid adhesion is equivalent to a contact angle
greater than zero as the thermodynamic criterion of
Soatability. However, the former provides insight
into the competition between various phases. Analy-
sis on work of adhesion and cohesion for a Sotation
system has contributed to the improved understand-
ing of Sotation physics. To this end, the concept of
critical surface tension proposed by Zisman was
adopted in a so-called Gamma Sotation process,
although its practical application has been limited to
coal Sotation or the recycling of low surface energy
polymeric materials. The role of collector adsorption
in Sotation is to reduce the work of adhesion of target
solids by exposure of weakly interactive hydrocarbon
tails to water. Such a system of high solid}liquid
interfacial tension is thermodynamically unfavour-
able, making the particles Soatable.
The use of the electrostatic double layer and van
der Waals forces considered in the classical colloidal
stability theory (known as Deryagin}Landau}
Verwey}Overbeek, or DLVO theory) has been suc-
cessful in accounting for the stability of some collo-
idal systems. It is now generally accepted that addi-
tional forces need to be considered to understand
fully the observed phenomena in Sotation systems.
A typical example is that alumina is not Soatable in
the absence of surfactant, although a strong electric
double-layer attraction between air bubbles and the
solids is predicted. On the other hand, quartz dehy-
drated at a temperature above 10003C is readily Soa-
table without any collector, yet the classical DLVO
theory would predict repulsive van der Waals and
electrostatic forces between the two. It is clear that
the additional force can be either attractive or repul-
sive, depending on the hydrophobic or hydrophilic
nature of the solid surfaces.
Thanks to a recent breakthrough in measuring sur-
face forces directly, the presence of additional long
range attractive forces between hydrophobic surfaces
and short range repulsive forces between strongly
hydrated surfaces has been conRrmed. The former
has contributed signiRcantly to comprehending the
thin Rlm rupture phenomena which occur in most
Sotation systems. The force between an air bubble
and a solid surface has been directly measured with
an atomic force microscope and results, shown in
Figure 7, conRrm the existence of additional attract-
ive forces. It should be noted that the direct force

Figure 7 Forces between an air bubble and a silica particle in (A) an electrolyte solution with (B) added surfactant, measured directly
with an atomic force microscope. A much greater jump in distance than predicted from the classical DLVO theory (A: shown by arrows)
confirms the existence of additional attractive forces. In contrast, the presence of 3 mmol L\1 SDS changed the forces from a long
range attraction to a long range repulsion, well-described by the DLVO theory (B), manifesting the role of surfactant in flotation.
measurement between an air bubble and a hydropho-
bic solid surface remains an unresolved challenge,
even though it is most relevant to Sotation.
Following recent advances in scientiRc instrumen-
tation, such as the atomic force microscope, Rlm
balance and surface forces apparatus, Sotation re-
search has gone through a period of thermodynamic
analysis of bubble}particle attachment to the under-
standing of intermolecular forces involved. These ad-
vances have allowed Sotation subprocesses to be ana-
lysed from Rrst principles. Bubble}particle adhesion,
and hence Sotation, for example requires par-
ticle}bubble contact (sliding) time greater than Rlm
rupture (induction) time controlled by wetting kinet-
ics. An attempt has been made to derive a Sotation
rate equation from Rrst principles by considering
both surface forces and system hydrodynamics. The
practical application of the derived equation in Sota-
tion process development remains to be explored.
The empirical relations, outlined in a review by
Radoev and Alexandrova (1992), remain the main
source for process design and simulation.
It is important to note that chemistry, physics and
mechanics, which form the three foundations of a So-
tation system and determine the process dynamics,
are interrelated among themselves. This is empha-
sized in Figure 1 by a triangular relation on the base
of the tetrahedron. Bubble size in a Sotation system,
for example, is determined by chemistry (addition of
frother) which affects the physics of Rlm stability
(surface forces) balanced by mechanical forces and
liquid viscosity. Only when these three factors are
considered simultaneously can the Sotation dynamics
be optimized. An important area in Sotation develop-
ment is the innovations in: (i) on-stream analysis
II / FLOTATION / Historical Development 1535
(X-ray Suorescence analyser, ash analyser, nuclear
magnetic resonance and image analysis); (ii) sensors
(redox, ion-selective Eh and conductivity probes);
(iii) dynamic process modelling (expert systems
and artiRcial neural networks to mimic control ac-
tions by human operators); (iv) control (fuzzy logic
and self-organizing controller); and (v) instrumenta-
tion. Details on these developments are not included in
this article, and interested readers are referred to the
recent review articles by Sastry and Fuerstenau (1988),
Mavros and Matis (1991) and the Proceedings of the
XIX International Mineral Processing Congress
(1995).
Recent Advances in Flotation Practice
It appears unnecessary to list all of the minerals pro-
cessed by Sotation, since almost all minerals mined
today can be separated effectively by froth Sota-
tion. In addition to developments in the three founda-
tion areas of Sotation (Figure 1), signiRcant progress
has also been made in circuit design. Following the
introduction of reverse, bulk and differential So-
tation, the multifeed circuit, shown in Figure 8 and
practised in China for processing copper sulRde, is
considered to be one of the most recent advances in
this regard. With the multifeed circuit, the improved
recovery has been attributed largely to the auto-
genous carrier (piggyback) effect. The reduced
reagent consumption, which creates a starving re-
agent addition, may have contributed to the im-
proved concentrate grade (i.e. selectivity). In the Cli-
max Mill (USA), the use of the multifeed circuit
improved the molybdenite grade from 14 to 34%
MoS2 at comparable molybdenite recovery and
1536 II / FLOTATION / Historical Development

Figure 8 Comparison of (A) conventional rougher}cleaner circuit with (B) multifeed circuit. The dotted lines represent a variation of
the circuits with recirculating loads. An added advantage with a multifeed circuit includes improved concentrate grade at reduced
reagent consumption while maintaining the valuable recoveries.

reduced by half collector and frother consumptions.
Another recent development is Sash Sotation which
uses a coarse Sotation cell (Skim-Air) in a grinding
circuit to produce Rnal concentrate from a cyclone
underSow stream before further grinding (Figure 9).
This approach minimizes overgrinding of valuable
material richer in the cyclone underSow than in the
feed stream. In addition to increased recovery, an
improved throughput for the subsequent unit opera-
tions and the high dewatering efRciency of Rnal
concentrate are recognized with Sash Sotation. In
tackling the challenge of Rne particle Sotation, high
intensity conditioning has been developed. The im-
proved process performance has been attributed to
mechanical surface cleaning of slimes, shear-induced
aggregation of target particles and, yet to be con-
Rrmed, in situ bubble formation on hydrophobized
particles by hydrodynamic cavitation and resultant
bubble}particle aggregation. Online pulp potential
control assisted by nitrogen as the carrier gas has also
been engineered in Sotation machines and columns as
a means of improving selectivity of sulRde mineral
separation.
Concluding Remarks
With generations of research efforts, Sotation
has matured into a process of choice for many separ-
ation tasks, including mineral separation, bitumen
extraction from tar sands, soil remediation, materials
recycling, de-inking, de-oiling, de-colouring, biolo-
gical species fractionation and industrial efSuent
detoxiRcation in the form of either froth Sotation or
absorptive bubble separation. Both inventions and
innovations have played an indispensable role in So-
tation development in an evolutionary, rather than
a revolutionary, fashion. Although Sotation practice
has always been ahead of Sotation science, the gaps

Figure 9 Schematics of a flash flotation circuit with a coarse flotation before grinding. The circuit minimizes overgrinding of valuables
and improves the recovery and product quality at an increased throughput.
 II / FLOTATION / Hydrophobic Surface State Flotation
between the two have narrowed signiRcantly. Im-
proving fundamental understanding of the Sotation
process remains the main focus of research for the
future. The areas where signiRcant advances are
anticipated include: (i) design and synthesis of more
effective, environmentally friendly Sotation re-
agents (mainly collectors, frothers and depressants);
(ii) engineering of a pulp potential monitor (mineral
electrodes) and control in sulRde Sotation practice;
(iii) development of new Sotation cells to maximize
separation efRciency and minimize energy con-
sumption; (iv) understanding and utilization of bio-
treatment to replace both collectors and depressants;
and (v) design of a better and reliable process control
system based on further development of sensors and
simulators. The main challenge that Sotation engin-
eers and scientists are facing is to develop viable
process alternatives for Rne particle Sotation. Four
areas of immediate interests are: (i) the development
and understanding of high intensity conditioning ; (ii)
hydrodynamic cavitation in Sotation machines; (iii)
selective aggregation by coagulation, Socculation or
oil agglomeration; and (iv) practical conditions for
collectorless Sotation of sulRde ores.
Further research is needed in the area of Sotation
chemistry and implementation of the outcome into
process development. All of these are driven by the
depletion of rich and simple mineral resources, reduc-
tion of metal prices and the increase of environmental
pressures. The processing of tailings with a gravity
concentrator at Laurium, from 1864 to 1920, left
tailings containing 3% lead, these were reprocessed
again in 1955 by Sotation with a resulting tailings
assay of 0.3% lead. It is not unrealistic to suggest that
Hydrophobic Surface State Flotation
J. D. Miller, University of Utah, Salt Lake City, UT, USA
Copyright ^ 2000 Academic Press
Introduction
The essence of particle separation by Sotation is the
creation of a hydrophobic surface state, i.e. a surface
that is not wetted by water, a particle surface at which
bubble attachment will occur leading to Sotation due
to the buoyancy of the particle}bubble aggregate. (Par-
ticle Sotation can also, however, be accomplished by
bubble entrapment rather than by bubble attachment.
For example, entrapment of air during particle ag-
gregation/Socculation can lead to the Sotation of aero-
Socs.) In many instances this hydrophobicity must be
1537
the resultant tailings may be reprocessed in the future
with further innovative developments, such as
integration of biotreatment in Sotation. To conclude,
there is a long-awaited need to widen the range of
Sotation applications to nonmineral-processing ap-
plications, such as in material recycling and waste
remediation, with revolutionary changes in Sotation
technology.
Further Reading
Ives KJ (ed.) (1984) The ScientiTc Basis of Flotation. The
Hague: Martinus Nijhoff Publishers.
Jones MH and Woodcock JT (eds) (1984) Principles of
Mineral Flotation. Victoria: AIMM.
Laskowski JS (ed.) (1989) Frothing in Flotation. New
York: Gordon Breach Science.
Matis KA (ed.) (1995) Flotation Science and Engineering.
New York: Marcel Dekker.
Mavros P and Matis KA (eds) (1991) Innovations in Flota-
tion Technology. Dordrecht: Kluwer Academic Pub-
lishers.
Parekh BK and Miller JD (eds) (1999) Advances in Flota-
tion Technology. Littleton: SME.
Sastry KVS and Fuerstenau MC (eds) (1989) Challenges in
Minerals Processing. Littleton: SME.
Schulze HJ (1984) Physico-chemical Elementary Processes
in Flotation. Amsterdam: Elsevier.
Somasundaran P and Moudgil BM (eds) (1988) Reagents in
Mineral Technology. New York: Marcel Dekker.
Souninen EJ, Forssberg KSE and Buckley AN (eds) (1997)
Application of Surface Science to Advancing Flotation
Technology. Amsterdam: Elsevier.
Wood R, Doyle FM and Richardson P (eds) (1996) Elec-
rochemistry in Mineral and Metal Processing, Vol. IV.
Pennington: Electrochemistry Society.
established in a selective manner, frequently by col-
lector (surfactant) addition, so that one particle type
can be separated from other particle types which are
maintained in a hydrophilic state.
The extent to which a surface is hydrophobic can
be described in various ways. Two of the most com-
mon laboratory methods are contact-angle measure-
ment and bubble attachment time measurement. The
contact angle measurement tends to be an equilib-
rium, or pseudo-equilibrium, measure of hydro-
phobicity, while the bubble attachment time
measurement is a kinetic measure of hydrophobicity.
Other measures of hydrophobicity are also possible
and include bubble pick-up and microSotation
experiments.
1538 II / FLOTATION / Hydrophobic Surface State Flotation

Figure 1 Bubble attachment. Sequence of events.

Bubble Attachment ported for chalcopyrite. The very strong effect of

Bubble attachment at a hydrophobic surface occurs
due to the instability of the aqueous Rlm that separ-
ates the bubble from the surface. As the bubble ap-
proaches the surface, to such a separation distance
that the bubble may be distorted, there is a thinning
of the aqueous Rlm to the point at which rupture
occurs. This time of Rlm thinning is called the ‘induc-
tion time’. After rupture, the Rlm is displaced as it
recedes across the hydrophobic surface to establish
the equilibrium contact angle. The total time of Rlm
thinning and Rlm displacement is the bubble attach-
ment time. The sequence of events at a polished
surface is depicted in Figure 1, where the bubble
attachment time is shown to consist of the Rlm thinn-
ing (induction) time and the Rlm displacement time.
Thus the bubble attachment time is, in part,
a measure of hydrophobicity and can vary from less
than a millisecond to several seconds in magnitude.
Although the hydrophobicity should be an intrinsic
property of the system, the bubble attachment time
measurement is signiRcantly inSuenced by the experi-
mental method. For example, the bubble attachment
time for a sample of naturally hydrophobic bitumi-
nous coal was found to vary by a factor of more than
50 when the results obtained for a polished surface
are compared with those obtained for a particle bed
as revealed in Table 1. Similar results have been re-
contact area, hydrodynamics, and surface morpho-
logy are revealed from these data. For a given experi-
mental teachique, the shorter the bubble attachment
time, the greater the hydrophobicity.
Contact Angle
The equilibrium state for the attached bubble is de-
scribed by the contact angle,
, as indicated in
Figure 2. The contact angle for this three-phase equi-
librium is related to the respective interfacial tensions
by Young’s equation,
SG"SL#LG cos

The attachment process should be spontaneous for
all Rnite contact angles, but generally a contact angle
of at least 203 is required for bubble attachment and
Sotation. The greater the contact angle, the greater
the hydrophobicity. Of course contact angles much
greater than 203 are desired in order to make ef-
fective Sotation separations. Generally the character-
istic contact angles for Sotation systems rarely exceed
1003. Typical values for naturally hydrophobic min-
erals are given in Table 2. Larger contact angles are

Table 1 Measured bubble attachment times for a low-volatile

bituminous coal at a polished surface and at a bed of particles
(100;200 mesh)

Mode of attachment Gas phase Attachment time (ms)

Polished surface Air 180}200

N2 170}190
CO2 140}150

Particle bed Air 3

N2 }
CO2 3 Figure 2 Equilibrium state for water drop at a hydrophobic
surface.
 II / FLOTATION / Hydrophobic Surface State Flotation 1539

Table 2 Naturally hydrophobic minerals and respective contact ite, coal, elemental sulfur, and iodyrite. Even the
angles surfaces of metal sulRde minerals are reported to be
hydrophobic in the absence of oxygen and can be
Mineral Composition Surface Contact angle
plane (degrees) considered to be intrinsically hydrophobic. Of course,
exposure to even parts per billion of oxygen can lead
Graphite C 0001 86 to oxygen Rxation and subsequent complex electro-
Coal Complex hydrocarbon 20}60 chemical reactions, the surface products of which
Sulfur S 85
may or may not be hydrophilic depending on solution
Molybdenite MoS2 0001 75
Stibnite Sb2S3 010 chemistry and the extent of oxidation. In general,
Pyrophyllite Al2(Si4O10)(OH)2 001 simply the Rxation of oxygen at sulRde mineral sur-
Talc Mg3(Si4O10)(OH)2 001 88 faces can provide sufRcient surface polarity to
Iodyrite AgI 20 create a hydrophilic state. Nevertheless, under anaer-
obic conditions the sulRde surface is expected to be
hydrophobic due to its limited ability to hydrogen
possible for specially prepared surfaces which are bond with interfacial water molecules.
highly water repellant. For example, water contact In addition to the elemental composition of
angles exceeding 1503 have been observed for the surface, the crystal structure and bonding
specially prepared surfaces as shown in Figure 3. inSuence the polarity of mineral surfaces. In some
cases, speciRcally surfaces that are created by break-
age of weak van der Waals bonds, a nonpolar surface
Nonpolar Surfaces is created even containing elements that normally
It is evident that the hydrophobic surface state is would hydrogen bond and be hydrated by interfacial
established by nonpolar surfaces which are not exten- water molecules. Examples include pyrophyllite, talc,
sively hydrated. Now the nonpolar surface criterion and boric acid. In this way it has been established that
for hydrophobicity is well known and has been estab- the hydrophobic nonpolar surface state can arise
lished for some time. Such characteristics of the hy- from the intrinsic properties of the elements of which
drophobic surface state have been known since the the surface is composed and from bonding consider-
mid-1950s. In some cases the hydrophobic surface ations associated with the crystal structure. Finally it
state is due to the elemental composition of the sur- should be noted that hydrophobic surfaces can be
face; the surface is composed of elements of low charged just as hydrophilic surfaces are and that gen-
polarity that do not hydrogen bond with water erally maximum hydrophobicity is found at the
molecules. These elements include C, H, S, and large isoelectric point, or the point of zero charge, of the
atoms of low polarizability. Examples include graph- surface.

Figure 3 Water contact angle for a sessile drop of water at the surface of a newly developed water repellant material.
1540 II / FLOTATION / Hydrophobic Surface State Flotation
Water Film Stability
Of course the hydrophobic surface state must not
only be described in terms of the elemental surface
composition and structure but also must be described
in terms of the interfacial water structure; in fact the
instability of the interfacial water Rlm accounts for
bubble attachment at a hydrophobic surface. The
characteristic features of interfacial water and its in-
stability at a hydrophobic surface have not been so
well described until recently. Now with the use of
atomic force microscopy, surface spectroscopy, and
a laser optical cavity technique, these features of
interfacial water have been revealed in greater detail.
Direct force measurements during the 1980s and
1990s have revealed that attractive hydrophobic
forces are usually 10 to 100 times larger than those
expected from van der Waals interactions. These
forces extend to distances of as much as 100}200 nm
from the surface. The extent of attraction between
hydrophobic surfaces is related to the degree of
hydrophobicity but seems to be also independently
effected by discrete features of the surface like
roughness and heterogeneity.
At the same time, during the 1990s, in situ surface
spectroscopy (sum frequency generation (SFG) and
Fourier transform infrared/internal reSection spectro-
scopy (FTIR/IRS)) of water at hydrophobic surfaces
has revealed important characteristics of interfacial
water. The SFG spectral information clearly shows
a distinction between water at a hydrophobic surface
and water at a hydrophilic surface. Interfacial water
at a hydrophobic surface is distinguished by a stron-
ger absorption band at 3600 cm\1 characteristic of a
dangling free OH bond. In contrast, interfacial water
at a hydrophilic surface is distinguished by a dimin-
ished absorption band at 3600 cm\1 and a stronger
signal at 3200 cm\1 characteristic of an ice-like struc-
ture with complete tetrahedral coordination. Based
on these surface spectroscopy studies, it appears that
interfacial water at a hydrophilic surface can be
viewed as organized dipoles in tetrahedral coordina-
tion and oriented with respect to the polarity of the
hydrophilic surface, whereas interfacial water mol-
Figure 4 Schematic picture of cavitation phenomena during approach of hydrophobic sphere and hydrophobic plane in water.
(A) Layers of lower medium density (adsorbed gas molecules), (B) nanobubbles formation, (C) bridging cavity formation, and
(D) multiple bridging cavities, leading to film rupture and attachment.
ecules at a hydrophobic surface are not so well organ-
ized at the surface and have incomplete tetrahedral
coordination with dangling free OH bonds.
It might be assumed that this in situ spectral data
can then be used to account for Rlm instability at
a hydrophobic surface. Unfortunately, it seems that
the phenomenon is not that simple. It is expected that
the interfacial water structure will extend only a dis-
tance of a few molecular diameters, not more than
a few nanometers or so. On the other hand, the
hydrophobic attractive forces can extend to 100 nm,
and even more. Thus it would seem that Rlm instabil-
ity at a hydrophobic surface involves more than just
the hydrogen bonding characteristics of interfacial
water.
Some researchers have attributed Rlm instability to
cavitation phenomena. The presence of nanobubbles
or defects in the interfacial water region at a hydro-
phobic surface has been reported based on experi-
mental results using a laser optical cavity technique.
Also it should be noted that surface force measure-
ments reveal that the range of the attractive hydro-
phobic force is signiRcantly greater in gas-saturated
solution then in degassed solution. It is expected that
slight perturbations in the pressure Reld would cause
these nanobubbles to coalesce and form cavities
which upon further coalescence would lead to cavita-
tion and failure of the water Rlm at a hydrophobic
surface as shown in Figure 4. In some cases, discon-
tinuities during force measurements were observed
which may be attributed to the phase transition (cav-
ity formation) between approaching surfaces. Finally,
recent FTIR/IRS spectroscopic evidence, indeed,
shows that dissolved gas is accommodated at a hydro-
phobic surface but not so at a hydrophilic surface.
Thus the presence of nanobubbles in the interfacial
water region of a hydrophobic surface is supported by
these spectroscopic results.
Summary
In summary, the hydrophobic surface state must be
considered both with regard to the particle surface
and with regard to the adjacent interfacial water

region. The particle surface must be of low polarity
which is determined by elemental composition and/or
structural bonding considerations. Water Rlm insta-
bility at a hydrophobic surface arises not only from
a disrupted interfacial water structure but also from
a cavitation phenomenon which involves coalescence
of nanobubbles in the interfacial water region. Such is
the nature of the hydrophobic surface state.
Further Reading
Arbiter N, Fuji U, Hansen B and Raja A (1973) Surface
properties of hydrophobic solids. In: Somasundaran
P and Grieves RB (eds) Advances in Interfacial Phe-
nomena of Particulate/Solid/Gas Systems, AIChE Sym-
posium Series 150, vol. 71, pp. 176}182. New York:
AIChE.
Bunkin NF and Bunkin FV (1993) Laser Physics 3: 63.
Bunkin NF, Kiseleva OA, Lobeyev AV, Movchan TG, Nin-
ham BW and Vinogradova OI (1997) Langmuir 13:
3024.
Chander S (1999) Fundamentals of sulRde mineral Sota-
tion. In: Parekh BK and Miller JD (eds) Advances in
Flotation Technology, p. 129. Denver: Society for Min-
ing, Metallurgy and Exploration.
Drelich J, Miller JD, Li JS and Wan RY (1997)
Bubble attachment time measurements at a chalcopyrite
surface using a high-speed video system. In: Proceedings
of the XX International Mineral Processing Congress,
Vol. 3: Flotation and other Physical Chemical Processes,
Aachen, Germany, 21}26 September 1997, pp.
53}64.
Drost-Hansen W (1969) Industrial Engineering Chemistry
61: 331.
Du Q, Freysz E and Shen YR (1994) Science 264: 826.
Fokkink LGJ and Ralston J (1989) Colloids Surfaces
36: 69.
Intensive Cells: Design
G. J. Jameson, University of Newcastle, Callaghan,
NSW, Australia
Copyright ^ 2000 Academic Press
Introduction
In conventional Sotation practice, the particles to be
treated are dispersed in a suspension in water. Re-
agents are added to make the particles to be Soated
hydrophobic or nonwetting. The particles which are
to be left behind remain in a wettable state. Air
bubbles are then introduced into the slurry or pulp in
a contacting device or cell, and collide with the non-
wetted particles, carrying them to the surface where
II / FLOTATION / Intensive Cells: Design 1541
Fuerstenau DW, Rosenbaum JM and Laskowski J (1983)
Colloids Surfaces 8: 153.
Fuerstenau MC, Miller JD and Kuhn MC (1985) Chemistry
of Flotation, p. 35. New York: Society of Mining
Engineers.
Gaudin AM (1957) Flotation, 2nd edn, p. 218. New York:
McGraw-Hill.
Israelachvili J and Pashley R (1982) Nature 300: 341.
Kitchener JA (1984) Surface forces in Sotation } a critique,
principles of mineral Sotation. In: Jones MH and Wood-
cock JT (eds) The Wark Symposium, Series No. 40,
pp. 65}71. Parkville, Victoria, Australia: Australian In-
stitute of Min. Met.
Laskowski J (1986) The relationship between Sotability
and hydrophobicity. In: Somasundaran P (ed.) Advances
in Mineral Processing. Littleton, CO: Society of Mining
Engineers.
Laskowski J and Kitchener JA (1969) Journal of Colloids
and Interfacial Science 29: 670.
Meagher L and Craig VSJ (1994) Langmuir 10: 2736.
Miller JD (1988) The SigniTcance of Electrochemistry in
the Analysis of Mineral Processing Phenomena. Seventh
Australian Electrochemistry Conference, Sydney, Aus-
tralia, February 14}19.
Miller JD, Hu Y, Veeramasuneni S and Lu Y (1999) Collo-
ids Surfaces 154: 137.
Nalaskowski J, Hupka J and Miller JD (1999) Physchemi-
cal Problems in Mineral Processing 33: 129.
Parker JL, Claesson PM and Attard P (1994) Journal of
Physical Chemistry 98: 8468.
Rabinovich YI and Yoon RH (1994) Langmuir 10: 1903.
Shibuichi S, Yamamoto T, Onda T and Tsujii K (1998)
Journal of Colloid Interface Science 208: 287.
Yamauchi G, Miller JD, Saito H, Takai K, Ueda T,
Takazawa H, Yamamoto H and Nislhi S (1996) Colloid
Surfaces A 116: 125.
Ye Y, Khandrika SM and Miller JD (1989) International
Journal of Mineral Processing 25: 221.
they form a froth. The froth concentrate Sows over
a weir and out of the Sotation cell, while the un-
wanted tailings Sow out of the bottom.
The effectiveness of this type of cell lies in the
ability of the bubbles rising in the liquid to collide
with particles in suspension. Because the concentra-
tion or hold-up of air in the liquid is not very high
} typically less than 10% by volume } the probability
of a collision is correspondingly low. The low fre-
quency of useful collisions between an individual
bubble and the particles in a Sotation machine can be
overcome by increasing the residence time of the
suspension. In this way, by using long residence times
which can sometimes be as much as an hour in a
1542 II / FLOTATION / Intensive Cells: Design
Sotation bank, it is possible to achieve high recoveries
of a Soatable material.
In recent times, a number of Sotation machines
have been introduced which seek to reduce the
residence time, by using new ways to bring about
the contact between particles and bubbles. These
are referred to as intensive Sotation cells. Although
the way in which the air is introduced } and the
bubbles are made } differs from one type to
another, they share a common feature. The collision
between particles and bubbles does not take place
within a liquid with a low number concentration of
bubbles, or with a low gas hold-up. Rather, the air is
introduced in such a way that contact is made in
a device with a high air void fraction } a high ratio of
gas volume to a given liquid volume. Once contact
has been made, the bubbly mixture, which resembles
a dense foam, passes to another vessel where the
bubbles can disengage from the liquid, bearing their
load of Soatable particles to the supernatant froth
layer.
The formation of bonds between bubbles and par-
ticles after collision is an essential step in Sotation,
and the topic has received much attention in Sotation
theory and practice. It has not always been appreci-
ated that phenomena which take place in the froth
phase above the liquid can also have a large
effect on overall Sotation performance. While
the yield or recovery of Soatable material obviously
requires an efRcient mechanism for contacting
particles and bubbles, the grade or purity of the
product is largely determined by froth-phase phe-
nomena. When a continuous cloud of bubbles, whose
surface contains selectively adsorbed hydrophobic
particles, rises upwards through the froth}liquid in-
terface, some of the liquid is trapped between the
bubbles and is entrained into the froth layer. This
liquid is the same composition as the main liquid in
the Sotation cell, so the concentration of gangue or
waste particles in the liquid in the froth is approxim-
ately the same as in the liquid layer from which it
arose, at least in the Rrst instance. The presence
of nonselective particles in the entrained water will
reduce the grade of the concentrate product. While
the bubbles are rising in the froth, the liquid layers
between bubbles are draining, and gangue particles
are carried downward, returning to the pulp layer.
Over the last 20 years, it has become commonplace to
apply clean water to the top of the froth layer, causing
a continuous downward Sow through the froth which
tends to wash out the entrained gangue. With froth
washing, Sotation products of very high grade can
easily be produced. (This assumes that the valuable
material is completely liberated from the gangue by
grinding. Any gangue which is locked into valuables
will generally Soat with the latter, thereby reducing
the concentrate grade.)
The main objectives in the design of froth Sotation
equipment are always the same: to produce a device
capable of achieving high grades and recoveries, with
small size, minimum capital and operating costs, ease
of operation and maintenance. To meet these objec-
tives, many new cells have been tried over the years.
This review will concentrate on the limited range of
such cells which can genuinely be described as inten-
sive, in that the contact time between bubbles and
particles is very short, and the Sotation cells are
correspondingly quite small relative to the through-
put. These are the air-sparged hydrocyclone (ASH),
the Jameson cell, and the Ekof cell.
The Air-Sparged Hydrocyclone
The ASH was invented by Professor Jan Miller of the
University of Utah, and was patented in the USA in
1981. It makes use of the centrifugal forces which
arise when air is sparged through the walls of a
hydrocyclone. The device consists of a cylinder
with a porous wall enclosed in an external chamber
(Figure 1). The feed slurry enters tangentially through
a conventional hydrocyclone header at the top of the
cyclone, to form an annular liquid layer on the inner
surface of the porous wall. The slurry moves down-
wards through the cylinder with a strong swirling
motion. Bubbles are generated at the surface of the
porous wall and, because of the swirling motion,
bubbles which are produced on the porous surface
experience an inwardly directed centrifugal force
which carries them away from the wall, to pass quick-
ly through the annular layer, collecting Soatable par-
ticles on the way, forming a froth layer in the core of
the cyclone. The froth leaves through the vortex
Rnder in the top of the cylinder, while the tailing
particles whose density is greater than that of water
move towards the wall and are discharged through an
annular gap in the bottom of the vessel.
An important feature of the ASH is the froth ped-
estal in the base. This stabilizes the froth and prevents
it from passing out in the tailings. The froth zone is
forced to move upwards through the vortex Rnder,
carrying the hydrophobic particles. The hydrophilic
particles are carried out in the tailings slurry.
The performance of the ASH is dictated by the Suid
motion in the swirl layer adjacent to the porous wall,
which in turn is controlled by the kinetic energy in the
inSowing slurry, and the physical dimensions of the
header and the vertical cylinder.
The bubble contact time in the hydrocyclone is of
the same order as that of the pulp residence time,
around 10 s. There is a correspondingly high capacity

Figure 1 Schematic of the air-sparged hydrocyclone.
per unit volume, which is of the order of 100}
600 tons day\1 ft\3 of cell volume (3600}21 500
tonne day\1 m\3) as against 1}2 tons day ft\3 (35}
70 tonnes day\1 m\3) for mechanical cells and col-
umns. To date, the cells are not very large, but the
capacity is quite high. Thus an ASH of diameter 5 cm
and height 50 cm has a capacity of 3}18 tpd of solids.
The feed enters at conventional hydrocyclone pres-
sures of 5}25 psi (35}170 kPa) and the air is supplied
at a relatively high pressure of around 65 psi
(440 kPa), which is necessary to force the air through
the porous wall at the required Sow rate.
An important parameter which limits the perfor-
mance of Sotation cells is the superRcial velocity Jg,
which is the volumetric Sow rate of the Sotation air
divided by the cross-sectional area of the pulp normal
to the direction of the air Sow. A high Jg will lead to
a high concentrate production rate, other things being
equal. In conventional cells, the only force acting on
the liquid in the froth is that of gravity. Because of the
centrifugal Reld in the ASH, the drainage force on the
liquid in the froth is enhanced, and high Jgs are
possible. Thus, the typical air velocity in an ASH is
around 1 standard L min\1 cm\2 of cylinder wall,
which corresponds to a superRcial velocity Jg of
17 cm s\1. This Rgure may be compared with typical
values for Sotation columns, which are of the order of
II / FLOTATION / Intensive Cells: Design 1543
0.5}4 cm s\1, and mechanical cells where the Rgure is
generally lower still } around 1 cm s\1. The conse-
quence is that the ratio of air-to-pulp Sow rates can
be very high, leading to high recoveries despite the
short residence time. Reported values of the air-to-
pulp ratio are as high as 16 : 1. In mechanical cells
and Sotation columns, the ratio is usually 1 : 1.
As far as contact between particles and bubbles
is concerned, the ASH is clearly a very intensive
Sotation device. However, it is not so effective at
handling the froth-phase requirements. Ideally, to ob-
tain high grades, it is necessary to be able to apply
clean washwater, which can drain through the froth
and Sush the gangue into the tailings stream, while
leaving the hydrophobic material attached to the bub-
bles. For this to occur, the velocity at which water can
drain through the froth under gravity must be greater
than the superRcial upward froth velocity in the core.
Using published data, it is possible to calculate that
the axial upward velocity of the froth core in an ASH
is in the range 180}1300 cm s\1. The diameter of
Sotation columns is Rxed to allow for froth washing
and the maximum working superRcial air velocity
Jg is about 4 cm s\1 } far below the values attained in
the ASH. Evidently it is not possible to design an ASH
which can allow both intensive contact between bub-
bles and particles, and effective control of the
1544 II / FLOTATION / Intensive Cells: Design
froth to obtain high grades. Accordingly, the ASH is
most effective in applications where grade is un-
important, and where high recovery is desired. It is
not surprising that the Rrst large scale applications
have appeared in the paper industry, for the removal
of toner particles from recycled paper.
The Jameson Cell
The Jameson Sotation cell was invented by Professor
Graeme Jameson at the University of Newcastle,
Australia, in 1986. It was developed to a practical
reality at Mount Isa Mines, Mount Isa, Queensland,
and was licensed to MIM Holdings of Brisbane in
1989. To date, there are 187 installations worldwide,
in 19 countries. The cell is used for roughing, scav-
enging and cleaning, and also for removing oil haze
from solvent extraction liquors. The distribution by
Reld of use is coal 37%; copper 29%; other minerals
17%; and solvent extraction 17%.
In this cell, contact between particles and bubbles
takes place in a dense foam which is produced in
a vertical downcomer, as depicted in Figure 2. The
pulp is introduced to the top of the downcomer as
a conRned liquid jet, and air is entrained into the feed
and broken up into Rne bubbles by the jet. A dense
foam with a high void fraction is created in the
Figure 2 Schematic of the Jameson cell. The height overall is
approximately 2}3 m. The cross-sectional area of the cell is dir-
ectly proportional to the desired feed flow rate.
downcomer, creating a very favourable environment
for collision of particles and bubbles. In fact, because
of the high void fraction, of the order 50}60% by
volume, the pulp is distributed in the form of thin
liquid Rlms between the bubbles, and collection oc-
curs by migration of particles within the thin Rlms,
which are not much thicker than the diameter of the
particles.
The dense mixture of bubbles and pulp discharges
at the base of the downcomer, and the bubbles disen-
gage from the pulp, rising into the froth layer. The
bubble-free pulp discharges as tailings from the bot-
tom of the cell. The froth behaves like that on top of
a Sotation column, in that grade and recovery can be
strongly inSuenced by the froth depth and the ap-
plication of washwater, and the upward superRcial
air velocity Jg. From the point of view of collection,
the downcomer operates best when the ratio of air
rate to feed rate is less than one-to-one on a volume
basis.
The froth is treated much as in conventional col-
umns. Washwater is usually applied if a high grade
product is required. As with columns, when the air
rate is altered, both steps in the Sotation process
} particle/bubble contact and froth entrainment } are
affected. Thus an increase in air rate may cause
an increase in recovery because more bubble surface
area is created on which to capture particles, and
because changes in the ratio of bubble to particle sizes
will affect the probability of collision. At the
same time, there may be an increase in entrainment of
the gangue into the froth which may lead to a de-
crease in grade, unless steps are taken to remove the
entrained gangue by changes in froth depth and wash-
water rate. Thus the optimum performance of the cell
is related to the air superRcial velocity, Jg.
The key features of the cell are:
1. The contacting environment is highly intensive,
so that only short residence times are required.
The total cell residence time is 1}2 min; the resi-
dence time in the downcomer is around 10 s.
A short column is therefore produced which is
ideal for retroRt, or installation in cramped head-
room. The Soor area is, however, similar to that
required by conventional columns for the same
throughput.
2. The bubbles formed by the impinging jet are very
small, offering enhanced carrying capabilities
for Rne concentrate particles.
3. Air is drawn in from the atmosphere and no air
compressor or blower is needed.
4. In the cleaning zone, with the use of washwater,
the levels of concentrate grade approach the max-
imum levels possible.

The size of bubbles produced in the Sotation cell is
an important determinant of cell capacity. The mass
of particles which can be carried out on the surface of
the bubbles is dependent on the gas}liquid interfacial
area. For a given gas Sow rate, the interfacial area is
inversely proportional to the bubble size, so there is
an advantage in making small bubbles. However, it
must be kept in mind that in the disengagement zone,
the buoyancy of the bubbles must be sufRcient to
lift particles of the largest size in the pulp to the
surface of the liquid. The best compromise appears to
be to make bubbles in the range 0.35}1 mm. Bubble
sizings on full scale operating cells and test cells show
that the Jameson cell produces an arithmetic mean
bubble diameter of the order of 300}600
m, while
the Sauter (volume-to-surface) mean diameter, dvs, is
of the order 360}950
m. These sizings compare very
favourably with conventional columns where the
Sauter mean bubble size is typically 2}3 mm.
Some general operating characteristics of the
Jameson cell are now discussed.
Air Velocity in the Cell
The superRcial gas velocity is the upward superRcial
velocity of air in a Sotation cell, calculated by divid-
ing the downcomer air rate (cm3 s\1) by the cross-
sectional area (cm2) of the riser part of the cell. The
cell is normally circular or rectangular in section, and
the appropriate cross-sectional area is simply the area
normal to the direction of the Sow of the froth,
excluding the area occupied by the downcomer(s).
The superRcial velocity Jg is conveniently expressed in
units of cm s\1 because values typically range from
0.5 to 4 cm s\1 in practice.
The recovery and concentrate carrying rate
(g min\1 cm\2) tend to increase with increasing Jg. As
in conventional columns, there is a limiting upward
Sux of bubble surface through the pulp above which
froth Sooding occurs, resulting in the loss of
froth}pulp interface, a very wet froth and total loss of
selectivity. There is consequently a maximum air rate
Jgmax deRned by this limiting Sux and bubble size. In
Sooding, the entire cell Rlls with froth as the only
stable phase, and there is no pulp phase.
The operating Jg used in the sizing of the Jameson
cell depends strongly on the application, and on the
residual reagent concentrations from any upstream
processes. Generally speaking, low values ( Jg"
0.4}0.8 cm s\1) are employed in cleaning appli-
cations, and high values ( Jg"1.0}2.0 cm s\1) are
employed in roughing or scavenging applications.
In cleaning operations, a high proportion of the
feed reports to the concentrate, and the froth loading
tends to be high. Consequently, the bubbles are well
II / FLOTATION / Intensive Cells: Design 1545
coated with particles, which tend to stabilize the froth
by reducing the froth coalescence rate. The drainage
rate of the interstitial liquid in the froth is retarded by
the relatively high concentration of particles, which
has the effect of increasing the apparent viscosity
of the interstitial liquid. Accordingly, it is necessary to
design for lower values of Jg to allow time for the
gangue to drain from the froth to obtain the required
high grade. In roughing applications, however, only
a small fraction of the feed reports to the concentrate
and the froths formed tend to be less stable as a conse-
quence. Also, gangue entrainment is not such a seri-
ous problem, because it can be dealt with in the
downstream cleaning circuit. As a consequence of the
higher coalescence rate the froth bed is shallower
than that of the cleaners and a lower froth residence
time will give good drainage. It is therefore usual to
design a Jameson cell for a roughing application with
a higher Jg than in the cleaners.
In some circumstances, high residual concentration
of reagents in the feed necessitates the use of low
values of Jg to avoid froth Sooding. Although frother
concentration is of primary importance to bubble size
and hence the advent of froth Sooding, circumstances
have arisen where collector and frother interaction
has been observed. In such a case, the frother dose
should be decreased if collector dose is increased, and
vice versa. Too high a frother or collector concentra-
tion can lead to froth Sooding while too low a dose
can lead to loss of froth stability.
Particle size can also have an inSuence on the max-
imum Jg, due to its effect on froth stability through
bubble-bridging. Small particles (less than 100
m) are
easily collected at low gas rates, while recovery of
coarser particles may be assisted by higher rates.
A complex system of liquid and air recirculation
patterns forms in the bottom of the cell. The cell
design is based on downcomer Sows and downcomer
placement to optimize this system to produce best
grade and recovery. There is no limit on cell volume,
providing the net downwards velocity of pulp, JL , is
sufRciently low to avoid the entrainment of bub-
bles in the underSow. When the froth and disengage-
ment zones have the same cross-sectional area, the
two important velocities are the rate of rise of the
bubbles in the pulp, and the rate of drainage of liquid
in the froth. The former is usually greater than the
latter, so that a column sized to give the correct Jg will
also give the correct JL, and bubble entrainment in the
downward Sow will not be a problem.
Froth Depth
The froth phase in a Jameson cell can be controlled as
in conventional columns. Shallow froth depths (less
1546 II / FLOTATION / Intensive Cells: Design
than 200 mm) are used where high recovery is neces-
sary and grade is of secondary importance, while
deeper froths (up to 1 m) are employed to obtain
maximum concentrate grade. Shallow froths result in
signiRcant entrainment of very Rne ((10
m) gangue
mineral which accompany the pulp phase. With
deeper froths, signiRcant drainage of hydrophilic gan-
gue will take place, producing a higher grade concen-
trate, and a higher percentage of solids in the concen-
trate. Under some circumstances, the addition of
washwater to the froth will assist froth mobility, and
assist an otherwise immobile froth to keep moving to
the overSow lip.
Air/Pulp Ratio
The air-to-pulp ratio (the ratio of the volumetric Sow
rates of air and pulp) in Jameson cells is usually in the
range 0.3}0.9. Experiences with large (2}3 m dia-
meter) Jameson cells indicate that operation at a low
air/pulp ratio does not detract from metallurgical
performance providing the superRcial gas velocity
Jg is maintained above 0.4 cm s\1. Operation at lower
air/pulp ratios has a stabilizing effect producing
a Rner, more uniform bubble size. A signiRcant ad-
vantage of operation at lower air/feed ratios is that
lower concentrations of frother are required.
The Sux of interfacial area for a given gas rate
varies directly as the gas Sow rate and inversely as the
bubble size. Thus the Sux of bubble surface area,
interfacial area per unit area of column cross-section
per unit time, can be maintained with reduced super-
Rcial gas rate, providing the bubble size decreases
accordingly.
The Effect of Washwater
Clean water can be applied to the top of the froth, to
Sush entrained material downwards, preventing it
from Sowing out with the Sotation product. There
are two measures which are used to measure and
control washwater addition: the bias and the wash-
water ratio.
Bias is the absolute excess of the washwater applied
to the froth, over the quantity of water being re-
covered in the concentrate, expressed as a superRcial
velocity Jb (cm s\1): Jb"(QWW!QWC)/AC, where
QWW, QWC are the volumetric Sow rates of wash-
water and water in concentrate, and AC is the cross-
sectional area of the column.
The washwater ratio is deRned as the ratio of the
washwater addition rate, to the Sow rate of water in
the concentrate: W"QWW/QWC. The washwater ra-
tio is a relative measure of the amount of washwater
applied. If no washwater is used, the washwater ratio
is zero and the bias is negative. When Jb"0, W"1.
A positive bias corresponds to washwater ratios
greater than unity.
Although the bias does give an indication of the
absolute amount of washwater being added, its use
can be misleading because it does not take into ac-
count the wide variation in the absolute values of the
rate of water entrainment in the concentrate. It is
preferable to use the washwater ratio, which is a rela-
tive Rgure. In practice, it has been found that best
performance is achieved when the washwater ratio is
greater than 1.
Scale-Up
Scale-up of the Jameson cell is relatively simple, since
the Sotation capacity is proportional to the cross-
sectional area of the cell, and the Sow capabilities of
the downcomer. Downcomers range in size from 0.2
to 0.36 m, and a large installation will have multiple
downcomers. Large cells are typically 5 m in dia-
meter, with 12}16 downcomers, and handle Sow
rates of 1200 m3 h\1. Extensive testing has shown
that the results obtained in small test units, of dia-
meter 0.3 m, give an accurate picture of the perfor-
mance of a large cell on the same feed. In many cases,
test work is not required, because of the availability
of data from operational plants which will allow
a design to be established for a new application with
an ore of similar characteristics.
Because there is a limit on the amount of air which
can be supplied to a given amount of feed slurry in the
downcomer, the Jameson cell can become limited by
carrying capacity. Thus, if there is an excess of par-
ticles in the feed above the mass which can be carried
by the available surface area of bubbles, some of the
hydrophobic material will not be transferred to the
concentrate. In such cases, it may be necessary to
install a second cell in series with the Rrst, to ensure
a high recovery of the values. An alternative which is
increasingly being used is to recycle part of the tail-
ings. The feed pump is then sized so as to be well
above the normal operating capacity. The feed pump
draws from a pump box in the circuit ahead of the
Jameson cell, and receives Sow from two sources:
new feed and recycled tails. It has been found that the
recovery with recycle in the range 30}50% of the feed
Sow rate is equivalent to the addition of a second cell
in series.
The Ekof Cell
The Ekof cell, also known as the PneuSot cell is
marketed by KHD Humboldt Wedag, of Bochum,
Germany. It arose from a cell invented by Professor
Albert Bahr, of Clausthal Technical University,
Germany, in 1974. The Bahr cell consists of a vessel
 II / FLOTATION / Intensive Cells: Design 1547

of inverted conical form. The feed is premixed with

air in a series of aerators distributed around the per-
iphery of the vessel, in which air is injected through
a porous wall into the transversely moving pulp. The
bubbly pulp mixture is then fed to the vessel, where
the bubbles rise and make contact with the Soatable
particles and carry them to the surface.
In the Bahr cell, the air}pulp mixture is introduced
through pipes which enter through the wall of the
vessel part-way up from the bottom of the cone,
pointing vertically upwards to form a jet or plume.
The jets are equispaced about the periphery of the
cell. The idea was that a jet would spread out and mix
with the pulp in the tank, bringing the bubbles into
contact with the particles. Each jet would increase in
area with vertical height, until at the surface the
cross-sectional area of the jets would be about the
same as that of the cell. The problem with this con-
cept is that, as the size of the vessel and the design
throughput increases, the height of the cell must
increase as well. This difRculty was overcome in
a later design in which the pipes delivering the Figure 3 The essential features of the Ekof or Pneuflot cell of
air}pulp mixture are directed tangential to the cell the new form.
wall, so that a low speed swirl develops. This design
has been referred to as a PneuSot cell. In both the
Bahr and PneuSot cells, the froth discharges over a lip back into the tank. From a plot of the concentration
into an annular launder which surrounds the vessel. of the element of interest in the froth concentrate, the
Because of the way the bubbles are generated, feed and the tailings, all as a function of time, it is
a high air pressure is needed to drive the air through possible to deduce the number of stages required by
the porous wall, and a relatively high feed pressure is a stepping procedure.
needed to accelerate the pulp to the required speed in The PneuSot cell is in use in a number of applica-
the aerator. In early models, froth washing was not tions, on magnesite, copper, galenaSuorite, etc.
provided, but it is available in later versions. The
Bahr cell has had success especially in coal Sotation.
In PneuSot cells of the new form (Figure 3) the feed
Future Developments
enters through a vertical pipe, and compressed air is Further intensiRcation of the Sotation process is likely
introduced through small openings in an aerator unit to come about in two directions. The processes of
at the top of the pipe. A model is also available in bubble contacting, bubble transfer to the pulp}liquid
which the air is introduced into a Venturi. The interface and froth drainage, are all responsive to
aerated pulp is led through a central pipe to a low a force Reld which can induce body forces on
point in the Sotation cell, and is diverted upwards by the liquid. In normal circumstances, the main body
a distributor. The purpose of the distributor is to force is that of gravity. Accordingly, if the effec-
create an upward Sow of bubbles which promotes tive Sotation rate per unit volume is to be increased,
Sotation of coarse particles. the logical step is to subject both the liquid and the
The latest form of PneuSot cells is Rtted with froth to a centrifugal Reld. This will undoubtedly
a froth crowder, which is of value when the loading of increase the mechanical complexity of the apparatus,
particles in the froth is low, and the froth is relatively and the saving in Sotation cell volume may not war-
unstable. The use of washwater is also possible in this rant the extra cost of building, running and maintain-
design, as depicted in Figure 3. ing the equipment. Another possible direction for
Scale-up of the PneuSot cell is not possible without increased intensiRcation is in the design of the initial
testing on a pilot plant. There is no Sotation time as in gas}liquid contacting device. We have seen that, in
mechanical cells. The number of stages needed for both the Jameson cell and the Ekof cell, contacting
a particular application can be determined by tests in takes place in a pipe or downcomer which must be
which a PneuSot cell is fed from an agitated tank at of sufRcient length to ensure efRcient contact
6}10 m3 h\1, with recycle of tailings from the cell between bubbles and particles. It is likely that other
1548 II / FLOTATION / Oil and Water Separation
contactors could be devised which could bring about
essentially instantaneous contact between particles
and bubbles, effectively eliminating the down-
comer. However, the over-riding objectives men-
tioned earlier } achieving high grades and recoveries,
with small size, minimum capital and operating costs,
in equipment which is easy to operate and maintain
} will always remain the prime concerns of industrial
users. In any new approaches, these objectives must
be kept in view.
Further Reading
Bahr A and Ludke H (1982) The development and intro-
duction of a new coal Sotation cell. In: Proceedings of
the XIV International Mineral Processing Congress,
Toronto, Canada, October 17}23, pp. VII-5.1}
VII. 5-14. Montreal: Canadian Institute of Mining and
Metallurgy.
Bahr A, Imhof R and Ludke H (1985) Applications and
sizing of a new pneumatic Sotation cell. In: Proceedings
of the XV International Mineral Processing Conference,
Cannes, France; OrganiseH sous l’ED gide de la SocieH teH de
l’industrie MineH rale et du Bureau de Recherches
GeH ologiques et Minières. St-Etienne, France: Edition
GEDIM, 1985}1986.
Bahr A, Legner K, Ludke H and Mehroff F-W (1987)
Five years of operational experience with pneumatic
Sotation in coal preparation. Aufbereitungs Technik 28:
1}10.
Bahr A, Imhoff R, Changgen L and Muller W (1991)
Development and progress in the application of the
pneumatic Sotation cell. In: Agar GE, Huls BJ and
Hyma DB (eds) Column ’91, Proceedings of an Interna-
tional Conference on Column Flotation, Sudbury, On-
tario, Canada, 2}6 June, p. 703. Montreal: Canadian
Institute of Mining, Metallurgy and Petroleum.
Cordes H (1997) Development of pneumatic Sotation cells
to their present day status. Aufbereitungs-Technik 38:
69}78.
Dawson W, Yannoulis GF, Atkinson BW and Jameson GJ
(1996) Applications of the Jameson cell in the Australian
Oil and Water Separation
B. Knox-Holmes, Baker Hughes Process Systems,
Rugby, UK
Copyright ^ 2000 Academic Press
Development
The process of Sotation needs a gas bubble to
collide with, and attach to, an oil droplet; because
of the hydrophobic nature of the oil droplet, a
coal industry. In: Gomez CO and Finch JA (eds) Column
’96: Proceedings of the International Symposium on
Column Flotation, Montreal, Canada, August 26}28,
pp. 233d246. Montreal: Canadian Institute of Mining,
Metallurgy and Petroleum.
Evans GM, Atkinson BW and Jameson GJ (1996)
Recent advances in Jameson cell technology. In:
Gomez CO and Finch JA (eds) Column ’96: Proceedings
of the International Symposium on Column Flotation,
Montreal, Canada, August 26}28, pp. 39}49.
Montreal: Canadian Institute of Mining, Metallurgy
and Petroleum.
Finch JA (1995) Column Sotation } part 4: novel Sotation
devices. Minerals Engineering 8: 587}602.
Finch JA and Dobby GS (1990) Column Flotation. Oxford:
Pergamon Press.
Harbort GJ, Jackson BR and Manlapig EV (1994) Recent
advances in Jameson Sotation cell technology. Minerals
Engineering 7: 319}332.
Imhof R (1991) Device for carrying out pneumatic Sota-
tion. German patent application DE41 16 645.0.
Jameson GJ (1988) A new concept in Sotation
machine design. Minerals and Metallurgical Processing
5: 44}47.
Jameson GJ (1990) Column Sotation method and appar-
atus. US patent 4,938,865.
Jameson GJ (1994) Column Sotation method. US patent
5,332,100.
Miller JD (1981) Air-sparged hydrocyclone and method.
US patent 4,279,743.
Miller JD, Ye Y, Pacquet E, Baker MW and Gopala-
krishnan S (1988) Design and operating variables in
Sotation separation with the air-sparged hydrocyclone.
In: Forssberg KSE (ed.) Proceedings of XVI Interna-
tional Mineral Processing Congress, Stockholm, pp.
499}510. Amsterdam: Elsevier.
Sanchez SP, Rojos FT, Fuentes GB et al. (1997) Ekof pneu-
matic Sotation technology: the alternative for rougher,
scavenger or cleaner Sotation of metallic ores. In:
Hoberg H (ed.) Proceedings of XX International
Mineral Processing Congress, Aachen, Germany.
Clausthal-Zellerfeld: GMDB Gesellschaft fuK r Bergbau,
Metallurgie, Rohstoff und Umwelttechnik.
stable gas}oil matrix is formed. The buoyancy of
the oil droplet is increased by the attachment of the
gas bubble, causing the oil droplet to rise rapidly
through the water. Typically, one gas bubble will
attach to one similar-sized oil droplet as described by
Leech. As the bubbles in the froth phase burst, an oil
layer is formed on the surface of the water. Oil and
froth are then removed from the surface on an inter-
mittent or continuous basis, depending on the mecha-
nism used.

Flotation is a kinetic process. While a number of
Sotation models exist, Klimpel’s Rrst-order rate equa-
tion has been demonstrated to provide modelling
Sexibility, ease of physical interpretation and a good
Rt to the experimental data. The Klimpel model is
written as:


exp(!kt)
R0(t)" 1! 1!
kt
where R0(t) is the fractional recovery of oil at time t(s)
and k is the characteristic Rrst rate constant (s\1). The
key to Sotation is the production of air or gas bub-
bles. The two major techniques are known as dis-
solved air Sotation (DAF) and induced air Sotation
(IAF). IAF can be further subdivided into mechan-
ically and hydraulically induced Sotation.
Mechanically induced Sotation has now been used
for about 100 years to separate from a suspension in
water, particles of valuable mineral from gangue. In
the mining industry this process is known as beneRci-
ation. It is now the main method of concentrating
copper, molybdenum, iron, phosphate, lead and zinc
ores. In the minerals industry the air is dispersed as
bubbles, either through an impeller in subaeration
cells or through spargers in Sotation columns. These
and other variants of the air addition to Sotation
pulps are classiRed as IAF. In the petroleum industry
natural gas, carbon dioxide or nitrogen may be used
as the Sotation gas, hence the process is termed in-
duced gas Sotation or IGF. The use of these gases
signiRcantly reduces downstream corrosion problems
and possible hydrocarbon degradation caused by the
use of air.
Flotation techniques can remove dispersed, but not
dissolved oil. As environmental legislation speciRes
both the limiting oil concentration and also the biolo-
gical oxygen demand (BOD) in the discharged water,
Sotation is often the second stage of a three-stage
efSuent treatment process of gravity separation,
Sotation and biological treatment. As an example,
British Petroleum’s Grangemouth facility has
achieved discharge concentrations of 2}3 mg L\1 oil-
in-water using gravity separation in American Petro-
leum Institute (API) separators, followed by IAF with
a Rnal biological treatment stage to remove dissolved
BOD materials including ammonia, phenols and sul-
Rdes. Limits for oil-in-water discharges vary around
the world, typically between 15 and 40 mg L\1.
A maximum BOD of 216 mg L\1 daily, with
a monthly average of 53 mg L\1 is considered to be
achievable via the application of the best available
technology by the United States Environmental Pro-
tection Agency.
II / FLOTATION / Oil and Water Separation 1549
Process Techniques
Dissolved Air Flotation (DAF)
In this method compressed air or gas (nitrogen, car-
bon dioxide or methane) is dissolved into all or part,
of a process liquid under pressure in a retention
vessel. The gas}oil}water mixture is then sent to
a Sotation cell, where the pressure is reduced, causing
bubbles to come out of solution. The bubbles then
attach to and are possibly nucleated on the oil and
suspended particles.
The solubility of a gas in water is proportional to
its partial pressure and inversely proportional to the
water temperature. Solubility may be characterized
by Henry’s law:
p
X G"
HG
where XG is the molecular fraction of the gaseous
component in the liquid, p is the partial gas pressure
and HG is a constant. The release of gas following
a reduction in pressure is proportional to:
1

XG" (p!pamb)
HG
The Henry’s law constants at 253C for some gases
used in DAF are given in Table 1.
Flotation efRciency depends on the gas used.
The effectiveness of the various gases in terms of
their bubble release increases in the following order:
nitrogen, oxygen, natural gas (methane) and carbon
dioxide. Solubility is also reduced as the dissolved
solids content is increased. The amount of gas that
can be dissolved ranges from 50% to 90% of its
equilibrium solubility depending on the design of the
pressurization system.
There are a number of ways of dissolving gas under
pressure. The gas can be sparged into the liquid in
a pressure vessel, liquid can be trickled over a packed
bed or sprayed into an unpacked vessel, gas can be
entrained with ejectors or gas can be injected into the
suction side of the recycle pump.
Table 1 Henry’s law constants at 253C for various gases used
in DAF
Gas H (atm/mol fraction)
Nitrogen 9.08;104
Oxygen 4.38;104
Methane 4.13;104
Carbon dioxide 1.64;103
1550 II / FLOTATION / Oil and Water Separation
Full Uow pressurization transfers gas to the whole
feed Sow at a relatively low pressure of between 30
and 40 psi. This technique is suitable when sufR-
cient gas can be dissolved to effect Sotation, and
the passage of the whole Sow through a centrifugal
pump will not impair the subsequent Sotation process
through Soc shearing.
Partial Uow pressurization passes a proportion of
the full Sow through the pressurization system at
60}75 psi. This method reduces the size of the pressur-
ization system, resulting in signiRcant cost savings.
This technique is suitable when sufRcient gas can
be dissolved to effect Sotation, and passage of the
partial Sow through the pump will not impair Sotation.
Recycle Uow pressurization is used when a natural
or chemically formed Soc is to be separated from
a wastewater. A portion of the clariRed Sotation
efSuent is recycled and used as the carrier of the
dissolved gas. This latter technique is the most ef-
Rcient and accounts for the majority of installations.
Figure 1 shows a Sowsheet of a system operating
with recycle Sow, a Sotation cell operating at atmos-
pheric pressure, a pressurized retention vessel, feed
and recycle pumps and a backpressure valve.
A level controller controls the Sow into the pressur-
ization vessel and excess gas may be vented. There is
adequate residence time (typically 1}3 min) in the
pressurization vessel for sufRcient gas dissolution
to take place (50}90% saturation). The pressurized
liquid from the vessel is mixed with fresh feed, and is
discharged through a back pressure valve to the Sota-
tion tank.
In the Sotation tank, which may be circular or
rectangular, the pressure is typically reduced to atmos-
pheric pressure, and this reduction or let-down causes
bubbles between 1 and 120
m in diameter to come
out of solution. Bubble size depends on the operation
of the pressure let-down valve.
Figure 1 Flow diagram for a dissolved air flotation system. Courtesy of Baker Process.
Gas bubbles may form by nucleation on an oil
droplet or solid particle, or they may come out of
solution and then attach to oil droplets and sus-
pended solids by collision, or they may become trap-
ped in a solid}chemical or oil droplet}chemical Soc.
Chemical usage is determined by the total chemistry
of the system, and a series of bottle tests at site will be
necessary to optimize performance.
Floated oil and suspended solids are removed
by skimmers, while non-Soatable settlings are re-
moved from the bottom of the cell by a grit scraper.
The efRciency of the removal process depends on
the ratio of air to solids and/or oil in the water. Too
little air and separation will not be achieved. Too
much air and the additional turbulence may actually
reduce separation performance by causing Soc re-
entrainment, resulting in a reduction in energy ef-
Rciency.
The DAF machine can be characterized by being
a relatively quiescent, high retention time device
(15}30 min), using small volumes of dissolved gas
(35}180 L m\3 of throughput). Depending on gas
type, dissolution pressure, stream temperature and
suspended solids loading, DAF may achieve 80}95%
removal of free and emulsiRed oil and suspended
solids.
Induced Gas Flotation (IGF)
In the induced gas Sotation (IGF) (or IAF) process,
which resembles the design of a subaeration minerals
Sotation cell (see Figure 2), bubbles are induced
mechanically.
IGF uses a star-shaped impeller to generate intense
local turbulence. This results in subatmospheric pres-
sures being generated in the region surrounding the
impeller which causes gas to be induced from the gas
space at the top of the compartment via gas inlet

Figure 2 Transverse cross-section of mechanical induced gas flotation machine. Courtesy of Baker Process.
ducts. Impeller rotation also causes an upward circu-
lation from the bottom of the vessel. The gas and
liquid mix to form a relatively homogenous two-
phase (gas}liquid) dispersion which leaves the impel-
ler with a mainly tangential velocity. The impeller is
shrouded by a perforated cylindrical disperser, which
separates the intensely turbulent inner zone from the
relatively quiescent outer region of the rest of the cell.
The dispersion passes through the disperser, which,
because of the shear resistance to the Sow through its
wall, reduces the size of the bubbles and improves the
uniformity of their radial distribution through the
main cell volume, thereby increasing the probability
of a bubble}droplet collision. After attachment of the
bubble to the droplet or suspended particle, separ-
ation of oil and solids occurs by Sotation. The surface
of the cell remains relatively quiescent as a result of
the bafSing effect of the disperser and hood,
which minimize re-entrainment of Soated oil and
solids. Since the upward surface Sow is uniform in the
outer quiescent region surrounding the impeller, the
loaded bubbles which form a froth layer at the upper
surface of the cell are usually removed simultaneously
from both sides of the cell.
A reduction in bubble size reduces gas Sow require-
ments because of the more favourable surface area-to-
volume ratio of the smaller bubbles. There is, however,
an optimum bubble size of about 10
m as the colli-
sion efRciency is reduced below this size. Design
considerations require a balance between impeller
power input and hence total gas Sow, mixing region
shear turbulence, surface and Sotation zone quiesc-
ence, oil droplet re-emulsiRcation and gas bubble size.
II / FLOTATION / Oil and Water Separation 1551
An IAF usually has four or more cells in series with
a 1 min residence time per cell. This reduces Sow
short-circuiting, thereby improving separation ef-
Rciency. Water enters the Rrst active Sotation cell via
a feedbox, and passes from cell to cell via underSow
weirs in the connecting bulkheads. Floated oil is re-
moved separately from each cell. Dispersed oil drop-
let concentration in the inSuent should typically be no
greater than 500 ppm on a long-term basis of which
approximately 50% is removed by each cell, the per-
centage removal efRciency increasing with the
inSuent oil droplet concentration for a Rxed residence
time. The treated water Rnally enters a quiescent
discharge cell with an approximately 1 min residence
time where separation continues as gas bubbles, still
entrained in the water leaving the last cell, rise to the
surface. Individual IGF machines are typically avail-
able in different sizes to treat feed Sows of be-
tween 50 and 5000 gpm (Figure 3).
The performance of the IGF process has been ex-
tensively investigated, and proprietary predictive
mathematical models derived. In the context of the
trials, the most inSuential variables in IGF, listed in
order of decreasing importance, are:
E water-treating chemical concentration;
E feed water Sow rate (relates to residence time);
E impeller speed (increasing the speed will ingest
more air and increase power consumption with
relatively little change in Suid circulation);
E impeller submergence (distance between the liquid
surface and the top of the rotor. Increased sub-
mergence increases power draw, whilst reducing
1552 II / FLOTATION / Oil and Water Separation
Figure 3 Longitudinal cross-section of mechanical induced gas flotation machine. Courtesy of Baker Process.
gas injection and increasing the liquid recirculation
rate. Typically 200 mm);
E impeller engagement (distance from bottom of the
impeller to top of draft tube. Engagement is posit-
ive if the rotor is in the tube and negative if it is
above the draft tube. InSuences liquid circulation.
Typically 50 mm.
The order of this list may vary for different
applications since removal efRciency is a strong
function of the type of crude oil and the chemical
composition of the feed water.
The IGF is a relatively low retention time (4}6 min)
device using relatively large volumes of gas (com-
pared to DAF) at near ambient pressure. The gas
dispersion in these cells is so effective that reten-
tion times are relatively short, allowing a reduction in
equipment size compared to the DAF method. The
process operates at near atmospheric pressure. Both
these factors give the IGF process a signiRcant eco-
nomic advantage over the DAF.
Induced Static Flotation Unit (ISF)
Since the IGF was originally developed for the mining
industry, the power required by the impeller needed
to be sufRcient to suspend solids of perhaps
0.2}0.3 mm in diameter at concentrations of
30}40% solids. As solids removal is not the dominant
process in oily water treatment, solids suspension
capability can be reduced in order to achieve low
turbulence in the Sotation vessel at reduced power.
This has led to the development of hydraulic rather
than mechanical induction of bubbles (see Figure 4).
The ISF generates bubbles hydraulically using an
eductor operating under pressure, usually 60 psig
rather than mechanically. Feed is added directly to
the Sotation vessel. ClariRed efSuent water is
recycled through a header, and gas is drawn from the
vapour space by a Venturi effect. The resultant
gas}liquid mixture is directed against a striker plate,
which causes the formation of numerous small bub-
bles that are distributed across the full cross-section
of the cell. The circular cross-section of the vessel
improves the uniformity of the bubble distribution in
the Sotation vessel, which improves the probability of
bubble}droplet collision.
Unlike the IGF, the ISF operates under pressure.
Pressure operation has the advantage that hazardous
(hydrogen sulRde) or environmentally sensitive gases
(hydrocarbons, carbon dioxide) are contained. It has
also eliminated the need for transfer pumps for the
clariRed water efSuent and the requirement for
mechanical skimmers to remove the Soated oil.
The ISF design reduces the number of moving
parts, while maintaining performance similar to the
industry-standard mechanical units. Operation of the
ISF with a centrally mounted skim trough and skim
cycle timers has provided the means to reduce the
skim volume to less than 1% of the forward Sow in
most cases. It also uses less power than the mechan-
ical IGF. All machine adjustments are external to the
vessel, thereby ensuring operator safety in hazardous
applications.
The use of eductors for inducing gas to generate gas
bubbles in the Sotation process was Rrst patented in
the mid-1970s. Initially, ISF performance was poor
compared to mechanical Sotation units. The design
was improved by employing a cylindrical pressure
vessel, centrally mounted skimmings trough for use
on Soating platforms and an improved eductor
conRguration. The most common version of the ISF
has four cells, hydraulically connected in series. In-
creases in the clean water recycle rate will cause
 II / FLOTATION / Oil and Water Separation 1553

Figure 4 Eductor for hydraulic induced static flotation machine. Courtesy of Baker Process.

a progressive reduction in residence time throughout
the unit, which may result in reduced bubble}droplet
collision efRciency. The clean water recycle rate
may be reduced by operating at a high nozzle pressure
(50}100 psi).
The key features of the hydraulic Sotation machine
are shown in the cut-away view in Figure 5.
Comparison of DAF and Induced Processes
Dissolved and induced processes differ in a num-
ber of key parameters.
1. The amount of gas transferred to the process water
is relatively small in the DAF process, approxim-

Figure 5 Hydraulic induced static flotation machine. Courtesy of Baker Process.

1554 II / FLOTATION / Oil and Water Separation
ately 20}40 times less than that used by the IGF
process.
2. Because the DAF works by dissolution of gas in
water it is sensitive to water temperature (in-
creased temperature reducing gas solubility). By
comparison, induced processes are relatively tem-
perature insensitive. In Relds where hot water or
steam Soods are used to recover oil, temperatures
above 403C will signiRcantly reduce the perfor-
mance of a DAF unit.
3. The means of bubble generation and mixing dif-
fers (pressure let-down in DAF vs mechanical or
hydraulic induction in the IAF and ISF, respective-
ly). Gas dissolved in the water is insigniRcant in
process terms in the induced mechanisms since
there is no pressure let-down.
4. When scaling up the process, residence time is the
key variable for the IGF. For the DAF it is hydrau-
lic loading (total process water feed divided by the
vessel surface area). Hydraulic loading is related to
plant Soor space.
The advantages of DAF are:
1. Often lower power requirement than IGF.
2. Lower volume of skimmings as a percentage of the
forward Sow (1}5% for DAF systems as com-
pared with 2}10% for mechanically induced gas
systems. ISF skimmings volume compares favour-
ably with that of DAF, being approximately 1% of
the forward Sow).
3. Smaller volumes of sludge may be formed than
with IGF, especially where organic rather than
inorganic (metallic) polyelectrolytes are used to
aid Sotation.
4. Better handling of suspended solids. DAF will allow
the use of metal salts for coagulation and Soccula-
tion. The DAF Sotation cell can also be Rtted with
a bottom skimmer for solid that can settle.
Advantages of induced processes are:
1. Relatively low capital cost.
2. Relatively small equipment footprint (single skid-
mounted unit vs a multicomponent system).
3. Volatile organic carbons (VOC) are contained
within the Sotation cell. Not all DAF units use
closed top Sotation cells.
4. Varied Sows can be handled easily.
5. Due to their high mechanical reliability, no
standby capacity is required.
6. Loss of one cell will not signiRcantly reduce separ-
ation efRciency, allowing on-line maintenance
of impeller mechanisms.
7. Lower chemical consumption, since Socculation is
not necessary.
In certain applications, such as removal of mineral
oils from a steel rolling mill efSuent, induced and
dissolved process have been used in series.
Chemical Selection
From a chemical addition standpoint the two types of
Sotation differ due to the size of the bubble each
creates:
E induced gas typically creates a relatively large gas
bubble in the 10}2000
m range;
E dissolved gas typically creates a relatively small gas
bubble in the 1}100
m range.
Dissolved Gas
In this instance, chemical treatment is used to Soccu-
late the oil droplets and solid particles. The larger the
Soc, the more gas bubbles are trapped underneath the
Soc structure. The Sotation process is signiRcantly
more quiescent than the induced processes, hence does
not tend to break up Socs. Separation is therefore
based on the amount of bubbles that are trapped
underneath the Soc, and the maintenance of uniform
Sow distribution to avoid Soc shearing. The inSuent/gas
bubble carrier Sow distributor and mixing chamber
design determines the efRciency of a DAF unit.
Ferric compounds work in a similar way to cationic
polymers except where polymers are speciRc to
the charge on the oil droplet; metal salts swamp the
bulk solution leading to charge destabilization by dis-
associating on addition to water. As a result, the dose
rate is typically higher than for a solution polymer.
The solubility of metal ions is limited and hence they
tend to generate weak Socs, which means that they are
better used as part of a two-stage treatment, usually in
conjunction with an anionic polymer. As an example,
one facility described by Berne and Cordonnier uses
10 mg L\1 aluminium sulfate with 1 mg L\1 anionic
polyelectrolyte. However, this does result in a large
volume of sludge, 10}30 m3 day\1 in this case. Use of
organic coagulants alone can signiRcantly reduce
sludge volumes (3}5 m3 day\1 in this example).
Induced Gas
The objective of chemical treatment is to change the
surface charge of either the gas bubble or the oil
droplet in order to cause adhesion between them after
collision. The size ratio of the bubble-to-oil droplet is
usually 41 : 1. The only solids that are Soated are
those whose surface charge is opposite to that on the
bubble surface, or those that are associated with, and
contained in, the oil droplet. For this reason, induced
processes may only remove 55}75% solids from the

forward Sow as compared with 95}98% oil droplet
removal. No attempt is made to Socculate oil droplets
and solid particles, because the mixing intensity in the
units will tend to break up any Soc structure that has
been formed.
In general, oil droplets and other suspended mater-
ial will be negatively charged. Addition of cationic
polymer neutralizes the charge on the oil}gas species,
while a long-chain polymer collects the contaminant
in preparation for removal. Most cationic polymers
work best in the pH range 6}9. If the pH falls outside
this range, pH adjustment of the wastewater may be
required. Overall removal rates are around 90%
without chemicals up to 98% with chemicals. In
order to size a unit for commercial use, laboratory
studies are conducted to determine the effect of
variables such as rotor speed, chemical addition and
feed rate on oil removal, all of which have an impact
on the rate constant K. The data are then fed into
proprietary models based on the Klimpel model,
which contain correction factors for scaling up the
equipment. The models have been validated using
numerous sets of data from commercial installations.
Oil}water separation is enhanced by using an emul-
sion breaker, a cationic high charge density, low
molecular weight coagulant polymer. The polymer is
distributed throughout the continuous (water) phase
where it neutralizes the anionic charge at the
oil}water interface. This destablizes the emulsion,
allowing oil droplets to coalesce by collision with one
another.
Figure 6 Single cell ISF. Courtesy of Baker Hughes Process Systems.
II / FLOTATION / Oil and Water Separation 1555
Cationic solution polymers would usually be ap-
plied as a 1}10% solution with a dose rate of
2}30 ppm. Emulsion polymers (tightly coiled poly-
mer molecules entrapped in solvent, activated by dilu-
tion in water) are applied at concentrations no greater
than 1}2% at dose rates of 0.5}5.0 ppm.
Future Developments
Hydrocyclones and Flotation
Flotation was practised extensively on Rxed off-
shore platforms throughout the 1970s and 1980s to
clean-up produced water prior to overboard dis-
charge. As the volumes of produced water have in-
creased with Reld life, water-handling facilities have
become constrained. Operators have retroRtted pro-
duced water processing capacity using hydrocyclones
rather than Sotation machines, because the former
have a smaller footprint per volumetric Sow rate of
produced water treated. However, Sotation oil}water
separation technology has a place on offshore
platforms as a polishing stage for produced water
clean-up following initial treatment by hydrocyc-
lones. One design uses what is in principle a dissolved
air Sotation vessel downstream of the oil}water sep-
aration hydrocyclones.
Motion Insensitive Flotation
With the increased use of Soating production facili-
ties, the motion experienced by Sotation devices
1556 II / FLOTATION / Pre-aeration of Feed
inhibits separation. A 23 tilt results in one end of
a 12 m long Sotation machine being approximately
350 mm higher than the other end of the vessel. This
results in one end of the machine Sooding while the
other end will not remove Soated oil. To overcome
this problem, one manufacturer has designed a Sota-
tion column ISF machine that reduces the impact of
pitch and roll to 10 mm at 23 tilt.
The method of bubble generation (eduction) is the
same as that used for the four-cell ISF. Sparging was
investigated as an alternative, however the bubbles
were larger and moved in a linear fashion to the
surface. Educed air forms smaller bubbles and ex-
hibits random movement. Both these latter character-
istics are desirable as they increase the chance of an
air bubble}oil droplet collision.
Since the single-cell ISF has one rather than four
cells, theory would predict that its contaminant re-
moval efRciency would be reduced from approx-
imately 90}98% for a four-cell unit to 80}90% for
a single cell unit, primarily because the residence time
in the single cell unit is some 65% of that in the
traditional four cell design. In trials, however, the
unit has operated successfully with inlet oil concen-
trations of between 50 and 1300 ppm, and suspended
solids concentrations of 10}50 ppm, and has
achieved oil removal efRciencies comparable to
the four-cell model, taking into account the reduced
residence time in the single cell. In service, actual
efRciencies will depend on optimization of Sow
regime and chemicals (see Figure 6).
See Colour Plate 45.
See also: II/Flotation: Column Cells; Cyclones for
Oil/Water Separations; Historical Development.
Further Reading
Arnold K and Stewart M (1998) Surface Production Opera-
tions, 2nd edn, Vol. 1, pp. 218}223. Houston, TX: Gulf.
Pre-aeration of Feed
M. Xu, Inco Technical Services Limited, Mississauga,
Ontario, Canada
Z. Zhou and Z. Xu, University of Alberta,
Edmonton, Alberta, Canada
Copyright ^ 2000 Academic Press
Introduction
Aeration of slurry is a key element in a Sotation
system. The extent of aeration inSuences the perfor-
BerneH F and Cordonnier J (1995) Industrial Water Treat-
ment, pp. 80}89. Paris: Gulf.
Bradley BW (1987) Two OilTeld Water Systems, pp.
171}200. Malabar, FL: Robert E Krieger.
Degner VR (1975) Dispersed air Sotation. Cell design and
operation. Water, AIChE Symposium Series, Vol. 51,
No. 151, pp. 257}264.
Degner VR and Winter MK. Recent advances in waste-
water treatment using induced air Sotation. Baker Pro-
cess Internal Report F8-PR-1.
Eckenfelder WW (1989) Industrial Water Pollution Con-
trol, pp. 71}83. Singapore: McGraw-Hill.
Gordon RD (1995) ReRnery efSuent treatment. In:
Hull JB et al. (eds) Strategies for Monitoring, Control
and Management of Waste, pp. 59}65. London: Mech-
anical Engineering Publications.
Leech CA (1987) Oil Sotation processes for cleaning oilReld
produced water, pp. 1}43. Petroleum in the Ocean En-
vironment Conference, Oily Water Clean-up 1 Session.
American Institute of Chemical Engineers Meeting,
Houston, Texas.
Leech CA and Radhakrishnan S (1978) Performance evalu-
ation of induced gas Sotation (IGF) machine through
math modelling, pp 2513}2522. Tenth Annual Off-
shore Technology Conference, Houston, Texas.
Liebermann NP (1997) A Working Guide to Process
Equipment, pp. 913}939. New York: McGraw-Hill.
Schulz J (1993) Evolution of induced Sotation in oil}water
separation } an historical perspective. American Filtra-
tion Society Conference, Houston, Texas.
Stacy MO and Wolfenberger EE (1997) Development of
a single cell induced gas Sotation machine. Produced
Water Management Technical Forum & Exhibition
American Petroleum Institute TECHE Chapter,
Lafayette.
United States Environmental Protection Agency (1999)
Federal Register, Vol. 64(8), pp. 147. Document ID
FR13JA99-23.
Zabel ThF (1992) Flotation in water treatment. In: Mavros
P and Matis KA (eds) Innovations in Flotation Techno-
logy, pp. 431}454. Dordrecht: Kluwer.
mance of Sotation machines and the overall recovery
process. Flotation can, in general, be divided macro-
scopically into two subprocesses: selective collection
of hydrophobic particles by air bubbles, and separ-
ation of bubble/particle aggregates from the pulp
containing hydrophilic particles. The method and lo-
cation of aeration or bubble generation control the
mechanism of particle collection by either collision
with and subsequent attachment to bubbles, or in situ
bubble formation on hydrophobic particle surfaces.
A Sotation machine should be designed to provide an

optimal aeration condition for efRcient particle col-
lection and a suitable hydrodynamic environment for
effective transfer of bubble/particle aggregates from
the remaining pulp. Unfortunately, conSicting hy-
drodynamic environments are usually required for
these two sub-processes. It is often difRcult } if not
impossible } to evaluate theoretically the relative con-
tributions of individual collection mechanisms in
a particular Sotation device. The limited understand-
ing of the aeration mechanisms in Sotation processes
is partly responsible for the development of more
than 200 Sotation cell designs over the years. Many
of these designs are not subtle variations in basic
hardware, but variations in design principles. There-
fore, knowing where and how collection occurs and
which aeration method is suitable for a particular
application is an important step in a more scientiRc
approach to Sotation cell design.
Aeration methods used in Sotation can be conve-
niently categorized as air dispersion and air dissolu-
tion. In the air dispersion approach, a stream of air is
dispersed into slurry to achieve suitable sizes and
population of bubbles. This is accomplished by shear-
ing the air stream into bubbles under mechanical
agitation as in mechanical Sotation machines, or
using in-line static mixers as in Microcels and packing
materials as in packed columns. Air can also be dis-
persed through porous spargers, as used in pneumatic
Sotation machines or conventional Canadian Sota-
tion columns.
With the air dissolution method, on the other hand,
the air is dissolved under a pressure of 3}5 atm into
slurry for subsequent gas nucleation (or gas precipita-
tion) and cavitation. Bubble formation is then
achieved by either releasing gas-supersaturated slurry
to atmospheric pressure as in dissolved air Sotation,
or decreasing the pressure of slurry by aspiration as in
vacuum Sotation.
Dispersed air Sotation is widely used in minerals
processing with relatively coarse particles (larger than
20
m) and high slurry densities (greater than 30%
solids). Other areas of applications include solid
cleaning, de-inking from recycled paper and bitumen
recovery from oil sands. Dissolved air Sotation is
suitable for municipal water and industrial efSuent
treatment, due to its capability of generating relative-
ly Rne bubbles of less than 100
m required for re-
covering particles Rner than 10
m at a slurry density
of less than 0.5% solids.
An emerging trend is to integrate the useful features
of dissolved air Sotation into dispersed air Sotation.
The combination of the two bubble-generating mech-
anisms has led to a new Sotation cell design. Tradi-
tionally, slurry aeration and Sotation separation are
performed in the same vessel. Feed aeration followed
II / FLOTATION / Pre-aeration of Feed 1557
Figure 1 Schematic illustration of the concept of a flotation
system consisting of a reactor and a separator.
by Sotation separation in a separation vessel (a reac-
tor}separator design), has been evolved with demon-
strated higher Sotation kinetics. The concept of a So-
tation system consisting of a reactor and a separator
is illustrated in Figure 1. The reactor is a vigorous
bubble/particle contacting device where particle col-
lection takes place with bubbles formed by both air
dispersion and nucleation/cavitation mechanisms.
The separator is a quiescent bubble/pulp separation
device where the hydrodynamics favour the separ-
ation of bubble/particle aggregates from the pulp
with essentially no or little turbulence.
With continuing improved understanding of par-
ticle/bubble collection mechanisms and the role of
aeration in Sotation, it is anticipated that pre-aer-
ation of feed will become an important component in
modern Sotation circuits as a means of increasing
Sotation kinetics and improving selectivity of Rne
particles. This article focuses on the fundamentals
and recent developments in pre-aeration of feed used
in mineral Sotation.
Fundamental Basis of Feed Aeration
A theoretical analysis of aeration in a Sotation system
is complicated. As a result, the development of aer-
ation techniques in Sotation is largely based on phe-
nomenological correlations. The feed aeration and
subsequent particle collection during the aeration are
the combination of features of dispersed and dis-
solved air Sotation. Particle collection by air bubbles
in Sotation is a multi-step process, involving three
phases with interactions among solid/liquid, solid/gas
and liquid/gas in the presence of various inorganic
and organic species under economical and mechan-
ical constraints. At least two particle collection mech-
anisms have been considered in Sotation.
Direct Contact of a Particle with a Bubble
In this collection process, a particle encounters
a bubble, either by relative motion or the turbulence
1558 II / FLOTATION / Pre-aeration of Feed
in a Sotation system. The probability of the particle
being captured by the bubble (P) can be expressed as:
P"Pc Pa Pd [1]
where Pc , Pa and Pd are the probability of
bubble/particle collision, attachment and detach-
ment, respectively. For a hydrophobic particle,
bubble/particle collision determines the particle col-
lection rate governed by hydrodynamic conditions.
Hydrodynamic analysis showed that the collision
probability between a descending particle with an
ascending bubble is given by:
Pc"a(dp/db) n
[2]
where parameters a and n are a strong function of
hydrodynamic characteristics of the system. Eqn [2]
shows that the probability of particle}bubble colli-
sion is proportional to the nth (n51) power of the
solid particle size (dp) and inversely proportional to
the same power of the bubble size (db). For Rne
particles, small bubbles have to be used to obtain
sufRcient particle}bubble collisions. The direct con-
tact was analysed between a descending Rne particle
of dp"10
m and a rising swarm of bubbles in a So-
tation column. A collection zone as tall as 10 m was
determined to be essential to ensure at least one
collision of the particle with a bubble. This implies an
inefRcient collection process under conventional col-
umn Sotation conditions.
To increase the particle}bubble contact frequency,
a relatively high turbulence or energy dissipation rate
is required, as in mechanical Sotation machines. The
number of particle}bubble collisions per unit volume
and time in a highly turbulent Sowing Suid (Zpb) can
be expressed as:
Zpb"5NpNb[(dp#db)/2]2(V2p#V2b)0.5 [3]
where Np and Nb are the number concentrations of
particles and bubbles in the pulp. Vp and Vb are the
mean relative velocities of the particles and bubbles
(with reference to Suid), which are given collectively
by:
Vi"0.33
4/9di7/9(/)2/3/1/3 [4]
In eqn [4], subscript i refers to bubble or particle,
is
the speciRc energy dissipation rate,  is the differ-
ence in densities of particle i and liquid medium,  is
the medium density, and  is the kinematic viscosity.
This equation shows that a high energy dissipation
rate favours particle}bubble collision. However, vig-
orous agitation as in mechanical Sotation machines
may break particle-bubble aggregates, thereby in-
creasing the probability of particle detachment from
bubbles and hence decreasing the overall collection
rate. In addition, the back mixing (or liquid circula-
tion) caused by increased turbulence may hinder the
transport of bubble}particle aggregates out of the
turbulent zone, contributing to low Sotation kinetics.
The incentive to separate the two functions of a Sota-
tion machine, i.e. aeration and separation, is evident,
as reSected in the reactor}separator design.
In Situ Bubble Nucleation on Hydrophobic Particles
With this mechanism, gas nucleates and bubbles form
selectively on hydrophobic particles. The theoretical
basis of Sotation by gas precipitation or nucleation
was proposed in the 1960s and has recently been
extended to hydrodynamic cavitation. The gas nu-
cleation mechanism has been used to account for
particle}bubble collection in dissolved air Sotation.
An advantage of this mechanism over a conventional
particle}bubble contact mechanism is the elimination
of a collision stage, a rate-limiting step in Rne particle
Sotation. However, direct adoption of this technique
to mineral Sotation faces a number of challenges.
Clearly, tiny bubbles in the micro and submicron
range generated solely by a gas nucleation mechanism
are not sufRcient to Soat coarse mineral particles
effectively. However, the collision probability of lar-
ger bubbles with the particle}tiny bubble aggregates
in the quiescent region may increase. The limited
number of bubbles that can be generated by gas
nucleation from a supersaturated system does not
provide sufRcient carrying capacity to Soat large
amounts of solids. To improve solid recovery rates,
large volume slurry saturation tanks are needed, pre-
senting extra capital and operating costs.
Alternatively, tiny bubbles and cavities can be for-
med by the reduction of pressure in a fast-Sowing
Suid, as indicated by Bernoulli’s equation:
P1#(1/2)U21"P2#(1/2)U22"C (constant) [5]
in which U is the water Sow velocity at a point where
the pressure is P, and  is the density of liquid. If the
liquid Sow velocity exceeds a critical value, the pres-
sure in the liquid stream reduces to a value where the
liquid pressure falls below its vapour pressure, at
which point cavities form which expand to relieve the
differential pressure, a phenomenon called hydro-
dynamic cavitation. The presence of solids enhances
hydrodynamic cavitation due to the increased turbu-
lence and pressure Suctuations around particles in the
stream. As in gas-supersaturated systems, cavities
would form preferentially on hydrophobic particles

relative to energetically unfavourable hydrophilic
solid}liquid interfaces. The principal advantage of
exploring hydrodynamic cavitation in Sotation is that
gas supersaturation of slurry is not required and addi-
tional air can be introduced into the system for air
dispersion. As a result, hydrodynamic cavitation can
be readily implemented in mineral Sotation systems.
A convenient way of aiding bubble nucleation and
cavitation is by aeration in the feed slurry line. The
existence of gas nuclei in water has been demon-
strated in coagulation, sedimentation and Rltration
tests using Rne coal and silica with a medium particle
size of 5 and 1.5
m, respectively. The size of gas
nuclei in natural water was estimated to be 10
m.
When forcing the water through the tip of a cavita-
tion tube at a Sow velocity above 8}15 m s\1, micro-
size bubbles were observed to form. Numerical simu-
lation conRrmed that, at this Sow velocity, a pressure
close to liquid vapour pressure was attained inside the
tip of the cavitation tube, suggesting the formation of
bubbles by the expansion of the pre-existing gas
nuclei and subsequently Rlled with liquid vapours.
Using a light attenuation method, the onset velocity
of bubble formation by hydrodynamic cavitation was
found to be dependent on the diameter and length of
the nozzle, slurry temperature and initial gas content.
With gas-supersaturated water, for example, the
onset velocity reduced from 15 to 7 m s\1. Adding
frother into liquid does not affect the onset of bubble
formation by cavitation, but it increases the bubble
stability. Sebba has reported the formation of stable
bubble swarms of approximately 25}50
m which he
called aprons, generated similarly. Adding a small
amount of air into the Sowing liquid stream enhanced
Figure 2 Schematic of a Venturi aerated column (VAC).
II / FLOTATION / Pre-aeration of Feed 1559
bubble formation at a reduced liquid Sow velocity,
which provides a direct justiRcation for feed aeration
by hydrodynamic cavitation.
Applications of Feed Aeration
There are many new Sotation devices that make use
of the combined features of dispersed and dissolved
air Sotation. Feed slurry aeration through the concept
of reactor and separator design fully exploits the
combined mechanisms of bubble generation (dis-
persed/dissolved). Some of these Sotation devices are
reviewed here.
Venturi Aerated Column
The Venturi aerated column (VAC) was designed
based on the concept of reactor and separator design
(Figure 2). In this case, the reactor is a Venturi tube
where air/slurry contact takes place. The separator is
a column of length 1}2 m. The partial recirculation of
tailings slurry was used to intensify the slurry jetting
action, facilitating bubble size control (typically
500}800
m) and bubble/particle interaction.
Figure 3 presents a direct comparison between a la-
boratory Denver cell and a single VAC cell in batch
tests for a nickel sulRde ore. The VAC cell gave
a higher concentrate grade at the same nickel recov-
ery. Two VAC cells in series (as a rougher}scavenger
conRguration) were tested in an operating plant treat-
ing nickel sulRde ores. Compared to the mill rougher
Sotation, superior metallurgy was obtained with the
VAC cells. With two VAC cells in series, similar
metallurgical performance to the plant multi-stage
1560 II / FLOTATION / Pre-aeration of Feed
Figure 3 A direct comparison in nickel grade/recovery perfor-
mance between a batch Denver cell and a batch VAC cell.
Squares, column; diamonds, column (repeat); triangles, Denver;
circles, Denver (repeat).
Sotation circuit was achieved, as shown in Figure 4.
The VAC cell has also been tested and found to be
successful in de-inking applications of recycled paper
pulp with the performance exceeding the existing
plant circuit in terms of Rbre recovery at comparable
brightness.
Cavitation Tube
Based on the principle of nucleation, a cavitation tube
(Figure 5) similar to the Venturi tube was developed
and tested in association with a mechanical cell. As
shown in Figure 6 for Rne silica Sotation (less than
5
m), silica recovery of 30% was obtained in the
mechanical cell under conventional operating condi-
tions (no cavitation tube). By installing a cavitation
tube in the feed slurry line, silica recovery was in-
creased signiRcantly, depending on the slurry velocity
Figure 4 Comparison in nickel metallurgy between a plant
rougher flotation circuit and two VAC cells in series. Squares, two
VAC cells, circles, mill overall daily; triangles, mill roughers.
Figure 5 Schematic of a cavitation tube on the feed line before
a mechanical cell.
through the nozzle. The improved recovery is clearly
due to the bubble formation and particle collection by
gas nucleation/cavitation. Addition of a small
amount of air into the feed slurry (less than 7%)
before the cavitation tube further increased silica re-
covery, indicating that the combined mechanisms of
bubble formation by dispersed/dissolved air were
beneRcial to Rne particle Sotation. In-plant testing
using a similar set-up demonstrated an improved So-
tation performance of Cu/Ni separation.
Other Devices and Processes Using Feed Slurry
Aeration
Jameson cell The innovative design of the Jameson
cell is based on the point of air addition and bubble
generation. It utilized the concept of reactor and
separator design. The downcomer into which air is
aspirated and particle collection occurs is the reactor,
while the cylindrical tank is the separator. The feed
under high pressure is introduced at the top of the
downcomer through a nozzle, producing a high speed
slurry jet which entrains air into the downcomer. In
Figure 6 Fine silica recovery vs slurry velocity through the
nozzle of a cavitation tube. Conditions: 1 wt.% silica (!5
m); 10
p.p.m. Dowfroth 250; 1.25;10\4 mol L\1 DAH; pH"7.5}7.8; air
flow rate in the mechanical cell !2 L min\1.

addition to its successful use in sulRde minerals and
coal Sotation, the Jameson cell has also been adopted
in de-inking applications for the paper industry.
Davcra cell The Davcra cell is a type of pneumatic
machine, employing feed slurry aeration. Air and feed
slurry are injected into the separating tank through
a cyclone-type dispersion nozzle. Air dispersion and
particle collection take place in the nozzle and in the
highly turbulent region in the separation tank, which
is separated by the quiescent zone. A limited applica-
tion has been reported.
Low energy extraction process for bitumen extract-
ion from oil sands In bitumen extraction from
Athabasca oil sands, Syncrude Canada recently
adopted a low energy extraction process with hydro-
transport of oil sand slurries. Oil sand slurry is trans-
ported from the mine site via pipelines (3}5 km) with
a relatively high slurry velocity (4 m s\1). Air is injec-
ted into the pipelines. The aerated oil sand slurry is
then introduced into primary separation vessels
through a tangential entry feed well. Bitumen drop-
lets attached to air bubbles Soat to the top of the
remaining slurry to form a primary froth. Wash water
is added under the froth to reduce the amount of
solids reporting to the bitumen froth. Here, hydro-
transport pipelines function as a reactor to increase
the bubble}bitumen contact frequencies. This tech-
nique has been implemented in operation.
The Microcel, German Bahr’s cell, rapid Sotation
cell, Contact cell, CentriSoat cell, and air-sparged
hydrocyclone are other examples employing feed
slurry aeration and the concept of reactor and separ-
ator design.
Concluding Remarks
With increased understanding of the role of aeration
in Sotation, a new trend in Sotation machine design
has been established. The selection of aeration in
Sotation is closely related to particle collection mech-
anisms and, to some extent, to operation and main-
tenance costs. Aeration in Sotation has evolved from
air dispersion using direct particle}bubble contact
mechanism, to in situ bubble formation on hydropho-
bic particles by gas nucleation and cavitation in dis-
solved air Sotation. The emerging trend is a combina-
tion of the two. The location of aeration has also been
evolved from slurry aeration in a Sotation vessel to
feed aeration in a virtual reactor followed by Sotation
separation in a separation tank.
Flotation is an energy-dependent process and, like
all the subprocesses } solid suspension, aeration, par-
ticle}bubble interaction and bubble formation } are
II / FLOTATION / Pre-aeration of Feed 1561
energy-dependent. Part of the energy dissipation in
Sotation systems can be attributed to the method of
aeration. Ideally, as much of the input energy as
possible should be directed to the main function of
Sotation: particle collection by bubbles. Feed aer-
ation is associated with signiRcant energy dissipation
efRciency as the input energy is distributed evenly in
slurry in contrast to mechanical cells, in which the
energy is concentrated in the impeller region. The
energy used for pumping, on the other hand, is a
major energy requirement not needed in the existing
collection processes. With the concept of feed aer-
ation, pumping energy is utilized to force feed slurry
through a hydrodynamic cavitation tube (or a Ven-
turi tube), which facilitates the in situ generation of
bubbles on hydrophobic particles. It is an exciting
design challenge to develop energy-efRcient feed aer-
ation Sotation systems.
Further Reading
Amelunxen RL (1993) The contact cell } a future genera-
tion of Sotation machines. Engineering and Mining
Journal 194: 36}39.
Arbiter N (1984) The Sotation cell } a critique. In: Jones
MH and Woodock JT (eds) Principles of Mineral Flota-
tion, pp. 301}311. Victoria, Australia: The Australasian
Institute of Mining and Metallurgy.
Arbiter N (1989) Flotation machine dynamics. In: Chander
S and Klimpel RP (eds) Advances in Coal and Mineral
Processing Using Flotation, pp. 369}372. Colorado:
AIME-SME.
Bahr A (1985) Application and Sizing of a New Pneumatic
Flotation Cell, pp. 314}326. XV International Mineral
Processing Congress, Cannes.
Finch JA (1995) A selected review } part IV: novel Sotation
devices. Minerals Engineering 8(6): 587}602.
Flint LR (1973) Factors affecting the design of Sotation
equipment. Mineral and Science Engineering 5(3):
232}241.
Hu H, Zhou ZA, Xu Z and Finch JA (1998) Numerical and
experimental study of a cavitation tube. Metallurgical
and Materials Trans B 29B: 911}917.
Jameson GJ (1988) A new concept in Sotation column
design. In: Sastry KVS (ed.) Column Flotation ’88,
pp. 281}285. AIME.
Jordan CE and Susko FJ (1992) Rapid Sotation using
a modiRed bubble-injected hydrocyclone and a shallow-
depth separator for improved Sotation kinetics. Mineral
Engineering 5 (10}12): 1239}1257.
Mankowski P, Ng S, Siy R et al. (1999) Syncrude’s low
energy extraction process: commercial implementation,
pp. 154}181. In: Edwards C (ed.) Proceedings of 31st
Annual Meeting of CMP. Ottawa: CMP.
Miller JD, Ye Y, Pacquet E et al. (1988) Design and operat-
ing valuables in Sotation separation with the air-sparged
hydrocyclone, pp. 499}510. In: Forssberg KSE (ed.) XVI
IMPC. Stockholm, Sweden: Elsevier.
1562 II / FLOTATION / Reagent Adsorption on Phosphates
Rubinstein J (1995) Column Flotation: Processes, Designs
and Practices. New York: Gordon and Breach.
Schubert H and Bischofberger C (1979) On the optimiza-
tion of hydrodynamics in Sotation processes. In: Las-
kowski J (ed.) Proceedings of the 13th International
Mineral Process Congress, pp. 1261}1287. Warsaw:
Elsevier.
Wills BA (1992) Introduction to Mineral Processing Techno-
logy, 5th edn, pp. 558}575. Oxford: Pergamon Press.
Xu M, Quinn P and Stratton-Crawley R (1994) Graph-
ite/chalcopyrite separation using a rapid column cell.
In: Yalcin T (ed.) Innovations in Mineral Processing,
pp. 181}186. Sudbury, Ontario, Canada.
Xu M, Quinn P and Stratton-Crawley R (1996) A feed-line
aerated Sotation column. Minerals Engineering 8(10):
1159}1173.
Yang DC (1988) A new packed column Sotation system,
column Sotation ’88. In: Sastry KVS (ed.) Proceedings
of the International Symposium on Column Flotation,
pp. 257}265. Phoenix, USA: SME.
Yoon RH, Adel GT and Luttrell GH (1988) A process and
apparatus for separating Rne particles by microbubble
Sotation together with a process and apparatus for
generation of microbubbles. US patent no. 5761008.
Yoon RH and Luttrell GH (1989) The effect of bubble size
on Rne particle Sotation. Mineral Processing and Ex-
tractive Metallurgy Review 5: 101}122.
Reagent Adsorption on Phosphates
P. Somasundaran and L. Zhang, Columbia University,
NY, USA
Copyright ^ 2000 Academic Press
Introduction
Adsorption of surfactants on minerals is the basic
process governing Sotation. It is controlled by various
physicochemical processes in the pulp involving inter-
actions among the mineral particles, surfactants, dis-
solved inorganics, solvent species and other additives
such as polymers. Adsorption can be considered as
selective partitioning of the surfactant adsorbate into
the interfacial region, resulting from the more ener-
getically favourable interactions between the adsor-
bate and the solid than those between the former and
the species in the bulk solution. The interactions
leading to adsorption include chemical bonding, elec-
trostatic interaction, desolvation of the surfactant po-
lar group and the mineral surface species, hydrogen
bonding, van der Waals interactions, etc.
Water chemistry plays an important role in the
adsorption process by affecting the surfactant}
solution equilibria, the mineral}solution equilibria
Young FR (1989) Cavitation. London: McGraw-Hill.
Zhou ZA, Xu Z and Finch JA (1994) On the role of
cavitation in particle collection during Sotation } a
critical review. Minerals Engineering 7 (9): 1073}
1084.
Zhou ZA, Xu Z and Finch JA (1995) The minimum
recovery zone height in Sotation columns from par-
ticle}bubble collision analysis. Transactions of the Insti-
tution of Mining and Metallurgy 104: C102}C106.
Zhou ZA, Xu Z and Finch JA (1995) Fundamental study of
cavitation in Sotation. In XIX International Mineral
Processing Congress, vol. 3, pp. 93}97. San Francisco,
USA: SME.
Zhou ZA, Xu Z and Finch JA (1996) Effect of gas nuclei on
hydrophobic coagulation. Journal of Colloid Interface
Science 179: 311}314.
Zhou ZA, Xu Z, Finch JA and Liu Q (1966) Effect of
gas nuclei on the Rltration of Rne particles with
different surface properties. Colloids & Surfaces 113:
67}77.
Zhou ZA, Hu H, Xu Z et al. (1997) Role of hydrodynamic
cavitation in Rne particle Sotation. International Jour-
nal of Mineral Processing 51: 139}149.
Zhou ZA, Langlois R, Xu Z et al. (1997) In-plant testing of
a hydrodynamic reactor in Sotation. In: Finch JA,
Rao SR and Holubec I (eds) Processing of Complex
Ores, pp. 185}193. Sudbury, Canada: CIM.
and subsequently the interactions between the surfac-
tants and the mineral particles. The interactions in
mineral}solution system include dissociation, micell-
ization and precipitation of the surfactant, dissolu-
tion of a small amount of solids followed by hydroly-
sis, complexation and precipitation of the dissolved
species, and the interactions between dissolved min-
eral species with surfactant in the bulk in various
forms. The dissolved species, including those intro-
duced due to dissolution from all the minerals present
in the ore and those from the water source, fresh and
recycled, are the major elements that affect the
water chemistry. While impurities introduced from
water can be controlled to some extent, the chemical
species released into the system due to dissolution
from the minerals cannot be avoided. In systems
containing soluble or sparingly soluble minerals
where the extent of dissolution is markedly higher
than that in most oxide/silicate systems, the ef-
fect of dissolved mineral species can be drastic. Un-
derstanding the mineral}solution}surfactant chem-
ical equilibrium under different physicochemical
conditions is critical for developing reagent and pro-
cessing schemes for separation.
 II / FLOTATION / Reagent Adsorption on Phosphates
Phosphate is one of the most important minerals
processed by Sotation techniques. Flotation is ef-
Rcient for the beneRciation of phosphate ores with
silicate gangues, but those with carbonaceous gan-
gues are difRcult to separate by Sotation tech-
niques. The low selectivity has been attributed to the
similarities in the surface chemical properties of the
constituent minerals. These properties, in turn, are
inSuenced by the water chemistry of the surfactant}
mineral systems. In this section the effects of
water chemistry on the surfactant}solution equilib-
rium, the mineral}solution equilibrium, the surfac-
tant}mineral interactions in the separation of phos-
phate and associated minerals are discussed. Methods
to manipulate and control the solution chemistry to
achieve selectivity in Sotation are also examined.
Water Chemistry of Flotation
Reagents
Long chain fatty acids such as oleic acid are among
the commonly used reagents for the Sotation of ox-
ides, silicates and salt-type minerals. Flotation of
these minerals using fatty acids is affected greatly
by solution properties such as pH, since weakly acidic
fatty acids undergo association interactions that can
inSuence their adsorption and Sotation properties.
For example, oleic acid species will undergo dissocia-
Figure 1 pH of oleic acid precipitation. (From Morgan LJ, Anan-
thapadmanabhan KP and Somasundaran P (1986) Oleate ad-
sorption on hematite: problem and methods. International Journal
of Mineral Processing 18: 39. Copyright: Elsevier Science.)
1563
tion to form ions (Ol\) at high pH values and exist as
neutral molecules (HOI) at low pH value. In the
intermediate region, the ionic and the neutral molecu-
lar species can associate to form ion}molecule com-
plexes ((Ol)2H\). As the surfactant concentration
is increased, micellization or precipitation of the
surfactant can occur in the solution. In addition,
surfactant species can associate to form other
aggregates such as the dimer (Ol22\) in premicellar
solutions. Also, long chain fatty acids such as oleic
acid have very limited solubility, which is a sensitive
function of pH. The pH of precipitation of oleic acid
calculated as a function of total oleate is shown in
Figure 1.
The solution equilibria of oleic acid (HOl) are
expressed as below:
HOl(liquid)"HOl(aq)
pKsol"7.6 (Ksol: solubility product)
HOl(aq)"H##Ol\
pKa"4.95 (Ka: acid dissociation constant)
2Ol\"(Ol)22\
pKd"!3.7 (Kd: dimerization constant)
HOl#Ol\"(Ol)2H\
pKad"!5.25 (Kad: acid}soap formation constant)
The species distribution of oleic acid as a function of
pH based on the above equilibria at a given concen-
tration is shown in Figure 2. It can be seen from this
Rgure that:
1. The pH of the precipitation of oleic acid at the
given concentration is 7.45.
2. The activities of oleic monomer and dimer remain
almost constant above the precipitation pH and
decrease sharply below it.
3. The activity of the acid}soap (Ol)2H\ exhibits
a maximum in the neutral pH range.
The surface activities of the various surfactant spe-
cies can be markedly different from each other. It
has been estimated that the surface activity of the
acid}soap (Ol)2H\ is Rve orders of magnitude higher
than that of the neutral molecule (HOl) and about
seven orders of magnitude higher than that of the
neutral molecule (HOl) and about seven orders of
magnitude higher than that of the oleate monomer
Ol\.
The existence of salt will also affect the surfac-
tant}solution equilibria by changing the surface
activities of the various surfactant species, the critical
micelle concentration and the solubility of the surfac-
tant, and the solvent properties of the solution.
1564 II / FLOTATION / Reagent Adsorption on Phosphates

trol the dissolution of calcite and apatite in water are

Figure 2 Oleate species distribution as a function of pH. Total
oleate concentration"3;10\5 mol L\1. (From Ananthapad-
manabhan KP and Somasundaran P (1980) Oleate chemistry
and hematite flotation. In: Yarar B and Spottiswood DJ (eds)
Interfacial Phenomena in Mineral Processing, p. 207. New York:
Engineering Foundation.)
It is clear that, to understand the adsorption of
reagents on solids, the effects of concentration,
pH, ionic strength and activities of the various poss-
ible reagent species on the adsorption process need to
be taken into account.
The Effect of Water Chemistry
on Mineral^Solution Equilibrium
When mineral particles are in contact with water,
they undergo dissolution, the extent of which is de-
pendent on the type and concentration of chemicals
in solution. The dissolved mineral species can under-
go further reactions such as hydrolysis, complexa-
tion, adsorption and precipitation. The complex
equilibria involving all such reactions can be expected
to determine the interfacial properties of the minerals
and their Sotation behaviour. The equilibria that con-
given in Table 1.
In the case of carbonaceous phosphate minerals,
apatite, calcite and dolomite will dissolve in water,
followed by pH-dependent hydrolysis and complexa-
tion of the dissolved species. Since these minerals are
sparing soluble, the dissolved species have a marked
effect on their interfacial properties.
It should be noted that, from theoretical consider-
ations, depending on the solution conditions, the sur-
face of apatite can be converted to calcite and vice
versa through surface reactions or bulk precipitation
of the more stable phase. The stoichiometry of the
equilibrium governing the conversion of apatite to
calcite can be written as:
Ca10(PO4)6(OH)2(S)#10CO23\
"10CaCO3(S)#6PO24\#2OH\
It can be seen from this equation that, depending
on the pH of the solution, apatite can be converted to
calcite if the total carbonate in solution exceeds a cer-
tain value. In fact, the amount of dissolved carbonate
from atmospheric CO2 does exceed that required to
convert apatite to calcite under high pH conditions.
Surface conversion due to the reaction of the dis-
solved species with the mineral surface can be pre-
dicted using stability diagrams for heterogeneous
mineral systems. This is illustrated in Figure 3 for the
calcite}apatite system. The activity of Ca2# in equi-
librium with various solid phases shows that the point
of interception for calcite and apatite is pH 9.3.
Above this pH, apatite is less stable than calcite and
hence conversion of apatite to that of calcite can be
expected in the calcite}apatite system. Similarly,
apatite is more stable than calcite below pH 9.3. It is
to be noted that Ca2# in equilibrium with calcite in
an open system (open to atmospheric CO2) is signiR-
cantly different from that in a closed system. Also,

Table 1 Equilibria controlling the dissolution of calcite and apatite in water

Ksp Ksp

Calcite
CaCO3 (S) 8 Ca2##CO23\ 10\8.4 Ca2##HCO\ 3 8 CaHCO#3 100.8
CO23\#H# 8 HCO\3 1010.3 Ca2##CO23\ 8 CaCO3(aq) 103.3
#
HCO\ 3 #H 8 H2CO3 106.3 Ca2##H2O 8 CaOH##H# 10\12.9
CO2(g)#H2O 8 H2CO3 10\1.5 Ca2##2H2O 8 Ca(OH)2#2H# 10\22.8
Apatite
Ca10(PO4)6(F,OH)2(S) 8 10 Ca2##6 PO34\#2 (F, OH)\ 10\118
PO34\#H# 8 HPO24\ 1012.3 Ca2##HPO24\ 8 CaHPO4(aq) 102.7
HPO24\#H# 8 H2PO\4 107.2 CaHPO4(aq) 8 CaHPO4(s) 104.3
#
HPO\ 4 #H 8 H3PO4 102.2 Ca2##H2PO\ 4 8 CaH2PO# 4 101.1
Ca2##H2O 8 CaOH##H# 10\12.9 Ca2##2F\ 8 CaF2(s) 1010.4
Ca2##2H2O 8 Ca(OH)2#2H# 10\22.8 Ca2##F\ 8 CaF# 101.0
F\#H# 8 HF 103.1
 II / FLOTATION / Reagent Adsorption on Phosphates
Figure 3 pH dependence of activity of Ca2# in equilibrium with
calcium oleate (dotted line: OlT"10\4 kmol m\3), calcite (open
(closed lines) and closed (dots and dashes) systems) and apatite
(dashed lines). (From Ananthapadmanabhan KP and Somasun-
daran P (1984) The role of dissolved mineral species in calcite-
apatite flotation. Mineral and Metallurgical Processing 1: 36.)
in the absence of atmospheric CO2, apatite has a wider
stability region than in the open system. Atmospheric
CO2 can thus be expected to play an important role in
these types of mineral}solution equilibria and in op-
erations dependent on interfacial properties.
Figure 4 Illustration of the effect of supernatants on the zeta
potential and isoelectric point of calcite and apatite:
2;10\3 kmol m\3 KNO3. Open circles, calcite in water; open
triangles, apatite in water; filled triangles, apatite in calcite super-
natant; filled circles, calcite in apatite supernatant. (From Anan-
thapadmanabhan KP and Somasundaran P (1984) The role of
dissolved mineral species in calcite-apatite flotation. Mineral and
Metallurgical Processing 1: 36.)
1565
The surface conversions in the calcite}apatite system
have been proved experimentally; electrokinetic data
obtained for the calcite}apatite system in water and in
the supernatant of each other are shown in Figure 4.
When apatite is in contact with calcite supernatant, its
zeta potential is seen to shift to that of calcite and vice
versa, suggesting surface conversion of apatite to cal-
cite and calcite to apatite, respectively.
The zeta potential data obtained in mixed super-
natants of calcite and apatite also show the effect
of dissolved mineral species. If supernatants of calcite
and apatite are combined as a 1 : 1 mixture, the two
minerals have almost identical surface charge charac-
teristics in the basic pH range (Figure 5).
The surface conversion of apatite and calcite is fur-
ther supported by the result of electron spectroscopy
for chemical analysis (ESCA) measurements. The re-
sults in Figure 6 show that, when apatite is condi-
tioned in the supernatant of calcite at pH&12, its
surface exhibits spectroscopic properties characteristic
of both calcite and apatite. This behaviour is attributed
to the precipitation of calcite on the apatite.
Dissolution equilibria of sparingly soluble min-
erals play a major role in determining the surface
properties of these minerals and in turn, adsorption of
reagents on them.
Figure 5 Illustration of the similarity in zeta potentials of calcite
(circles) and apatite (triangles) in mixed supernatants. (From
Ananthapadmanabhan KP and Somasundaran P (1984) The role
of dissolved mineral species in calcite-apatite flotation. Mineral
and Metallurgical Processing 1: 36.)
1566 II / FLOTATION / Reagent Adsorption on Phosphates

in equilibrium with Ca2#-oleate, Ca2#-oleate can be

expected to precipitate.
Depletion isotherms of oleic acid on both francolite
and dolomite has been observed to be a two-region
linear isotherm with a change of slope at about
10\4 kmol m\3 (Figure 7). Simultaneous analysis
of the dissolved mineral species in the supernatants
of the samples used in the adsorption experiments
(Figure 8) shows a sharp decrease in the concentrations
of both Mg and Ca species when oleate concentration
exceeds 1.0;10\5 kmol m\3 in the case of francolite
and 3.0;10\5 kmol m\3 in the case of dolomite.
This suggests that bulk precipitation of calcium and
magnesium species can occur under such conditions.
Major chemical equilibria for the precipitation
of Ca and Mg species by oleate can be given as
follows:

Ca2##2 Ol\"Ca(Ol)2 KCa(Ol)2"3.81;10\13

Mg2##2 Ol\"Mg(Ol)2 KMg(Ol)2"1.58;10\11

The onset of the precipitation of Ca(Ol)2 and Mg(Ol)2

is calculated from the solubility products given above
and marked in Figure 8. The calculated oleate con-
centrations at the onset of precipitation are in good
agreement with experimental observations.
It is postulated that, in the case of oleate adsorption
on dolomite and francolite, different mecha-
nisms govern the adsorption process. In the low con-
Figure 6 ESCA spectra of C(1s) peak of apatite conditioned in
centration range ((10\4 kmol m\3), the adsorption
calcite supernatant at pH&12. (A) Apatite in water; (B) calcite in
water; (C) apatite in calcite supernatant. (From Ananthapad-
manabhan KP and Somasundaran P (1984) The role of dissolved
mineral species in calcite-apatite flotation. Mineral and Metallurgi-
cal Processing 1: 36.)

The Effect of Water Chemistry on

Adsorption of Reagents on Minerals
Chemical equilibria in aqueous solutions containing
both the minerals and the surfactants can be expected
to be much more complex than in either of the indi-
vidual systems discussed above. In addition to surfac-
tant adsorption at the solid}liquid interface, interac-
tions between dissolved mineral species with various
surfactant species can be expected. All these interac-
tions can affect the surfactant adsorption and the
subsequent Sotation.
As indicated earlier, oleic acid has a very low solu-
bility and adsorption of oleate, in some cases, is in Figure 7 Depletion isotherms of 14C-labelled oleic acid on fran-
colite (squares: pH"8.2) and dolomite (circles: pH"9.2). Tem-
fact precipitation of the surfactant in the interfacial
perature, 253C; S/L"0.3; I"3;10\2 kmol m\3 KNO3. (From
region. In Figure 3, the activity of Ca2# in equilib- Somasundaran P, Xiao L and Wang D (1991) Solution chemistry
rium with various solid phases is plotted. If, at any of flotation of sparingly soluble minerals. Mineral and Metallurgical
stage, activity of Ca2# in solution is greater than that Processing 8: 115}121.)
 II / FLOTATION / Reagent Adsorption on Phosphates
Figure 8 Dissolved Ca (squares) and Mg (circles) levels from (A) francolite (pH"8.2) and (B) dolomite (pH"9.2) suspensions as
a function of oleate concentration. (From Somasundaran P, Xiao L and Wang D (1991) Solution chemistry of flotation of sparingly
soluble minerals. Mineral and Metallurgical Processing 8: 115.)
of oleate on both minerals occurs mainly due to
chemical bonding on surfaces without any precipita-
tion. At an intermediate concentration of about
10\4 kmol m\3, the solubility limit of Ca and Mg
oleate can be reached in the interfacial region but not
in the bulk solution, suggesting surface precipitation
of oleate on both minerals. In the high concentration
range ('5 ; 10\4 kmol m\3), oleate depletion may
be dominated in the case of both minerals by the
precipitation of Ca and Mg species with oleate, on the
mineral surface and in the bulk solution.
From the above discussion on apatite}calcite con-
version, it is clear that a Sotation separation scheme
designed on the basis of the surface properties of
a single mineral is not likely to perform satisfactorily.
The effect of dissolved species of calcite and
apatite on fatty acid Sotation of both minerals has in
fact been studied using mineral supernatant solutions
containing various dissolved species. The Sotation
results are shown in Figure 9. Both supernatants of
calcite and apatite are found to depress the calcite
Sotation by oleic acid in the tested pH range, with
apatite supernatant exhibiting a greater depressing
effect. Similar results have also been obtained for
apatite Sotation. The supernatants of calcite and apa-
tite depress the apatite Sotation under all tested pH
conditions.
Studies on the dissolved species responsible for the
observed effect revealed that, for calcite Sota-
tion, the depression role of apatite supernatant results
from the combined effects of calcium species and
the phosphate species in solution, while the depress-
ion role of calcite supernatant is mostly that of the
calcium ion and possibly some carbonate ions. The
depression due to calcium ion is caused by the de-
1567
pletion of oleate owing to the precipitation of calcium
oleate. In the case of apatite Sotation, the depression
is due to phosphate and carbonate species in solution.
The adsorption of these ions on the surface calcium
sites reduces the sites available for oleate adsorption
which, in turn, lowers the hydrophobicity of the sur-
face and so depresses the apatite Sotation. Calcium
oleate precipitation, in this case, does not occur to
a signiRcant extent due to the low concentration of
oleic acid used in Sotation. The above observations
clearly show that water chemistry plays a crucial role
in the Sotation of apatite}calcite systems.
In addition to reagent complexation and precipita-
tion, other reactions that occur in the bulk solution
can take place in the interfacial region. For example,
hemimicellization at a solid}liquid interface is a
phenomenon that drastically affects the adsorp-
tion of collector reagents on solids.
Flotation is a dynamic process. In addition to
the equilibrium effects associated with the water
chemistry, it can also inSuence the adsorption kinet-
ics of surfactants on the solid surfaces. Anionic condi-
tioning is a unit operation that precedes rougher
Sotation and skin Sotation of phosphates in Florida
Sotation plants. The effect of water chemistry on
oleic acid adsorption on francolite during anionic
conditioning has recently been studied in detail. In
order to identify the effect of process variables
on the adsorption, the experiment was carried out
under both laboratory and plant conditions (Table 2).
The kinetics of oleic acid adsorption on francolite
under both laboratory and plant conditions, using
distilled water and plant water, is shown in Fig-
ure 10. The adsorption density and kinetics are quite
different depending on the conditions and the
1568 II / FLOTATION / Reagent Adsorption on Phosphates

Figure 9 (A) Effect of apatite supernatant (squares) on calcite flotation. K oleate 10\4 kmol m\3; I"3;10\2 kmol m\3 KNO3.
Circles, water. (B) Effect of calcite supernatant (squares) on calcite flotation. K oleate 10\4 kmol m\3; I"3;10\2 kmol m\3 KNO3.
Circles, water. (C) Effect of calcite supernatant (squares) on apatite flotation. K oleate"2;10\6 kmol m\3; I"3;10\2 kmol m\3
KNO3. Circles, water. (D) Effect of apatite supernatant (squares) on apatite flotation. K oleate"2;10\6 kmol m\3; I"3;
10\2 kmol m\3 KNO3. Circles, water. (From Ananthapadmanabhan KP and Somasundaran P (1984) The role of dissolved mineral
species in calcite-apatite flotation. Mineral and Metallurgical Processing 1: 36.)

water. Under laboratory conditions, the adsorption in
plant water is signiRcantly lower than that in the
distilled water. It is proposed that this is due to
reagent loss resulting from the dissolved species in
plant water precipitating the oleic acid. In contrast,
under plant conditions, the adsorption behaviour of
oleic acid in plant water and distilled water is similar
and adsorption densities are lower than those under
laboratory conditions. The high solid/liquid ratio un-
der plant conditions will reduce the adsorption den-
sity on the solids because of the much greater solid
surface on to which the reduced total amount of

Table 2 Comparison of laboratory and plant conditions

Laboratory conditions Plant conditions

Conditioner Wrist-action shaker Lightnin Labmaster L1U08, four-bladed

cruciform propeller operating at 350 rpm
pH 9.1}9.5 9.1}9.5
Water Distilled and plant water Plant water
Solid (%) 10 (2 g sample) 72 (1000 g sample)
Time (min) 120 (except for kinetics) 3 (except for kinetics)
 II / FLOTATION / Reagent Adsorption on Phosphates 1569

Figure 10 Kinetics of oleic acid adsorption on francolite in distilled water and plant water under laboratory and plant conditions. Open
squares, distilled water in laboratory conditions; filled squares, plant water in laboratory conditions; open circles, distilled water in plant
conditions; filled circles, plant water in plant conditions. Oleic acid concentration: 8.1;10\3 mol L\1; pH 9.1}9.5. (From Maltesh C,
Somasundaran P and Gruber GA (1996) Fundamentals of oleic acid adsorption on phosphate flotation feed during anionic conditioning.
Mineral and Metallurgical Processing 13: 157.)

reagent in the water adsorbs. This will also result in
a lower reagent concentration in solution reducing
the precipitation eject. The intense agitation in the
plant conditioner may also remove some of the bound
reagent from the surface.
The adsorption isotherms of oleic acid on fran-
colite under laboratory and plant conditions are
compared in Figure 11. Adsorption is markedly
higher under laboratory conditions than under plant
conditions. On the other hand, under plant condi-
tions the adsorption is similar in distilled water and
plant water. This suggests that the effect of dis-
solved species is reduced under plant conditions.
From the above discussion, it can be seen that the
adsorption of surfactant on a mineral is a compli-
cated process involving interactions such as surfac-
tant self-association, mineral dissolution, bulk
precipitation, adsorption and surface precipitation.
The interactions are further complicated by the kin-
etic effects of the various reactions.
Understanding the effect of the water chem-
istry on reagent adsorption offers opportunities
to manipulate such processes by optimizing the con-
tributing factors such as alteration of the surface
properties, complexation of ions which cause precipi-
tation of the surfactant, prevention or enhancement
of collector adsorption and changes in the adsorption
kinetics to achieve the desired selectivity in Sotation.
In the anionic Sotation of phosphate, Ca2#
affects the grade of phosphate by activating the
quartz through formation of calcium-bearing precipi-
tates at high pH. This detrimental effect can be
prevented by adding sodium silicate, which can inter-
act with Ca2# and form calcium silicate. Since cal-
cium silicate and quartz are negatively charged, de-
tachment of calcium silicate from quartz can occur
and thus quartz Sotation can be depressed.
It has been found that in carbonate/phosphate sys-
tems, with fatty acid as collector, apatite is depressed
in the acid medium (pH 5.5}6.0) while carbonate is
Soated. The depression of phosphate at this pH is
possibly due to the adsorption (or formation) of aque-
ous CaHPO4 on its surface, preventing surfactant
ions from approaching the surface of the phosphate
particles. Free Ca2# in solution can affect the
formation of aqueous CaHPO4. From thermody-
namic considerations it can be predicted that the
selective Sotation of carbonates from phosphates in
acid media can be enhanced by minimizing free Ca2#
in solution and by increasing HPO24\ in the system.
This can be done by (1) decreasing free Ca2# concen-
tration in the system to low values by adding suitable
chemical reagents such as sulfuric acid or chelating
agents such as oxalic acid, and (2) adding soluble
phosphate salts to enhance the depression of the
phosphate minerals. Results from experiments with
1570 II / FLOTATION / Reagent Adsorption on Phosphates

Figure 11 Adsorption isotherms of oleic acid adsorption on francolite in distilled water and plant water under laboratory and plant
conditions. Squares, distilled water in laboratory conditions; open circles, distilled water in plant conditions; filled circles, plant water in
plant conditions. (From Maltesh C, Somasundaran P and Gruber GA (1996) Fundamentals of oleic acid adsorption on phosphate
flotation feed during anionic conditioning. Mineral and Metallurgical Processing 13: 157.)

natural phosphate ores are in agreement with the
theoretical predictions.
Based on the oleic acid solution chemistry, a two-
stage conditioning process for the Sotation of dol-
omite from apatite has been proposed. The mixed
minerals are Rrst conditioned at pH 10 with oleic acid
collector. The system is then reconditioned below pH
4.5 where dolomite is Soated. The selectivity of dol-
omite from apatite is attributed to two factors in this
process.
1. High adsorption of oleate on dolomite during the
Rrst stage at pH 10, which is maintained after
reconditioning at lower pH.
2. Oleate to oleic acid transformation upon recon-
ditioning, reducing its efRciency, and this re-
duction being more severe for apatite than for
dolomite.
In the high pH range, oleate adsorbs on to apatite
and calcite through speciRc interactions, while at
low pH, when oleic acid is the major species, the
adsorption is through weaker physical interaction.
Thus, oleic acid is a poor collector compared to
oleate.
ModiRcation of collector adsorption on minerals
can be used to control their Sotation response. In one
study, Alizarin Red S, a dye that stains calcite,
was tested as a modifying agent in calcite}apatite
system due to its preferential adsorption on these
minerals. Even though Alizarin Red S adsorbs
more on apatite than on calcite, it depresses the Sota-
tion of apatite using oleate as collector more than that
of calcite (Figure 12). In the absence of the dye, both
calcite and apatite Soat with oleate at pH 10.5. When
the dye concentration increases to 5;10\6 mol m\3,
the Sotation of calcite is very little affected with
a recovery of about 90%, while apatite Sotation is
depressed to 5}10%. Calcite Sotation is only
affected at higher concentrations of dye. Alizarin
Red S or its derivatives are hence promising reagents
for the beneRciation of phosphate with carbonaceous
gangues.
Summary
Mineral}solution equilibria, surfactant}solution
chemistry, as well as interactions among dissolved
species, surfactant and solids, can have a drastic ef-
fect on surfactant adsorption and Sotation separation
of sparing soluble minerals. Studies on the effects
of water chemistry on adsorption of surfactant on
phosphate minerals such as apatite and francolite and
associated minerals such as calcite and dolomite show
that these interactions have marked effects on
the reagent adsorption as well as Sotation. Surfactant
can exist in different forms in solution depending
on the solution pH and the surfactant concentration.
Minerals can undergo dissolution, with the extent of
 II / FLOTATION / Reagent Adsorption on Phosphates 1571

Figure 12 Flotation of calcite (triangles) and apatite (circles) from their mixture (1 : 1) at pH 10.5 as a function of Alizarin red
S concentration. Alizarin red S conditioning time"1 min; K oleate"9;10\5 kmol m\3; KCl"3;10\2 kmol m\3; pH"10.5$0.2.
(From Fu E and Somasundaran P (1986) International Journal of Mineral Processing 18: 287, with permission from Elsevier Science.)

dissolution depending upon solution conditions such Further Reading

as pH, ionic strength and concentration of constitu-
ent ions. The dissolved mineral species can further
interact with mineral solids, leading to surface con-
version of the minerals. They can also interact with
surfactant, leading to surface and bulk precipitation.
All these processes can signiRcantly affect the
adsorption of surfactant on minerals. A full under-
standing of the various interactions in surfactant}
solid}solution system is essential for developing
efRcient separation schemes. Indeed, desired se-
lectivity can be achieved by using appropriate addi-
tives to control dissolved species or modifying
collector adsorption and by optimizing solution con-
ditions as well as the kinetics involved.
Acknowledgement
The authors acknowledge Rnancial support of
the National Science Foundation (CTS-9622781 and
EEC-94-02989)
Amankonah JO, Somasundaran P and Ananthapadma-
nabhan KP (1985) Effects of dissolved mineral spe-
cies on the dissolution/precipitation characteristics of
calcite and apatite. Colloids and Surfaces 15: 295.
Amankonah JO, Somasundaran P and Ananthapadma-
nabhan KP (1985) Effects of dissolved mineral
species on the electrokinetic behavior of calcite and
apatite. Colloids and Surfaces 15: 335.
Ananthapadmanabhan KP and Somasundaran P (1980)
Oleate chemistry and hematite Sotation. In: Yarar B
and Spottiswood DJ (eds) Interfacial Phenomena in
Mineral Processing, p. 207, New York: Engineering
Foundation.
Ananthapadmanabhan KP and Somasundaran P (1984)
The role of dissolved mineral species in calcite-
apatite Sotation. Mineral and Metallurgical Processing
1: 36.
Ananthapadmanabhan KP and Somasundaran P (1985)
Surface precipitation of inorganics and surfactants
and its role in adsorption and Sotation. Colloids and
Surfaces 13: 151.
1572 II / ION EXCHANGE / Catalysis: Organic Ion Exchangers
Ananthapadmanabhan KP and Somasundaran P (1988)
Acid}soap formation in aqueous oleate solutions. Jour-
nal of Colloid Interface Science 122: 104.
Dho H and Iwasaki I (1990) Role of sodium silicate in
phosphate Sotation. Mineral and Metallurgical Process-
ing 7: 215.
Elgillani DA and Abouzeid A-ZM (1993) Flotation of car-
bonates from phosphate ores in acidic media. Interna-
tional Journal of Mineral Processing 38: 235.
Fu E, Somasundaran P (1986) Alizarin red S as a Sotation
modifying agent in calcite-apatite systems. International
Journal of Mineral Processing 18: 287.
Leja J (1982) Surface Chemistry of Froth Flotation. New
York: Plenum Press.
Maltesh C, Somasundaran P and Gruber GA (1996) Funda-
mentals of oleic acid adsorption on phosphate Sotation
feed during anionic conditioning. Mineral and Metallur-
gical Processing 13: 157.
Morgan LJ, Ananthapadmanabhan KP and Somasundaran
P (1986) Oleate adsorption on hematite: problem and
ION EXCHANGE
Catalysis: Organic Ion
Exchangers
R. L. Albright, Albright Consulting,
Southampton, PA, USA
Copyright ^ 2000 Academic Press
Introduction
The ion exchange polymers most often used in cataly-
sis are insoluble materials that can be constructed
from inorganic or organic monomer units. This
article will present only catalysis performed by the
organic ion exchangers that are insoluble solids.
There are commercial ion exchangers that are liquids,
but to date they have been used very little in
catalysis and, therefore, will not be included in this
discussion.
Insoluble ion exchangers carry out their catalytic
work in a heterogeneous rather than a homogeneous
fashion and are, therefore, part of the group called
heterogeneous catalysts. Heterogeneous catalysts
have three very signiRcant advantages over homo-
geneous catalysts: Rrst, they are not corrosive; sec-
ond, they are very readily separated from the reaction
mixture by a simple Rltration; and third, they are
methods. International Journal of Mineral Processing
18: 139.
Moudgil BM and Chanchani R (1985) Selective Sotation of
dolomite from francolite using two-stage conditioning.
Mineral and Metallurgical Processing 2: 19}25.
Somasundaran P (1969) Adsorption of starch and
oleate and interaction between them on calcite in
aqueous solutions. Journal of Colloid Interface Science
31: 557.
Somasundaran P and Ananthapadmanabhan KP (1986)
Solution chemistry of Sotation. In: Somasundaran P.
(Ed.), Advances in Mineral Processing, p. 426. New
York: AIME.
Somasundaran P, Amankonah JO and Ananthapadma-
nabhan KP (1985) Mineral}solution equilibria in spar-
ingly soluble mineral systems. Colloids and Surfaces 15:
309.
Somasundaran P, Xiao L and Wang D (1991) Solution
chemistry of Sotation of sparingly soluble minerals.
Mineral and Metallurgical Processing 8: 115}121.
most often recyclable from one run to the next with-
out any added treatment. Potentially, they suffer
one major disadvantage over homogeneous catalysts.
Intimate contact between reactants and the catalytic
site is not achieved simply by mixing the heterogen-
eous catalyst with the reactants. In a stirred reactor,
intimate contact between reactants and the homo-
geneous catalyst is very easily achieved and mass
transport of reactants to catalyst is very rapid and
almost never rate-limiting. With heterogeneous cata-
lysis, mass transport of reactants to the catalytic site
may often be the rate-limiting element, especially if
the activation energy for the reaction is small and the
chemical reaction is rapid. There are excellent texts
and monographs on the issues surrounding hetero-
geneous catalysis, and the reader is referred to these
for the development of a fuller understanding
(see Further Reading).
Nature of Organic Ion Exchange
Polymers
Chemical Composition
Organic ion exchangers are made by polymerization
of organic monomers into large molecules which are
made insoluble by crosslinking with a polyfunctional
monomer. The nature and the level (concentration) of
 II / ION EXCHANGE / Catalysis: Organic Ion Exchangers
the crosslinking inSuences the elasticity, the dimen-
sional stability or strength of the copolymer particle,
and the available space surrounding the ionogenic or
catalytic site within the solvent-swelled gel phase.
Many monomers have been transformed into insol-
uble ion exchangers by various polyfunctional mono-
mers, but an unabridged listing of these will not be
given here. Instead, this discussion will consider only
those monomer systems most used.
The most prominent insoluble copolymer matrices
for constructing organic ion exchangers are those
derived by the free radical copolymerization of
styrene with divinylbenzene. Other matrices that
have been used and are presently used to a much
lesser extent are those made by the condensation
polymerization of phenol (including the other hy-
droxylated aromatic derivatives of phenol such as
catechol, resorcinol, hydroquinone, etc.) with formal-
dehyde and the copolymer matrices made by the free
radical polymerization of the acrylate and meth-
acrylate monomers with divinylbenzene.
The functional groups that perform the catalytic
work are attached to the preformed crosslinked poly-
meric matrix. A vast array of chemistry allows the
attachment of many different functional groups
for anchoring the catalytic agent. With the aromatic
polymers, electrophilic substitution reactions provide
the means of functional group attachment, and with
the aliphatic acrylic and methacrylic resins, the car-
boxyl group provides the means of functional group
attachment by nucleophilic substitution reactions.
A plethora of chemistry is available to build a hetero-
geneous catalyst upon the polymeric matrices em-
ployed to make ion exchangers.
Much of the chemistry for designing effective
catalysts built upon crosslinked polymers, however,
has not been pursued to a fruitful outcome. Many of
the special heterogeneous catalysts have been built
upon crosslinked polymers with poor mass transport
in the solvent systems necessary for effective
chemical transformations and, therefore, have had
inferior performance to the corresponding homo-
geneous catalysts. This inferior performance resulting
from poor mass transport has partially quenched the
commercial development of what could be excellent
heterogeneous catalysts when built upon the properly
designed structures of the crosslinked polymers. As
a result, most of the commercial effort to use ion
exchangers as catalysts has been in two areas } acid-
and base-catalysed reactions. Solid acids and solid
bases are the two major ion exchangers employed
in water demineralization and puriRcation and have
been most explored as catalysts.
The strong acid ion exchangers are sulfonated
polymers of styrene crosslinked with divinylbenzene.
1573
The solid bases are copolymers of styrene and/or
vinylpyridine crosslinked with divinylbenzene and
functionalized to give either a quaternary ammonium
hydroxide group or a tertiary amine group. Solid
bases are also prepared from copolymers of the acryl-
ate and methacrylate monomers by crosslinking with
divinylbenzene followed by attachment of the amino
group to the polymer via an amide linkage.
Physical Structure
The geometry of ion exchange particles as manufac-
tured today is spherical. The bead diameters can
be varied by the method of manufacture, but the
standard size of commerce is a Gaussian distribution
of beads ranging in diameter from 250
m (60
mesh US Sieve Series) to about 1000
m (18 mesh
US Sieve Series). The condensation polymers of
phenol with formaldehyde may still be supplied as
irregularly shaped particles, but even these polymers
can be made in spherical bead form, if desired. The
spherical geometry arises from the method of manu-
facture which is by stirring a suspension polymeriz-
ation of monomer droplets dispersed in an immiscible
liquid. The immiscible liquid most used is water
properly formulated to maintain droplet integrity
throughout the transformation of monomer into
polymer.
Recently a number of manufacturers of ion ex-
changers have developed technology for making
monosized particles in which the range of size is very
narrow with a uniformity coefRcient of less than
1.12. Monosized particles may have an advantage in
some catalytic applications if the ion exchanger being
used is a gel resin. The monosized gel beads will have
an advantage over a Gaussian distribution of beads if
the average diffusional path length is shorter for
the monosized beads than that for the Gaussian dis-
tribution. For macroporous polymers, the bead
diameter has a very small impact upon the mass
transport because the ingress and egress is through
a continuous pore system rather than through a sol-
vated polymer network as in a gel polymer.
Ion exchange beads have two internal polymer
morphologies: one is a gel in which the network of
polymer chains is continuous throughout the bead
volume; the other is a macroporous structure in which
the bead is constructed from small microgel particles
tending towards spherical symmetry and packed to-
gether into clusters and arrays of clusters. The macro-
porous bead has both a continuous pore phase and
a continuous gel phase, whereas the gel bead has only
a continuous gel phase. Within the gel bead, there are
no pores. Porosity develops only as the polymer chains
are solvated by the reaction medium and become
1574 II / ION EXCHANGE / Catalysis: Organic Ion Exchangers

solvent separated. Within the macroporous polymers,
there are two subgroups: those with a small speciRc
surface area (SM ) less than about 400 to 500 m2 mL\1
and those with a large speciRc surface area greater than
about 600 m2 mL\1.
The macroporous polymers with a small speciRc
surface area have good accessibility into the core of
the bead but the number of catalytic sites on the pore
surface is insufRcient to provide acceptable rates
of catalysis. Consequently, the working phase in these
beads is primarily the gel phase of the microgel. The
macroporous polymers with a large speciRc surface
area (SM ) have sufRcient catalytic sites on the inter-
nal pore surface to give acceptable rates of catalysis
and are, therefore, true surface phase catalysts.
Table 1 shows these relationships for a family of
sulfonated macroporous polymers.
In the macroreticular synthesis of macroporous
polymers, large surface areas are achieved only by
increasing the level of crosslinking in the polymeriz-
ing monomer mixture. The microgel of the resulting
polymer is so tightly crosslinked that it is impen-
etrable even to molecules as small as methylene di-
chloride (CH2Cl2). For effective catalysis, the sur-
face phase must be the working arena since the gel
phase is impenetrable and also not functionalized.
Consequently, mass transport and catalytic ef-
fectiveness are inSuenced quite differently within
these three physical structures by the following:
Mass Transport: A Critical Element
in Performance
With ion exchangers as with other heterogeneous
catalysts, mass transport of reactants into the
catalytic site and mass transport of products from the
catalytic site can become totally rate controlling.
A qualitative tabulation of these interacting relation-
ships is provided in Table 2.
For estimating the suitability or design of a macro-
porous polymer for effective mass transport, the
equation given below, derived from the studies of
HalaH sz and Martin, has been found very useful. The
appropriate pore system for good mass transport can
be selected by simply knowing the molecular weight
of the expected product or the largest reactant
molecule.
dp"5 dM
dM"0.2457 (MWM)0.588
where: dp"pore diameter in A> of the pore system at
50% of the total pore volume of the macroporous
polymer; dM"random coil diameter in A> of either
the product molecule or the largest reactant molecule;
MWM"molecular weight of product or largest
reactant.

1. Level of crosslinking
Functional Groups ^ Catalytic Agents
2. Bead diameter The most studied catalytic functional group is the sul-
3. Solvating nature towards the polymer by the reac- fonic acid group (}SO3H) attached to styrene}
tion medium divinylbenzene copolymers of both gel and macro-
4. Size of the reactants and/or products. porous morphologies. Many reactions catalysed by

Table 1 Intrinsic properties of a family of sulfonated porous aromatic polymers: the relationship of surface capacity, specific surface
area, crosslinking density, and the working arena

Sulfonated Theory Measured Rings on Calculated Crosslinking Specific surface area, S
Working
porous weight weight internal theory wt. density phase in
polymer capacity, capacity, surface, cap. on int. (wt.%DVB) (m 2 g\1) (m 2 mL\1) catalysis
(meq g\1) (meq g\1) (No.%) surface
(meq g\1)

Amberlyst
XN-1008a 5.299 5.26 2.76 0.146 12 40 60 Gel phase
Amberlyst 15 5.210 5.00 3.72 0.197 20 55 82 Gel phase
Amberlyst
XN-1005b 4.854 3.50 11.95 0.683 50 120 180 Gel phase
Amberlyst
XN-1010 4.749 3.60 37.92 1.827 85 615 850 Surface
phase

a
Amberlyst XN-1008 is no longer marketed commercially. It has been replaced by Amberlyst 16 which has a similar specific surface
area but is slightly more porous.
b
Amberlyst XN-1005 is no longer commercially available.
 II / ION EXCHANGE / Catalysis: Organic Ion Exchangers 1575

Table 2 Qualitative relationship between polymer morphology of ion exchanger and change in reaction system

Polymer morphology Change in the reaction system Effect on the kinetics of catalysis

Gel polymer Increasing level of crosslinking from 2 to
12 wt.% divinylbenzene
Increasing bead diameter from 250 to
1000
m
Poorly solvating reaction medium for
catalyst
Increasing size of reactants and/or products
Decreasing rate of reaction. Gel copolymers of
styrene}divinylbenzene with greater than about 6%
divinylbenzene perform poorly as catalysts even in
good swelling solvents
Decreasing rate of reaction. In good solvating
reaction medium, the smaller the bead diameter,
the better the catalyst performance
Ineffective catalysis at any particle diameter and
any crosslinking level
Decreasing effectiveness as a catalyst

Macroporous polymer with
small specific surface area (SM )
(SM (400}500 m2 mL\1 bead)
Increasing level of crosslinking from 6 to 25
wt.% divinylbenzene
Increasing bead diameter from 250 to
1000
m
Poorly solvating reaction medium for
catalyst
Increasing size of reactants and/or
products
Moderate decline in catalytic effectiveness from
compensating changes. As the crosslinking level
increases, the SM increases and the microgel
diameter decreases
For small to moderate sized molecules, almost no
change in catalyltic effectiveness
Very small impact on catalytic performance
Provided the pore system is sufficient for ingress of
reactants and egress of products, only a moderate
decline in effectiveness

Macroporous polymer with
large specific surface area (SM )
(SM '600 m2 mL\1 bead)
Increasing level of crosslinking from 60 to
100 wt.% divinylbenzene
Increasing bead diameter from 250 to
1000
m
Poorly solvating reaction medium for
catalyst
Increasing size of reactants and/or
products
Increasing catalytic effectiveness, provided pore
dimensions remain large enough to accommodate
entrance of reactants and exit of products. As level
of DVB increases, the surface area increases
Little or no impact on catalytic effectiveness
No impact upon catalytic effectiveness
No impact upon catalytic effectiveness provided
pore system allows influx of reactants and efflux of
products

homogeneous strong acids have been examined for
effective catalysis by strong acid ion exchangers
because of their lack of corrosiveness and their ease of
separation from the reaction liquor by Rltration
(Table 3). Strong acid resins with crosslinking by
divinylbenzene of 8 wt.% or less work well only
where the reaction medium is very polar (water,
dimethylformamide, N-methylpyrrolidinone or
dimethyl sulfoxide) and the reactants or resulting
products are small (MW(250 Da).
Strong acid macroporous polymers work well in
essentially all reaction media, especially the large
speciRc surface area, strong acid macroporous poly-
mers where the surface phase is the catalytic arena.
Here solvation of the gel phase is unimportant. The
surface phase sulfonic acid groups, however, are not
as powerful in protonating reactants as those buried
in the gel phase } the surface sulfonic acid moiety is
a weaker acid than those within the gel. By placing
two sulfonic acid groups on each surface ring, by
sulfonation with fuming sulfuric acid, both the acid
strength and the thermal stability are increased. Sul-
fonated aromatic polymers with the sulfonic acid
group attached directly to the aromatic ring begin to
desulfonate at about 1203C, since sulfonation is acid
catalysed and reversible. By attaching an electron
withdrawing group, such as }SO2}, }SO3H, Cl, Br, F,
etc., to the aromatic ring, in addition to the sulfonic
acid group, the thermal stability of the strong acid
resin is boosted to about 1503C.
Ion exchangers have their functional groups an-
chored in space relative to each other and the neigh-
bouring groups can be used to enhance the rates of
reactions of appropriately structured molecules. As
an example of this effect, the rate of hydrolysis of
oleRnic esters is greatly enhanced by loading silver
cations on to a portion of the sulfonic acid groups. At
50% loading of the cation exchanger with silver ca-
tions, the maximum rate of hydrolysis of allyl acetate
is observed even though the concentration of acid
1576 II / ION EXCHANGE / Catalysis: Organic Ion Exchangers
Table 3 Transformations catalysed by ion exchangers
A. Reactions catalysed effectively by strong acid resins
Acetal and ketal synthesis
Addition of carboxylic acids to olefins
Alkylation of aromatic molecules, especially activated rings
such as phenols, toluene, etc.
Cumene hydroperoxide conversion to phenol and acetone
Dehydration of alcohols into olefins
Epoxidation of olefins with H2O2
Esterification
Etherification
Hydrolysis
Hydrolysis of starch, cellulose and saccharides
Olefin acylation
Olefin alkylation
Olefin hydration
Olefin isomerization
Olefin oligomerization
Solvolysis of epoxides
Transesterification
B. Reactions catalysed effectively by anion exchangers
Active methylene condensation reactions
Aldol condensation
Cannizzaro reaction
Cyanoethylation
Epoxide addition to carboxylic acids
Michael addition reactions
Nitrile hydrolysis to amides
C. Reactions catalysed effectively by heterogenized solid phase
transition and noble metal catalysts
Epoxidation of olefins
Hydroformylation
Hydrogenation
sites is halved. The silver ion with its propensity to
complex with double bonds pulls into the resin phase
a higher concentration of allyl acetate, thereby in-
creasing the rate of hydrolysis over that of the strong
acid resin without silver cations. This is a polymeric
matrix effect which is not possible with a homo-
geneous catalyst. Polymeric matrix effects are an
added advantage of solid phase over liquid phase
catalysts for enhancing the catalytic effectiveness
of ion exchangers.
Anion exchangers are the second most studied
group of solid phase catalysts. The catalytic agents
that are attached to the polymeric matrices are quat-
ernary ammonium and tertiary amine groups. The
positively charged nitrogen of the quaternary am-
monium group is effective in catalysing some
reactions, but most often it is the associated anion
that is varied to achieve an effective catalytic
agent. The positively charged nitrogen of the quater-
nary group is effective in catalysing epoxide ad-
dition reactions, as one example. Base catalysis is
carried out with either the quaternary ammonium
hydroxide or the tertiary amine group. For reactions
whose products are sensitive to the base strength,
a less powerful basic anion such as carbonate or
acetate can be employed when the quaternary am-
monium agent is the catalyst. The quaternary am-
monium resin in the hydroxide form begins to
decompose when used above 603C. Other anionic
forms are stable to about 1503C. The tertiary amine
resins are thermally stable to about 1503C.
Phase transfer catalysis is accomplished by anion
exchangers with any one of a number of appropriate
nucleophilic anions associated with the quaternary
ammonium group. The appropriateness of the nuc-
leophilic anion is controlled by the nature of the
chemical reaction undergoing catalysis. Phase trans-
fer catalysts can also be designed by attaching a quat-
ernary phosphonium group to the polymeric matrix
in place of the quaternary ammonium group. Spacer
arms that move the onium group further from the
crosslinked polymer backbone enhance catalytic ac-
tivity in phase transfer catalysis.
Crosslinked styrene}divinylbenzene copolymers
have been and continue to be actively investigated as
solid supports to heterogenize homogeneous catalytic
agents. These solid phase catalysts have transition
metals and noble metals anchored to the solid poly-
meric matrix through appropriate ligands. In hetero-
geneous form, they promote the same chemical
reactions as in solution, albeit with the imposed
mass transport limitations of the solid support.
See also: II/Ion Exchange: Historical Development; Inor-
ganic Ion Exchangers; Novel Layered Materials: Phos-
phates; Novel Layered Materials: Non-Phosphates;
Organic Ion Exchangers; Theory of Ion Exchange.
III/Catalyst Studies: Chromatography.
Further Reading
Albright RL (1987) Basic principles of catalysis by func-
tionalized porous organic polymers: theoretical concepts
and considerations. In Stiles AB (ed.) Catalyst Supports
and Supported Catalysts, pp. 159}186. Boston:
Butterworths.
Chakrabarti A and Sharma MM (1993) Cation exchange
resins as catalyst. State-of-the-Art Report, Reactive
Polymers, Vol. 20, pp. 1}45. Amsterdam: Elsevier.
Ford WT (ed.) (1986) Polymeric Reagents and Catalysts.
ACS Symposium Series 308. Washington, DC: American
Chemical Society.
Gates BC and Katzer JR (1979) Chemistry of Catalytic
Processes. New York: McGraw-Hill.
Hodge P and Sherrington DC (eds) (1980) Polymer-sup-
ported Reactions in Organic Synthesis. New York: Wiley.
Jakovac IJ (1987) SpeciRc reactions catalyzed by function-
alized porous organic polymers. In Stiles AB (ed.) Cata-
lyst Supports and Supported Catalysts, pp. 187}200.
Boston: Butterworths.
 II / ION EXCHANGE / Historical Development 1577

Neier W (1991) Ion exchangers as catalysts. In Dorfner
K (ed.) Ion Exchangers, pp. 981}1027. New York:
Walter deGruyter.
SatterReld CN (1991) Heterogeneous Catalysis in Indus-
trial Practice, 2nd edn. New York: McGraw-Hill.
Sherrington DC and Hodge P (eds) (1988) Synthesis and
Separations Using Functional Polymers. New York:
Wiley.
Thomas JM and Thomas WJ (1996) Principles and Practice
of Heterogeneous Catalysis. New York: VCH Publishers.

Historical Development
I. Grafova, Institute for Sorption and Problems of
Endoecology, National Academy of Sciences of
Ukraine, Kiev, Ukraine
Copyright ^ 2000 Academic Press
Development of Ion Exchange
Concept, Materials and Methods
The main stages in the development of ion exchange
are shown in Table 1 and 2. Ion exchange gradually
became an important separation method in water
treatment, waste water puriRcation, analytical chem-
istry, medicine, the food industry and many other
areas of application.
The Rrst systematic studies of ion exchange occur-
ring in natural inorganic materials were performed
during the period 1850}80, clays, sands and zeolites
became objects of investigation and it was shown that
soil treated with ammonium salts absorbs these ions,
releasing an equivalent amount of calcium ions.
Later, some natural materials found application for
puriRcation of water as well as for other purposes. At
that time, the evidence for existence of ions in solu-
tion had not yet been elucidated and the concept of
a double electric layer had not yet been proposed.
Despite this the stoichiometry of ion exchange and its
connection with aluminosilicates present in the soil
were established. It was demonstrated that the degree
of exchange increased up to a limiting value with the
increase of concentration of salt solution, while the
inSuence of temperature on ion exchange was shown
to be less signiRcant.
At the beginning of the twentieth century com-
plementary investigations in the areas of synthesis
and application of ion exchangers took place. Indus-
trial production of synthetic amorphous aluminosili-
cate ion exchange materials was started. These mater-
ials (permutites) were used for water softening and in
the treatment of sugar syrups. In the Rrst artiRcial
sodium aluminosilicates a substitution of sodium to
calcium occurred, but the ion exchanger could be
regenerated in a column by treatment with saturated
sodium chloride solution.
Ion exchange materials can also be obtained by
oxidation and sulfonation of coals. Some types of
charcoal, soft and hard brown coals, are suitable for
this purpose. They can be converted into cation ex-
changers after treatment with fuming sulfuric acid. As
a result, sulfonic and carboxylic groups (resulting
from oxidation) are introduced into the coal struc-
ture, playing the role of Rxed ions. Furthermore, the
coal is transformed to a gel due to polycondensation
reactions. The total exchange capacity of such mater-
ials is about 1.5 meq g\1.
Organic Ion Exchange Materials
Later, the ion exchange properties of some organic
materials were discovered, which led to the creation

Table 1 Principal practical achievements in the field of ion exchange

Year Milestone

1850}52 Discovery of ion exchange phenomenon in soil (Thompson, Way and Roy)
1903 The first synthetic inorganic ion exchanger (Harms, RuK mpler, Gans)
1935 The first ion exchange resin possessing high capacity (Adams and Holmes)
1944 Development of ion exchange resin synthesis by means of copolymerization (d’Alelio)
1947 Synthesis of zeolites (Barrer)
1950 Synthesis of ion exchange membranes (Wyllie, Sollner)
1958 Synthesis of inorganic ion exchanger based on zirconium phosphate (Amphlett)
1964 Synthesis of the first crystalline zirconium phosphate of -type structure (Clearfield and Stynes)
1975}79 Development of ion chromatography (Small, Gjerde)
1980}present New layered materials of - and -types, organic ion exchangers; improvement of ion exchange chromatography
method
 II / ION EXCHANGE / Historical Development 1577

Neier W (1991) Ion exchangers as catalysts. In Dorfner
K (ed.) Ion Exchangers, pp. 981}1027. New York:
Walter deGruyter.
SatterReld CN (1991) Heterogeneous Catalysis in Indus-
trial Practice, 2nd edn. New York: McGraw-Hill.
Sherrington DC and Hodge P (eds) (1988) Synthesis and
Separations Using Functional Polymers. New York:
Wiley.
Thomas JM and Thomas WJ (1996) Principles and Practice
of Heterogeneous Catalysis. New York: VCH Publishers.

Historical Development
I. Grafova, Institute for Sorption and Problems of
Endoecology, National Academy of Sciences of
Ukraine, Kiev, Ukraine
Copyright ^ 2000 Academic Press
Development of Ion Exchange
Concept, Materials and Methods
The main stages in the development of ion exchange
are shown in Table 1 and 2. Ion exchange gradually
became an important separation method in water
treatment, waste water puriRcation, analytical chem-
istry, medicine, the food industry and many other
areas of application.
The Rrst systematic studies of ion exchange occur-
ring in natural inorganic materials were performed
during the period 1850}80, clays, sands and zeolites
became objects of investigation and it was shown that
soil treated with ammonium salts absorbs these ions,
releasing an equivalent amount of calcium ions.
Later, some natural materials found application for
puriRcation of water as well as for other purposes. At
that time, the evidence for existence of ions in solu-
tion had not yet been elucidated and the concept of
a double electric layer had not yet been proposed.
Despite this the stoichiometry of ion exchange and its
connection with aluminosilicates present in the soil
were established. It was demonstrated that the degree
of exchange increased up to a limiting value with the
increase of concentration of salt solution, while the
inSuence of temperature on ion exchange was shown
to be less signiRcant.
At the beginning of the twentieth century com-
plementary investigations in the areas of synthesis
and application of ion exchangers took place. Indus-
trial production of synthetic amorphous aluminosili-
cate ion exchange materials was started. These mater-
ials (permutites) were used for water softening and in
the treatment of sugar syrups. In the Rrst artiRcial
sodium aluminosilicates a substitution of sodium to
calcium occurred, but the ion exchanger could be
regenerated in a column by treatment with saturated
sodium chloride solution.
Ion exchange materials can also be obtained by
oxidation and sulfonation of coals. Some types of
charcoal, soft and hard brown coals, are suitable for
this purpose. They can be converted into cation ex-
changers after treatment with fuming sulfuric acid. As
a result, sulfonic and carboxylic groups (resulting
from oxidation) are introduced into the coal struc-
ture, playing the role of Rxed ions. Furthermore, the
coal is transformed to a gel due to polycondensation
reactions. The total exchange capacity of such mater-
ials is about 1.5 meq g\1.
Organic Ion Exchange Materials
Later, the ion exchange properties of some organic
materials were discovered, which led to the creation

Table 1 Principal practical achievements in the field of ion exchange

Year Milestone

1850}52 Discovery of ion exchange phenomenon in soil (Thompson, Way and Roy)
1903 The first synthetic inorganic ion exchanger (Harms, RuK mpler, Gans)
1935 The first ion exchange resin possessing high capacity (Adams and Holmes)
1944 Development of ion exchange resin synthesis by means of copolymerization (d’Alelio)
1947 Synthesis of zeolites (Barrer)
1950 Synthesis of ion exchange membranes (Wyllie, Sollner)
1958 Synthesis of inorganic ion exchanger based on zirconium phosphate (Amphlett)
1964 Synthesis of the first crystalline zirconium phosphate of
-type structure (Clearfield and Stynes)
1975}79 Development of ion chromatography (Small, Gjerde)
1980}present New layered materials of
- and -types, organic ion exchangers; improvement of ion exchange chromatography
method
1578 II / ION EXCHANGE / Historical Development

Table 2 Important theoretical advances, elucidating the essence of ion exchange

Year Milestone

1879 Helmholtz theory of electrical double layer

1911 Donnan theory of membrane equilibria
1950s Statistical ion exchange models of Gregor, Kachalsky, Harris and Rice
1958 First edition of Helferrich’s monograph devoted to ion exchange was published
1940s (2nd half)}1960s Theories of ion exchange dynamics are developed
1960s (2nd half)}present Theoretical models and description of new crystalline layered materials possessing ion exchange
properties
1980}present Theoretical background of ion exchange chromatography

of an ion exchange resin by Adams and Holmes.
These new materials were characterized by their high
capacity (5}10 meq g\1) relative to inorganic ion ex-
changers. Resins were obtained by polycondensation
of phenols or amines with formaldehyde, and their
large-scale production began. Owing to their high
degree of cross-linking the polymers had negligible
solubility. The resins were hydrophilic due to the
presence of ionic groups as an inseparable part of the
polymer matrix: for example, for anion exchange
resins amino groups inside the matrix were balanced
by an equivalent quantity of anions. For cation ex-
changers phenolic, sulfoxylic, carboxylic or phos-
phonato or phosphinato groups were present inside
the matrix, balanced by an equivalent quantity of
cations.
A discovery by d’Alelio had great industrial signiR-
cance. He invented a method of synthesis of ion
exchangers based on styrene}divinylbenzene
copolymers. This invention was anticipated by
Staudinger’s synthesis of reticular polystyrene. The
Rrst cation exchanger of this type was obtained in
1944, followed in 1948 by an equivalent anion ex-
changer. These resins possess high chemical and
mechanical stability; their distinguishable feature is
a certain degree of control over the synthetic process.
Moreover, such materials were characterized by
high exchange capacity and working exchange rate.
Different Rxed ions can be introduced into the
styrene}divinylbenzene matrix, offering a possi-
bility to obtain resins with different cross-linking
numbers and swelling behaviour. All these properties
make this kind of synthetic resin of major practical
signiRcance (Table 3).
In the area of water treatment ion exchange tech-
niques occupy a leading position worldwide and due
to their increasing importance they are under con-
tinuous development. In 1951 Reents was the Rrst to
apply a mixed layer of anion and cation exchangers
for the ultra-puriRcation of water.
Emergence of cross-linked polymer electrolyte}
ion exchange resins has allowed a new approach to
the solution of problems of analytical and preparative
chemistry: puriRcation and separation of compounds
possessing similar chemical properties. However,
water treatment and waste water puriRcation remain
the main areas of application of ion exchange resins.
Here exchangers capable of being universal absorb-
ents for a wide variety of ions are mainly needed.
Parallel to investigations aimed at enhancing sorption

Table 3 Main fields of application of ion exchange resins

Resin type Matrix type Type of fixed groups Application fields

Strongly acidic cation exchange resins Gel Macroreticular Sulfonic
Weakly acidic cation exchange resins
Strongly basic anion exchange resins
Weakly basic anion exchange resins
Water treatment; separation of rare earth
elements; separation of amino acids, etc.
Carboxylic Decarbonizing of industrial water, water
softening and deionization
Gel Macroreticular Purification of antibiotics, copper and
nickel recovery
Gel Macroreticular Quaternary ammonium Different water conditioning processes;
elimination of organic compounds with
high molecular weight (macroreticular)
Gel Macroreticular Tertiary amine or polyamine Industrial water treatment; decolorization
of sugar syrups (macroreticular)

capacity, improving the exchange kinetics, thermal
stability, mechanical properties and chemical resist-
ance, there has been a considerable development of
selective ion exchangers. This requirement arose in
the 1950s in connection with both the analytical
problem of direct selective determination of elements
in a complex mixture and the problem of extraction
of metals from technological solutions during com-
plex ore processing. The selectivity of ion exchangers
is determined by two factors. The Rrst consists of an
exact correlation between the dimension of the
sorbent’s pores and the radius of the hydrated ion to
be absorbed. The second factor is related to the
formation of a coordination bond supplementary to
an ionic one between the ion and the functional
groups within the matrix.
Besides development of polymeric acids and bases
of different strengths, ion exchangers containing
functional groups able to form chelate complexes
with speciRc ions have also been obtained. This class
of ion exchangers is characterized by high ion ex-
change constants and high selectivity owing to do-
nor}acceptor interaction between adsorbent and
adsorbate. In 1940, Skogseid made a macromolecular
ion exchanger selective towards potassium, contain-
ing moieties analogous to dipicrylamine. Iron, ura-
nium and rare-earth elements readily form complexes
with oxygen-containing ligands. Several early
transition metals like cobalt, nickel and copper give
rise to stable amino complexes. Thus, anion ex-
changers containing amino groups (preferably those
of primary amines) are selective for the latter group of
elements, while exchangers containing phenolic
groups are suitable for iron. Resins with carboxylic
and phosphonic groups are suitable for uranium and
rare-earth metals.
However, some researchers consider the ion ex-
change procedure of water treatment a kind of ‘eco-
logical boomerang’, bearing in mind the fact that
wastewater after ion exchange still contains many
mineral compounds. Regeneration solutions contain
them in quantities greater by an order of magnitude
than the level of contaminants to be extracted. Ion
exchange membranes avoid this disadvantage. The
increase of mass of compounds in wastewater with
respect to the quantity of extracted substances does
not take place during membrane puriRcation, provid-
ing a signiRcant advantage for this method, when com-
pared to distillation or sorption on ion exchangers. In
1950, the Rrst samples of heterogeneous membranes
were obtained by Wyllie and Patnode based on com-
mercial ion exchangers reinforced by inert polymer
Rbres in order to provide high mechanical stability. In
1952, Manecke and Sollner reported the Rrst homo-
geneous membrane, and in 1957 Gregor made the
II / ION EXCHANGE / Historical Development 1579
Rrst inter-polymer membranes, where the ion-ex-
change and binder components are linked to each
other by chemical bonds. Ultra-pure water cannot be
obtained by the sole use of ion exchange membranes:
nowadays one has to assemble a combination of sep-
aration methods such as Rltration, microRltration,
ultraRltration, reverse osmosis, electrodialysis with
ion exchange membranes and a mono-layer of granu-
lated ion exchange resin, and electrodialysis with ion
exchange membranes Rlling the inter-membrane
space with a mixed layer of granulated ion exchange
resins.
Inorganic Ion Exchangers and Application of
Inorganic Layered Materials
A reawakening of interest in inorganic ion exchangers
was connected with the search for materials which
can withstand high temperature, ionizing radiation
and some aggressive chemicals. In 1943, Russel et al.
discovered that insoluble zirconium phosphate is suit-
able for the separation of uranium and plutonium
from nuclear Rssion products. Thus, a new class of
inorganic ion exchangers was synthesized on the basis
of group 4 elements, mainly of titanium and zirco-
nium. Various kinds of functional groups can be
attached to the polymer chains consisting of Ti or Zr
atoms bonded to oxygen, producing different
types of ion exchangers. It is a well-known fact that
zirconium readily forms chains:
in solution. The behaviour of titanium is analo-
gous. The most intense polymerization occurs
in a range of pH values close to that of the
hydroxide sedimentation. A polymer containing zir-
conyl groups together with residues of acid is ob-
tained by addition of salt or acid to a zirconium (or
titanium) salt solution. If the acid is polyprotic, then
one obtains a cation exchanger containing an ex-
change site like:
The Rrst material of this kind was zirconium phos-
phate obtained by Amphlett in 1958. It possessed
a capacity of 1 meq g\1 at pH 3, and 5 meq g\1 at pH
11. Amorphous inorganic sorbents had a great ad-
vantage over organic resins owing to their ease of
preparation.
1580 II / ION EXCHANGE / Historical Development
In 1947 Barrer realized a synthesis of zeolite for the
Rrst time. He became a founder of zeolite chemistry,
studying synthesis, structure, sorption and ion ex-
change properties of this new class of materials.
In 1964, ClearReld and Stynes synthesized the Rrst
crystalline zirconium phosphate and established its
layered
-type structure
-Zr(HPO4)2 . H2O, usually
referred to in the literature as
-ZrP. In 1968, the
same authors reported the Rrst -type zirconium
phosphate, referred to as
-ZrP. Since 1975, some
organic derivatives of the latter modiRcation have
been synthesized by Yamanaka et al. Further develop-
ment of this class of inorganic ion exchanger by
Alberti et al. (1978) resulted in M(IV) phosphonates
and organic phosphates with a layered structure of
zirconium bis-monohydrogen phosphate (
-ZrP).
Some years after the Rrst inorgano-organic sorbent
was reported, Dines and GrifRth described the
synthesis of diphosphonates of general formula
M(IV)(O3PdRdPO3). A series of covalently pillared
diphosphonates with a regular interlayer microporos-
ity was subsequently obtained. As far as -ZrP
(ZrPO4(H2PO4) . H2O) is concerned, a wide variety of
layered and pillared M(IV) phosphonates were ob-
tained in the period 1987}1990, since their structure
depends on the starting material and on the nature of
the O2PRR group that replaced O2P(OH)2.
Ion Exchange Chromatography
High performance in the separation of organic sub-
stances by liquid chromatography had already been
achieved by the 1970s, while techniques of chromato-
graphic separation of inorganic ions were developed
to a lesser extent. In 1971, the Rrst papers dealing
with automatic spectrophotometric detection of
metal ions separated by ion exchange chromatogra-
phy on cation exchange resins appeared. Similarly,
anions were separated on anion exchange resins.
Conductometric detection was established by Small
et al. in 1975 and Gjerde et al. in 1979. Finally it is
necessary to mention ion exclusion chromatography
applied to separate sugars and carbonic acids with an
ion exchange column, but without any ion exchange
interaction.
Development of Theoretical
Background in the Field of Ion
Exchange
Helmholtz developed the theory of the double electric
layer in 1879, and this had a fundamental signiRcance
for the explanation of many phenomena related to
ion exchange. According to his classic theory, the
double layer consists of two electric layers, analogous
to the plates of an electric capacitor. The concept was
later modiRed, as it had been proven that a double
layer is formed at the interface between a solid phase
and a solution. It consists of an immobile inner layer
and a diffuse outer layer. Charge separation in
this type of a system is of the order of molecular
dimensions. As a result, the solid phase surface ac-
quires an electric charge (positive or negative), while
the counterions are distributed along the interface.
These ions form a virtual outer plate of the dif-
fuse layer. In fact, there is no interface but a dynamic
equilibrium exists between the ions in the diffuse
layer and those of the environment (the bulk solu-
tion). The existing equilibrium is disturbed upon
changes in either pH or concentration of ions in the
bulk solution. In this case new ions enter the dif-
fuse layer, substituting ones already there and estab-
lishing a new ion exchange equilibrium.
The inSuence of valence, hydration and ion dimen-
sions on the degree of ion exchange transformations
has been established and ion selectivity series were
composed by Wiegner for ion retention on alumino-
silicate sorbents:
Li#(Na#(K#(Rb#(Cs#
Na#(Mg2#(Al3#(Th4#
Mattson demonstrated in 1927 that aluminosilicates
enriched in silica gave an increase in cation exchange
capacity. Jaeger demonstrated that at higher concen-
trations, as well as in mixed organo-aqueous solu-
tions, the values of the exchange potentials of ions
with the same valence could be quite close, with
possible inversion of the selectivity series.
Selectivity, i.e. preferable absorption of one of the
counterions, is an important property of ion ex-
changers. Hence, the ion exchanger becomes enriched
in counterions which possess small dimensions in the
solvated state or which are able to enter into speciRc
interactions with the Rxed ions or with the matrix.
The isotherm of ion exchange is a graphical repres-
entation of the dependence of an equivalent fraction
of the counterion A in the ion exchanger, versus its
equivalent fraction in solution.
In 1939 Nikolskii demonstrated that the law of
mass action could be applied successfully to ion
exchange:
(y1/Z
1
1
)/(y1/Z
2
2
)"K(a1/Z
1
1
)/(a1/Z
2
2
)
where y1 and y2 are the quantities of ions absorbed by
the resin (meq); Z1 and Z2 are charges of these ions in
solution; a1 and a2 are activities of the above ions in
the solution and K is the ion exchange constant.

In 1935 Kielland introduced activity coefR-
cients for the ion exchanger phase in order to estimate
deviations of ion exchange equilibria from a simple
form of the law of mass action. The most clear ther-
modynamic approach to these equilibria was pro-
vided by Gaines and Thomas in 1953, which was not
connected to model concepts.
A physical model of ion exchange processes has
been gradually formed together with the accumula-
tion of empirical data. For example, the theory of
membrane equilibria created by Donnan in 1911 pro-
moted a breakthrough after its application to ion
exchange processes. The essence of Donnan theory
can be brieSy described as follows. An electrolyte
RNa is supposed to be on one side of a membrane,
and an NaCl solution on the other side. Since the
membrane is not permeable for R\ ions, only sodium
and chloride ion redistribution will take place in the
system. However, in the above case the diffusion
process is unable to equalize concentrations of all
ions from both sides of the membrane. For ion ex-
change equilibria, an interface between liquid and
solid phases is considered as a membrane, while a col-
loidal particle bearing the exchangeable ion is as-
sumed to be a non-diffusing ion.
In the early 1950s, Gregor proposed an osmotic
theory of exchange. According to this model, the
matrix of ion exchange resin is expanded upon swell-
ing, and thus applies pressure to the liquid inside the
pores. In such a case the equilibrium is determined by
a difference in the osmotic pressure between the
external solution and that of the liquid inside the
pores together with the elastic forces of the matrix.
However, Gregor’s model does not consider the
formation of ion pairs, and hence does not explain
selective adsorption; moreover, its accuracy is not
sufRcient to describe exchange in dilute solutions,
which are typical for ion exchange processes.
Some years later another model of ion exchange
appeared, built on a molecular background rather
than the macroscopic Gregor model. The model
of Kachalsky (mid-1950s) presumes that the
energy of electrostatic interaction which imparts
changes in free energy of the system is uniformly
distributed over the polymer chain between the
ionogenic groups. The model describes the resin as
a linear polyelectrolyte and provides an accurate de-
scription of ion exchangers with a small number of
cross-linking bonds.
A similar approach was used in the model de-
veloped by Harris and Rice at about the same time. It
also uses a molecular level approach and the concept
of linear polyelectrolytes, but the authors concen-
trated on interactions of neighbouring functional
groups both in the same and in the adjacent polymer
II / ION EXCHANGE / Historical Development 1581
chains, taking into account the regular distribution of
Rxed ions in the ion exchange resin phase. Another
distinctive feature of this model is that it considers
ion pair formation between the Rxed ions and
counterions.
Pepper has studied swelling of different abso-
lutely dry resins in water and demonstrated that the
total volume of the system undergoes shrinkage at the
Rrst step and then remains unaltered. Thus initially,
the primary hydration shell is formed which consists
of the so-called ‘Rxed’ water, possessing speciRc prop-
erties. The water absorbed afterwards is called ‘free’
water and behaves like ordinary loosely bound water.
The degree of swelling depends on the structure and
cross-linking of the particular resin. Inorganic ion
exchangers usually have a rigid crystal structure and
their swelling is insigniRcant; layered ion exchangers
undergo intralaminar swelling.
Ion exchange interactions occur at different
rates in heterogeneous media; therefore studies deal-
ing with ion exchange kinetics are of great import-
ance. In 1947, Boyd and co-workers showed that the
exchange rate is determined either by the diffu-
sion rate inside the resin bed (gel diffusion) or by
the diffusion in the layer of liquid surrounding
the bed (Rlm diffusion). When the rates of gel
and Rlm diffusions are comparable, both compo-
nents determine the exchange rate. Determination of
the mechanism and the limiting stage of the ion ex-
change process presents a rather difRcult prob-
lem, because the kinetics simultaneously depend on
a number of parameters, such as the concentration of
adsorbate in solution, the nature of the ionic species,
the type and granular composition of the ion ex-
changer and the relative migration rate of the interac-
ting phases. The limiting stage of ion exchange can be
approximately estimated as follows:
for gel diffusion:
(cDM )/(cDr0) ) (5#2KA/B)1
for Rlm diffusion:
(cDM )/(cDr0) ) (5#2KA/B)1
and in case of mixed diffusion:
(cDM )/(cDr0) ) (5#2KA/B)+1
where cN "zicN i, c" zj cj are total concentrations
of exchanging ions in the solid and liquid phases
1582 II / ION EXCHANGE / Historical Development
respectively; DM and D are ionic diffusion
coefRcients in the solid and liquid phases, respec-
tively;  is the thickness of the diffusion inter-
layer; r0 is the radius of the ion exchanger bed; and
KA/B"cN AcB/cAcN B is the separation coefRcient for
both sorts of counterions in the equilibrium state: the
counterion A is present initially inside the bed, while
the counterion B is initially in solution.
An equivalent exchange takes place under the fol-
lowing condition:
n change properties. Each layer in their structure can
zRcN R" zAicN Ai
i
where cN R is the concentration of Rxed groups inside
the bed, cN Ai is the concentration of i-th counterion.
This is only an approximation, applicable to ion
exchangers of high capacity treated with relatively
dilute aqueous solutions. In all other cases, the ion
exchanger absorbs a signiRcantly higher quantity of
ions than is needed for equivalence, i.e. a super-
equivalent exchange occurs. An excess of counterions
penetrates into the exchanger bed accompanied by
a quantity of co-ions, necessary to compensate the
electric charge of the former, in order to preserve
a condition of electroneutrality of the bed:
n m
zRcN R#  zXjcXj"  zAicN Ai
j"1 i"1
The higher the concentration of external solution, the
greater is the contribution of super-equivalent ex-
change. Even for zeolites and highly cross-linked
resins it becomes apparent at concentrations of ex-
ternal solution of about 0.1 N and higher. For scarce-
ly reticulated, macroporous, highly swelling or weak-
ly charged ion exchange resins the super-equivalent
exchange has a much more pronounced effect
and it becomes distinguishable at considerably lower
concentrations.
A sorption without ion exchange is usually named
Donnan absorption of electrolyte, because the ther-
modynamic description of the process is principally
the same both for the system ‘ion exchanger}solution’
and for systems with a real semipermeable membrane
as an interface separating phases in Donnan theory.
However, the latter postulates that the dissociation in
both phases is complete. Another situation is ob-
served when the exchange occurs between the bed
and weakly dissociated electrolyte or when ionogenic
groups on the former are weakly dissociated. In that
case, one deals with sorption of fragments of undis-
sociated molecules and ion pairs and the exchange
cannot be described in terms of Donnan theory. The
concept of ‘super-equivalent exchange’ includes the
exchange of ions, absorption of electrolyte molecules
without ion exchange, and other processes which
could be characterized as between the above two
cases.
Recent Progress in Ion Exchange
In recent years extensive research has been carried
out on new crystalline inorganic and inorgano-
organic layered compounds which possess ion ex-
be considered as a planar macromolecule, while the
substance as a whole is assumed to be a molecular
crystal formed by these planar macromolecules. A re-
versible process of intercalation between the layers
occurs due to interactions of guest species with active
sites on the surface of the layer (lamella). However,
the layers are unable to move spontaneously in a di-
rection perpendicular to the plane. This is due to
a certain rigidity of layers that plays an important
role in intercalation reaction mechanism and ener-
getics. Like other ion exchange materials, the charged
layered solids may be strong, medium or weak
cationic (or anionic).
The exchange of protons of
-ZrP phase for Li#,
Na# and Ca2# occurs rapidly in acidic solutions,
while H# exchange for larger or strongly hydrated
cations like NH# # #
4 , Rb , Cs , Ba
2#
, Mg2#, Cu2#
3#
and Cr is quite slow at room temperature due to
the high activation energies of interlayer expansion.
Exchange can be facilitated in materials with large
interlayer distances like
-Zr(HPO4)(NaPO4) . 5H2O
(d"11.8 A> ) or in intercalation compounds with
ethanol or alkylamines. The compounds that can be
protonated are preferably used as guest species. For
example, an amino derivative of cyclodextrin has
been used for intercalation, increasing the interplanar
distance in
-ZrP up to d"35.6 A> . These distances
for other guest species are; 14.2 A> for ethanol, 20.4 A>
with benzimidazole, 22.8 A> with 1-hexylamine and
23.1 A> with lysine. The layered compounds under
discussion can swell upon introduction of water or
other solvents into the interlayer space. Sometimes
the process leads to delamination, i.e. destruction of
the crystal into separate lamellae. Withdrawal of the
solvent results in reaggregation of the lamellae in thin
Rlms or membranes. Inorgano-organic derivatives,
phosphonates of layered
-structure, can be obtained
by introducing the corresponding acid H2O3PR
(where R"}CH3, }C6H5, }O(CH2)nCH3 etc.) into
the reaction, instead of H3PO4. It is also possible to
synthesize those compounds by substitution of
existing OH groups in the
-ZrP structure by R or
OR. Another interesting group of compounds is
 II / ION EXCHANGE / Historical Development 1583

covalently pillared zirconium diphosphonates of gen- A nontraditional application of ion exchangers

eral formula MIV(O3P}R}PO3). If the R group is in nonpolar organic media is the ultra-puriRcation
small, then a low degree of interlayer microporosity is of organometallic compounds used as precursors in
observed, while for pillared compounds containing chemical vapour deposition. These precursors are
fragments of 3,3(5,5)-tetramethylbiphenyldiphos- widely utilized for synthesizing materials possessing
phonic acid the value of interlayer porosity is raised valuable properties for micro-, opto- and acousto-
to 375 m2 g\1 (an average pore size of 5 A> ). Inor- electronics, and protective and optical coatings.
gano-organic derivatives have also been obtained for The organometallics in question react readily
-ZrP by substitution of the interlayer O2P(OH)2 with atmospheric oxygen and moisture, while at
groups for O2PRR. Pillared phases of -ZrP with, for the same the requirements on their purity are quite
example, biphenylphosphonate groups have a volume rigorous (less then 1;10\3% of the sum of con-
of micropores of 320 m2 g\1 and an average size of taminants). The above requirements can be met by
5.8 A> . treatment with a sorbent composition containing in-
A limited number of inorganic anion exchangers is organic ion exchangers based on titanium and zirco-
known. Layered double hydroxides (or hydrotalcite- nium phosphates, thus replacing energy-intensive and
like anionic clays) can exchange a large number of expensive traditional methods (sublimation or distil-
inorganic and organic anions, while layered ZrPO4Cl lation).
can selectively replace chloride anions with other
monodentate anionic ligands.
Currently, ion exchange is of extreme importance Future Developments
for processing of irradiated nuclear fuel and treat-
Undoubtedly, the future development of ion ex-
ment of spent fuel elements of nuclear power stations,
change as a method of separation will be directed
where it is often combined with other techniques, e.g.
towards ecological and biotechnological problems.
extraction. The processes of sorption play an impor-
The development of society parallel to scientiRc
tant role in deactivation of nuclear industry wastes
and technical progress will promote greater regard
and in puriRcation of cooling water from nuclear
for natural resources. Hence, particular attention
reactors. Different kinds of ion exchangers are
will be drawn to the application of renewable tech-
widely used for the clean-up of the world’s worst
nologies and closed technological cycles including
nuclear accident at Chernobyl. For example, Strelko
ion exchange stages or applying ion exchange mater-
et al. are carrying out both research and application
ials mainly in the area of water treatment and waste-
of highly selective inorganic granulated ion exchange-
water puriRcation, as well as in several other Relds.
rs for elimination of radioactive isotopes from drink-
Certainly, the use of ion exchangers in medicine will
ing water, milk, etc. Ion exchange is extensively used
increase.
in medicine for haemosorption (or haemoperfusion)
New, advanced ion exchange materials possessing
} the method of blood puriRcation from toxic com-
desirable properties will be obtained by targeted
pounds by direct contact of the sorbent with the
synthesis; computer modelling and simulations
patient’s blood. This method was applied for the Rrst
as well as molecular design will be increasingly
time by Muirhead and Reid in 1948; they directed the
applied.
blood Sow through a mixture of cation and anion
The speciRcity and selectivity of ion exchange will
exchangers taken in a ratio of 9 : 1. Haemosorption
grow; i.e., the most suitable materials from the view-
can be applied alone or in combinations with
point of their origin, matrix type, the type of
haemodialysis (when the toxins are distributed be-
ionogenic groups etc. will increasingly be applied for
tween two liquid phases, separated by a semiperme-
speciRc cases.
able membrane). The ion exchangers are used to
The phenomenon of ion exchange discovered
regenerate dialysate from the artiRcial kidney appar-
a century and a half ago, as well as processes estab-
atus. Further improvement of the haemosorption
lished on the basis of it, are still in a process
method is connected with the necessity to resolve
of dynamic development. The potential of ion ex-
problems of selective blood puriRcation, as well as the
change both in practical applications and from the
problem of better sorbent compatibility with biolo-
elaboration of theoretical concepts related to ion ex-
gical Suids. Another possible medical application
change is not yet complete.
of ion exchangers consists of the creation of drugs
and pharmaceuticals with prolonged activity, of-
fering the possibility to release an active component See also: II/Ion Exchange: Novel Layered Materials:
inside the patient’s body over time and maintaining Phosphates; Novel Layered Materials: Non-Phosphates;
its necessary concentration. Organic Ion Exchangers; Theory of Ion Exchange.
1584 II / ION EXCHANGE / Inorganic Ion Exchangers
Further Reading
Alberti G, Casciola M, Costantino U and Vivani R (1996)
Layered and pillared metal(IV) phosphates and phos-
phonates. Advanced Materials 8(4): 291.
Amphlett CB (1964) Inorganic Ion Exchangers. Amster-
dam: Elsevier.
ClearReld A (ed.) (1982) Inorganic Ion Exchange Mater-
ials. Boca Raton, FL: CRC Press.
Fritz JS, Gjerde DT and Pohlandt C (1982) Ion Chromato-
graphy. Heidelberg: HuK thig.
Greig JA (ed.) (1996) Ion Exchange Developments and
Applications. Cambridge: Royal Society of Chemistry.
Inorganic Ion Exchangers
E. N. Coker, BP Amoco Chemicals,
Sunbury-on-Thames, Middlesex, UK
Copyright ^ 2000 Academic Press
Summary
In the Rrst part of this chapter, the origins of
ion exchange in inorganic materials are discussed
in relation to the structure of the exchanger.
Thereafter, the various types of inorganic ion
exchangers are introduced and categorized according
to their ion exchange properties. Descriptions of
particular materials follow, with special emphasis
on some structure-speciRc and composition-speciRc
ion exchange properties. The materials which are
discussed include zeolites and zeolite-like materials,
clays and other layered materials, zirconium
phosphates, heteropolyoxometalates and hydrous
oxides.
Types of Ion Exchange Sites in
Inorganic Materials and their
Origin
For the purposes of this chapter, ion exchange
interactions will be deRned as those involving the
interchange of positively or negatively charged
species (atomic or molecular) at an ion exchange
site.
There are two types of chemical species which
constitute the vast majority of ion exchange sites in
inorganic materials:
1. structure-terminating, covalently bonded groups
such as }OH
Helfferich F (1962) Ion Exchange, 2nd edn. New
York: McGraw-Hill.
Hwang S-T and Kammermeyer K (1975) Membranes in
Separations. New York: Wiley.
Marinsky JA and Marcus Y (eds) (1973) Ion Exchange
and Solvent Extraction. New York: Marcel
Dekker.
Osborn GH (1961) Synthetic Ion-Exchangers: Recent De-
velopment in Theory and Application. London: Chap-
man & Hall.
Weiss J (1994) Ion Chromatography, 2nd edn. Weinheim:
Wiley.
2. charge-compensating groups, electrostatically as-
sociated with, and not covalently bonded to,
a charged moiety
Type 1 sites, illustrated in Figure 1A, are respon-
sible for the ion exchange properties of materials
such as hydrous oxides and single-layer clays.
All oxidic materials have these sites to some degree,
at the surfaces of particles or crystals or at defect
sites within the structure. Ion exchange reactions
involving these types of sites may be regarded as
chemical reactions, which may display amphoteric
nature.
Type 2 sites, illustrated in Figure 1B, are respon-
sible for most of the ion exchange capacity of zeolites,
double-layer clays and zirconium phosphates. These
sites arise in structures possessing, for instance,
charged layers or charged porous frameworks. The
exchangeable ions are present to retain overall elec-
troneutrality. When materials such as zeolites are
concerned, a mixture of Type 1 and Type 2 sites is
available, although Type 2 sites will usually greatly
outnumber Type 1 sites, and the latter are often
ignored. Exchange interactions involving Type 2 sites
are physical in nature, as chemical bonds are neither
made nor broken.
Types of Inorganic Ion Exchange
Material
An important distinction between ion exchange ma-
terials is whether they exhibit capacity for cations,
anions, or both. Cation exchangers, and in particular
zeolites, clays and zirconium phosphates, are the
most common and best understood of the ion ex-
changers. Anion exchangers are also important but

Figure 1 The two major types of ion exchange site. (A) Type 1,
structure-terminating and defect groups; (B) Type 2, charge-com-
pensating groups. M is an oxide-forming metal with oxidation
state 4; T is an oxide-forming metal with oxidation state 3. The
regions enclosed in dotted lines are those giving rise to ion
exchange where Z# (or Z}O\) is exchangeable. Shaded areas
represent a continuation of the oxidic network.
the exchange of anions is often not fully reversible,
thus the exchangers cannot be easily regenerated and
the reactions are more difRcult to treat thermo-
dynamically. Multiply charged anions, in particular,
may be held tenaciously by the exchanger. Examples
of anion exchangers are certain clays such as hydroxy
double salts (e.g. [CuNi(OH)3]Cl) and layered
double hydroxides (e.g. hydrotalcite, Mg6Al2(OH)16
(CO3) ) 4H2O). Amphoteric ion exchangers possess
predominantly Type 1 exchange sites, e.g. hydrous
oxides.
While ion exchange properties may be exhibited by
both amorphous and crystalline solids, studies of the
ion exchange properties of amorphous solids are of-
ten hampered by difRculties in preparing mater-
ials reproducibly and the difRculties in character-
izing them fully. With crystalline materials, however,
reproducible preparations can be easily veriRed and
well-deRned structural data aids in the interpretation
of the results of ion exchange experiments.
II / ION EXCHANGE / Inorganic Ion Exchangers 1585
Most crystalline inorganic ion exchangers are por-
ous. This porosity may arise through the presence of
void space between the layers in clay materials and
layered double hydroxides, or through the intrinsic
microporosity present in zeolitic materials. Many of
the layered materials have the versatility to (revers-
ibly) change their interlayer spacing and hence the
size of the voids, which allows the ion exchange
properties to be adjusted. The more rigid zeolite
structures give rise to exchange reactions which may
show extremely high selectivity to certain cations, or
perform ion sieving.
Zeolites
Zeolites are microporous crystalline aluminosilicate
minerals which occur naturally and may be syn-
thesized easily in the laboratory. An introduction
to the structures and properties of zeolites is given
in the article by Dyer. Zeolites are used on a large
scale as ion exchangers in many Relds; most notable
are their use as ‘builders’ or water softeners for laun-
dry detergents, and their use in the decontamination
of various types of waste streams. Typical applica-
tions of zeolites as ion exchangers are given in
Table 1. Additionally, the ion exchange capability of
zeolites can be used as a tool to modify their catalytic
and sorptive properties. Some attention will be paid
to structural parameters which inSuence the ion ex-
change properties of zeolites in the following para-
graphs.
Besides the conditions under which an ion ex-
change reaction is performed, a number of factors
may inSuence the ion exchange properties of zeolites,
including:
E the structure of the zeolite, particularly the dia-
meters of the windows allowing access to the pores
and cavities
E the location of the ion exchange sites; different
cation environments lead to different ion ex-
change properties. The number of charge-balanc-
ing cations required for an electroneutral material
is often less than the number of available ion ex-
change sites, thus partial occupancy of sites is com-
mon. Some of the possible cation positions in
zeolites A and X (two of the most widely used
synthetic zeolite ion exchangers) are indicated in
Figure 2
E the composition of the zeolite framework;
varying the Si : Al ratio or changing the frame-
work substituent elements may change, for
example, the density of exchange sites, the electric
Reld strength or the hydrophobicity of the sample
as a whole
1586 II / ION EXCHANGE / Inorganic Ion Exchangers

Table 1 Principal applications of zeolites as ion exchangers

Application Type of zeolite frequently used Ion exchange process

Detergent building A (synthetic) Removal of Ca2# and Mg2# from solution

Wastewater treatment
Nuclear waste treatment
Animal food supplement
MAP (synthetic)
X (synthetic)
Clinoptilolite (natural) Uptake of NH#
4 and heavy metals from waste
Chabazite (natural) streams
Mordenite (natural)
Phillipsite (natural)
Clinoptilolite (natural) Uptake of 137Cs#, 90Sr2# and other radionuclides
Chabazite (natural)
Phillipsite (natural)
Mordenite (natural)
Mordenite (synthetic)
Ionsiv IE-96 (synthetic)
Ionsiv A-51 (synthetic)
Various (natural) Regulation of NH#
4 and NH3 levels in stomach

Animal food supplement Various (natural) Scavenging of radionuclides following contamina-

tion of livestock
Fertilizer Various NH#4 forms (natural), often those Slow release of NH#
4 (and other cations)
used to remove NH#4 from wastewater

The empirical structural formula
aluminosilicate zeolite may be given as
x/n[(AlO2)x(SiO2)y] ) wH2O
M(n)
where the framework is constructed from the
entities within the square brackets and the water
molecules and charge-balancing cations (M)
for an occupy the interstitial space. The x/n Mn# cations
are present to counterbalance the x units of
negative charge on the framework due to the presence
of x AlO2 groups. In many cases, ion exchange
reactions in zeolites may reach completion, that is,
all of the charge-balancing cations (M) initially
present are capable of being replaced by the ingoing
cation.

Figure 2 A representation of some of the possible positions of exchangeable cations in the structures of zeolites A (A) and X (B).
Note: the two structures are not shown on the same scale. Reproduced with permission from Stucky GD and Dwyer FG (eds) (1983)
Intrazeolite Chemistry. ACS Symposium Series, vol. 218, p. 288. Washington, DC: American Chemical Society.
 II / ION EXCHANGE / Inorganic Ion Exchangers 1587

Incomplete ion exchange reactions In some cases,

some of the cations are constrained within the struc-
ture and are nonexchangeable. Such cations are intro-
duced into small cavities in the structure during
growth of the zeolite crystal. This situation is
common with feldspars and feldspathoids, which are
similar in composition to zeolites, but possess more
limited porosity. Even in instances when all charge-
balancing cations in the zeolite are physically ex-
changeable, the total theoretical exchange capacity
might not be obtained practically.
There are several reasons for incomplete ion ex-
change; the three most important of these are given
below and illustrated schematically in Figure 3.
1. The most obvious cause of partial or nonexistent
exchange is ion-sieving, where the cation to be
exchanged into the zeolite is too large, or has
a hydration sphere which is too large and robust
for it to have unrestricted access to the pores of the
zeolite. Univalent cations will typically reach
100% exchange, except in limiting cases such as
large cations combined with small-pore zeolites.
Ion-sieving is more commonly observed with
multiply charged cations, which tend to have lar-
ger hydration spheres on account of their higher
charge densities. Zeolites which possess more than
one ion exchange site (see Figure 2) may display
ion-sieving properties depending on the thermo-
dynamics of the exchange reactions occurring at
the various sites. The sites which offer the
greatest thermodynamic advantage are exchanged
Rrst, while the less favourable sites may not ex-
change at all.
2. Volumetric exclusion may occur if bulky (organic)
cations are exchanged into zeolites of high charge
density. Here, the volume occupied by the cations
Figure 3 The principal reasons for limitations to ion exchange
may reach that available in the pores of the crystal
reactions found in zeolites. (A) Ion-sieving; (B) volume exclusion;
before complete exchange has occurred. (C) low charge density (with multivalent cations). The lightly
3. A third reason for limited exchange to be observed shaded regions represent an extract of the zeolite framework. For
is when multivalent cations are exchanged into clarity, only ingoing cations are shown.
zeolites of low charge density. As the density of
ion exchange sites decreases, the mean separation
exchange is kinetically limited but still capable of
between adjacent sites increases, until a point is
reaching 100% of the theoretical capacity is the sof-
reached where multivalent cations are unable to
tening of water.
satisfy two or more cation exchange sites because
Zeolites are used in vast quantities in the detergent
of the distance between them. Table 2 illustrates
industry as a water-softening additive for laundry
this point by listing the maximum exchange limits
detergents } up to 30% by weight of most modern
observed for several multivalent cations in samples
washing powders is zeolite. The zeolite is added prin-
of zeolites ZSM-5 and EU-1 possessing a range of
cipally to remove calcium and magnesium and thus
Si/Al ratios.
prevent their precipitation with surfactant molecules.
It is easy to visualize the limiting factors of ion Zeolite A is most commonly used, due to its high ion
exchange under equilibrium conditions; however, exchange capacity, which is a consequence of the
practical ion exchange may have also kinetic limita- framework possessing the maximum possible number
tions. A particular example of when the desired ion of aluminium atoms (Si : Al"1 : 1). Recently, zeolite
1588 II / ION EXCHANGE / Inorganic Ion Exchangers

Table 2 Ion exchange limits (mole fraction) for various multivalent cations and temperatures in samples of zeolites ZSM-5 and EU-1
with varying numbers of aluminium atoms in the framework. In all cases, the ingoing cation replaces sodium

Zeolite type Al per Ca 2# Sr 2# Ba 2# La 3# Ca 2# Sr 2# Ba 2# La 3#

u.c.a (253C) (253C) (253C) (253C) (653C) (653C) (653C) (653C)

ZSM-5 1.1 0.28 0.31 0.36 0.50 0.51 0.52

ZSM-5 2.0 0.31 0.36 0.56 0.54 0.64 0.76
ZSM-5 2.4 0.36 0.48 0.67 0.39 0.50 0.67 0.77 0.48
ZSM-5 4.2 0.37 0.42 0.90 0.62 0.85 0.93
EU-1 1.2 0.54 0.56 0.56
EU-1 2.1 0.62 0.67 0.67 0.85 0.89 0.89
EU-1 3.8 0.86 0.93 0.93 0.96 0.97 0.97

a
Number of aluminium atoms in framework per unit cell.

MAP (Maximum Aluminium P), also with
Si : Al"1 : 1, has been introduced into some deter-
gents. Although the Mg2# ion (radius 0.07 nm) is
considerably smaller than the Ca2# ion (radius
0.1 nm), its exchange into the zeolite is far less
facile than that of Ca2#, due to its large, tight
hydration sphere (the radii of the hydrated Ca2#
and Mg2# cations are estimated to be 0.42 and
0.44 nm, respectively). Figure 4 shows the kinetics of
exchange of Ca2# and Mg2# into Na-A zeolite.
The major restriction to the hydrated Mg2# cation
is the 0.42 nm window in zeolite A through which
it must pass to gain access to the exchange sites
within the structure. In order for the ion exchanger
to be effective as a water softener for detergents,
it must reduce water hardness within a few minutes of
beginning the wash cycle. While zeolites A and MAP
perform well at removing calcium from hard water
quickly, their performance towards magnesium is
generally poor. Despite the kinetic limitations, Ca2#
and Mg2# are fully exchangeable into zeolite A, al-
though selectivity is greater for Ca2# (Figure 5). De-
tergent-grade zeolites possess small crystallite sizes in
order to provide acceptable kinetics of Ca2#
exchange.
Materials closely related to zeolites
Semicrystalline zeolites Some interest has been
shown in the ion exchange properties of zeolite pre-
cursors, which are obtained by quenching a zeolite
synthesis mixture before it has fully crystallized. In
these semicrystalline materials, some larger windows
and pores are present than in the crystalline counter-
part because the structure has not fully formed. This
leads to ion exchange selectivities which are dif-
ferent from the crystalline material. Also, their ion
exchange capacities are lower than the corresponding
crystalline zeolites. The materials typically show
weak zeolite X-ray diffraction patterns, and are

Figure 4 Kinetics of exchange of Ca2# and Mg2# for 2Na# in Figure 5 Isotherms for Ca2#/2Na# and Mg2#/2Na# exchange
zeolite A. Circles, Ca2# exchange; triangles, Mg2# exchange. in zeolite A. Circles, Ca2# exchange; triangles, Mg2# exchange.
Data were determined at 253C, pH 10 and at a solution concentra- Data were determined at 253C, pH 10 and at a solution concentra-
tion of 0.05 mol equiv. L\1. tion of 0.05 mol equiv. L\1.
 II / ION EXCHANGE / Inorganic Ion Exchangers
Figure 6 Kinetics of exchange of Ca2# and Mg2# for 2Na# in the semicrystalline precursor to zeolite A. Circles, Ca2# exchange;
triangles, Mg2# exchange. Data were determined at 253C, pH 10 and at a solution concentration of 0.05 mol equiv. L\1.
thus not totally amorphous, but possess some short-
to-medium range order. Semicrystalline precursors to
zeolites have been investigated as potential water
softeners with enhanced magnesium performance for
detergent use. The materials show slightly limited
capacities for both calcium and magnesium, but the
selectivity ratio of Mg : Ca is higher than that in the
fully crystalline counterpart. In the kinetics of ex-
change, one sees the inSuence of the population of
larger windows and pores. The rate of Mg2# ex-
change approaches that of Ca2# exchange, since the
openness of the semicrystalline structure presents less
limitation to the diffusion of large hydrated ca-
tions (see Figure 6 and compare with Figure 4). Des-
pite the improvement in Mg2# exchange properties
relative to Ca2#, the performance of such zeolite
precursors is probably too poor for detergent
applications.
Materials with nonaluminosilicate frameworks
Zeolite-like structures composed partially or
wholly of oxides other than those of Al and Si such
as silicoaluminophosphates (SAPOs), metal alumino-
phosphates (MeAPOs), stannosilicates, zincosilicates,
titanosilicates and beryllophosphates are expected
to possess ion exchange properties, although few
data exist in the literature. Of these materials,
the titanosilicates have received the most attention.
Recently, the titanosilicate TAM-5 has been de-
veloped; this exhibits high selectivity for Cs# in
the presence of high concentrations of other alkali
cations and over a pH range from below 1 to above
14. Also, high selectivity of this material for Sr2#
in basic media has been observed. These high
selectivities, and its stability to solutions covering
this pH range, has led to commercialization of
1589
the material by UOP as Ionsiv IE-910 (powder) and
Ionsiv IE-911 (granules) for use in nuclear waste
treatment.
Particularly interesting ion exchange properties are
shown by materials possessing high electric Reld
strengths, which may arise with frameworks com-
posed of oxides of elements with valencies differ-
ing from each other by more than one unit. An
example is the beryllophosphate Na8[(BeO2)8(PO2)8]
) 5H2O, which has the same structure as the alumino-
silicate zeolite gismondine (or synthetic zeolite P).
Beryllium and phosphorus are strictly alternating in
the structure and have valencies of #2 and #5
respectively, giving rise to a framework with alternat-
ing !2 and #1 nominal charges (on Be and P), as
opposed to !1 and 0 for Al and Si in the aluminosili-
cate analogue. Due to the high electric Reld gradient,
hard cations tend to be favoured over soft ones. Thus,
magnesium is favoured kinetically over calcium; the
diffusion coefRcient for exchange of Mg2#
into Na8[(BeO2)8(PO2)8] ) 5H2O is more than three
times higher than that of Ca2# under the same condi-
tions (Figure 7), which is a reversal of the situation
seen in the aluminosilicate zeolites (compare Fig-
ures 7 and 4). The relatively slow kinetics of ex-
change may be attributed to the small window size of
the beryllophosphate material (the beryllophosphate
unit cell is smaller than the aluminosilicate one).
Univalent cations also exhibit unusual exchange char-
acteristics with Na8[(BeO2)8(PO2)8] ) 5H2O, due in
part to the relatively short Be}O and P}O bonds and
the rigidity of the structure. High resistance is experi-
enced by ingoing cations and large hysteresis loops
are seen in, for instance, the exchange of K# for
Na#, while the same reactions in the aluminosilicate
analogue do not exhibit hysteresis (compare
1590 II / ION EXCHANGE / Inorganic Ion Exchangers

Figure 7 Kinetics of exchange of Ca2# and Mg2# for 2Na# in Na8[(BeO2)8(PO2)8] ) 5H2O. Circles, Ca2# exchange; triangles, Mg2#
exchange. Data were determined at 253C, pH 10 and at a solution concentration of 0.05 mol equiv. L\1. Interdiffusion coefficients (D):
D(Ca)"2.0;10\18 m2 s\1; D(Mg)"6.5;10\18 m2 s\1. (Reproduced with permission from Coker EN and Rees LVC (1992) Ion
exchange in beryllophosphate G. Part 2. Ion exchange kinetics. Journal of the Chemical Society, Faraday Transactions 88: 273}276.)

Figures 8 and 9). Hysteresis occurs when the two
end-members of exchange (in this case, the pure
K and Na forms) are mutually immiscible, and form
separate phases which can usually be differenti-
ated by X-ray diffraction. The two phases will be
present simultaneously over a range of cation com-
positions (in intermediate Na/K forms), depending on
the degree of immiscibility of the two end-members.
Solid-state ion exchange in zeolites The exchange of
cations from one solid to another, probably mediated
by the presence of small quantities of water, is refer-
red to as solid-state ion exchange. This is a technique
which is useful for the preparation of catalysts, that is,
the introduction of cations which are only sparingly
soluble, or which processess hydration spheres which
are too large to allow easy diffusion into the

Figure 8 Isotherm for K#/Na# exchange in Na8[(BeO2)8
(PO2)8] ) 5H2O. Circles, forward exchange; triangles, reverse ex-
change. Data were determined at 253C, pH 10 and at a solution
concentration of 0.05 mol L\1. (Reproduced with permission from
Coker EN and Rees LVC (1992) Ion exchange in beryllophos-
phate G. Part 1. Ion exchange equilibria. Journal of the Chemical
Society, Faraday Transactions 88: 263}272.)
Figure 9 Isotherm for K#/Na# exchange in zeolite P. Circles,
forward exchange; triangles, reverse exchange; Ks, cation frac-
tion in solution; Kz, cation fraction in the solid. Data were deter-
mined at 253C and at a solution concentration of 0.1 mol L\1.
(Reproduced with permission from Barrer RM and Munday BM
(1971) Cation exchange reactions of zeolite NaP. Journal of the
Chemical Society A 2909}2914.)

cavities of the zeolite from solution. The technique
may involve thermal treatment (at temperatures up to
5003C) of an intimate mixture of the zeolite and the
salt containing the cation to be exchanged (or another
zeolite) although, in some instances, exchange has
been observed to occur under ambient conditions.
Another advantage of the solid-state approach to
preparing catalysts is the avoidance of generating
large quantities of waste exchange solution.
Clays and Other Layered Materials
Clays are one of the most abundant materials present
on the earth’s surface. They constitute a large com-
ponent of soil, while many ceramic and building
materials as well as industrial adsorbents and cata-
lysts contain clay. Soils owe their ability to sustain
plant life largely to clays which have the ability to
exchange ions with their surroundings. Clays are typ-
ically composed of sheets of linked SiO4 tetrahedra,
which are connected to Al(OH)6 octahedra. If one
sheet of silica interacts with a plane of Al(OH)6, then
a two-tier sheet (Al2Si2O5(OH)4) typical of kaolinite is
obtained. If the octahedral plane is sandwiched be-
tween two silica sheets, then a three-tier sheet is
obtained (Al2Si4O10(OH)2), as found in the smectite
and mica clays. The sheets are bonded to one another
via covalent bonds between the silica and alumina
sheets to yield a layer. It is how these layers
stack together (via electrostatic and van der Waals
forces only) which give clays many of their interesting
properties, and gives a large degree of Sexibility to
the structures. Clay-like materials may be composed
of oxides of elements other than silicon and
aluminium.
Table 3 Examples of layered materials
Layer charge
Neutral (no intrinsic ion exchange capability)a
Positive (anion exchange properties)
Negative (cation exchange properties)
a
Neutral layered materials may undergo a type of ion exchange reaction via redox intercalation, whereby a neutral species is
intercalated, followed by a transfer of electrons between the layer and the guest species. Thus both the layer and the intercalated
species become charged.
II / ION EXCHANGE / Inorganic Ion Exchangers 1591
The three principal types of clay } single-layer,
nonexpandable double-layer and expandable double-
layer } have been introduced by Dyer. Clays may
be either cationic (exhibiting cation exchange
properties) or anionic (anion exchangers). The
former type is more common, accounting for the
majority of naturally occurring clays; typical exam-
ples are montmorillonite and bentonite. Anionic
clays, such as hydrotalcite, occur rarely in nature, but
may be synthesized in the laboratory. Layered mater-
ials composed of neutral layers also exist, although
they possess little or no intrinsic ion exchange capa-
bility. Table 3 lists some common types of layered
material possessing cationic, anionic and neutral
layers.
Pillared clays Expandable cationic clays may be
converted into pillared clays by exchanging some or
all of their charge-balancing cations with bulky inor-
ganic species such as [Al13O4(OH)24(H2O)12]7# or
[Zr4(OH)14(H2O)10]2# and then calcining the com-
posites to dehydrate and dehydroxylate the pillaring
species, leaving hydroxy/oxide pillars. An interesting
pillaring process is that involving ion exchange with
a cationic ‘templating’ agent (cetyltrimethylam-
monium), followed by the synthesis of a mesoporous
silica phase around the template cations. The resultant
materials, in which the clay layers are propped apart
by the mesoporous silica, possess surface areas up to
800 m2 g\1 and interlayer spacings of 3.3}3.9 nm.
For layered materials with anion exchange proper-
ties, like layered double hydroxides, species such as
[V10O28]6\ and [H2W12O40]6\ may be exchanged
with anions residing between the layers to increase
the interlayer spacing.
Example
TaS2
MoO3
Layered double hydroxides:
[MII1 xMIII (OH)2]x#[Xnx/n]x\ ) zH2O
\ x
Hydroxy double salts:
[MII(1 x)MII’(1#x)(OH)3(1 y)](1#3y)#[Xn(1#3y)/n](1#3y)\ ) zH2O
\ \ 2
(Xn\"Cl\, NO\ 3 , SO4\, CO3\, H5C2O\, etc.)
2
Smectite clays (low charge density)
Micas
MIVH-phosphates (high charge density, e.g.
-ZrP, -ZrP)
Layered titanates
Silicic acids
1592 II / ION EXCHANGE / Inorganic Ion Exchangers
While pillared clays usually offer advantages
over normal clays in terms of their higher surface
areas, higher sorptive capacities and greater ion
exchange capacities, these properties begin to be
diminished when the density of pillars becomes too
great and the interlayer space becomes Rlled with
pillars. Pillared clays are seldom employed as ion
exchangers; their main applications lie in the Relds of
catalysis and adsorption.
Metal Phosphates
The most important and widespread of the
metal phosphates is
-zirconium phosphate
(Zr(HPO4)2 ) H2O, or
-ZrP), which has an expand-
able layer structure. Each layer possesses a central
plane of octahedral Zr atoms linked to two outer
sheets of monohydrogen phosphate groups. The hy-
drogen form has an interlayer spacing of 0.76 nm,
corresponding to a void space with diameter 0.26 nm.
Although the calculated surface area of
-ZrP ap-
proaches 1000 m2 g\1, in the unexpanded H form the
surface area available to N2 is only 5 m2 g\1.
Another crystalline form of zirconium phosphate
-ZrP (Zr(PO4)(H2PO4) ) 2H2O), is formed by a cen-
tral zirconium phosphate sheet in which the PO4
groups are linked solely to octahedral Zr atoms; this
sheet is linked to dihydrogen phosphate groups to
yield the -ZrP structure. The complex interlinking
results in a more rigid framework in which only c.
50% of the theoretical ion exchange capacity is nor-
mally obtained.
Swelling of zirconium phosphates The interlayer
cavities in
-ZrP of 0.26 nm are accessible to only
small and poorly hydrated cations. A certain degree
of expansion of the interlayer distance may occur
concomitantly with these exchanges. Larger or more
strongly hydrated ions do not readily exchange with

-ZrP. However, since the layers are held together
principally by electrostatic forces, the distance be-
tween them can be increased to allow access of larger
ions according to the following mechanism.
The acid form of an
-ZrP possesses H# cations
which stabilize the negative charge on the Zr(PO4)2
units. A number of these protons may be neutralized
by addition of hydroxide ions via the solution phase.
This causes negative charge to build up on the layers,
causing electrostatic repulsion and forcing the layers
apart. Once the material has swelled, access to the
exchange sites by larger and more strongly hydrated
cations is possible. This view may be slightly oversim-
pliRed, since migrating OH\ ions would naturally be
accompanied by cations (to preserve electroneutrality
in both the solid and solution phases). It is more likely
that the above two-step process actually occurs
as a one-step process driven by the neutralization
reaction.
‘Catalytic’ exchanges in
-ZrP The interlayer spac-
ing of
-ZrP may be too small to allow large cations
access (a situation anomalous to ion-sieving in
zeolites). For instance, the Mg2# ion will not ex-
change with the protons in
-ZrP directly. However,
in the presence of sodium, some magnesium exchange
does occur. The process is shown conceptually below.
The hydrated Mg2# ion is too bulky to reach the
exchange sites between the layers of the acid form,
while the smaller hydrated Na# ion is not. The par-
tial exchange of Na# for H# causes a swelling of the
interlayer spacing to a point which allows the hy-
drated Mg2# to exchange.
Heteropolyoxometalates
Heteropolyoxometalates, or heteropolyacids (HPAs)
and their salts are materials which are Rnding wide-
spread applications as acidic and/or redox catalysts.
The most common examples are those with the
Keggin structure, composed of a central hetero spe-
cies, typically PO34\ or SiO44\, surrounded by 12
transition metal oxide octahedra, typically MoO6 or
WO6, as depicted in Figure 10. The octahedra and
central hetero species are linked via shared oxygens to
yield materials with the formula [XM12O40]n\ where
X"P (n"3) or Si (n"4) and M"Mo or W. Many
other structure types are known, with up to 40
transition metal octahedra per molecule. The nega-
tive charge is balanced by protons in an HPA and by
certain cations in HPA salts. The charge-balancing
cations are in many cases partially or wholly ex-
changeable, and physical properties such as solubil-
ity, surface area and porosity may vary widely de-
pending on the nature of the cation (Table 4).
Heteropolyoxometalates are principally used as
catalysts. Due to the high solubility of many of the
cationic forms of heteropolyoxometalates in aqueous
media, their application as ion exchangers has been
limited. Apart from ammonium phosphomolybdate
and ammonium phosphotungstate which possess low
solubility and have been used to scavenge radioactive
caesium, and [NaP5W30O110]14\, which has been
shown to have high selectivity for lanthanide and
certain multivalent ions, comparatively few data are
 II / ION EXCHANGE / Inorganic Ion Exchangers 1593

present as a necessity to terminate the structure

Figure 10 The structure of [XM12O40]n\ where X (P or Si) is
located at the centre and is surrounded by 12 metal oxide oc-
tahedra. (Reproduced with permission from Klemperer WG and
Wall CG (1998) Polyoxoanion chemistry moves towards the fu-
ture: from solids and solutions to surfaces. Chemical Reviews 98:
297}306.)
available concerning the ion exchange properties of
the HPAs.
Hydrous Oxides
Hydrous oxides are amorphous metal oxides, on
the surface of which exist hydroxyl groups which are
(see Figure 1A). The general formula for a hydrous
oxide is [M(n)O(n x)/2(OH)x ) wH2O]m, where the cen-
\
tral cation, M, is n-valent (n is typically *3). Most
of the metals in the periodic table are able to form
hydrous oxides which exhibit ion exchange proper-
ties. However, for the material to be applied as an ion
exchanger, it must be stable under the conditions
used for exchange. In particular, solubility can be
a deciding factor in the utility of hydrous oxides;
stability to pHs extending from strongly alkaline to
strongly acidic may be necessary. Those hydrous ox-
ides comprised of large, low valent cations or small,
multivalent cations tend to be soluble, while those
intermediate between the two extremes are stable.
Typical examples of acid- and alkali-stable hydrous
oxides are those of AlIII, GaIII, InIII, SiIV, SnIV, TiIV, ThIV,
ZrIV, NbV, BiV, MoVI and WVI. Many of the materials
are amphoteric, that is, they can act as either cation or
anion exchangers depending on, principally, the pH of
the electrolyte solution and the basicity of the metal
forming the hydrous oxide (the strength of the
metal}oxygen bond relative to the oxygen}hydrogen
bond).
The change of a commercial alumina from cation
exchanger to anion exchanger with varying pH is
shown in the chapter by Dyer (Figure 8). The am-
photeric nature of hydrous oxides may be illustrated
schematically thus:
Cation exchange M}O}H P M}O\ # H#
Anion exchange M}O}H P M# # \O}H

Table 4 Changes in surface properties of phosphomolybdates and phosphotungstates upon ion exchange

Approximate composition Surface area by N2 BET (m 2 g\1)b Pore volume ;103 (cm 3 g\1) Mean pore radius (nm)
of HPA salt a

HPMo, NaPMo, Essentially nonporous

(MeNH3)PMo
(NH4)PMo 193 52 1.3
KPMo 40 15 0.9
CsPMo 145 6 1.4
HPW, NaPW, AgPW, Essentially nonporous
(MeNH3)PW, (Me4N)PW
(NH4)PW 128 50 1.0
KPW 90 31 0.9
CsPW 163 34 1.4
HSiW, NaSiW, KSiW Essentially nonporous
(NH4)SiW 117 40 1.0
CsSiW 150 52 1.0
RbSiW 116 40 1.0

a
PMo, PW and SiW represent (PMo12O40)3\, (PW12O40)3\ and (SiW12O40)4\ respectively. The charge-balancing cation indicated is
assumed to be fully exchanged into the HPA, although some variation of composition is inevitable. Note that the surface properties will
vary slightly depending upon the preparation and exact composition of the HPA.
b
Surface area determined using the Brunauer, Emmett and Teller isotherm approach.
1594 II / ION EXCHANGE / Inorganic Ion Exchangers
Cation exchange typically takes place in alkaline
solution, while anion exchange is preferred in acidic
solution. Dissociation of M}O}H near to its isoelec-
tric point allows both exchange mechanisms to oper-
ate simultaneously.
Silica, the most common and extensively studied of
the hydrous oxides, is a weakly acidic cation ex-
changer. The physical properties of silica, particularly
the porosity and surface area, vary widely depending
upon the method of preparation. Generally, multi-
valent cations interact more strongly with the silica
surface than do univalent ones, while in all cases the
interactions are relatively weak and ion exchange is
facile. Silica possesses between 0.5 and 0.8 hydroxyl
groups per nm2 on its surface.
Miscellaneous Materials
A number of speciRc materials have been discussed in
this chapter. There are, however, numerous inorganic
materials possessing ion exchange properties which
have not been mentioned. In this section, a few of
those materials which exhibit interesting ion ex-
change properties are introduced brieSy. The list is
far from complete, but serves to illustrate the diver-
sity of ion exchange materials.
E Hydroxyapatites may undergo limited ion ex-
change reactions. While the calcium form
(Ca10(PO4)6(OH)2) is the most common (it is a ma-
jor component of teeth and bones), pure exchange
end-members of Sr2#, Cd2# and Pb2# are known,
while various cations may form intermediate
mixed-cation phases. The Sr2# end-member, due
to a slight lattice expansion, possesses superior ion
exchange properties compared to Ca-hydroxyapa-
tite. Of the Sr-hydroxyapatites, that with a (non-
stoichiometric) Sr/P ratio of 1.73 has the highest
ion exchange capacity of those measured. It is
interesting that the presence of HCl may assist the
ion exchange reaction by formation of a chlorapa-
tite phase. This may be an example of simultaneous
anion and cation exchange.
E Copper hexacyanoferrates, CuII2FeII(CN)6 ) xH2O
and related compounds show quite promising ex-
change properties for Cs#, and have been investi-
gated as agents for nuclear waste treatment. On
passing caesium-containing waste through a column
of CuII2FeII(CN)6 ) xH2O at room temperature, de-
contamination factors (ratios of pre-column to post-
column Cs# concentrations) of 103 can be achieved.
E Lithium manganate containing mixed-valence
manganese ions exhibits unusual ion exchange
properties, in that it undergoes combined ion ex-
change and redox reactions. Upon acid treatment
of LiMnIIIMnIVO4, the MnIII is oxidized to MnIV
and Li is displaced from the structure thus:
4 LiMnIIIMnIVO4#8 H#
#
P 3 MnIV
2 O4#4 Li #2 Mn
2#
#4 H2O
The resulting spinel structure (-MnO2) is highly
selective for Li, and will readily re-insert Li# to
regain the Li-manganate spinel:
2 O4#(n)LiOH P Li(n)Mn(n)Mn(2 n)O4
MnIV III IV
\
#(n/2)H2O#(n/4)O2
This type of exchange reaction is often referred to
as the ion memory effect.
E Iodide ions may be efRciently exchanged for
nitrate ion using BiPbO2NO3 in solutions of
pH*13. Under such conditions, the theoretical
exchange capacity of 2 mmoL g\1 is approached.
Conclusions
As with any commercial venture, improvements to
large scale ion exchange processes will always be
sought. With the advances made in structural charac-
terization and synthetic methods, it is becoming in-
creasingly possible to tailor the ion exchange proper-
ties of materials to speciRc needs. Thus, the strive for
water-softening zeolites for detergents with greater
capacity, selectivity and rate of exchange for Ca2#
and Mg2#, or for exchangers with better stability
over wide pH ranges coupled with high selectivity for
certain ions present in waste streams will be ever-
present. Recent advances have made some signiRcant
steps in these particular directions:
E The Reld of nuclear waste clean-up has spawned
a number of interesting materials; inorganic ex-
changers are now available which have good struc-
tural stability in waste streams and exhibit high
selectivities for Cs# and Sr2# in the presence of
large excesses of other ions over wide pH ranges.
E Zeolites continue to be used in vast quantities as
water softeners in detergents. A signiRcant recent
development has been the introduction of a new
detergent zeolite MAP, which offers improved
performance over zeolite A.
Interesting ion exchange properties are exhibited
by framework materials possessing high electric Reld
gradients, such as the beryllophosphates. However,
this particular area is deserving of more extensive
exploration.
The prediction of ion exchange behaviour for
a particular material is possible given data
for exchange reactions in that material under
 II / ION EXCHANGE / Novel Layered Materials: Phosphates
different conditions. However, the prediction of
ion exchange properties on the basis of the structure
of the exchanger alone may become more readily
possible through the use of computer modelling.
The study of ion exchange behaviour under the
inSuence of microwave radiation is an area which
preliminary research has suggested may be interest-
ing.
See also: II/Ion Exchange: Historical Development;
Novel Layered Materials: Non-Phosphates; Organic Ion
Exchangers; Theory of Ion Exchange.
Further Reading
ClearReld A (ed.) (1982) Inorganic Ion Exchange Mater-
ials. Boca Raton, FL: CRC Press.
Multispecies Ion Exchange Equilibria
See II / ION EXCHANGE / Surface Complexation Theory: Multispecies Ion Exchange
Equilibria
Non-Phosphates: Novel Layered Materials
See II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
Novel Layered Materials: Phosphates
U. Costantino, Università di Perugia, Perugia,
Italy
Copyright ^ 2000 Academic Press
It has long been known that many polyvalent cations
can be precipitated as amorphous phosphates from
dilute solutions and these salts are useful in gravimet-
ric analysis. More recently it has been recognized that
many of these precipitates contain exchangeable acid
protons and behave as inorganic ion exchangers.
Phosphates of tetravalent metals such as Zr(IV),
Ti(IV) and Sn(IV) have been found to possess high
ion-exchange capacity and good stability in acid and
oxidizing solutions and when exposed to high tem-
peratures and ionizing radiation. Because of these
properties, their potential uses for the puriRcation of
nuclear reactor cooling water or for the treatment of
radioactive waste were investigated during the late
1595
Dyer A, Hudson MJ and Williams PA (eds) (1993) Ion
Exchange Processes: Advances and Applications. Cam-
bridge, UK: Royal Society of Chemistry.
Dyer A, Hudson MJ and Williams PA (eds) (1997) Progress
in Ion Exchange: Advances and Applications. Cam-
bridge, UK: Royal Society of Chemistry.
Greig JA (ed.) (1996) Ion Exchange Developments and
Applications. Cambridge, UK: Royal Society of Chemistry.
Helfferich F (1962) Ion Exchange. New York, USA:
McGraw-Hill.
Slater MJ (ed.) (1992) Ion Exchange Advances. London,
UK: Elsevier Applied Science.
van Bekkum H, Flanigen EM, Jacobs PA and Jansen JC
(eds) (2000) Introduction to Zeolite Science and Prac-
tice, 2nd edn. Amsterdam: Elsevier.
Williams PA and Hudson MJ (eds) (1990) Recent Develop-
ments in Ion Exchange 2. London, UK: Elsevier Applied
Science.
1950s and early 1960s, especially in nuclear centres.
The ion-exchange properties of amorphous zirco-
nium, titanium and tin phosphates were reviewed by
Amphlett in 1964. However, the beginning of the
chemistry of layered phosphates may be dated back
to 1964, when ClearReld and Stynes reSuxed zirco-
nium phosphate gel in phosphoric acid solutions in an
attempt to produce a material which was more resis-
tant to hydrolytic attack than the original gel. The
microcrystals obtained were found to possess
a layered structure, called the
-type, and with the
composition Zr(HPO4)2 ) H2O. This compound was
indeed more resistant to hydrolytic attack than the
amorphous analogue. It possesses two exchangeable
protons per formula weight and is an excellent inter-
calating agent of protophilic species and a pure solid-
state protonic conductor. Moreover, it is possible to
correlate the observed properties with the structural
1596 II / ION EXCHANGE / Novel Layered Materials: Phosphates

features. These Rndings stimulated research on the
synthesis of layered phosphates of other polyvalent
metals. The progress made up to 1982 was reviewed
by ClearReld, and by Alberti and Costantino, and
from that date the Reld of layered phosphates has
been continuously expanding with the discovery and
resolution of the structure of new crystalline phases.
It was found that zirconium (or titanium) phosphate
has an isomorphous modiRcation, named -type, and
the composition Zr(PO4) (H2PO4) ) 2H2O. This com-
pound, as well as having cation exchange and interca-
lation properties, undergoes a topotactic anion ex-
change reaction of the dihydrogenphosphate groups
with other anions. Most recently, the preparation of
a new crystalline layered phase, named -type, and
having the composition Zr(PO4)Cl(CH3)2SO, has
opened new research possibilities. This article deals
with the preparation, structure, ion exchange and
intercalation properties of layered phosphates and
phosphonates of polyvalent metals, mainly zirco-
nium, and with their application. Exfoliation of
layered phosphates which allows the preparation of
mixed layered phosphates or thin-layer coatings on
substrates such as silica and alumina or of micropor-
ous pillared layered phosphates will be described.
However, before discussing in more detail the
above-mentioned materials, it is worth commenting
brieSy on the preparation, ion exchange properties
and application of amorphous zirconium phosphate,
because of its commercial availability and renewal of
interest in its use in nuclear waste treatment.
Amorphous zirconium phosphate is easily prepared
by adding a solution of zirconium salts to a solution of
phosphoric acid in acid media (2d4 mol dm\3 HCl).
The precipitate can be appropriately treated to obtain
the exchanger in glassy, granular or powdered form.
The composition is best described by the formula
Zr(HPO4)2 (OH)2 nH2O, x ranging between 0 and
\V V
0.2. The material is stable up to 180oC (temperature at
which condensation of phosphates to pyrophosphates
starts) in acidic medium (e.g. 6 mol dm\3 HNO3), and
has a remarkable resistance to strong doses of ionizing
radiation. The ion exchange capacity ranges from 4 to
6 mequiv. g\1. At low loading, the exchanger prefers
cations with lower hydrated ionic radius and higher
charge. Its use for the selective removal of 137Cs and
89
Sr radioisotopes from aqueous nuclear wastes in
ultraRltration and Suidized bed systems has been pro-
posed. Amorphous zirconium phosphate, because of
its bio-compatibility and high insolubility, is used to
Rll cartridges for the removal of urea from blood in
haemodialysis machines.
Layered Phosphates of Groups 4
and 14, 5 and 15 elements
Preparation
Numerous layered phosphates of the elements of the
groups 4, 5, 14 and 15 of the periodic table have been
synthesized and many of them are listed in Table 1,
together with their interlayer distance, the free area

Table 1 Formulae and some properties of layered phosphates of groups 4, 14 and 5, 15 elements

Formula Density (g cm\3) Ion exchange Interlayer distance Free area

capacity (As ) (As 2)*
(mmol H#g\1)

-Ti(HPO4)2 ) H2O 2.61 7.76 7.56 21.6

-Zr(HPO4)2 ) H2O 2.72 6.64 7.56 24.0

-Hf(HPO4)2 ) H2O } 5.15 7.60 23.7
-Ti(PO4)(H2PO4) ) 2H2O 2.37 7.25 11.60 16.5
-Zr(PO4)(H2PO4) ) 2H2O 2.43 6.27 12.20 17.8

-Si(HPO4)2 } 8.90 7.4 }

-Ge(HPO4)2 ) H2O } 7.07 7.75 }

-Sn(HPO4)2 ) H2O 3.12 6.08 7.80 21.4

-Pb(HPO4)2 ) H2O } 4.79 7.95 21.5
ZrPO4Cl(CH3)2SO } } 10.2 }
VOPO4 ) 2H2O 2.4 } 7.41 38.5
VO(HPO4) 0.5H2O 2.8 5.81 5.70 35.7
NbOPO4 ) 3H2O } } 8.04 }
HNb(PO4)2 } 3.52 } }
HTa(PO4)2 ) 2H2O } 2.45 9.48 46.0
HAs(PO4)2 2.88 3.52 7.98 37.1
KSb(PO4)2 3.50 } 8.47 19.6
HSb(PO4)2 } } } }
SbOPO4 4.42 } 6.34 }

*Area associated to each }OH group on the plane.

 II / ION EXCHANGE / Novel Layered Materials: Phosphates
surrounding the surface phosphate groups, density
and calculated ion exchange capacity. It may be seen
that, except for carbon,
-type phosphates of all the
elements of groups 4 and 14 have been obtained.
They are prepared with procedures similar to those
used to obtain Zr(HPO4)2 ) H2O, that is, by reSuxing
the amorphous precipitates in phosphoric acid
(10}12 mol dm\3). An alternative procedure, espe-
cially used for Zr and Ti hydrogenphosphates, in-
volves direct precipitation from solutions containing
phosphoric acid and Zr (or Ti) Suoro-complexes. The
degree of crystallinity of the precipitates may be con-
trolled by modifying the velocity of removal of the
complexing agent, as gaseous HF. With this method,
crystals of millimetre dimensions have been obtained.
Note that only Zr(IV) and Ti(IV) can form phosphate
dihydrogenphosphates of -type. The preparation in-
volves the slow decomposition of Suoro-complexes
in an NH4H2PO4 solution. The precipitate, e.g.
[Zr(PO4)(NH4HPO4)], is then converted into its hy-
drogen form by treatment with HCl solution.
Tetravalent elements with large dimensions, such
as Ce(IV) and Th(IV) do not give rise to layered
phosphates of
- or -type. The acid phosphates of
these elements have been obtained in Rbrous form
suitable for the preparation of fully inorganic, self-
consistent papers, thin Rlms or membranes. The acid
phosphates of groups 5 and 15 elements have been
obtained by dissolving the oxides in concentrated
phosphoric acid and heating to 2703C. HSb(PO4)2
can be obtained by treating the potassium salt with
a strong acid solution. Group 5 elements also pro-
duce non-acid layered phosphates of formula XOPO4
(X"V, Nb, Ta) and structure similar to that of
Zr(PO4)Cl(CH3)2SO. Vanadyl phosphate is one of the
rare examples of a layered phosphate which has a low
electronic conductivity and is capable of redox inter-
calation reactions similar to those shown by graphite
or layered dichalcogenides. Generally speaking,
layered phosphates possess good chemical and ther-
mal stability. Layered Zr(HPO4)2 ) H2O is a very in-
soluble compound, stable even in highly concentrated
non-complexing acid solutions. The interlayer water
is lost after prolonged heating at 1103C while the
condensation water of monohydrogenphosphates to
pyrophosphates is lost at 450}5003C. Molybdenum
and some divalent cations such as Fe, Cd and Mn also
form layered phosphates but their physical and chem-
ical properties have not been investigated thoroughly.
Structural Aspects
Layered solids are molecular crystals formed by the
packing of giant planar macromolecules called layers
or lamellae. The bonds between the atoms present in
1597
the layer are strong, primarily covalent, while those
between the atoms of adjacent lamellae are weak,
essentially of the van der Waals type. Thus, layered
solids generally exhibit a high anisotropy in their
physical properties. The reactivity of layered solids is
shown by the intercalation reaction, that is, the re-
versible insertion of guest species into the interlayer
region without appreciable modiRcation of the struc-
ture of the lamellae which move apart to accommo-
date the guest species. Hence, the structural aspects
of a layered solid are closely connected with the
bidimensional structure of the layers. The Greek let-
ter preRx that often indicates a layered phosphate is
related to the layer structure. The structures of the
layered phosphates listed in Table 1 will be illustrated
with reference to the
-, - and -zirconium phos-
phates, but the phosphates of other elements have
similar structures. Geometrical considerations indi-
cate that bidimensional structures can be easily for-
med by concatenation through the vertices of MO6
octahedra (M being the polyvalent metal) of suitable
dimension, and of PO4 tetrahedra. In the present case
different concatenation gives rise to different
layer structures.
Crystals of
-Zr(HPO4)2 ) H2O are monoclinic with
a"9.060(2) A> , b"5.297(1) A> , c"15.14(3) A> , and
"101.71(2) A> , space group P21/n. The sequence of
two layers is shown in Figure 1. Each layer may be
described as the concatenation of ZrO6 octahedra
and O3POH tetrahedra. Note that each tetrahedron
bridges three different octahedra and these, in
turn, bridge six tetrahedra. The layer is a planar
macromolecule bearing acid P}OH groups on the
Figure 1 Computer-generated representation of the sequence
of two layers of
-Zr(HPO4)2 ) H2O. (Crystal data from Clearfield
A and Smith GD (1969) Inorganic Chemistry 8: 431I436.)
1598 II / ION EXCHANGE / Novel Layered Materials: Phosphates

Figure 2 Computer-generated representation of the sequence

of two layers of HSb(PO4)2. (Crystal data from Piffard Y, Oyetola
S, Courant S and Lachgar S (1985) Journal of Solid State Chem-
istry 60: 209I213.)

surfaces. The distance between adjacent phosphate

groups on one side of the layer is 5.3 A> and the ‘free
area’ around each P}OH group is 24 A> 2. The inter-
layer distance is 7.56 A> and the arrangement of the
pendant phosphate groups creates six-sided cavities,
each containing one water molecule, in the interlayer
region. This layered structure is common to the other

-layered phosphates and it is very similar to that of
HSb(PO4)2, shown in Figure 2.
The second layer structure, in which two dif-
ferent tetrahedral species are used at the same time,
is present in the -compound with formula
Zr(IV)(PO4)(H2PO4) ) 2H2O. The -layer consists of
two ideal planes containing zirconium atoms bonded
by tetrahedral PO4 and H2PO4 groups (see Figure 3).
The PO4 group shares all four oxygens with zirco-
nium atoms while the H2PO4 shares two oxygens
with two different Zr atoms and points the re-
maining two OH groups towards the interlayer re-
gion. The interlayer distance is 12.2 A> , and the free
area surrounding the P(OH)2 groups on the surface of
the layers is 35 A> 2.
A third layer structure of great interest can be
formed by bridging four different zirconium
atoms with a tetrahedral PO4 group in a slightly
different manner from -zirconium phosphate,
and then by balancing the residual positive charge
and completing the octahedral conRguration of each
zirconium atom with a monovalent anionic ligand,
Cl\ and a neutral monodentate ligand, (CH3)2SO, as
Figure 3 Computer-generated representation of the sequence
of two layers of -Zr(PO4)(H2PO4) ) 2H2O. (Crystal data from
Christensen A, Andersen EK, Andersen IGK, Alberti G, Nielsen
N and Lehman MS (1990) Acta Chemica Scandinavica 44:
865I872.)
illustrated in Figure 4. Note that this structure is
essentially the same as that of layered vanadyl phos-
phate (see Figure 5) and of uranyl phosphate.
Chemical Reactivity
Ion Exchange Properties
The protons of layered acid phosphates are able to
diffuse in the interlayer region and these com-
pounds behave as inorganic cation exchangers and
proton conductors. Mainly the ion-exchange proper-
ties of
-Zr(HPO4)2 ) H2O will be considered as these
have been investigated extensively. However, the
Rndings apply to the other members of the class. All
these compounds are solid acids and the simplest way
to completely replace the protons with other cations
is by titrating the microcrystals with solutions of
the hydroxide of the cation to be exchanged. The
 II / ION EXCHANGE / Novel Layered Materials: Phosphates
Figure 4 Computer-generated sequence of two layers of
Zr(PO4)Cl(CH3)SO. (Data from Alberti G, Bartocci M, Santarelli
M and Vivani R (1997) Inorganic Chemistry 36: 3574I3575.)
titration curves of
-Zr(HPO4)2 ) H2O with alkaline
metal hydroxides, in the presence of the correspond-
ing metal chloride, are shown in Figure 6.
It may be seen that the exchange process occurs
stepwise. In each plateau of the titration curve the
composition of the solution, and hence also the pH,
is constant. Since temperature and pressure are also
Figure 5 Computer-generated structure of the sequence of two
layers of VOPO4. (Crystal data from Tietze HR (1981) Australian
Journal of Chemistry 34: 2035I2038.)
1599
Figure 6 Potentiometric titration curves of
-Zr(HPO4)2 ) H2O
with the hydroxides of the indicated alkali metal ions, in the
presence of the corresponding metal chlorides. (Reproduced with
permission from Alberti G and Costantino U (1974) Journal of
Chromatography 102: 5I29. Copyright: Elsevier Science Publish-
ing, Amsterdam.)
constant, the phase rule requires the presence of two
solid phases. The X-ray diffraction patterns of
samples with increasing metal ion loading indeed
indicate the presence of two solid phases, one trans-
forming into the other as the exchange reaction pro-
ceeds. According to a model developed by Alberti, ion
exchange in the
-phases takes place by diffusion
of the cations from the external part of the layered
crystals towards the bulk with an advancing phase
boundary with the co-existence in the same crystallite
of two phases. To illustrate the model consider
H#/Na# exchange (see Figure 6 and the scheme in
Figure 7). Initially we observe the formation of
a phase of composition ZrHNa(PO4)2 ) 5H2O and
interlayer distance 11.8 A> , according to the reaction
(the number in parentheses refers to the interlayer
distance):
Zr(HPO4)2 ) H2O (7.6 A> )#Na##4H2OP
ZrHNa(PO4)2 ) 5H2O (11.8 A> )#H#
The composition of the exchanged phase does not
change until half the protons of the original hydrogen
1600 II / ION EXCHANGE / Novel Layered Materials: Phosphates

Figure 7 Schematic representation of the phases formed during the H#/Na# ion-exchange process in
-Zr(HPO4)2 ) H2O micro-
crystals.

form have been replaced by Na# ions. At the end of By using suitable precursors, a large number of
the process only the monosodium form is present and cations of the periodic table, as well as organic ca-
the pH of the solution rises to a value at which the tions or cationic complexes, have been intercalated
following reaction takes place: via ion exchange processes into zirconium phosphate
and other layered phosphates. Table 2 states the
ZrHNa(PO4)2 ) 5H2O(11.8 A> )#Na#P composition and interlayer distance of a selected
number of phases, some of them prepared for practi-
Zr(NaPO4)2 ) 3H2O(9.9 A> )#2H2O#H> cal applications. These layered phosphates possess
good thermal resistance and are stable even when
Examination of the ion-exchange process exposed to high doses of ionizing radiation.
Zr(HPO4)2 ) H2O#M##(n!1)H2OP Zirconium phosphate has been used to perform ion
exchange processes in molten salts at high temper-
ZrHM(PO4)2nH2O#H# atures. Figure 9 shows the Na#/K# forward and
reverse isotherms obtained in molten NaNO3}KNO3
where M is an alkaline metal, shows the following mixtures at 4503C.
selectivity sequence: K#'Na#'Li#Rb# Good resistance to radiation makes these phos-
Cs#, since potassium uptake occurs at pH 2 while phates particularly suitable for the uptake of danger-
H#/Cs# exchange occurs at pH 7. The differ- ous radionuclides such as 137Cs#, 89Sr2# and 60Co2#.
ent selectivity towards K# and Cs# is a direct conse-
quence of the structural features of the host. The
zeolitic cavities present in the interlayer region of

-Zr(HPO4)2 ) H2O, are interconnected by windows
of 2.64 A> width. Therefore, cations such as Rb# and
Cs# that have an ionic diameter greater than 2.64 A> ,
as well as highly hydrated divalent and trivalent ca-
tions, are not taken up unless energy is supplied to
spread the layers apart. Accordingly, a facile ex-
change of large monovalent ions or of highly hy-
drated divalent or trivalent cations takes place if
precursors with a high interlayer distance such as
polyhydrate zirconium phosphate, Zr(HPO4)2 ) 4H2O
(d"10.4 A> ) or the monosodium form ZrHNa
(PO4)2 ) 5H2O (d"11.8 A> ) or some intercalation
compounds with alkanols or amines (see below) are
employed. A study of the ion exchange isotherms of
ZrHNa(PO4)2 ) 5H2O with different monovalent
and divalent cations (see Figure 8) revealed the fol- Figure 8 Forward Na#/Ca2#, Na#/Cs# and Na#/Li# ion-ex-
lowing selectivity order Ba2#'Ca2#'Cs#'K# change isotherms on
-ZrNaH(PO4)2 ) 5H2O. Concentration:
'Mg2#'Na#'Li#. 0.1 equiv dm\3, temperature 253C.
 II / ION EXCHANGE / Novel Layered Materials: Phosphates 1601

Table 2 Formulae and interlayer distances of some anhydrous and hydrated salt forms of
-zirconium phosphate. Some exchanged
forms with cationic complexes are also listed

Compound Interlayer distance (As ) Compound Interlayer distance (As )

ZrHLi(PO4)2 ) 4H2O 10.1 Zr(UO2)0.9H0.2(PO4)2 ) 5H2O 10.5

Zr(LiPO4)2 ) 4H2O 10.0 ZrH0.4Mg0.8(PO4)2 ) 4H2O 9.8
Zr(LiPO4)2 7.05 ZrH0.4Mg0.8(PO4)2 7.9
ZrHNa(PO4)2 ) 5H2O 11.8 ZrBa(PO4)2 ) 2.5H2O 9.5
Zr(NaPO4)2 ) 3H2O 9.8 ZrMn(PO4)2 ) 4H2O 9.7
Zr(NaPO4)2 8.42 ZrCo(PO4)2 ) 4H2O 9.6
ZrHK(PO4)2 ) H2O 8.02 ZrNi(PO4)2 ) 4H2O 9.55
Zr(KPO4)2 ) 3H2O 10.7 ZrCu(PO4)2 ) 4H2O 9.6
Zr(KPO4)2 9.0 ZrZn(PO4)2 ) 4H2O 9.6
ZrHCs(PO4)2 ) 2H2O 11.3 Zr[Cr(NH3)6]0.25H1.25(PO4)2 10.8
Zr(CsPO4)2 ) 6H2O 14.2 Zr[Co(C5H5)2]0.5H1.5(PO4)2 12.0
Zr(CsPO4)2 9.5 Zr[Pt(NH3)4]0.5H(PO4)2 10.6
Zr(AgPO4)2 8.4 Zr[Cu(bpy)]0.5H(PO4)2 14.5
ZrZn(PO4)2 7.66 Zr[Cu(phen)]0.5H(PO4)2 15.8
ZrH0.5Cr0.5(PO4)2 ) 4H2O 11.6 Zr[Pd(dmp)]0.5H(PO4)2 17.3
ZrRh0.66(PO4)2 ) 4H2O 11.6 Zr[Fe(C5H5)2]0.2H1.8(PO4)2 11.6
Zr(VO)0.5H(PO4)2 ) 3H2O 9.75 Zr[Cu(NH3)4]0.6H0.8(PO4)2 9.6

bpy"bipyridyl; phen"phenantroline; dmp"dimethylphenantroline.

In addition, zirconium phosphates exchanged with
transition metal ions are heterogeneous catalysts and
supports for chromatographic separation. For the lat-
ter application it should be noted that layered acid
phosphates are usually obtained as small platelets
(&1
m) and very compact chromatographic col-
umns are usually obtained. The Sows are therefore
slow while some particles tend to be released into the
external solution. This problem may be overcome by
using larger particles, even though the rate of ex-
change decreases, or by supporting the layers on
a suitable support such as silica gel.
Topotactic anion exchange reactions We have seen
above that the majority of layered phosphates are
inorganic cation exchangers. Layered phosphates of
the -type show a typical reaction, which formally
represents an anion exchange process. A topotactic
exchange reaction is deRned as the replacement of
one group by another without alteration of the host
matrix. If we consider the structure of the -phases we
observe that the dihydrogenphosphate groups, pres-
ent on the surface of the lamellae, have a net charge of
!1, delocalized over two oxygen atoms. The H2PO\ 4
is weakly bonded to the central tetravalent atom (Zr
or Ti) and it may be easily replaced by other suitable
groups, when the layered phosphate is equilibrated
with a solution containing such groups. Topotactic
exchange reactions with phosphites, hypophosphites,
phosphonates or phosphinates, according to the gen-
eral reaction:

Zr(PO4)(H2PO4) ) 2H2O#RPO2OH\P
Zr(PO4)(RHPO3) ) nH2O#H2PO\
4 #(2!n)H2O

are particularly efRcient. R is an aliphatic or

Figure 9 Forward and reverse Na#/K# ion-exchange iso-
therms on layered
-zirconium phosphate in molten
NaNO3}KNO3 mixtures at 4503C. (Reproduced with permission
from Alberti G and Costantino U (1974) Journal of Chromatogra-
phy 102: 5I29. Copyright: Elsevier Publishing Science,
Amsterdam.)
aromatic organic moiety that may bear a functional
group. This simple procedure has allowed the prep-
aration of a large number of new layered phos-
phate}phosphonates of -type with very interesting
properties.
1602 II / ION EXCHANGE / Novel Layered Materials: Phosphates

Intercalation Properties Table 3 Interlayer distances and guest contents of intercalation

compounds of
-zirconium phosphate
Over time, research on layered phosphates has moved
from the study of their ion exchange properties to Guest molecule mol Guest/ Interlayer
that of their intercalation properties in great part mol
-ZrP distance (As )
determined by the presence of Br+nsted acid groups in
Methylamine 2.0 12.1
the interlayer region. Both
- and -zirconium phos- Ethylamine 2.0 14.8
phates are excellent intercalating agents of Lewis Propylamine 2.0 17.6
bases. The intercalation chemistry of the former has Penthylamine 2.0 21.5
been more widely investigated and we will be mainly Diethylamine 1.0 12.7
Dipropylamine 1.0 15.7
concerned with
-zirconium phosphate. Many mol-
Dioctylamine 0.8 26.8
ecules belonging to various classes of organic com- Aniline 2.0 18.4
pounds (alkanols, glycols, alkyl and aryl amines, p-Methoxyaniline 2.0 21.7
heterocyclic bases, aminoacids and dyes) have been Benzylamine 2.0 19.1
intercalated. The corresponding intercalation com- Benzylethylamine 2.0 22.4?
Ephedrine 2.0 22.0
pounds have been characterized for composition and
Histamine 1.9 20.5
arrangement of the guest molecules in the interlayer Pyridine 0.95 10.9
region. Table 3 gives the interlayer distance and com- Pyrazole 0.75 10.8
position of some typical examples of intercalation Imidazole 0.95 10.7
compounds. 3-Methylpyrazole 0.98 12.1
1-Methylimidazole 0.58 10.4
Let us examine in more detail the intercalation of
Benzimidazole 1.90 20.4
n-alkylamines that leads to the formation of com- Pyridazine 0.64 10.8
pounds containing two moles of guest per mole of Pyrimidine 0.71 11.1
host, according to the reaction Pyrazine 0.78 10.8
2,2-Bipyridyl 0.25 10.9
1,10-Phenantroline 0.5 13.6

-Zr(HPO4) ) H2O#2RNH2P
 2,9-Dimethylphenantroline 0.5 14.6

-Zr(HPO4) ) 2RNH2 ) H2O Ethanol 14.2
 1-Propanol 16.6
1-Butanol 18.7
where R is the n-alkyl-chain. The reaction proceeds 1-Octanol 26.7
stepwise with the formation of different phases. Isopropanol 15.6
At low amine loading we observe the formation of 2-Methyl-1-propanol 17.5
a phase with interlayer distance 10.4 A> and the alkyl- 3-Methyl-1-butanol 19.2
Benzyl alcohol 21.0
chain axis is almost parallel to the layer plane (see
Diethylene glycol 10.5
Figure 10A). At half intercalation, the alkyl chains are Acetone 9.9
arranged as a monolayer of extended molecules with Acetylacetone 13.5
the chain axes inclined by 553 with respect to the layer Acetonitrile 11.3
plane (see Figure 10B). At full intercalation n-al- Urea 0.9 9.9

-Alanine (DL) 0.5 12
kylamines give rise to compounds in which the inor-
Phenylalanine (DL) 1.7 23.2
ganic layer regularly alternates with a bilayer of al- Histidine (DL) 0.9 16.2
kylamines with the n-alkyl chain in trans}trans confor- Crystal violet 0.5 22
mation (see Figure 10C). The terminal }NH2 groups Rhodamine 0.66 24.7
are protonated by the hydrogenphosphate groups.
Intercalation compounds with
,-alkyldiamines
contain one mole of guest per formula weight. The Heterocyclic bases give rise to non-stoichiometric
guest molecules are arranged as a monolayer of ex- intercalation compounds and the heterocyclic ring is
tended chains and the terminal }NH2 groups interact positioned parallel to the layer plane. For the arrange-
with the P}OH groups belonging to two-faced layers. ment of other intercalated guests the reader is referred
Alkanols and glycols produce intercalation com- to recent reviews given in the Further Reading sec-
pounds whose composition and arrangement of guest tion. Materials with special properties have also been
species are similar to those found in alkyl mono- obtained by intercalation. Porphyrins and
amines and diamines, respectively. However, direct metalloporphyrins, thionine, methylene blue and
intercalation is prevented by the lower basicity of the rhodamine have been intercalated in
-zirconium
alkanol OH group, compared to that of the NH2 phosphate and the materials obtained have been in-
group. It is necessary to use as precursors pre-swelled vestigated for their optical properties. The possibility
zirconium phosphates. of intercalating dyes and of controlling, at least to
 II / ION EXCHANGE / Novel Layered Materials: Phosphates 1603

Figure 10 Arrangement of n-alkyl monoamines intercalated into
-Zr(HPO4)2 ) H2O: (A) alkyl-chain axis parallel to the layer plane.
(B) Monolayer of extended molecules in trans}trans conformation. (C) Bilayer of extended molecules in trans}trans
conformation.

some extent, molecular orientation is of interest in the conduction of the host and some of the compounds
preparation of new composite materials for non- obtained have been used as active components in
linear optic applications. Intercalation of weak solid-state electrochemical gas sensors. Molecular
Br+nsted bases was found to enhance the proton and chiral recognition properties have been induced
1604 II / ION EXCHANGE / Novel Layered Materials: Phosphates
in layered zirconium phosphate by the intercalation
of suitable receptors such as aminated -cyclodex-
trins, crown ethers or a Pirkle receptor.
Much attention is presently being paid to the
possibility of performing reactions in the interlayer
region. Polymerization, induced by chemical, thermal
or photochemical treatment, of pyrrole, aniline,
propargylamine or -aminocaproic acid intercalated
in layered phosphates, produces interesting com-
posite materials in which the inorganic layers regular-
ly alternate with the polymers formed in the inter-
layer region.
Exfoliation Process
We have seen that layered polyvalent metal phos-
phates are obtained as molecular crystals built up by
the packing of the layers which are planar macro-
molecules. These bidimensional macromolecules are
usually very thin (5}15 A> ), whereas planar dimen-
sions are of the order of
m2, depending on the
conditions of synthesis. If a layered crystal is ex-
foliated in single layers, materials with a very large
surface area and with enhanced reactivity are ob-
tained. For example, the complete exfoliation of 1 g
of
-zirconium phosphate will produce material with
a surface area of 950 m2. Furthermore, the suspension
of the layers may be used to obtain thin Rlms and
pellicles or to cover suitable supports.
It is well known that layered smectite clays undergo
so-called ‘inRnite swelling’, that is, they disintegrate
into single layers or packets of a few layers, when
suspended in water. This phenomenon has never been
observed in layered phosphates probably because of
stronger layer}layer interactions. However, inter-
calation has made it possible to exfoliate both
-
and -zirconium phosphates. In the case of
-
Zr(HPO4)2 ) H2O and of the other
-type layered
phosphates a good exfoliation has been obtained by
the intercalation of short-chain alkylamines, such as
methylamine or propylamine at 100% and 50%
loading, respectively. This exfoliation process is
shown schematically in Figure 11.
-Zr(PO4) ) (H2PO4) ) 2H2O is best exfoliated when
treated with dimethylamine. Colloidal dispersions
containing highly anisotropic particles of
nanoscale dimensions have a number of potential
applications. After treatment with acids, Socculation
allows the formation of completely inorganic pellicles
or Rlms useful in assembling the sensor layer of solid-
state gas sensors, or to cover glass surfaces for
chromatographic application. Composites of layered
phosphates and silica gels or pillared layered phos-
phates have also been prepared from colloidal disper-
sions.
Solid dispersions of layered phosphates in silica gel
Solid dispersions of
- or -zirconium phosphates in
porous silica can be prepared starting from mixtures
of a tetrapropylammonium oligosilicate solution and
zirconium phosphates, previously exfoliated with
amines. They are formed after geliRcation of the mix-
ture with acetic acid and subsequent calcination at
6503C to remove the organic moieties. At this tem-
perature zirconium phosphates are transformed into
layered pyrophosphates, but non-condensed phos-
phate groups are still present on the free surfaces of
the lamellae. Accordingly, the composites obtained
have a large surface area (350}500 m2 g\1), good
surface ion-exchange capacity and acid catalytic
properties. Such composites may Rnd application as
stationary phases in chromatography.
Pillared layered phosphates The success obtained in
the pillaring of clays to obtain microporous solids
with larger pore diameters than those found in
zeolites has stimulated research in preparing pillared
layered structures based on metal(IV) phosphates.
Synthetic strategy requires the insertion of large or-
ganic or inorganic cations (pillars) between the layers
to prop them apart. If the pillars are sufRciently
spaced, a microporous structure is obtained and the
dimensions of the channels or diffusion paths are
determined by the size of the pillars and their spacing
in the interlayer region (see Figure 12).
Inorganic pillars are preferable to organic pillars
because of their much higher thermal stability. To
obtain thermally stable structures, pillaring has been
performed with highly charged polyoxycations such
as the Al13 Keggin ion [Al13O4(OH)24(H2O)12]7#, or
[Zr(OH)2(H2O)4]8# 4 , or inorganic clusters such as
[Nb6Cl12]2#. After suitable thermal treatment, the
layered phosphates contain as pillars, aggregates of
inorganic oxides which have considerable thermal
stability. The problem of inserting such large pillars
has often been overcome by contacting the solution of
the pillaring species with colloidal dispersions contain-
ing single layers, or packets of a few layers, of tet-
ravalent metal(IV) phosphates. This provides access to
the surface POH groups, the exchange reaction and the
Socculation of the pillared material. However, the
problem of achieving uniform pillar spacing to obtain
a narrow distribution of micropores of predictable
dimensions has not been completely resolved. The
topic is of great interest since materials for molecular
sieving and for shape-selective catalysis might result.
Metal(IV) Phosphonates
A fundamental step in the development of the
chemistry of layered phosphates was made in 1978
 II / ION EXCHANGE / Novel Layered Materials: Phosphates 1605

Figure 11 Schematic representation of the exfoliation of
-Zr(HPO4)2 ) H2O microcrystals by intercalation of n-propylamine. The
formation of completely inorganic films or the coating of solid surfaces is also reported.

when the Rrst Zr(IV) phosphonates and Zr(IV)

Figure 12 Schematic representation of a pillared layered struc-
ture showing the microporosity and the diffusion paths.
organophosphates with formula Zr(RPO3)2 or
Zr(ROPO3)2 respectively, were prepared (R being an
organic group). These compounds are organic deriva-
tives of
-Zr(HPO4)2 ) H2O in which the }OH groups
attached to the phosphorous atoms have been
replaced by organic R groups, leaving the inorganic
structure of the
-layer essentially unchanged.
A further development in layered metal(IV) phos-
phates was achieved with the resolution of the struc-
ture of the -phases and with the discovery that it is
possible to replace interlayer dihydrogenphosphate
groups by monovalent phosphonate or phosphinate
anions by simple topotactic anion exchange reactions
(see above).
Nowadays, a large number of metal(IV) phosphon-
ates of
- and -type are known and many others can
be prepared for special purposes, constituting a very
large and versatile class of layered materials. A brief
account of preparation procedures, structural
1606 II / ION EXCHANGE / Novel Layered Materials: Phosphates
Figure 13 Computer-generated representation of the sequence of two layers of
-Zr(C6H5PO3)2. (Data from Alberti G, Costantino U,
Allulli S and Tomassini N (1978) Journal of Inorganic and Nuclear Chemistry 40: 1113I1117, with permission from Elsevier Science.)
features and the chemistry of
- and -zirconium
phosphonates is given.
Zirconium Phosphonates of
-Type
The preparation of Zr phosphonates is closely related
to the methods employed for the preparation of

-Zr(HPO4)2 ) H2O, i.e. reSuxing of amorphous
precipitates with solutions containing the chosen
phosphonic acids, and the direct precipitation
method in the presence of Zr Suorocomplexes and the
suitable H2O3PR acid.
As already mentioned, the layer structure arises
from the concatenation of ZrO6 octahedra and O3PR
tetrahedra similar to that present in
-zirconium
phosphate. Due to the short lateral distance between
adjacent O3P}R groups on each side of the
-layer
(5.3 A> ) interpenetration of the R-groups belonging to
adjacent layers cannot occur for steric reasons and
a double Rlm of R-groups is expected for all the
members of this class. Therefore these organic deriva-
tives have a layered structure similar to that of zirco-
nium benzenephosphonate (see Figure 13) or zirco-
nium carboxyethanphosphonate (see Figure 14), two
typical compounds of the class.
A list of selected
-layered phosphonates is given in
Table 4. Note that the compounds contain a variety
of functional groups. By choosing appropriate or-
ganic groups attached to the phosphorus atom, it is
possible to vary the acid properties of the phosphon-
ates from neutral (e.g. P}CH3) or weakly acid (e.g.
P}CH2COOH) to strongly acid (e.g. P}C6H4SO3H)
or even to basic (e.g. P}C2H4NH2), or to anchor the
amino acid chiral group. The nature of the covalently
 II / ION EXCHANGE / Novel Layered Materials: Phosphates 1607

Figure 14 Computer-generated representation of the sequence of two layers of
-Zr(HOOCCH2CH2PO3)2. (Data from Alberti G,
Costantino U, Casciola M, Vivani R and Peraio A (1991) Solid State Ionics 46: 61I68, with permission from Elsevier Science.)

attached groups depends on the imagination and abil- present on the surfaces of the layers of the parent
ity of the chemist to synthesize the appropriate phos-
-Zr(HPO4)2. However, more voluminous groups
phonic acids. may be attached to the
-layers if their dimensions are
The only limitation to synthesis is the use of or- compensated by introducing small groups R (R be-
ganic groups with a cross-section equal to or less than ing H, OH, CH3) to obtain compounds of formula
24 A> 2. This is the free area around each P}OH group Zr(RPO3)2 x(RPO3)x. These mixed component
\
phases are of great interest since a very special type of
complexing agent or redox couple may be Rxed to the
Table 4 Interlayer distances of some zirconium bis-monophos-
layers. A selection of the multicomponent phases pre-
phonates and organophosphates with
-layered structure pared to date is given in Table 5.
Zirconium diphosphonates, of general formula
Compound Interlayer distance (As ) Zr(O3P}R}PO3), in which adjacent inorganic layers of
the
-type are covalently joined to each other by
Zr(O3PCH3)2 8.9
Zr(O3PCH2OH)2H2O 10.1
divalent organic groups, may also be obtained. These
Zr(O3PCH2Cl)2 10.1 zirconium phosphates do not possess interlayer micro-
Zr(O3PCH2CN)2 10.8 porosity, because the distance between adjacent pillars
Zr(O3PC3H7)2 14.0 is 5.3 A> and the van der Waals diameter of the alkyl or
Zr(O3P(CH2)2COCl)2 13.5 aryl pillar is about 4 A> . It is however possible to create
Zr(O3PCH2COOH)2 11.3
Zr(O3P(CH2)2COOH)2 13.0
microporosity in the interlayer region if some
Zr(O3P(CH2)3COOH)2 15.0 pillars are replaced by small O3P}H groups, and if
Zr(O3PCH"CH2) 10.6 the pillar has been suitably designed. By using a pillar
Zr(O3PCH2SO3H)2 15.4 with bases, such as 3,3,5,5-tetramethylbiphenyl-
Zr[(O3PO)(CH2CH2O)nPO3] diphosphonic acid, a pillared compound exhibiting
Zr[(O3PO)(CH2CH2NH2)2] ) 2HCl 14.3
Zr[HOOCCH(NH2)CH2OPO3]2 14.5
a high phosphite percentage and interlayer micro-
porosity of 375 m2 g\1, has recently been prepared.
1608 II / ION EXCHANGE / Novel Layered Materials: Phosphates
Table 5 Compositions and interlayer distances of some deriva-
tives of
-zirconium phosphate with two different pendant groups
Compound Interlayer
distance (As )
Zr(O3POH)0.66(O3PH)1.34 6.5
Zr(O3POH)1.15(O3PC6H5)0.85 12.4
Zr(O3POH)(O3PC2H4COOH) 12.9
Zr(O3PCH2OH)(O3PH) 7.0
Zr(O3PC2H4COOH)1.25(O3PCH2OH)0.75 13.6
Zr(O3PC6H5)(O3PH) 10.5
Zr(O3PC6H4SO3H)0.85(O3PC2H5)1.15 ) 3.7H2O 18.5
Zr(O3PC6H4SO3H)0.97(O3PCH2OH)1.03 ) 4.9H2O 19.6
A computer-generated structural model of this micro-
porous pillared compound is shown in Figure 15.
Zirconium Phosphate Phosphonates of -Type
The structure of the -layer differs from that of
the
-layer since the ZrO6 octahedra are placed in two
different planes and joined to each other by PO4
tetrahedra. Due to the fact that only three oxygens
are available in phosphonate groups, pure -
zirconium phosphonates cannot exist. However, it is
possible to replace the interlayer H2PO4 groups by
monovalent phosphonate or phosphinate anions to
obtain layered inorganic}organic derivatives in which
the inorganic layer regularly alternates with organic
Figure 15 Computer-generated structural model of a microporous
-zirconium phosphite-diphosphonate. (Data from Alberti G,
Costantino U, Marmottini F, Vivani R and Zappelli P (1993) Angew. Chem. Int. Ed. Engl. 32: 1357I1359.)
layers. These materials are obtained from a simple
topotactic reaction by contacting the original zirco-
nium phosphate microcrystals with a solution of a suit-
able phosphonic acid. As we have already seen this
reaction is similar to an anion exchange process. The
texture of the -layer remains practically unchanged
and it is therefore possible to predict the arrangement
of the organic groups in the interlayer region by con-
sidering the interlayer distance and the dimension of
the groups. Figure 16 shows the probable structure of
-zirconium phosphate benzene-phosphonate.
Many organic derivatives have been prepared with
this simple procedure including pillared compounds
with regular interlayer porosity obtained by partial
replacement of the dihydrogenphosphates with bi-
valent diphosphonate groups. A selected number of
recently prepared compounds is reported in Table 6.
Monophosphonic or biphosphonic acids containing
crown ethers have also been used for the topotactic
reaction and compounds with crown ethers covalent-
ly attached to the inorganic layers have been ob-
tained. These materials show promise for interesting
applications in ionic or molecular recognition and
hence for performing selective separations. The -sys-
tem is thus very versatile and the interlayer region can
easily be engineered with a large variety of organic
groups to obtain materials for application in several
Relds including the preparation of new stationary
phases for chromatographic separation.
 II / ION EXCHANGE / Novel Layered Materials: Phosphates 1609

Figure 16 Computer-generated representation of the sequence of two layers of -zirconium phosphate-benzenephosphonate. (Data
from Alberti G, Vivani R, Biswas RK and Murcia-Mascaros S (1993) React. Polym. 19: 1I12, with permission from Elsevier Science.)

Conclusion a rich chemistry. Many of them are inorganic ion

exchangers that support the most common organic
The layered phosphates of polyvalent metals are ob- resins in processes which occur at high temperatures
tained with different layer structures and exhibit or in the presence of strong oxidizing solutions and

Table 6 Composition and interlayer distances of some organic derivatives of -zirconium phosphate obtained by topotactic exchange
reactions

Acid used Composition Interlayer distance (As )

H3PO3 ZrPO4O2PHOH ) 2H2O 12.2

H3PO2 ZrPO4O2PH2 ) H2O 8.8
H2O3PCH3 ZrPO4O2POHCH3 ) 2H2O 12.8
H2O3PC3H7 ZrPO4O2POHC3H7 ) 1.2H2O 15.1
HO2P(CH3)2 ZrPO4(H2PO4)0.33(O2P(CH3)2)0.67 ) H2O 10.3
H2O3PC6H5 ZrPO4(H2PO4)0.33(O2POHC6H5)0.67 ) 2H2O 15.4
H2O3P(C6H11) ZrPO4(H2PO4)0.33(O2POHC6H11)0.67 ) H2O 16.9
H2O3PC6H5 ZrPO4O2PHC6H5 15.1
H2O3PC10H21NO*
3 ZrPO4(H2PO4)0.71(C10H21NO3PO3)0.29 16.2

*N-(phosphonoethyl)aza crown; (12)crown-4.

1610 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
strong doses of ionizing radiation. Furthermore, these
inorganic ion exchangers possess a high ion exchange
capacity and some peculiar selectivities. Layered
phosphates are good intercalating agents of ionic or
polar species. This allows the construction in the
interlayer region of supramolecular assemblies with
special functionalities in the Relds of chromato-
graphic supports, chemical and electrochemical sen-
sors, ion exchange membranes, ionic and molecular
recognition and catalysts. The delamination of
layered phosphates has permitted the preparation of
thin Rlms and coatings and pillared layered structures
with accessible microporosity. There are many more
possibilities in layered phosphonate chemistry be-
cause functional groups may be inserted on alkyl
chains or on aryl rings. The Reld of layered phos-
phates and phosphonates is in continuous expansion
and these materials will Rnd many applications as
soon as their potential is realized.
See also: II/Ion Exchange: Catalysis: Organic Ion
Exchangers; Historical Development; Inorganic Ion
Exchangers; Novel Layered Materials: Non-Phosphates;
Organic Ion Exchangers; Theory of Ion Exchange.
Further Reading
Alberti G (1978) Syntheses, crystalline structure, and ion-
exchange properties of insoluble acid salts of tetravalent
metals and their salt forms. Accounts in Chemical
Research 11: 163d170.
Alberti G and Costantino U (1982) Intercalation chemistry
of acid salts of tetravalent metals with layered structure
and related materials. In: Whittingham MS and Jacob-
Novel Layered Materials: Non-Phosphates
R. Mokaya, University of Cambridge, Cambridge, UK
Copyright ^ 2000 Academic Press
Introduction
In this paper the structure and composition of layered
materials (excluding those which contain phosphates)
and their modiRed variants are described. Layered
materials are made up of sheets or planes of atoms
held together by interplanar forces which are weaker
than intraplanar binding forces. This structural set-up
allows the insertion of atomic or molecular guest
species between the layers. Such insertion (or inter-
calation) provides a means for controlled variation of
the physical and chemical properties of the host
son JA (eds) Intercalation Chemistry, Chapter 5,
pp. 147d180. New York: Academic Press.
Alberti G and Costantino U (1984) Recent progress in the
intercalation chemistry of layered a-zirconium phos-
phate and its derivatives, and future perspectives for
their use in catalysis. Journal of Molecular Catalysis
27: 235d250.
Alberti G and Costantino U (1991) Intercalates of zirconium
phosphates and phosphonates. In: Atwood JL, Davies
JED and MacNicol DD (eds) Inclusion Compounds, Vol.
5, Inorganic and Physical Aspects of Inclusion, Chapter
5, pp. 136d176. Oxford: Oxford University Press.
Alberti G and Bein T (eds) (1996) Solid State Supramolecu-
lar Chemistry: Two and Three Dimensional Networks,
Vol. 7, Comprehensive Supramolecular Chemistry,
Chapters 4 and 5, pp. 107d187. Oxford: Pergamon.
Alberti G, Casciola M, Costantino U and Vivani R (1996)
Layered and pillared metal(IV) phosphates and phos-
phonates. Advanced Materials 8: 291d303.
Alberti G, Bartocci M, Santarelli M and Vivani R (1997)
Zirconium phosphate chloride dimethyl sulfoxide, a re-
active precursor of a large family of layered compounds.
Inorganic Chemistry 36: 3574.
Amphlett CB (1964) Inorganic Ion Exchangers. Amster-
dam: Elsevier.
Cheetham AK and Day P (eds) (1992) Solid State Chem-
istry: Compounds, Chapter 6, pp. 182d223. Oxford:
Clarendon Press.
ClearReld A (ed.) (1982) Inorganic Ion Exchange Materials,
Chapters 1}3, pp. 1d132. Boca Raton, FL: CRC Press.
ClearReld A (1990) Layered phosphates, phosphites and
phosphonates of groups 4 and 14 metals. Comments in
Inorganic Chemistry 10: 89d128.
ClearReld A (1998) Metal phosphonate chemistry. In:
Karlin KD (ed.) Progress in Inorganic Chemistry,
Vol. 47. New York: John Wiley.
layered material over wide ranges to yield new vari-
ants of novel layered materials. The intercalated
layered materials are also described.
Layered materials may be broadly classiRed into
three groups according to the composition of their
layers and the forces that hold the layers together.
The interlayer forces determine the inherent ability of
he layers to resist distortions involving displacements
transverse to the layer planes.
1. Type I layered materials are made up of layers of
atomically thin sheets. The neutral layers are held
together by van der Waals forces. Examples are
graphite and boron nitride. In graphite the layers
tend to be ‘Soppy’ and are easily separated with
respect to distortions transverse to the layer
 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
Figure 1 Schematic classification of layered solids (di"inter-
layer distance).
planes. Graphite is however rigid against longitu-
dinal in-plane distortions.
2. Type II layered materials, such as dichalcogenides
and lamellar oxyhalides, have layers composed of
a few (usually three) distinct planes of strongly
bonded atoms held together by van der Waals
forces.
3. Type III layered materials have layers made up of
dense (up to seven) assemblies of strongly bonded
atoms. The layers may be charged in which case
the interlayer forces are ionic resulting in layered
structures, such as silicate clays and layer double
hydroxides, which are quite rigid to interlayer
distortion or expansion.
Figure 1 gives a schematic illustration of the three
classes of layered materials.
In all cases the intralayer forces are much stronger
than the interlayer forces and therefore guest species
can be inserted into the interlayer region between the
host layers without any change to the layers them-
selves. This attractive feature of layered materials has
been extensively exploited. Indeed, the bidimensional
character of many layered materials can be gradually
modiRed by intercalation, grafting, or pillaring with
a variety of guest species to yield new classes of novel
layered materials. Type I intercalation materials, such
as those of graphite, form stages in which n-multi-
layers of the host are separated by monolayers of
guest intercalant to form expanded n-stage materials.
Type II materials are able to accept guest species into
random interlayer sites and may ultimately form
a saturated stage-1 expanded material at sufR-
cient guest species concentrations. In contrast type III
materials always form intercalation compounds with
a stage 1 stacking sequence in which the host layers
are separated by one or more layers of the guest
1611
species. Novel layered materials that have found use
in separation processes are mainly (intercalated or
otherwise modiRed) type III materials and this paper
is therefore devoted to such materials with only
a brief mention of type I and type II intercalated
materials (using graphite and dichalcogenides as
examples) given below.
Type I: Graphite
Graphite is known to be intercalated by both electron
donors and acceptors and to a large extent the driving
force for intercalation is electronic in nature. Thus
depending on the guest species, positively charged
carbon layers or negatively charged carbon layers
may be obtained. Graphite intercalation compounds
usually exhibit a high degree of ordering and are
unique among layered host materials in that the inter-
calation occurs such that, depending on the extent of
guest species incorporation, it is possible to observe
the staging phenomenon (Figure 2). The staging phe-
nomenon is deRned by a periodic arrangement of
n graphite layers (where n is the stage index) between
sequential intercalant layers. Well staged graphite
intercalated materials can be prepared up to n&10.
Strong interatomic intercalant}intercalant binding
relative to the intercalant}graphite binding favours
a close-packed in-plane intercalant arrangement and
is the driving force for the staging phenomenon.
Graphite intercalation compounds (GICs) have found
use as catalysts, electric conductors, recording mater-
ials (in inks and coloured leads) and as lubricants and
low friction coatings.
Type II: Dichalcogenides
Dichalcogenides, sometimes denoted TX2, have
layers made up of a sheet of metal atoms (T)
Figure 2 The staging phenomenon as exhibited by graphite
(for stage 3, di"carbon interlayer distance, dic"intercalate dis-
tance and dr"repeat basal distance).
1612 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
Figure 3 Schematic illustration of intercalant concentration
dependent intercalation of type II layered materials.
sandwiched between two sheets of chalcogen
(X) atoms. T is usually a transition metal and X may
be S, Se or Te. The layers are largely neutral and
separated by a van der Waals gap. As mentioned
above, dichalcogenides are able to accept guest spe-
cies into random interlayer sites and may ultimately
form a saturated stage 1 expanded material at suf-
Rcient guest species concentrations (Figure 3).
During intercalation, the guest species are inserted
in the van der Waals gap and in most cases occupy
interstitial sites. The intercalation is generally
accompanied by charge transfer between the interca-
lant species and the host layers and therefore interca-
lation complexes are formed with electron donor
species only. Such species include alkali metal atoms,
transition metal atoms and organic molecules. The
intercalation of metal atoms (especially alkali metal
atoms) results in efRcient transfer of electrons to
the host compound resulting in unique electronic
properties. For metal atom intercalation the increase
in layer separation is not large but the weak host layer
interactions are replaced by strong Coulomb
(alkali metal) and covalent (transition metal)
interactions yielding a quasi three-dimensional solid.
The intercalation of organic molecules results in
much larger layer separations. An example is the
intercalation of amines in which the orientation of the
amines in the van der Waals gap depends on the
number of carbon atoms. Short chain amines, such as
methylamine, pack parallel to the layers whereas in-
termediate chain amines (e.g. C4}C9) orient at an
angle to the layers with the nitrogen with its lone pair
of electrons adjacent to the layer. The angle of incli-
nation generally increases with chain length and for
chain lengths'C16, the amines are arranged perpen-
dicular to the host layers and form bilayers resulting
in layer separations as high as 57 A> for stage 1 inter-
calation.
Type III: Layer Silicates (Clays) and
Layer Double Hydroxides (LDHs)
Nature of Layer Silicates (Clay Minerals)
Clays are by deRnition Rne grained solids with particle
size generally (2
m and many of their properties
result from their small particle size. The layers of clays
are formed by condensation of sheets of linked
Si(O,OH)4 tetrahedra with those of linked M2}3(OH)4
octahedra, where M is a divalent or trivalent cation.
A 1 : 1 condensation gives two sheet minerals such
as kaolinite with a general layer formula of
M2}3Si2O5(OH)4. A 2 : 1 condensation results in the
octahedral sheet being sandwiched between two sheets
of tetrahedra giving the mica type layer structure with
a layer formula of M2}3Si4O10(OH)2. In both cases the
tetrahedral sheets are linked in the unit structure to
octahedral sheets and to groups of coordinated cations
or individual cations. The apical oxygen at the fourth
corner of the tetrahedron, which is directed normal or
nearly normal to the sheet, forms part of an immedi-
ately adjacent octahedral sheet in which octahedra are
linked by sharing edges. The junction plane between
tetrahedral and octahedral sheets consists of the shared
apical oxygens of the tetrahedra and unshared OH
groups that lie in projection at the centre of each six-
fold ring of tetrahedra. Figure 4 shows a three-dimen-
sional schematic illustration of layer silicates. Also pos-
sible, for example in chlorite, are four sheet clays in
which the trimorphic units alternate with M(OH)2}3
sheets of octahedrally coordinated M2# or M3# ions.
Smectite clays, which exhibit the property of inter-
calation, are made up of negatively charged layers
and therefore possess an ion exchange capacity which
distinguishes them from the mica and pyrophillite-
talc groups of minerals (see below). The layer charge
arises generally from isomorphous substitution of
Si4# by Al3# in the tetrahedral sheet and/or Al3# by
Mg2#, Fe2# in the octahedral sheet. Some charge
may also arise from broken bonds at edges of the clay
crystal. Following below are ideal structural formulae
of some clay silicates showing, where appropriate,
isomorphous substitution:
1. Dioctahedral smectites (two-thirds of octahedral
sites are occupied by trivalent cations)
(a) Pyrophyllite
[(Si8)(Al4)O20(OH)4] No layer
charge.
(b) Montmorillonite
[Mx(Si8)[Al4 xMgx]O20(OH)4 ) nH2O] Octahedral
\
substitution.
(c) Beidellite
[MxSi8 xAlx(Al4)O20(OH)4 ) nH2O] Tetrahedral
\
substitution.
 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates 1613

Pillared Clays
Pillared clays are usually smectite clay minerals in
which the interlayer cations are three-dimensional
species which in some cases, after appropriate treat-
ment, are Rxed to the layers of the host clay. The
shape and size of these cations allows them to func-
tion as molecular pillars which keep the layers apart
at a Rxed distance. The pillaring phenomenon there-
fore exposes much of the intercrystal basal surfaces
for adsorption and molecular sieving purposes. Per-
manent porosity may be introduced in montmorillon-
ite by replacing the interlayer alkali or alkaline earth
cation with a variety of species such as tetraalkylam-
monium ions, tris-metal chelates, bicyclic amine ca-
tions and polymeric oxymetal cations. Clays pillared
by oxycations or metal oxides are of greatest interest
because they exhibit thermal stability in excess of
5003C and, depending on preparation methods,
materials with large pore diameters and surface area
(Table 1).
The most extensively studied pillared clays are
those containing polymeric hydroxy-aluminium spe-
cies as the pillaring cation. In this paper such Al
pillared clays are used to illustrate the nature and
properties generally possessed by metal oxide pillared
clays. In the non-calcined so-called precursor pillared
clay, layer charge is balanced by the pillaring polyca-
tions which in the case of Al pillared clays is the
Keggin-like [Al13O4(OH)24(H2O)12]7# ion. On calci-
nation this ion is converted into an oxide with the
layer charge balanced by the release of an equivalent
number of protons, i.e.
Figure 4 Three-dimensional illustration of the structure of sili-
cate clays.
2[Al13O4(OH)24(H2O)12]7#P13Al2O3#14H#
pillar
#41H2O
2. Trioctahedral smectites (all octahedral sites are
occupied by divalent cations) The formation of pillars Rxed to the layers of the
(a) Talc host clay is dependent on the calcination temperature.
[(Si8)(Mg6)O20(OH)4] No layer In general the basal (0 0 1) spacing of the precursor-
charge. Al pillared clay decreases to a Rxed value upon
(b) Hectorite
[Mx(Si8)(Mg6 xLix)O20(OH)4 ) nH2O] Octahedral
\ Table 1 Pillar type and corresponding basal spacing and sur-
substitution. face area for montmorillonite pillared clays
Smectite clays can intercalate other compounds in a Pillar type Basal spacing (As ) Surface area (m2 g\1)
three component system:
Alumina 18}19 250}400
1. Host layer with an overall negative layer Iron oxide 17}18 &280
charge. Chromia 19}21 350}400
2. Exchangeable intercalates (ions) which compen- Zirconia 18}22 250}300
sate for the overall negative charge. Titania 18}20; 25}29 300}350
3. Neutral molecules (e.g. water) which occur be- Silica 12}13; 16}20 40}200; 150}400
Silica/alumina 16}19 350}500
tween the layers and are associated with the inter- Titania/silica 38}40 250}400
layer cations and the layers.
1614 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
heating at 5003C. Heating to temperatures up to
4003C causes some contraction but does not
prevent re-expansion of the clay upon exposure to
moisture. However, the pillared clay obtained by
calcination at 5003C usually shows no tendency to
expand. This is because in the temperature range
400}5003C an irreversible contraction in the layer
spacing occurs, during which the pillars are held
within the host aluminosilicate sheets resulting in
cross-linked materials. Therefore the precursor pillar-
ing species dehydroxylates progressively on heating
to 4003C, releasing protons which migrate into the
clay structure and at 5003C condensation takes place
of terminal hydroxy groups present on the polymeric
ions with the lattice hydroxy groups on the clay. The
oxide pillars formed become linked directly via oxy-
gen to the aluminium and magnesium atoms in the
octahedral layer resulting in a rigid cross-linked struc-
ture resistant to expansion. These changes are illus-
trated in Figure 5.
The microstructure of pillared clays is controlled
by the wet chemistry of synthesis and, to a large
Figure 5 Schematic description of pillaring. In the case of an
alumina pillared clay prepared from Ca-montmorillonite,
d1"14.4 A> , d2"20.5 A> , d3"19.0 A> .
extent, the method used to dry the precursor pillared
clay. The basal spacing of the pillared clays depends
on the age of the pillaring reagent, the degree of
hydrolysis (polymerization) of the pillaring reagent,
the amount of reactants (i.e. Al/clay ratio) and the
temperature of pillaring. Pillaring of clays increases
their surface area from as low as 30 m2 g\1 to
500 m2 g\1 (Table 1) and generates a microporous
structure similar but less constrained than that of
zeolites. The volume created can be used for adsorp-
tion purposes; the adsorption characteristics are
known to vary with the method employed in drying
the pillared clay. Air-dried pillared clays are zeolite-
like products which cannot adsorb molecules of kin-
etic diameter 9.2 A> (e.g. 1,2,5-triethylbenzene) but
can adsorb molecules of kinetic diameter 6.0 A> .
Freeze-dried pillared clays can, however, adsorb ap-
preciable amounts of molecules with kinetic diameter
of 10.0 A> . Freeze-dried pillared clays therefore con-
tain a signiRcant fraction of pore openings '10.0 A>
whereas all the pore openings of air-dried pillared
clays are (9.0 A> . This is related to the mechanism of
layer aggregation during drying. The aggregation
may be face to face (for air-dried materials) or edge to
face and edge to edge layer contact for freeze-dried
materials. Air-dried pillared clays therefore exhibit
long range lamellar order with a regular and relative-
ly narrow pore size distribution while freeze-dried
pillared clays, on the other hand, exhibit less lamellar
order and a broad pore size range.
Metal oxide pillared clays in general tend to pos-
sess pores in both the micropore and mesopore size
range. The ratio of micropore to mesopore volume
largely depends on the interlayer spacing (pillar
height) and the interpillar distance. The interpillar
distance may be controlled by varying the ion ex-
change capacity of the host clay; this in turn deter-
mines the number of pillaring polycations required to
balance the host layer charge. A low exchange capa-
city favours a low pillar density and vice versa. The
interlayer spacing, on the other hand, may be control-
led by varying the pillar type. Figure 6 gives a dia-
grammatic representation of two common pillar
types and Table 1 gives some examples of pillar type
and basal spacing for montmorillonite clay.
The porosity of pillared clays may also be varied
by combining the pillaring process with other treat-
ments such as competitive ion exchange with mono-
cations or acid activation. Indeed acid activation
of clays (see below) prior to pillaring yields a
different class of materials, generally referred to
as pillared acid-activated clays, with quite distinct
properties.
An important characteristic of pillared clays (and
clays in general) which is in some cases crucial to their
 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
Figure 6 Diagrammatic illustration of polymeric hydroxy-Al
(A) and -Ti (B) pillaring cations.
use in separation processes is that they possess con-
siderable acidity and may be classiRed as solid acids.
For example non-calcined precursor-alumina pillared
clay possesses Br+nsted acidity which arises through
the following mechanisms:
1. Polarization of interlamellar water by initial ex-
changeable cations not replaced by the hydroxy-Al
polycations. This is especially the case if the initial
exchangeable cation is acidic.
2. The pillaring polymer may hydrolyse to release
protons, i.e.
[Al13O4(OH)24(H2O)12]7#
P[Al13O4(OH)24#x(H2O)12 x]7##xH#
\ nature the charge-balancing hydrated anion is usually
3. The OH groups of the clay lattice and the pillar
may also act as Br+nsted acid sites.
However, for a pillared clay calcined at temperatures
above 4003C, Br+nsted acidity is weaker than Lewis
acidity. This is due to the migration of protons from
the interlayer region into the layer structure where
they neutralize the negative layer charge thus remov-
ing some Br+nsted acid sites.
Acid activated clays When ‘activatable’ clay
minerals are treated in acid, their chemical composi-
tion and physical properties are altered. The activa-
tion process enhances properties already present in
the clay minerals and gives them certain desirable
properties with respect to their applicability as
adsorbents and catalysts. The clays of choice for
acid activation are non-swelling bentonites contain-
ing montmorillonite as the major component. In gen-
eral terms the acid activation of montmorillonites
proceeds via the removal of octahedral ions and any
isomorphously substituted tetrahedral ions. The
changes that take place in an idealized montmoril-
lonite with no isomorphous substitution may be
1615
expressed as:
H#


#
(Al4)(Si8)O20(OH)4#3H P(Al3)(Si8)O20(OH)2
#Al3##2H2O
2H#


#
(Al4)(Si8)O20(OH)4#6H P(Al2)(Si8)O20(OH)2
#2Al3##4H2O
Layer Double Hydroxides (LDH)
Natural layer double hydroxides (or hydrotalcite-like
compounds as they are sometimes called) are, unlike
clay minerals, relatively rare. Where they occur they
are associated with metamorphic rock formations or
saline deposits. The structure of LDHs is very similar
to that of brucite, Mg(OH)2, in which magnesium is
octahedrally surrounded by six oxygen atoms in the
form of hydroxide with the octahedral units extend-
ing to form inRnite sheets through edge sharing. If
some of the magnesium in the sheets is isomorphously
substituted by a higher charge cation such as Al3#,
the resulting Mg2#}Al3#}OH layer gains a positive
charge. Sorption of an equivalent amount of hydrated
anions occurs so as to maintain electrical neutrality; in
carbonate. The OH groups of the positively charged
brucite-like sheet are linked to the CO23\ groups either
directly (via OH}CO3}HO linkages) or via intermedi-
ate water (i.e. OH}H2O}CO3}HO). The interlayer
carbonate anions adopt an orientation parallel to the
layers, i.e. they lie Sat surrounded by loosely bound
water (Figure 7). The resulting natural LDH may exist
in either of two dimorphic forms, i.e. as a rhombohed-
ral hydrotalcite or a hexagonal manasseite.
LDHs may be described by the general formula
[M2# 3#
1 xMx (OH)2]
x#
[Xm\]x/m ) nH2O
\
where M represents a metal cation and X represents
an anion. M2# may be Mg2#, Fe2#, Co2#, Ni2#,
Zn2# and M3# may be Al3#, Cr3# or Fe3#.
M2#/M3# ratios between 1 and 5 are possible but are
typically 0.254x40.33 and 04n46. Syntheti-
cally there is a wide range of variables such
as: (i) different combinations of M2# and M3#;
(ii) different charge balancing anions; (iii) dif-
ferent amounts of interlayer water; and (iv) crystal
morphology and size. To form LDHs, the M2# and
M3# cations must be of a size that can be contained in
1616 II / ION EXCHANGE / Novel Layered Materials: Non-Phosphates
Figure 7 Illustration of the top view of LDH
(OH)16CO3 ) 4H2O) lattice.
the holes (octahedral sites) between the close-packed
OH groups in the brucite-like layers. This limits the
possibilities to cations of ionic size between 0.5 and
0.8 A> and, in the main, excludes cations such as
Be2# (0.35 A> ), Ca2# (0.99 A> ) and Cd2# (0.97 A> ).
The formation of LDHs is not, however, limited
to M2#/M3# cations; it is, for example, possible to
incorporate monovalent cations (M#) such as Li# in
a Li/Al material, or to have divalent/tetravalent
materials such as Co/Ti.
The number of exchangeable anions in LDHs de-
pends on the charge density on the host layers. How-
ever there are no particular restrictions on the nature
of the anion. Inorganic charge-balancing anions in-
clude Cl\, OH\, NO23\, ClO\ 4 and SO4\. Organic
2
acids such as adipic, succinic, oxalic, malonic, sebacic
and terephthalic may also serve as charge-balancing
species. However, as mentioned above, nature fa-
vours the carbonate ion which is tenaciously held in
the interlayer region due to its relatively high polariz-
ability and synthesis of pure LDHs with other anions
requires special preparation procedures (see below).
LDHs may undergo swelling in a manner not unlike
that of silicate clays. For example sulfate-containing
LDH may be solvated with glycol or glycerol. In
general swelling of LDHs depends on the nature of
the interlayer anion (charge, mass, structure), nature
of the solvent (polarity, molecular dimensions) and of
course the layer charge.
Pillared Layer Double Hydroxides
Pillared LDHs which possess empty interlayer/inter-
pillar space are desirable but unlike pillared clays
are difRcult to prepare. The difRculty is large-
ly due to the afRnity of the layers for the carbon-
ate anion; if CO2 is present during synthesis, the
carbonate is preferentially incorporated and once in
the interlayer it is held tenaciously and not easily
replaced. Most of the pillaring strategies employ
a CO2-free environment and make use of the fact that
Cl\ or NO23\ anions are easier to displace. Thus the
Cl\ or NO23\ LDH is prepared, usually under nitro-
gen, and these anions are then replaced with larger
polyoxometalate anions such as, for example,
V10O628\, Ta6O18OH7\, Nb6O18OH7\. Another ap-
proach has relied on the use of LDH initially syn-
thesized with large intercalated organic anions, for
example the terephthalate dianion (Figure 8) as the
interlayer species. The organic anion is then displaced
by the polyoxometalate species. As in clays, the pillar-
ing of LDHs results in an increase in surface area and
(Mg6Al pore volume. The increases are however lower than in
pillared clays. This is due to the high layer charge
in LDHs which leads to a high pillar density which in
some cases yields materials in which the pillars are
‘stuffed’ into the LDH and do not exist as iso-
lated discrete pillars. An example is polyvanadate-
intercalated LDH which has a surface area of
ca. 35 m2 g\1 compared to 25 m2 g\1 for the unpil-
lared material. True pillaring does occur as in the case
of Zn2Al[-SiV3W9O40] which exhibits a surface area
of 155 m2 g\1.
Applications of Pillared (or Intercalated) Layered
Solids
The applications of expanded layered solids (LDHs
or clays) are largely due to their large surface area and
variation in their chemical and physical properties.
These properties may be enhanced by the ability to
Figure 8 Illustration of terephthalate intercalated LDH.

tailor them for speciRc uses. In general these materials
have found use as catalysts, ion exchangers and ad-
sorbents and are also useful in gas and liquid separ-
ation processes (where they exhibit molecular sieving
properties similar to those of zeolites). Some exam-
ples are:
1. Organoclays containing molecules such as
(CH3)4N# are especially suited for certain separ-
ation processes due to their hydrophobic nature and
high afRnity for certain organic compounds.
2. Clays and their oxide-pillared derivatives have
found use as: (i) scavengers for hazardous organics
(especially from efSuent streams); (ii) selective
adsorbents of heavy metals from efSuent
streams; and (iii) puriRers for edible oils where the
clays adsorb compounds such as carotenoids and
chlorophyll to give the oil its clear look and taste.
Indeed acid activated clays are the industry stan-
dard for the decolorizing of oil.
3. LDHs have found use as excellent acid residue
scavengers.
Greater use of layered materials in separation pro-
cesses can be achieved when the materials are used in
the form of membranes where they act as ionic and
molecular Rlters or sieves.
See also: II/Ion Exchange: Historical Development;
Novel Layered Materials: Phosphates; Organic Ion Ex-
changers; Organic Membranes.
Organic Ion Exchangers
C. Luca, ‘Petra Poni’ Institute of
Macromolecular Chemistry, Lasi, Romania
Copyright ^ 2000 Academic Press
Abstract
The deRnition and some characteristic concepts regard-
ing organic ion exchangers are pointed out. The devel-
opment of these ion exchangers, beginning with chemic-
ally modiRed natural products and continuing with the
synthetic ones, is further presented. A classiRcation of
organic ion exchangers is proposed according to sev-
eral criteria, such as the synthesis method, morpho-
logy of the three-dimensional network, their physical
shape and the nature of their functional groups.
Of the general characteristics of organic ion ex-
changers only the exchange capacity and selectivity
are brieSy discussed.
II / ION EXCHANGE / Organic Ion Exchangers 1617
Further Reading
Barrer RM (1978) Zeolites and Clay Minerals as Sorbents
and Molecular Sieves. London: Academic Press.
Bein T (ed.) (1992) Supramolecular Architecture: Synthetic
Control in Thin Films and Solids. ACS Symposia Series,
vol. 499.
Dresselhaus MS (ed.) (1986) Intercalation in Layered
Materials. New York: Plenum Press.
Dresselhaus MS, Dresselhaus G, Fischer JE and Moran MJ
(eds) (1983) Intercalated Graphite. MRS Symposia Pro-
ceedings, vol. 20.
LeH vy F (ed.) (1976) Crystallography and Crystal Chemistry
of Materials with Layered Structures. Dordrecht: Reidel.
LeH vy F (ed.) (1976) Structural Chemistry of Layer-type
Phases. Dordrecht: Reidel.
LeH vy FA (ed.) (1979) Intercalated Layered Materials.
Dordrecht: Reidel.
Lieth RMA (ed.) (1977) Preparation and Crystal Growth of
Materials with Layered Structures. Dordrecht: Reidel.
Mitchell IV (ed.) (1990) Pillared Layered Structures: Cur-
rent Trends and Applications. London: Elsevier.
Newman ACD (ed.) (1990) Chemistry of Clays and Clay
Minerals, 2nd edn. London: Longman and Mineralogi-
cal Society.
Birch R (1988) Pillared clays. Catalysis Today 2.
Sequeira CAC and Hudson MJ (eds) (1992) Multifunctional
Mesoporous Inorganic Solids. Dordrecht: Kluwer.
Whittingham MS and Jacobson AJ (eds) (1982) Intercala-
tion Chemistry. London: Academic Press.
Yamagishi A, Amarita A and Taniguchi M (eds) (1998) The
Latest Frontiers of the Clay Chemistry. Sendai: Smectite
Forum of Japan.
Preparation and structure } chemical property rela-
tionships, and some applications with reference to
strong and weak cation and anion exchangers, as well
as to chelating ion exchangers } are described in more
depth.
De\nition
An ion exchanger generally is a solid, insoluble ma-
terial that contains groups which ionize in aqueous
medium.
Organic ion exchangers are three-dimensional
covalent networks that contain exchangeable ions
associated with Rxed acid or basic groups. The term
‘ion exchange resins’ is also used to describe organic
ion exchangers.
The ion exchangers that have Rxed acid groups and
carry exchangeable cations (usually H# or Na#) are
cation exchangers described as in the H form and Na
1618 II / ION EXCHANGE / Organic Ion Exchangers
form, respectively. Those with Rxed base groups
and exchangeable anions (OH\ or Cl\) are
anion exchangers in the OH form and Cl form, re-
spectively.
In the accepted terminology, the three-dimensional
network with the Rxed groups is called the matrix or
framework and exchangeable ions of opposite sign,
which neutralize the Rxed ionic groups, are the
counter-ions that are responsible for the ion exchange
process.
Co-ions are mobile ions having the same sign as the
Rxed charges of the matrix. In fact, organic ion ex-
changers are crosslinked polyelectrolytes. Thus, a ca-
tion exchanger is an anionic polyelectrolyte while an
anion exchanger can be regarded as a cationic poly-
electrolyte.
General Aspects
The Rrst organic ion exchanger that found technical
application was a chemically modiRed natural prod-
uct, namely a sulfonated coal, described in many
patents during the 1930s.
Other exchangers were synthesized by sulfonation
or phosphorylation of wood, paper, cotton, lignin
and tannins, as well as by the crosslinking of pectins
with formaldehyde or epichlorhydrin.
In 1935 the discovery by Adams and Holmes of ion
exchange properties in the product of a reaction be-
tween phenol, or m-phenylenediamine, with formal-
dehyde started the development of synthetic organic
ion exchangers. These products have a greater im-
portance than those from a natural organic source
and have found much wider technical application
because of their greater chemical stability and mech-
anical strength as well as their very different
physical and chemical structures.
Synthetic organic ion exchangers are obtained by
the two principal reactions used to produce polymeric
materials, namely polycondensation or addition
polymerization of a mixture of co-monomers. In poly-
condensation, incorporation of a trifunctional co-
monomer is required while in polymerization the pres-
ence of a bifunctional co-monomer is sufRcient.
Most commercially available ion exchangers
are from polymerization processes which create struc-
tures with higher hydrolytic and oxidative stabilities
as well as better deRned physical features and cross-
linkings.
In the case of the polycondensation exchangers, the
reaction between a co-monomer that carries base or
acid groups and a crosslinking agent (formaldehyde,
epichlorohydrine, etc.) is used.
In 1944, D’Alelio found that sulfonated styrene}
divinylbenzene copolymers have ion exchange
properties. This Rnding was the beginning of the pol-
ymerization ion exchangers.
These structures are made by the polymerization of
a mixture of a monovinylic monomer with a basic or
acidic group and a divinylic monomer. The achieve-
ment of a neutral network, called the precursor or
starting material, followed by the introduction of basic
or acidic groups by suitable polymer-analogous reac-
tions, is often preferred.
Usually divinylbenzene (DVB) is used as the
divinylic monomer and the quantity added, in terms of
the percentage in the mixture of co-monomers, deRnes
the degree of crosslinking of the network, although
crosslinking side reactions can occur during the poly-
mer-analogous transformations.
The structures created are called ‘conventional’ or
‘gel’-type ion exchangers and generally have about 8%
DVB for crosslinking. This amount is required to
achieve a network with both mechanical strength
and easy diffusion of exchangeable ions as the
exchanger comes into contact with an aqueous phase
when swelling of the network occurs.
Meitzner and Oline found that the copolymerization
of styrene with DVB in the presence of an appropriate
inert compound, called ‘diluent’ or ‘porogene agent’,
gave a network with signiRcant and measurable phys-
ical porosity in the dried state, generally containing
internal pores having diameters larger than 3;10\9 m.
This discovery led to signiRcant progress in the Reld of
synthetic ion exchangers, namely the development of
macroporous resins. These exchangers offer the ad-
vantage that they can be used with non-aqueous solvents
and have much higher sorption rates of ions and non-
electrolytes than the conventional gel exchangers.
Polymerization produces exchangers in bead form,
with a relatively wide distribution of size, by the
suspension polymerization technique. More recently
ion exchangers with uniform and controlled bead size
have become available.
The polycondensation exchangers often appear as
irregular-shaped particles, because they are made by
bulk polycondensation followed by grinding of the
bulk polymer into smaller particles. However, poly-
condensation exchangers can also be made in bead
form by reverse-phase suspension polycondensation.
Ion exchangers in Rbre form are also known,
made by chemical modiRcation of natural and syn-
thetic Rbres. Ion exchanger Rbres have an improved
kinetic performance when compared with the same
structures in bead form.
Classi\cation
Scheme 1 is a summary of the classiRcation of
organic ion exchangers. Table 1 shows the most used

Scheme 1 Classification of organic ion exchangers.
acid and base functional groups on organic ion ex-
changers.
Characterization
Ion exchange capacity is the most appropriate charac-
teristic of organic ion exchangers.
The total capacity indicates the number of Rxed
acidic or basic groups per speciRed amount of ion
exchanger. It can be described as both weight capa-
city and volume capacity, having as units mil-
liequivalents per gram of dry exchanger (meq g\1)
and milliequivalents per cubic centimetre of fully
swollen exchanger (meq cm\3), respectively. If not
II / ION EXCHANGE / Organic Ion Exchangers 1619
otherwise stated, the capacity should be reported per
gram of H form for a cation exchanger or the Cl form
for an anion exchanger in the dry state. This capacity
is a constant for the material and does not depend on
the experimental conditions.
The effective capacity is the number of ex-
changeable counter-ions per speciRed amount of ex-
changer (the same units are used as above). This
capacity depends on the experimental conditions and
is lower than total capacity.
Another important characteristic is the selectivity
which has a major role in the ion exchange processes.
The selectivity is the preference of an ion exchanger
for a particular counter-ion over the others, when it is
1620 II / ION EXCHANGE / Organic Ion Exchangers

Table 1 Types of ion exchangers and their functional groups

Type Name of fixed functional group Chemical structure of functional group

Cation exchangers
Strong acid Aryl sulfonic }C6H5}SO3H
Weak acid Carboxylic acid }COOH
Phenolic hydroxyl }C6H5}OH
Intermediate acid Phosphonic }P(O)(OH)2
Phosphonous }P(O)H(OH)
Phosphoric }O}P(O)(OH)2
Anion exchangers
Strong base Quaternary ammonium

Phosphonium

Sulfonium
Weak base Primary amine }NH2
Secondary amine }NHR
Tertiary amine }NR2
Amphoteric Mixture of acid
exchangers and base groups

in contact with an electrolyte solution. The selectivity
has various physical causes.
An ion exchanger tends to prefer a counter-ion with
higher valence, lower solvatation, higher polarizabil-
ity, stronger interactions with the Rxed groups or the
matrix, and less participation in complex formation
with the co-ions. The selectivity of an ion exchanger is
improved by increasing degree of crosslinking and by
decreasing solution concentration and temperature.
Types of Synthetic Organic
Ion Exchangers
Strong Acid Cation Exchangers
The most important strong acid cation exchangers
are those of arylsulfonic acid type.
Polycondensation structures of this type can be
obtained as follows:
1. By the sulfonation of a phenol followed by the
condensation of the sulfonated product with for-
maldehyde.
2. By the sulfonation of a preformed phenol}
formaldehyde three-dimensional network.
In the Rrst method, the addition of unsulfonated
phenol to provide the trifunctionality is essential. The
structures created are illustrated in Figure 1.
A method for the synthesis of sulfonated condensa-
tion exchangers in bead form has been developed
using organic solvents as dispersion media. This is an
alternative to the grinding of bulk polymers as pre-
viously mentioned.
Most commercially available strong acid cation
exchangers are those based on styrene}DVB
copolymers with different morphologies of their
three-dimensional networks. These products have
higher capacities and better durabilities than their
polycondensation predecessors. The common method
for the production of these structures consists in sul-
fonation of the styrene}DVB copolymers with sul-
fonation agents such as sulfuric acid, sulfur trioxide,
oleum or chlorosulfonic acid.
From the point of view of the mechanism, the
sulfonation is an electrophilic substitution into an
aromatic ring whereby the }SO3H group is attached
in the para-position and a double sulfonation is prob-
ably impossible because of steric hindrance due to the
polymer chain.
During the sulfonation reactions, crosslinking side
reactions take place independent of the sulfonating
agent, however chlorosulfonic acid apparently leads
to the most crosslinks.
Side crosslinks are due to the inter-chain sulfone
bridges that appear by reaction between the already
attached }SO3H groups and the unreacted aromatic
rings. Intra-chain sulfone bridges also can appear.
The chemical structure of a sulfonated styrene}DVB
copolymer is illustrated in Figure 2.
The pre-swelling with organic solvents of the
copolymer beads before sulfonation reduces the num-
ber of sulfone bridges.

Figure 1 Preparation of sulfonated phenol}formaldehyde cation exchangers.
Addition to the styrene}DVB mixture of small
amounts of a polar monomer, such as acrylonitrile,
vinylpyridine, etc., improves the physical properties
of the resultant ion exchanger } especially its resist-
ance to osmotic shock because of the more uniform
sulfonation reaction.
Sulfonations with the agents previously mentioned
show some differences. Thus, reaction with sul-
furic acid used the acid itself as a reaction medium
hence a considerable excess of reagent is required.
Reactions with chlorosulfonic acid or sulfur trioxide
may be performed in an organic solvent, thus they
need only a small excess of reagent over the
stoichiometric quantities.
Sulfonations with the latter reagents take place at
lower temperatures than with sulfuric acid which
requires a temperature at about 1003C.
Figure 2 Chemical structure of sulfonated styrene}DVB copolymer-based cation exchanger.
II / ION EXCHANGE / Organic Ion Exchangers 1621
Post-sulfonation treatment of the sulfonated prod-
ucts is important to maintain whole beads. This can
be achieved by the prevention of the changes that
determine swelling, called ‘osmotic shock’, which
leads to the disintegration of the beads. The gradual
addition of water, or aqueous electrolyte solutions,
decreases this deleterious effect.
Macroporous copolymer beads, because of their
large internal surface areas, have a higher reactivity
towards the sulfonation agents. They also require
much lower quantities of organic swelling solvent,
and are less susceptible to degradation by osmotic
shock, both during preparation and in subsequent
usage. In addition, they have a higher oxidation stab-
ility than the sulfonated structures of the gel type.
Strong acid cation gel-type exchangers have re-
ceived major attention because of their utility in
1622 II / ION EXCHANGE / Organic Ion Exchangers
water softening which is their principal use. The
equivalent macroporous structures can also be used
as catalysts for certain reactions, particularly in non-
aqueous media, instead of sulfuric and toluene-4-sul-
fonic acids. The resin catalysts show some advantages
compared to low molecular weight acids, such as in
their regeneration and potential reuse.
Because of their very high acidity, the aryl}SO3H
groups are fully ionized throughout the pH domain of
aqueous solutions. The very low preference of the
sulfonic-type cation exchanger for the H ion requires
the use of large quantities of mineral acids for its
regeneration to the H form after the exhaustion cycle,
especially in water treatment processes.
The strong acid exchanger in its H form partici-
pates in ion exchange reactions with bases like NaOH
and with alkaline or neutral salts. The latter reaction
is called ‘salt-splitting’.
Weak Acid Cation Exchangers
The polycondensation exchangers can be prepared by
the reaction of salicylic acid, or 1,3,5-resorcylic acid,
with formaldehyde. In the former case, the addition
of phenol is required because one ortho-position is
not accessible to the aldehyde.
Several polymerization networks that contain
}COOH groups are known. Some structures together
with their preparative routes are illustrated in
Figure 3. Only those with acrylic networks are com-
mercially available.
Compared with the sulfonic group the }COOH
group, has a much lower acidity and is fully ionized
only in an alkaline medium as a salt form. The
}COOH group also shows a very considerable prefer-
ence for the H ion, unlike the }SO3H group. This
situation leads to easy regeneration of the weak acid
exchangers from salt form to H form using
stoichiometric quantities of mineral acids. These ex-
changers can react only with bases, like NaOH, and
alkaline salts; they show a strong preference for Ca
and Mg cations. The ‘salt-splitting’ reactions do not
take place in the case of the weak acid cation
exchangers.
The acrylic-type exchangers have a higher acidity
than the methacrylic ones and can be used for the
treatment of hard water containing large quantities of
bicarbonates. The methacrylic type is used for special
applications, such as the puriRcation of antibiotics,
where a mild pH is required.
Strong Base Anion Exchangers
Strong base anion exchangers are known only as
polymerization products. Those with quaternary am-
monium groups are the most common commercially
available exchangers. Their preparation is performed
by the chloromethylation of gel- or macroporous-
type styrene}DVB copolymers in bead form, followed
by the amination of the chloromethylated copolymers
with trimethylamine or dimethylethanolamine lead-
ing to the so-called strong base anion exchangers of
Types I and II, respectively. Their chemical structures
are shown in Figure 4.
Usually the chloromethylation is carried out with
monochloromethyl ether, in the presence of a Lewis
acid (ZnCl2, AlCl3, SnCl4, etc.) as catalyst. The reac-
tion takes place under mild conditions: temperature
about 503C and reaction times of 5}8 h. Generally,
the }CH2Cl groups are attached to over 90% of the
para-positions of the styrene aromatic rings, follow-
ing the chloromethylation of the mono alkylbenzene
derivatives.
The main chloromethylation reaction is usually
accompanied by a side alkylation reaction between
pre-attached }CH2Cl groups and non-functionalized
aromatic rings. Such a side reaction determines inter-
chain and/or intra-chain methylene bridges that de-
crease the amount of }CH2Cl groups as well as the
swelling capacity of the chloromethylated product.
The latter aspect is especially prevalent in the case of
gel-type copolymers. In most cases, the styrene}DVB
macroporous networks show a reduction of their
speciRc area and of the volume of their pores after
chloromethylation, but an increase of the average
diameter of the pores can be observed.
The use of a large excess of chloromethyl methyl
ether or mixtures of chloroform or carbon tetrachlor-
ide with the halogenated ether reduces the side
reaction.
An alternative route to obtain the chloromethylated
styrene}DVB network is via the free-radical polym-
erization of chloromethylstyrene (vinylbenzyl chlor-
ide) with divinylbenzene. The Rrst monomer is a
60 : 40 mixture of meta- : para-isomers.
The chemical structures of the two crosslinked
polystyrene-based chloromethylated compounds or
products are illustrated in Figures 5A and B. From
these two Rgures one can see that the chloromethyl-
styrene}DVB copolymer (Figure 5A) has a more
homogeneous chemical structure than the chloro-
methylated styrene}DVB copolymer (Figure 5B), but
the former structure has the drawback of a much
higher cost. For this reason chloromethylated
styrene}DVB copolymers are chosen as the precur-
sors to polystyrene-based anion exchangers.
Aminations of the chloromethylated styrene}DVB
copolymers with trimethylamine and dimethyl-
ethanolamine take place easily, because the benzylic
chlorine structure has a very high reactivity towards
these nucleophilic reagents. Amination is performed

Figure 3 Some methods for the preparation of weak acid cation exchangers.
in organic or aqueous media, at a temperature of
about 40}503C, and reaction times of 6}8 h. It must
also be mentioned that amination with the
two amines, in contrast to the chloromethylation
reaction, does not lead to crosslinking side reactions.
When the reactions are carried out in water, a side
reaction can occur at a very low level from the hy-
drolysis of a small number of }CH2Cl groups.
The chemical structures of strong base anion ex-
changers of Types I and II are not very stable in
alkaline media because of the well-known Hofmann
degradation, a property of quaternary ammonium
compounds; the Type II displays a lower stability in
alkaline media than Type I.
II / ION EXCHANGE / Organic Ion Exchangers 1623
Hofmann degradation of the two structures takes
place according to Figure 6. The degradation can lead
to both loss of exchange capacity (routes A and A in
Figure 6) and the appearance of a weak base capacity
caused by the presence of tertiary amine groups (B, B
and C in Figure 6).
Strong base anion exchangers have a lower thermal
stability than the cation exchangers.
Other commercially available strong base ex-
changers are those formed with an acrylic matrix.
They are usually made in bead form by free-radical
polymerization of 3-dimethylaminopropyl methac-
rylamide with DVB followed by a quaternization
reaction of the copolymer with alkyl halides as shown
1624 II / ION EXCHANGE / Organic Ion Exchangers

Figure 4 Classical structures of the structural units of Type I and Type II strong base anion exchangers.

in Figure 7. For the quaternization, gel- or macropor-
ous-type copolymers can be used.
Generally, the acrylic strong base anion exchangers
have a lower stability to hydrolysis, especially under
acid or alkaline conditions, compared to the polysty-
rene-based exchangers. The hydrolysis becomes more
signiRcant when the spacer between the amide group
and the quaternary group decreases in size. Thus, the
product with a spacer of only one methylene group
between the two functional groups has hydrolytic
instability. The same phenomenon occurs in the anion
exchanger prepared from 3-dimethylaminopropyl
methacrylate, CH2"C(CH3)COO(CH2)3N(CH3)2, in-
stead of the amide monomer.
In addition to the strong base anion exchangers
previously presented as commercially available prod-
ucts, other specialized strong base exchangers are
known.
In an effort to develop anion exchangers with
3 anion over the SO4\ anion
2
preference for the NO\
(an important factor for nitrate removal from potable
water which invariably contains sulfate), the design
of such a structure was conceived. It is the reaction
product of the chloromethylated styrene}DVB
copolymer with triethylamine, and can be described
as a strong base anion exchanger of Type III.
Gel or macroporous 4-vinylpyridine}DVB co-
polymers are the precursors for strong base

Figure 5 The two crosslinked polystyrene-based chloromethylated structures.


Figure 6 Hofmann degradation of Type I and Type II strong base anion exchangers.
exchangers. These exchangers are made by the well-
known quaternization reaction with alkyl halides as
shown in Figure 8.
The synthesis of this category of anion exchangers
takes place by a single chemical transformation step
which avoids crosslinking side reactions. However,
these exchangers cannot be utilized in many Relds of
application because of their very low chemical stabil-
ity in alkaline media.
Ion exchangers with benzyltrialkylphosphonium
groups, especially benzyltri-n-butylphosphonium
halide can be made. These structures are not used in
ion exchange processes but have special applications
as phase-transfer catalysts. For the improvement of
their properties, structures with a spacer larger than
one methylene group between the aromatic ring and
II / ION EXCHANGE / Organic Ion Exchangers 1625
the phosphonium group have been synthesized.
Figure 9 shows the phosphonium-type structures and
their preparative routes.
Commercially available exchangers of Types I and
II are fully ionized in the whole pH domain of the
aqueous medium, like the strong acid ones. The Type
I exchanger is such a strong base that a considerable
quantity of NaOH is required for its regeneration in
the OH form, while the Type II exchanger, a weaker
base, requires less. This aspect is an advantage of the
Type II structure over Type I.
The strong base anion exchangers in their OH form
react with both strong and weak acids. With the
latter, the strong base anion exchanger of Type I is
more effective than Type II. Because of this situ-
ation, Type I exchangers are used for soluble and
1626 II / ION EXCHANGE / Organic Ion Exchangers
Figure 7 Preparation of an acrylic strong base anion exchanger.
colloidal silica removal from natural waters. For re-
moval of the colloidal silica only, Type I strong base
anion exchangers with special macroporous struc-
tures are effective.
Certain Type I strong base exchangers in their Cl
form are used for adsorption of ionic organic com-
pounds and are called ‘scavenger’ ion exchangers.
Weak Base Anion Exchangers
Commercial weak base anion exchangers are pre-
pared by the condensation of m-phenylenediamine
Figure 8 Preparation of strong base anion exchangers based
on 4-vinylpyridine}DVB copolymer.
with formaldehyde. The chemical structure of this
exchanger is illustrated in Figure 10. In this structure,
the amine groups directly attached to benzene rings
have a very low basicity.
Polycondensation weak base exchangers with high-
er basicity were later obtained by the condensation of
other reagents. An example is the epoxy structures
formed by the condensation of aliphatic polyamines
with epichlorohydrin. This halo-epoxy compound
can react even with tertiary amine groups, thus anion
exchangers containing amine and quaternary
ammonium groups can be obtained as shown in
Figure 11.
The most readily available commercial weak base
exchangers are the polymerization structures based
on polystyrene or acrylic matrices containing pri-
mary, secondary or tertiary amine groups, or all these
groups together.
The polystyrene-based weak base exchangers are
obtained by the same reaction scheme as the strong
base ones with polystyrene matrices, but with the
difference that dimethylamine is used in the
amination step instead of trimethylamine or
dimethylethanolamine. When using the secondary
amine, in contrast with the tertiary amines, besides
the main amination reaction which leads to the terti-
ary amine groups, an undesirable side reaction can
also take place. This is the quaternization reaction
between the pre-attached tertiary amine and }CH2Cl
groups. It can take place intra- or inter-chain but both
situations can occur.
The chemical structures of aminated units are
shown in Figure 12. Both quaternization types (inter-
and intra-chain) lead to anion exchangers with mixed
 II / ION EXCHANGE / Organic Ion Exchangers 1627

Figure 9 Preparation methods of phosphonium-type strong base anion exchangers.

functional groups and the inter-chain ones also con-
trol the degree of crosslinking.
Amination takes place quantitatively under mild
conditions (30}403C; 4}6 h) in aqueous or organic
media. By using a large excess of amine, the side
quaternization reaction is greatly reduced.
The most common acrylic weak base exchangers
are made by the acylation of primary or secondary
amines. This reaction with esters, so-called ester
aminolysis, is the most frequently used method to
produce acrylic weak base exchangers.
The aminolysis of macroporous or gel-type ethyl
acrylate}DVB copolymers, in bead form, with 3-
dimethylamino-1-propylamine, ethylendiamine or
other aliphatic polyamines, is illustrated in Figure 13.
The same structures can also arise from amino-
lysis}hydrolysis reactions of acrylonitrile}DVB
copolymers. An example is shown in Figure 14.

Figure 10 Chemical structure of m-phenylenediamine}formaldehyde weak base anion exchanger.

1628 II / ION EXCHANGE / Organic Ion Exchangers
Figure 11 Chemical structure of an epoxy-type base anion
exchanger.
The aminolysis and aminolysis}hydrolysis reac-
tions take place under more stringent reaction
conditions (temperature over 1003C and reaction
time over 10 h) than the amination of the chloro-
methylated styrene}DVB copolymers.
Acrylic weak base exchangers synthesized from
aliphatic polyamines have much higher exchange
capacities than polystyrene-based structures.
The weak base exchangers cannot be regarded as
typical ionic polymers since their amine groups are
ionized only under certain conditions. Thus, an amine
group in an alkaline medium is in the free-base form,
but in a neutral medium it can exist in a partial
ionization level, which depends upon amine basicity.
Figure 12 Chemical structure of the functional groups which can exist in a polystyrene-based weak base anion exchanger.
In an acid medium a high level of ionization is present
and because of this, the weak base exchangers are
usually used to retain strong acids in water treatment.
They also can be used as insoluble acceptors of acids
in different chemical reactions such as the prep-
aration of esters from acid chlorides and alcohols, etc.
For the latter aim, the best known are the 4-vinyl-
pyridine}DVB copolymers with a low degree of
crosslinking.
Scheme 2 shows some more distinctive ion ex-
change reactions for the four above-mentioned types
of exchangers.
Amphoteric Ion Exchangers
Ion exchangers which contain both acidic and basic
groups are called amphoteric resins. Usually their
matrix has some structural units with acidic groups
and other units with basic groups.
Very interesting amphoteric resins are the so-called
‘snake-cage polyelectrolytes’. One feature distin-
guishes the snake-cage polyelectrolytes from other
amphoteric resins, namely the acidic and basic groups
are not attached to the same matrix. For example,
a snake-cage polyelectrolyte is prepared by polym-
erization of acrylic acid (snake) into a strong base
anion exchanger with quaternary ammonium groups
(cage). These resins are excellent reversible sorbents
for electrolytes and can be regenerated by rinsing
with water. Electrolyte sorption seems to be mainly
determined by the preference of the acidic groups for
the cation and of the basic groups for the anion. The
resins show preference for different electrolytes.
This phenomenon can be used for separating electro-
lytes one from another.
 II / ION EXCHANGE / Organic Ion Exchangers 1629

Figure 13 Acrylic-type weak base anion exchangers and their preparation methods.

The snake-cage polyelectrolytes are also used in the
technique of ‘ion retardation’. This technique is the
separation of strong electrolytes from weak electro-
lytes or non-electrolytes.
Chelating Ion Exchangers
The cation binding of transition, heavy and noble
metals can be performed by an ion exchange process
with strong and weak acid cation exchangers as well as
with strong base exchangers; the latter are used if the
metal cations are present in the form of complex anions.
However, the most promising technique for
the binding of metal cations is the use of ion ex-
changers which contain chelating functional groups.
Thus, exchangers with iminodiacetate groups can re-
move several ppm of Ca2#, Mg2# or Sr2# from
brine, in contrast to cation exchangers containing the

Figure 14 Preparation of an acrylic-type weak base anion exchanger from acrylonitrile}DVB copolymer.
1630 II / ION EXCHANGE / Organic Ion Exchangers

Scheme 2 Some distinctive ion exchange reactions.

Table 2 Some chelating acid and base functional groups

Functional group Chemical structure Utilizations

Amine }NH}(}C2H4}NH})x }H Removal of transitional metals

Iminodiacetic }N}(}CH2}COOH)2 Selective removal of heavy metals
Isothiouronium }S}C(NH2)"NH Selective removal of mercury and noble metals
Aminophosphonic }CH2}NH}CH2}P(O)(OH)2 Especially for decalcification of brine solutions
Phosphonic acid }P(O)(OH)2 Preconcentation of uranyl ions
Hydroxamic acid }CO}NHOH Selective retention for Fe(III) ions
Hydroxyamine }CH2}N(CH3)}CH2}(CHOH)4CH2OH Selective retention for boric acid

Figure 15 Classical binding of metal cations by some chelating functional groups. (P is a polymer structure unit that contains an acid
or base group.)
non-chelating common acid groups. The latter ex-
changers show a higher selectivity for divalent cations
than for Na#, but the difference is not very large,
so that they will not be able to bind Ca2#, Mg2# or
Sr2# cations present at very low concentrations.
The high selectivity of chelating ion exchangers is
attributed to the establishment of much stronger
bonds than the simple electrostatic attractive forces
present in the case of the common ion exchangers.
Because of these bonds, the chelating processes show
a high degree of irreversibility.
It must be mentioned that the chelating groups,
attached to polymer networks, are groups with two
or more electron donor elements, such as N, S, O and
P, and they can function like the model low molecular
weight chelating agents. Their preparation takes
place in two main ways } polymerization or polycon-
densation.
Table 2 shows some functional groups, which are
capable of chelating. The binding of metal cations by
some chelating functional groups is illustrated in
Figure 15.
The nature of metal cation binding can be
modiRed by changing the pH of the solution to cause
chemical modiRcation of the chelating groups. Thus,
Figure 16 Chemical structure dependence of amine and isothiouronium groups on the pH of aqueous medium.
II / ION EXCHANGE / Organic Ion Exchangers 1631
amine and isothiouronium groups can take part in the
equilibria, shown in Figure 16. These groups will
bind metal cations by a coordination process in neu-
tral or base media via the free base forms while in an
acid medium the same cations will be bound as com-
plex anions by an anion exchange process via the salt
forms.
Conclusions
The area of organic ion exchangers remains a very
active one with intensive technical and scientiRc
work on both conventional resins and entirely novel
systems. Thus, organic ion exchangers with dif-
ferent morphologies of their three-dimensional
networks and base and acid groups with various
chemical structures have been prepared.
There is also a growing interest in the development
of selective chelating ion exchangers for the possible
application of these resins in analytical chemistry,
metal recovery and wastewater treatment. The latter
two applications have a great importance in eco-
nomic and ecological domains. Requirements of the
properties of these ion exchangers include high capa-
city, high selectivity and fast kinetics. Most of the
1632 II / ION EXCHANGE / Organic Membranes
commercial resins show a high capacity but a poor
selectivity towards metal ions.
The combination of the physical strength of an
inorganic support with the higher ion exchange capa-
city and kinetics of the organic ion exchangers could,
in the future, lead to an interesting class of ion ex-
changers with special applications.
See also: II/Ion Exchange: Historical Development; In-
organic Ion Exchangers; Organic Membranes; Theory of
Ion Exchange.
Further Reading
Albright RL and Yarnell PA (1987) Ion-exchange Poly-
mers: Encyclopedia of Polymer Science and Engineer-
ing, 2nd edn, vol. 8. New York: John Wiley.
Camps M, Chatzapoulos JM and Montheard JP (1987}88)
Chloromethylation of polystyrene and styrene co-
polymers. Journal of Macromolecular Science } Review
of Macromolecular Chemistry and Physics C27: 505.
Organic Membranes
R. WoH dzki, Nicholas Copernicus University,
Torun& , Poland
Copyright ^ 2000 Academic Press
Organic ion exchange membranes are made of insol-
uble polymeric foils, tubes or hollow Rbres to which
ion exchange groups are covalently bound. The mem-
branes have all the properties typical of ion exchange
resins and the ability to keep two different solu-
tions physically separated. Thus, the main property of
an ion exchange membrane is a selective exchange of
ions and a selective permeability to ions, water or
other speciRcally membrane-soluble species. Al-
though optimizing all the properties would be dif-
Rcult in the case of one membrane used for a variety
of applications, the most desired property is always
high membrane selectivity, which allows the separ-
ation of ions with low energy consumption and high
transport rates. There are many additional require-
ments for ion exchange membranes, such as low
electric resistance, high permselectivity, low free dif-
fusion of salts (leakage), low osmotic water transport,
high mechanical strength, high selectivity between
ions of the same charge and high chemical stability.
Preparation
Ion exchange membranes can be classiRed as mono-
polar or bipolar. In parallel, some intermediate mem-
Dorfner K (1972) Ion Exchangers, 3rd edn. Michigan:
Ann Arbor Science Publishers.
Dorfner K (ed.) (1991) Ion Exchangers. Berlin: W de G de
Gruyter.
Frechet JMJ and Farrall MJ (1977) Chemistry and Proper-
ties of Crosslinked Polymers. London: Academic Press.
Helfferich F (1962) Ion Exchange. New York:
McGraw-Hill.
Hodge P and Sherrington DC (eds) (1980) Polymer-
supported Reactions in Organic Synthesis. New York:
John Wiley.
Holliday L (1975) Ionic Polymers. London: Applied Science.
Samuelson O (1963) Ion Exchange Separations in Analyti-
cal Chemistry. New York: John Wiley.
Sherrington DC (1988) Reactions of Polymers: Encyc-
lopedia of Polymer Science and Engineering,
2nd edn, vol. 14. New York: John Wiley.
Streat M and Naden D (eds) (1987) Ion Exchange and
Sorption Processes in Hydrometallurgy. Critical Reports
on Applied Chemistry, vol. 15, ch. 3 and ch. 4.
New York: John Wiley.
branes have been categorized as mosaic and am-
photeric. The monopolar membranes can be divided
into cation exchange membranes and anion exchange
membranes. Combination of these membranes results
in a bipolar ion exchange membrane.
The properties of any ion exchange membrane are
determined by the properties of its polymer matrix
and the type and concentration of the Rxed ionic
moieties. The polymer matrix of ion exchange mem-
branes is usually cross-linked. The degree of cross-
linking extensively inSuences the degree of swelling
(water sorption), chemical and mechanical stability
and membrane permeability by changing ionic mobil-
ity inside the membrane phase. Thus, the proper
selection of membrane components, their content and
the method of preparation signiRcantly affect the
membrane properties and structure. The subsequent
chemical modiRcation of the polymer matrix involves
the introduction of ionogenic groups, resulting typi-
cally in the following Rxed sites:
Cation exchange membranes:
3 , }COO\, }PO3\, }HPO\
}SO\ 2
2 , SeO\
3
Anion exchange membranes:
}NH# # # # #
3 , }RNH2 , }R3N ,"R2N , }R3P , }R2S
#

Monopolar Membranes
The simplest ion exchange membranes are composed
of derivatives of styrene-divinylbenzene copolymers
or vinylpyridine-divinylbenzene copolymers. In order
to maintain their mechanical strength, these homo-
geneous membranes are often reinforced by backing
materials. Other simple ion exchange membranes can
be synthesized from Rnely powdered ion exchange
resin and inert polymer powder applied as a bonding
material. Nowadays, more advanced methods of
membrane preparation are recommended. The most
widely used methods are:
1. Impregnation of a basic polymer component with
styrene and divinylbenzene for copolymerization
followed by chemical modiRcation (e.g. sulfona-
tion);
2. Casting the mixed solution of poly(styrene sul-
fonic) acid and other inert polymer to produce an
interpolymer or ‘snake-in-cage’ membrane;
3. Radiation grafting of a polymerizable monomer
into a conventional polymer Rlm and then intro-
ducing the ion exchange groups to the Rlm. The
Rlms used are generally polyethylene, polypropy-
lene and perSuoroethylene;
4. Chemical modiRcation: this method involves di-
rect introduction of ion exchange groups to a con-
densation-type polymer such as polysulfone,
poly(ethylene oxide), poly(vinyl chloride), poly(vi-
nylidene Suoride), poly(ether ether ketone). The
membrane can be formed subsequently by casting
the polymer solution, and then phase inversion.
The procedure results in an anisotropic membrane
structure with a thin skin layer and a supporting
layer of sponge morphology;
5. Plasma polymerization. This relatively new
method has been utilized to prepare: Rrst, an ani-
on exchange membrane by plasma polymerization
of
-aminopropylethoxydimethylsilane on a por-
ous polymer Rlm; second, a perSuorocarbon
sulfonic acid membrane, by plasma polymeriz-
ation of perSuorostyrene and SO2; and third,
a thin Rlm sulfonic-type membrane, by plasma
polymerization of ethylene and SO2, or acetylene
and SO2.
Bipolar and Mosaic Membranes
Preparation methods of bipolar ion exchange mem-
branes involve the introduction of cation exchange
groups into one side of the membrane and anion
exchange groups into the other side. In order to ob-
tain the simplest bipolar membrane a separate cation
exchange membrane and an anion exchange mem-
brane can be glued together by using poly(styrene
II / ION EXCHANGE / Organic Membranes 1633
sulfonic) acid as a binder. Other bipolar ion exchange
membranes have been prepared by graft polymeriz-
ation of acrylic acid on one side of the porous poly-
mer membrane and of N-(2-methacryoyloxyethyl)-
N,N,N-trimethylammonium chloride on the other
side, after oxygen plasma treatment of the porous
membrane. The mosaic membranes are usually pre-
pared by casting a multiblock copolymer composed
of styrene, butadiene and vinylbenzyl dimethylamine
and the subsequent introduction of cation and anion
exchange groups into the polymer.
Structure
The properties of ion exchange membranes depend
on their microstructure, which results from the dif-
ferences in properties of their ionic and neutral com-
ponents. The hydrocarbon ion exchange membranes
are generally composed of derivatives of styrene-
divinylbenzene copolymer and other inert polymers
such as polyethylene and poly(vinyl chloride). Ionic
derivatives of such copolymers are Rnely distributed
in the inert polymer-forming microdomains because
of their poor mutual compatibility. This inhomogene-
ity causes the distribution of ion exchange groups in
the membrane material to be of varied local concen-
tration. The local distribution of Rxed charges can
affect all the basic membrane properties, such as
electrolyte sorption, electrical conductivity and perm-
selectivity as well as mass transport rates. SpeciRcally,
a rather unusual structure is characteristic for per-
Suorocabon cation exchange membranes (with
sulfonic or carboxylic acid groups), which can be
classiRed as ionomer materials. Ionomers and ion-
containing polymers with ionic sites and counterions
spontaneously organized into dipole multiplets and,
for some materials, into larger ion clusters containing
50 or more pairs of ions. These clusters are embedded
in the perSuorocarbon membrane backbone, and are
connected by narrow channels forming a continuous
network.
Despite the heterogeneity resulting from the
hydrophilic and hydrophobic nature of membrane
components, certain additional morphology features
originate from the method of preparation. So-called
macroscopically homogeneous membranes can be
prepared either by the polymerization of functional-
ized monomers or by careful modiRcation of a homo-
geneous polymer Rlm. However, depending on the
method of preparation, the structural imperfection of
membranes increases in the following order: interpo-
lymer membranes, graft and block polymer mem-
branes and membranes composed of powdered ion
exchange resins embedded thereafter in an inert poly-
mer binder.
1634 II / ION EXCHANGE / Organic Membranes
General Physicochemical Properties
Ion exchange membranes are characterized by many
of the parameters and properties described below.
Independently, the basic properties describing equi-
librium and transport properties of any membrane
are as follows: ion exchange capacity (mol kg\1 of
dried membrane), swelling (wt.% of water or other
solvent), electrolyte sorption (mol kg\1 of water
sorbed into membrane), distribution coefRcient
(ratio of solute concentration in the membrane and
external solution), transference number for cations
and anions (determined under standard conditions),
electrical conductivity (or resistivity) and salt leakage
(both determined after contacting the membrane with
a standard electrolyte solution).
Co-ion Exclusion
In a cation exchange membrane, due to the system
electroneutrality, the Rxed anions are in equilibrium
with mobile cations (referred to as counterions). In
contrast, anions functioning as co-ions are more or
less excluded from the membrane because their
charge is identical to the Rxed ion charge. This phe-
nomenon, known as the Donnan exclusion of co-ions
or electrolytes, enables the ideal membrane to trans-
fer cations only. In the same way cations are excluded
from the anion exchange membrane. The Donnan
exclusion equilibrium, and thus the membrane selec-
tivity, depends quantitatively on the concentration of
the Rxed ions, the valence of co- and counterions, the
concentration of an equilibrating external solution,
and the afRnity of the exchange groups to respect-
ive counterions.
Transport Processes
The effectiveness of any membrane process is
determined by the Sux of species through the mem-
brane. High Suxes arise because of high permeability
of the internal aqueous membrane solution in respect
to sorbed solutes. The presence and properties of
transport-mediating functional ionic groups can
either facilitate or hinder transport processes depend-
ing on many speciRc interactions between mobile
species, the charged polyelectrolyte network and sol-
vent. In general, membrane transport phenomena can
be described by a general equation derived from the
linear thermodynamics of irreversible processes:
Ji" Li,kXk (i, k"1, 2, 3, 2 , m)
i
where Ji denotes the Sux of an individual permeant,
the volume of solvent, heat or electricity transferred
across a membrane, and Xi,k denotes the driving force
represented by a difference in chemical potential
(), temperature (T), pressure (p) or chemical
afRnity (A). Li,k is the phenomenological coef-
Rcient linearly relating the Sux and driving force. In
practice, a general Sux equation can be reduced to
a description of a speciRc process by neglecting coup-
ling transport phenomena. Depending on the solution
and assumptions, various mathematical relations
have been derived and applied to discuss transport
rates of ionic substances and water through ion ex-
change membranes.
Permselectivity
The characteristic property of ion exchange mem-
branes is their ionic permselectivity related to electri-
cally driven processes and quantiRed by the following
deRnitions:
t#,CEM!t# t ,AEM!t
PCEM" and PAEM" \ \
t t#
\
where t is the transference number for anions (!)
and (#) cations in free electrolyte or acting
as counterions (cation exchange membrane (CEM)
and anion exchange membrane (AEM)) in an
ion exchange membrane. An ideal permselective
membrane should have a P value of 1, and the per-
mselectivity is equal to zero when the transference
numbers within the membrane are the same as in an
electrolyte solution. Usually, transference numbers
for practical membranes attain values from 0.8 to
0.98.
Separation Ability
One of the most important uses of ion exchange
membranes is the selective separation of solutes. This
can be performed because of the differences in
the rates at which solutes permeate through the mem-
brane. The size of this difference depends on the
separation system, i.e. on the composition of feed
solution and the membrane. The criterion of separ-
ation is that the composition of the mixture emerging
from the membrane cell should be different from
that entering on the feed side. Consequently, the
separation factor AB measures the extent of selective
transport of A in relation to B:
JAXB
AB"
JBXA
where JA and JB are the Suxes (transport rates ex-
pressed as numbers of moles issuing from the mem-
brane per unit area and time) and XA and XB are the
mole fraction of A and B in the feed.

Membrane Processes and
Applications
Many applications of ion exchange membranes in
transport and separation processes have been made.
Although the driving force for ions and water to
penetrate through the membrane initially is primarily
an electrochemical and chemical potential, it has been
reported in recent years that the hydraulic perme-
ability, temperature difference and difference
in proton concentration can also be exploited. This
results in some sophisticated transport mechanisms,
the backgrounds and practice of which are reviewed
below. The latest applications are due to develop-
ments in ion exchange membrane composition and
structure. Ion exchange membranes with speciRc
properties have been produced to meet many indus-
trial requirements. For example, the following mem-
branes of speciRc properties have been developed and
commercialized: perSuorocarbon anion exchange
membranes for high temperature usage and chemical
stability in corrosive media, anion exchange mem-
branes for diffusion dialysis (treatment of corros-
ive acidic waste waters), anion exchange membranes
of high acid retention (electrodialytic concentration
of dilute acids), hydrogen ion permselective cation
exchange membranes (efRcient electrodialysis)
and monovalent cation or monovalent anion per-
mselective membranes.
Diffusion Dialysis
Diffusion dialysis results from the difference in
rates of permeation of salts and acids through anion
exchange membranes. According to the scheme shown
in Figure 1, this membrane allows the selective trans-
port of anions across the membrane, ideally remaining
impermeable to cations other than protons. Diffu-
sion dialysis has been exploited to remove acids from
Figure 1 Scheme of diffusion dialysis of salt and acid through
anion exchange membrane (AEM).
II / ION EXCHANGE / Organic Membranes 1635
solutions containing different cation species. High
quality anion exchange membranes allow this method
to be applied to the recovery of sulfuric and other acids
from waste solutions generated in steel, metal-reRn-
ing and electroplating industries.
Donnan Dialysis
Donnan dialysis, also referred to as ion exchange
dialysis, occurs after placing an ion exchange mem-
brane between two solutions containing different
electrolytes. According to the scheme depicted in Fig-
ure 2, ions can cross a membrane when their sign is
opposite to the sign of the membrane-forming poly-
electrolyte. At the same time, the permeation of co-
ions is hindered because of their electrostatic exclu-
sion from the membrane phase. Usually, divalent or
univalent metal cations are transported from the di-
lute feed solution into the highly acidic stripping
solution. The coupling of Suxes in this membrane
system makes it possible to reach a stable Sux of
cations from dilute to concentrate phase. Thus, uphill
transport (chemical pumping) arises as a result of the
interdiffusion of different counterions.
The Rnal distribution of ions between two mem-
brane-adjacent solutions corresponds to the Donnan
equilibrium principle, which in the typical case of
Mz#/H# Donnan dialysis, takes the following form:


ZM
CH,s
CM,s"kCM,f
CH,f
Figure 2 Scheme of Donnan dialysis of M# and H# cations
through exchange membrane (CEM). (From WoH dzki R,
SzczepanH ski P and Pawlowski M (1999) Recovery of metals from
electroplating waste solutions and sludge. Comparison of Donnan
dialysis and pertraction technique. Polish Journal of Environ-
mental Studies 8(2): 111}124.)
1636 II / ION EXCHANGE / Organic Membranes

Figure 3 Scheme of neutralization dialysis of M# cations and A\ anions through cation (CEM) and anion exchange membranes
(AEM).

Consequently, when the initial concentration (C) of
metal ions (M) in the feed (f ) is much lower than the
concentration of counter-transported protons (H) in
the stripping solution, it is possible to attain substan-
tial enrichment of metal ions in the receiver, or alter-
natively, almost complete removal of these cations
from the feed. Despite the many advantages, only
a pilot dialyser for the nuclear industry (recovery of
134
Cs, 90Sr and concentration of uranyl ions) has been
reported as large scale implementation of this tech-
nique. On the other hand, Donnan dialysis is widely
used in analytical laboratories as an efRcient
method for the preconcentration and separation of
various cations and anions or for the treatment of
complex matrices before analysis.
branes is possible. During this process (Figure 4) an
ion exchange membrane is kept in contact with
a stream of the mixture of water and miscible organic
solvent, whereas the second side of the membrane is
kept under vacuum. Due to the strong afRnity of
ionic sites to water and an inertness of the polymer
backbone to an organic component, water can per-
meate through the membrane. Thus, the permeation
of water can be optimized by changing the membrane
polymer, the type and content of the membrane-
forming polyelectrolyte (ionomer, cross-linked
copolymer, etc.), and the kind of counterions. For
instance, cation and anion exchange membranes of
various ionic forms have been examined for the separ-
ation of alcohol}water and pyridine}water mixtures.

Neutralization Dialysis
Fixed-site Mediated or Carrier Transport
Neutralization dialysis is a membrane process based
Analogous to liquid membrane transport mediated
on the coupling of two simultaneously occurring
by the mobile carriers, ion exchange sites can be
Donnan dialyses. According to the scheme shown in
exploited as Rxed carrier centres in the polymer mem-
Figure 3, in this case a salt solution is separated from
brane phase. The membranes exhibiting such a func-
an external acidic and basic solution with a cation
tion are referred as Rxed-site carrier membranes or
and anion exchange membrane, respectively. Protons
reactive membranes. In general, transport phe-
and hydroxide ions permeate into the desalination
nomena in such membranes are believed to occur as
compartment by the Donnan dialysis mechanism,
a sequence of exchange reactions between permeating
which generates the counter-Sow of other cations and
solute and reactive groups located along the polymer
anions. Under ideal conditions, i.e. without salt
leakage and with balanced Suxes of protons and
hydroxide ions, the overall process results in almost
complete desalination of the internal solution. The
practical use of neutralization dialysis for the de-
mineralization of mixtures containing some organic
substances (mono-, oligo- and polysaccharides) and
polyelectrolytes is recommended.

Pervaporation
In pervaporation processes the application of ion ex- Figure 4 Scheme of water pervaporation through ion exchange
change membranes instead of inert polymer mem- membrane (IEM).

chain rather than by pure diffusion phenomena.
To be effective the transport mechanism requires
Rxed sites to be mobile over a certain restricted area.
Moreover, their concentration should exceed a cer-
tain concentration, enabling instantaneous overlap-
ping of loaded and unloaded centres. Despite many
difRculties in understanding and in the theoretical
description of this type of membrane transport and
separation, a number of processes have been de-
veloped. The most interesting are processes dealing
with the separation of gaseous substances (CO2), ole-
Rns and sugars. For example, a cation exchange mem-
brane exchanged with ethylenediamine is selectively
permeable to CO2 (from a feed containing CH4 and
H2S). On the other hand, silver ions exchanged with
a cation exchange membrane form an Ag#}oleRn
complex in the membrane phase. Carrier transport of
1-hexene and 1,5-hexadiene from a decane phase
through a cation exchange membrane impregnated
with silver ions has been reported. Similarly, styrene
permeates selectively (compared with ethylbenzene)
through the Ag# form of cation exchange membrane.
A selective transport of sugars via complexation with
borate ions Rxed in an anion exchange membrane has
been performed and referred as carrier}relay trans-
port. In this way, the selective transport of D-glucose,
D-xylose, D-arabinose, D-mannose, D-galactose, D-
fructose, L-sorbose, sucrose and D-lactose can be
achieved.
Membrane Extraction^Hybrid Membrane
Systems
The combinations of liquid membrane extraction
(usually mediated by speciRc ion exchange extrac-
tant/carrier) with ion exchange membranes in some
Figure 5 Scheme of membrane hybrid system composed of
two cation exchange membranes and the liquid membrane con-
taining an ionic carrier (C). Coupled countertransport of M# and
H# cations. (From WoH dzki R, SzczepanH ski P and Pawlowski
M (1999) Recovery of metals from electroplating waste solutions
and sludge. Comparison of Donnan dialysis and pertraction tech-
nique. Polish Journal of Environmental Studies 8(2): 111}124.)
II / ION EXCHANGE / Organic Membranes 1637
relatively new membrane systems have been called
membrane hybrid systems (MHS). The scheme of
cation transport in a simple MHS is depicted in
Figure 5. The MHS operation involves a series of
ion exchange}diffusion processes (ion exchange
membrane) as well as permeation through a liquid
membrane. The presence of an ion exchange mem-
brane at a particular interface stabilizes the liquid
membrane and enhances the interfacial ion exchange
reactions because of high, and speciRc, sorption of
certain ions by the membrane polyelectrolyte. The
long term and stable operation of a liquid membrane
system can be achieved with the MHS idea. The
selective and active transport of metal cations, inor-
ganic anions and some carboxylic acids has been
reported. It is worth noting that these systems can be
regarded as biomimetic, which means that they cor-
respond to a cellular envelope of Gram-positive
bacteria composed of an ion exchange polymer
membrane (cell wall) and a quasi-liquid membrane
(cytoplasma membrane).
Membrane Permeation Coupled to External
Reaction
It is possible to accelerate chemical reactions by re-
moving some products from the reaction medium
before the reaction reaches its equilibrium state. Ion
exchange membranes are effective in these cases.
For example, fermentation processes producing ionic
materials such as acetic acid can be carried out con-
tinuously as an extractive fermentation by removing
the products by means of diffusion dialysis, per-
vaporation or electrodialysis. Ion exchange mem-
branes can be applied to improve the kinetics of
chemical reactions by the pervaporative removal of
water from the reaction medium. Such a technique
has been demonstrated for esteriRcation of oleic acid
with ethanol and of propionic acid with isopropanol
or propanol.
In more advanced systems, membrane transport
phenomena are coupled with enzyme reactions. The
most representative system is composed of a cation
exchange membrane layer, a porous membrane layer
containing entrapped urease and an anion exchange
membrane layer. Urea is decomposed into NH# 4 and
CO23\ in the enzyme layer and products permeate
through the respective membrane layers without ap-
plication of an electric Reld. A similar idea is utilized
in many enzymatic membrane sensors.
Ion-Exchange Membrane Separators in Power
Sources, Sensors and Electrodes
A typical example is the use of an ion exchange
membrane as a solid polyelectrolyte for a fuel
1638 II / ION EXCHANGE / Organic Membranes

cell. The composite, in which anode catalyst,

poly(Suorocarbon sulfonate) cation exchange mem-
brane and cathode catalyst are combined, has been
used for hydrogen-oxygen fuel cells. The only prod-
uct of this cell is water, which seems to be desirable
from an ecological point of view. Methanol has also
been reported for the generation of electricity and
utilizable industrial chemicals by the use of a mem-
brane-containing fuel cell. Various applications of ion
exchange membranes of low electrical resistance in
alkali batteries are possible. Modern cation exchange
membranes made of acrylic acid grafted onto poly-
ethylene Rlm are widely used as separators in alkaline Figure 6 Basic idea of membrane electrodialysis.
batteries, such as a Ni}Cd secondary battery.
Water content of the ion exchange membranes and
thus their physicochemical properties changes with electrodialysis is widely applied in environmental
humidity. Therefore, ion exchange membranes are protection (depolluting and recycling of chemicals),
usable as a working part of a hygrometer. When in bio-industries (food, pharmacy and biotechnology)
water content of the membrane increases with in- and in the treatment of drinking water. Some of these
creasing humidity, the increase in the current or ionic new applications have led to substantial improve-
conductivity between the electrodes on both sides of ments in membrane quality. For instance, special
the membrane can be measured and calibrated. membranes with low permeability to the divalent ions
The concentration of alcohol can be indirectly de- in respect to the monovalent ones, membranes with
termined by means of a membrane with Rxed alcohol- very low permeability to hydroxyl ions or very low
dehydrogenase or alcohol-oxydase. It has been re- permeability to protons, are currently produced.
ported that these sensors operate accurately in acidic
media after coating the enzyme-Rxed membrane by Membrane Electrolysis
a cation exchange membrane (organic acids cannot
A typical example of this kind of application is the
approach the enzyme membrane). A similar concept
membrane electrolysis of sodium chloride to produce
is used for constructing in vivo operating sensors for
chlorine, hydrogen and sodium hydroxide (mem-
glucose presence and concentration. Some modiRed
brane chlor-alkali process). A schematic diagram for
electrodes are constructed by coating classic elec-
this process is presented in Figure 7. PerSuorocarbon
trodes with layered ion exchange materials. A typical
cation exchange membranes for chlor-alkali pro-
example is the perSuorocarbon cation exchange
cesses should have an anisotropic structure in their
membrane-coated electrode to control the permeabil-
cross-section. The cathode side of the membrane has
ity of redox species such as Ru(bpy)3Cl3 or 1,1-dihyd-
a thin layer of carboxylic acid groups of a given ion
roxymethylferrocene.
exchange capacity, and the anode side of the mem-
brane has a thick sulfonic acid group layer or a car-
Industrial Applications boxylic acid group layer of high ion exchange capa-
city. Ion exchange membranes can also be applied for
Membrane Electrodialysis the electrodialysis of water to produce hydrogen and
oxygen. The technology exploiting a solid polymer
The basic principle of membrane electrodialysis is
electrolyte method is efRcient when perSuorocar-
presented in Figure 6. The membranes are arranged
bon cation exchange membranes are used in the form
in a series of anion and cation exchange membranes
of a stack with catalytic electrodes covering the mem-
placed between an anode and a cathode. The cations
brane surfaces. Ion exchange membranes are also
migrating towards the cathode permeate the nega-
used as separators in organic synthesis by electrolysis.
tively charged cation exchange membrane and are
A typical example is the hydrodimerization of acryl-
retained by anion exchange membranes. The anions
onitrile to produce adiponitrile.
are transported in the opposite direction. The Rnal
result is an increase in ion concentration and ion
depletion in alternate compartments. Membrane
electrodialysis has been developed within the last
Conclusion
several years, mainly for desalting brackish waters The technique of membrane separation can be con-
and concentrating brine from seawater. Nowadays, sidered as an energy-saving method because, in

Figure 7 Chlor-alkali process (membrane electrolysis) using a cation exchange membrane (CEM).
general, it does not cause phase conversion. From this
point of view, ion exchange membranes and their
application technology are one of the most advanced
methods enabling the embodiment of closed-loop
chemical processes. As solid polyelectrolytes, the
membranes are easy to regenerate, recycle and/or
promote continuous usage to improve industrial pro-
cesses. Environmental beneRts can be achieved as
well.
Since the use of ion exchange membranes has be-
come very diverse, the requirements for membranes
with new properties will increase. These requirements
should result in the development of new highly func-
tionalized ion exchange membranes. Photosynthetic
membranes, biomimetic membranes, complex and
specialized membrane sensors, polymerized phos-
pholipids and polymerized crown ethers with ionic
groups can be considered as new possible areas of
membrane science development.
See also: II/Ion Exchange: Catalysis: Organic Ion
Exchangers; Historical Development; Inorganic Ion
Phosphates: Novel Layered Materials
See II / ION EXCHANGE / Novel Layered Materials: Phosphates
II / ION EXCHANGE / Organic Membranes 1639
Exchangers; Organic Ion Exchangers; Theory of
Ion Exchange. III/Porous Polymer Complexes for Gas
Separations: Membrane Separations.
Further Reading
Drioli E and Nakagaki M (1986) Membranes and Mem-
brane Processes. New York: Plenum Press.
Helfferich F (1962) Ion Exchange. New York: McGraw-
Hill.
Ho WS and Sirkar KK (1992) Membrane Handbook. New
York: Chapman and Hall.
Kesting RE (1985) Synthetic Polymeric Membranes.
A Structural Perspective. New York: Wiley.
Lakshminarayanaiah N (1969) Transport Phenomena in
Membranes. New York: Academic Press.
Lloyd DR (1985) Material Science of Synthetic Membranes.
Washington DC: American Chemical Society.
Selegny E, Boyd G and Gregor HP (1976) Charged Gels and
Membranes, Part I. Dordrecht: D. Reidel.
Strathmann H (1990) Membranes and membrane separ-
ation processes. In: Arpe H-J (ed.) Ulmann’s Encyclo-
pedia of Industrial Chemistry, vol. A 16, pp. 187}263.
Weinheim: Verlag Chemie.
1640 II / ION EXCHANGE / Surface Complexation Theory
Surface Complexation Theory: Multispecies Ion Exchange
Equilibria
W. H. HoK ll, Forschungszentrum Karlsruhe,
Eggenstein-Leopoldshafen, Germany
J. Horst, Karlsruhe, Germany
Copyright ^ 2000 Academic Press
Introduction
Inorganic and organic ion exchangers consist of
either a crystalline or a polymeric matrix and func-
tional groups. Depending on the pH value of the
liquid phase, these groups can either be protonated or
dissociated. By this means exchangers are able to
interact with ions from the liquid phase. There is
a large variety of such interactions: they may be due
to electrostatic and van der Waals forces, heteropolar
and covalent binding, or coordination forces. The
resulting sorption phenomena take place at the inner
pores and/or surface of the exchangers.
The surface charge generates an electric potential
normal to the surface. Consideration of the resulting
electrostatic interactions has led to several theoretical
descriptions, e.g. the Helmholtz, Gouy}Chapman and
Stern models. Modern theoretical approaches consider
the adsorption of counterions as a result of chemical
interactions between the surface groups and dissolved
species. Sorption of protons or of any other kind of ion
is treated as a local equilibrium reaction. SpeciRc (for
protons and hydroxyl ions) and non-speciRc interac-
tions lead to the formation of ion pairs at the surface
that are designated as surface complexes. Differ-
ent kinds of surface complexes can be discriminated.
Spectroscopic investigations of surfaces have given
rise to the assumption that more than one single layer
has to be assumed to account for the uptake of
counterions. A typical approach used by many
authors is the triple-layer model consisting of surface,
inner and outer layers.
A further reRnement of the triple-layer model is the
approach developed by Horst. In this model individual
sorption layers are credited to each kind of counter-
ions. The respective theoretical approach allows the
description of pure cation or anion exchange equilib-
ria as well as of amphoteric equilibria encountered
with activated alumina or activated carbon.
Theory
Amphoteric Reactions at a Charged Surface
Derivation of the mathematical relationships uses
a generalized exchanger whose surface contains
S}OH groups as functional sites. In aqueous systems
these surface groups can be protolysed in two dif-
ferent ways.
In acid media the surface may be protonated ac-
cording to:
S}OH#H# 0 S}OH#
2 [1]
In order to maintain the condition of electroneutrality
the charge on the surface has to be balanced by the
negative charge of an anion, e.g.:
S}OH#
2 #Cl\ 0 S}OH2Cl [2]
Therefore, in sufRciently acid media, a sorption
of acids takes place similar to the uptake by weakly
basic exchange resins.
In alkaline solutions the surface hydroxyl groups
can dissociate and hand protons over to the liquid
phase:
S}OH 0 S}O\#H# [3]
The negative surface charge has to be balanced by
cations, e.g. by sodium ions:
S}O\#Na# 0 S}ONa [4]
Together the formal equation [3] and [4] represent
a cation exchange on an arbitrary cation exchanger.
Each of the above models holds for the uptake of
acids or the exchange of protons for other mono-
valent cations in arbitrary pure anion or cation ex-
change processes. If the state of equilibrium is con-
sidered for an amphoteric exchanger and a liquid
phase containing different cations and anions,
then the uptake of anions will decrease with increas-
ing pH value until the sorption is completely sup-
pressed. In the same direction the cation exchange of
protons for cations increases. The pH regions of al-
most exclusive cation and anion sorption are separ-
ated by a maximum of noncharged surface groups. In
this transition region sorption of both cations and
anions occurs. At the point of zero charge, therefore,
the surface is covered with equal amounts of equiva-
lents of positively and negatively charged counterions.
Surface Model Assumptions
During the protonation, or dissociation, of the sur-
face OH groups either a negative or a positive surface
 II / ION EXCHANGE / Surface Complexation Theory
potential is generated, depending on the pH and on
the ionic strength of the liquid phase. Therefore, the
respective counterions are subject to an attraction
while the co-ions are rejected. Due to the sorption of
counterions at functional groups, the surface poten-
tial is partly decreased. Unlike the description widely
used in the literature, it is assumed that each kind of
counterion is located at a characteristic distance from
the surface. Thus, ordered double or ‘Stern’ layers are
formed. Ion pairs between surface groups and
counterions in the ordered layer are designated as
surface complexes. Excess charges at the surface are
balanced by counterions in the diffuse layer,
which also contains co-ions. As a consequence, the
surface potential continuously decreases normal to
the surface to zero in the liquid phase. The equilib-
rium between solid and liquid phases is mainly dom-
inated by protolytic reactions. In this way, the pro-
tons are the potential-determining species.
Figure 1 Schematic arrangement of ions at the amphoteric surface of activated alumina and the respective development of electrical
potential.
1641
For a system with Cl\ and Na# ions the schematic
arrangement of counterions and co-ions and the cor-
responding development of the potential are plotted
in Figure 1. With respect to the radii of the hydrated
species, the anion layers are assumed to be closer to
the surface than those of the cations. (Cl\ and Na#
ions are used as model counterions. Individual prop-
erties are not taken into account. Furthermore, the
absolute values of the potential in the different
layers are of no importance for the further mathemat-
ical derivations.)
General Sorption Equilibrium Relationships
for an Amphoteric Surface
The derivation of equilibrium relationships makes use
of the following assumptions:
1. The functional groups are uniformly distributed
across a plane surface.
1642 II / ION EXCHANGE / Surface Complexation Theory

2. Activity coefRcients in the exchanger phase Table 1

are assumed to be 1.
3. Any ion exchange develops as the replacement of Symbol Unit Definition
one surface complex by a new one. As a conse- A0 m2 g\1 Specific surface area
quence, an oxide valence is deRned that is equal to c(i ) mol L\1 Concentration of species i
the smallest common multiple of the valences of c(i )s mol L\1 Concentration of species i in the
the counterions. Stern layer
c(i )x mol L\1 Concentration of species i at
For the derivation of the equilibrium relationships position x in an electrical field
a simple system with Cl\ and Na# ions is considered. C(i, j ) F m\2 Electric capacitance of capacitor
formed by the layers of species
The protolytic reactions at the surface are considered i and j
as local equilibria that can be described by the mass C(S, i ) F m\2 Electric capacitance of capacitor
action law. The formation of the two surface com- formed by the surface and the layer
plexes can be described by the respective formation of species i
constants, K (see Table 1 for explanation of symbols F As Faraday constant
K ij Equilibrium constant of surface
used). reaction
L L Volume of liquid phase
Protonation LS N left-hand side expression of eqns
[38] and [39]
m(H, j ) N Abbreviation, defined by eqn [22]
S}OH#H#
s 0 S}OH#
2 [5] q(i ) mol g\1 Oxide loading with species i
qmax mol g\1 Maximum exchanger loading
c(S}OH# Q ij N Generalized separation factor
2 )
K#
H " [6] R J mol\1 K\1 Gas constant
c(H#)s ) c(S}OH) S g Mass of sorbent
S} Abbreviation used to designate the
where c is the concentration of the species in paren- surface
T K Temperature
theses in mol L\1.
V N Sign of charge
y(i) N Dimensionless loading with species
Dissociation/neutralization i
y(i, i) Probability of the presence of two
# adjacent ions, i at the surface
S}O\#H s 0 S}OH [7] z(i) N Valence of species i

i A s m\2 Surface charge density of layer with
c(S}OH)
H"
K\ [8] ions i
c(S}O\) ) c(H#)s x V Electrical potential at position x

The sorption of counterions leading to two further

surface complexes is considered in an analogous way. liquid phase by means of the Poisson}Boltzmann rela-
tionship:
Sorption of anions

 
z(i)F
S}OH#
2 #Cl\ 0 S}OH2Cl [9] c(i)x"c(i);exp ! ;x [13]
RT
c(S}OH2Cl)
Cl "
K\ [10]
c(S}OH#2 ) ) c(Cl\)s Here c(i)x represents the concentration of species i at
position x in an electric Reld, z(i) is the valence of
Sorption of cations species i, F is the Faraday constant, x is the electrical
potential at position x, R is the gas constant, and T is
S}O\#Na# 0 S}ONa [11] the temperature in Kelvin.
The uptake of anions develops as the simultaneous
c(S}ONa) sorption of protons and chloride ions. Since we have
K#
Na" [12]
c(S}O\) ) c(Na#)s a sequence of reactions, the equilibrium of this com-
mon sorption can be expressed by the product of the
Symbols with subscript ‘s’ designate concentrations in respective formation constants:
the respective Stern layer. These unknown quantities
#
can be expressed in terms of the concentrations in the Cl"KH ;K\
KH Cl [14]
 II / ION EXCHANGE / Surface Complexation Theory 1643

The uptake of cations, however, develops as the com- of the surface with Cl\ and Na# ions. After resolving
petitive sorption of protons and sodium ions. As the equations for the generalized separation factor
a consequence, the equilibrium of the cation ex- one obtains:
change is expressed by the ratio of the corresponding
Cl"log KCl#m(H, Cl);[!y(Cl)#y(Na)]
formation constants: log QH H

[20]
K\
H
Na" #
KH log Q #m(H, Cl) ) y(Cl)
H
[15] Na
KNa
Na#m(H, Na) ) y(Na)]
"log KH [21]
By means of eqns [6]}[15] one obtains:
The terms m(H, i) are given by the following equa-
c(S}OH2Cl) tion:
Cl"
KH
c(S}OH) ) c(Cl\) ) c(H#) 1 qmax ) F 2
m(H, i)" ; [22]
12.303 A0 ) C(S, i) ) RT
 
F
;exp # (S!Cl) [16]
RT where A0 is the speciRc surface area and C(S, i) de-
c(S}OH) ) c(Na ) # notes the capacitance of the capacitor formed by the
Na"
KH surface and layer i. The derivation has been given in
c(S}ONa) ) c(H#)
the literature.

 
F For a system with an arbitrary number of mono-
;exp (S!Na) [17] valent counterions, the following relationship can be
RT
deduced from similar considerations:
Multiplying by the volume of the liquid phase and
dividing by the mass of sorbent, the concentrations of \
i 1

i "log Ki #
m(H, j ) ) V( j ) ) y( j )
log QH H
surface complexes are converted to exchanger load- j"2
ings q(OH), q(Cl\), and q(Na#) respectively; n
(q(OH)) denotes nonionized surface groups. #m(H, i)
V(i) ) y(i) [23]
As a consequence, the Rrst factor on the right-hand i

sides exclusively contains quantities that can be de-

rived from experiments. Both expressions are desig-
nated as generalized separation factors, Q. After
introducing dimensionless loadings with species
i, y(i)"q(i)/qmax, where qmax is the maximum ex-
changer loading, the following expressions are
obtained:
q(Cl)
Cl"
QH
q(OH) ) c(H#) ) c(Cl\)
y(Cl)
" [18]
y(OH) ) c(H#) ) c(Cl\)
q(OH) ) c(Na#) y(OH) ) c(Na#)
QH
Na " " [19]
q(Na) ) c(H#) y(Na) ) c(H#)
Similar relationships can be derived for any mono-
valent counterion.
The difference of electrical potentials in the
exponential terms of eqns [16] and [17] needs consid-
eration about the surface charge densities in the series
of electric capacitors formed by the surface, Stern and
diffuse layers. As has been shown in earlier pub-
lications, the unknown differences in the electri-
cal potential can be expressed in terms of the loading
i"2,32, n index of counterions
j"2,32, i!1 running index
Hydrogen ions are always taken as component ‘1’.
The factors V(j) and V(i) are the signs of the charge of
the ions having the values of !1 for anions and #1
for cations. The Rrst summation comprises counter-
ions from index ‘2’ to ‘i!1’ (closer to the surface
than species ‘i’). The second summation considers the
counterions with indices running from ‘i’ to ‘n’. For
a system with n counterions there are n!1 equations
[23]. By subsequent evaluation of data all constants
log KHi and m(H, i) can be derived from the multicom-
ponent system.
During the exchange of one divalent counterion for
two monovalent ions, one divalent ion replaces two
adjacent monovalent species at the surface, where at
least one divalent species of counterion is present, the
sorbent has to be assumed to have divalent functional
sites in the mathematical treatment (assumption 3).
As a consequence, a surface complex consisting of
two sites and two monovalent counterions is deRned
that is replaced by a surface complex which consists
of two sites with one divalent counterion. For a sys-
tem with H#, Na#, Cl\ and SO24\ the generalized
1644 II / ION EXCHANGE / Surface Complexation Theory

separation factors are then given by the following systems with only monovalent counterions have to be
expressions: converted corresponding to eqns [31] and [32].

Relationships for Pure Cation or Anion

c(S}OH2)2"SO4
IISO4"
QH [24] Exchange on Weak Electrolyte Resins
cS}OH, S}OH ) c(H#)2 ) c(SO24\)
In the case of weakly acidic exchange resins there is
cS}OH2Cl, S}OH2Cl no exchange of anions. Therefore, the terms for an-
IICl"
QH [25] ions in the general equation [24] vanish. For a simple
cS}OH, S}OH ) c(H#)2 ) c(Cl)2
system with H# and Na# ions, eqn [21] simpliRes to:
cS}OH, S}OH
IINa"
QH [26]
Na"log KNa#m(H, Na) ) y(Na)
log QH H
cS}ONa, S}ONa ) c(H#)2 [33]

Here the subscript II refers to calculation under the In a similar way, there is no exchange of cations by
assumption of divalent functional sites. weakly basic exchangers. For the uptake of hydro-
The probability of two adjacent monovalent chloric acid eqn [20] yields:
counterions i at the surface is given by:
Cl"log KCl!m(H, Cl) ) y(Cl)
log QH H
[34]


2
y(i)
y(i, i)" ;
y(j) [27] The uptake of anions by protonated functional

y(j)
groups is equivalent to the exchange of anions for
hydroxyl groups:
where
y( j ) is the sum of dimensionless loadings of
all monovalent counterions with valences #1 and S}OH ) H2O#Cl\ 0 S}O2Cl#OH\ [35]
!1.
By this means the above separation factors can be
A similar theoretical development of the exchange of
expressed as:
OH\ for Cl\ ions leads to the equilibrium relation-
ship:
y(SO4) ) y(H)#y(Cl)#y(Na)
IISO4"
QH [28]
y(OH)2 ) c(H#)2 ) c(SO24\) Cl "log KCl #m(OH, Cl) ) y(Cl)
log QOH OH
[36]

y(Cl)2 Relationships for Strong Electrolyte Exchangers

IICl"
QH ,[QH 2
ICl] [29]
y(OH) ) c(H#)2 ) c(Cl)2
2
Relationships for the description of equilibria with
strongly acidic and strongly basic resins can also be
y(OH)2 ) c(Na#)2
IINa"
QH ,[QH 2
INa] [30] derived from the basic equation [23]. The differ-
y(Na)2 ) c(H#)2 ence, however, is that hydrogen ions cannot be as-
sumed to be adsorbed in the innermost layer. In the
If divalent functional sites are assumed for systems case of the cation exchangers this can be explained by
with exclusively monovalent counterions, then the the fact that the exchanger is completely dissociated,
equilibrium parameters log KH IIi and mII(H, i) can be even in the H# form. Furthermore, hydrated protons
transformed to the parameters for monovalent func- have a rather large diameter and, therefore, cannot be
tional sites according to: assumed to be located close to the surface. Similar
considerations hold for strongly basic anion exchange-
log KH
IIi
rs and hydroxyl ions. For both cases, therefore, the
li "
log KH [31] sequence of layers starts with the kind of counterions
2
which in each is located the closest to the surface.
mII(H, i)
ml(H, i)" [32]
2 Liquid Phase Equilibrium Relationships
The liquid phase equilibrium is calculated by means
The subscripts I and II refer to the assumption of of:
either monovalent or divalent functional sites. If
a system contains monovalent and divalent counter- E the dissociation of water;
ions the entire calculation has to assume divalent E the condition of electroneutrality; and
functional sites. Equilibrium parameters derived from E mass balances for each component.
 II / ION EXCHANGE / Surface Complexation Theory 1645

Figure 2 Development of the generalized separation factor for the weakly acidic ion exchange resins (A) LEWATIT CNP 80 and
(B) AMBERLITE IRC 50 (by courtesy of Marcel Dekker Inc.).

Figure 3 Development of the generalized separation factor for the weakly basic ion exchange resin AMBERLITE IRA 93 (by courtesy
of Marcel Dekker Inc.).
1646 II / ION EXCHANGE / Surface Complexation Theory

Activity coefRcients in the liquid phase can be

calculated from the ionic strength using the extended
Debye}HuK ckel relationship.

Experimental Evaluation of
Equilibrium Data
General
Exchange equilibria have been determined for pure
cation and anion exchangers as well as for am-
photeric materials such as activated alumina and ac-
tivated carbon. The pretreatment of these materials
and the experimental conditions needed for obtaining
proper equilibrium data have been described in detail
in several publications. Cation concentrations can be
measured by titration, photometric methods or by
atomic absorption spectroscopy. Nitrate concentra-
tions are measured photometrically, chloride and sul-
fate concentrations by ion chromatography.
In the case of pure anion or pure cation exchange,
equilibria samples with increasing amounts of ex-
changer material are contacted with a Rxed volume of
the solution at constant temperature until equilibrium
is obtained. Experiments with amphoteric equilibria
are carried out at constant concentrations in the
liquid phase and constant quantities of activated
alumina or activated carbon in each of the samples.
The volume/mass ratio is adjusted so that there is
a substantial degree of sorption, as well as an equilib-
rium concentration that can be measured easily and
with sufRcient accuracy. Since the state of equi-
librium is mainly dominated by the pH of the liquid
phase, the pH value is varied in each series of
samples. Details are described in the literature.
Evaluation of Data
For evaluation of equilibrium parameters the general-
ized separation factors have to be determined from
equilibrium concentrations and resin loadings. These
are then plotted against the respective dimensionless
resin loading or the (algebraic) sum of resin loadings.
In the case of binary systems with pure cation or anion
exchange, the respective graphical representations dir-
ectly yield the equilibrium parameters. The intersec-
tion of the linear relationship leads to log K and the
slope of the straight line yields m(i, j). Figures 2}5
show examples for different types of exchangers.
The results show that linear relationships are ob-
tained in most cases. Systematic deviations are found
for small loadings. This is attributed to the neglect of
counterions in the diffuse layer. As has been
demonstrated by Horst for simple systems, when this
layer is considered, excellent agreement is obtained.
Figure 4 Development of the generalized separation factor for
the exchange of calcium for sodium ions on two different strongly
acidic resins. (A) Dowex HCRS; (B) LEWATIT S100.
atively larger than for other resin types. Evaluation of
data for the exchange of sulfate for chloride on
strongly basic exchangers shows that there are two
different straight lines. There are clearly two
different layers that are subsequently ‘Rlled’.
Evaluation of data for a broad range of total concen-
trations up to 100 mmol L\1 in the liquid phase re-
veals identical development of the linear relation-
ships. For the entire range of ionic strength investi-
gated the parameters can, therefore, be considered as
‘constant’ equilibrium constants.
Evaluation of amphoteric data requires a more
sophisticated approach. The Rrst step is the assump-
tion of a certain sequence of layers. Evaluation
of data starts with the evaluation of the equilibrium
for the uptake of protons and ions in the Rrst layer.
The respective parameters are required for evaluation
of the sorption of species in the next layer. Deter-
mination of the parameters for the third layer re-
quires the sets of parameters for both the inner layers.
This process has to be continued until all sets of
binary parameters are derived. The sequence of
a layers is assumed correctly if both the log K and
m(H, i) values show a steady increase with increasing
distance. If this conditions is not fulRlled the calcu-
lation must be repeated with a modiRed sequence.
Considering a system of layers H#, Cl\, NO\ 3 , Na
#

For strongly acidic exchangers this region is compar- on activated alumina as an example, the equilibrium
 II / ION EXCHANGE / Surface Complexation Theory 1647

Figure 5 Development of the generalized spearation factor for the strongly basic ion exchange resin AMBERLITE IRA 410
(Reprinted from HoK ll, Horst, Franzreb and Eberle (1993) by courtesy of Marcel Dekker Inc.).

parameters can be evaluated from the following rela- NO3!m(H, Cl) ) !y(Cl)
log QH
tionships:
"log KH
NO3#m(H, NO3) ) y(Na)!y(NO3) [38]
log Q "log K #m(H, Cl)
H
Cl
H
Cl
Na!m(H, Cl) ) !y(Cl)!m(H, NO3)
log QH
) y(Na)!y(Cl)!y(NO3) [37]
) !y(NO3)"log KH
Na#m(H, Na) ) y(Na) [39]

Figure 6 Development of the generalized separation factor for

the uptake of nitrate by activated alumina (COMPALOX
AN/V800) from a multicomponent system (cation"Na# or K#). Figure 7 Development of the generalized separation factor for
The symbols represent different experiments. the sorption of sodium on COMPALOX AN/V800.
1648 II / ION EXCHANGE / Surface Complexation Theory

Figure 8 Development of the generalized separation factor for

the sorption of carbonate species by activated carbon (NORIT
ROW 0.8 Supra) from a multicomponent system.

Results of the evaluation of equilibrium data for the

#
sorption of Cl\, NO\ 3 and Na on activated alumina
(COMPALOX AN/V800) are plotted in Figures 6
and 7. Although the results for the uptake of sodium
ions are not satisfactory, the data demonstrate that
linear relationships are obtained.

Prediction of Multicomponent
Equilibria
General
Prediction of arbitrary exchange equilibria requires
a set of conditions to be fulRlled:
E conservation of mass
E electroneutrality on the liquid phase
E electroneutrality on the resin phase
E adjustment of equilibrium for the exchange of ions
E dissociation equilibria in the liquid phase.
Figure 10 Comparison of experimental data and predicted
equilibria for the ternary system H#/Mg2#/Ca2# on the weakly
acidic resin LEWATIT OC 1046.
As can easily be shown, the number of equations is
always greater than the number of unknown quantit-
ies. Therefore, some conditions are fulRlled automati-
cally. The remaining set of coupled nonlinear equa-
tions must be solved by appropriate numerical
methods.
Respective results for the uptake of inorganic ions
by an activated carbon are plotted in Figures 8 and 9.
With few exceptions, the equilibrium data satisfac-
torily fall on linear relationships as predicted by the
theoretical approach.

Pure Cation or Anion Exchange

Figure 9 Development of the generalized separation
factor for the sorption of chloride ions by activated carbon
(NORIT ROW 0.8 Supra) from a multicomponent system.
LQ "log Q (H, Cl)# m II(H, CO 3) y (CO 3)! m II(H, Ca) y (Ca)!
mII(H, mg)y(mg).
For prediction of multicomponent equilibria it is con-
venient if the binary parameters for the exchange of
different counterions for the same reference ion
have been deduced. For pure cation or anion ex-
change systems consideration of a three-component
equilibrium with ions. A (reference ion), B and

Figure 11 Comparison of experimental data and predicted
3 /NO\
equilibria for the ternary system Cl\/HCO\ 3 on the strongly
basic resin AMBERLITE IRA 410.
C leads to the relationships:
log QAB"log KAB#m(A, B) ) [y(B)#y(C)]
log QBC"log KBC#m(B, C) ) y(C)
As has been demonstrated earlier, the binary para-
meters for the exchange of A versus B are the same as
in the pure binary case. Also, the binary parameters
for the exchange of A for C have been evaluated. In
the above system, therefore, the parameters for the
exchange of B for C are unknown. However, they can
easily be calculated: the unknown parameter log KBC
results from the deRnition:
log KBC"log KAC!log KAB
For derivation of the unknown slope we make use of
the fact that the sequence of layers can be considered
as a series of layers of an electric capacitor. Since the
II / ION EXCHANGE / Surface Complexation Theory 1649
deRnition of m(i, j) contains the electric capacitance,
C, we can make use of the well-known relationship
for capacitances in series:
1 1 1
" # [43]
C(A, C) C(A, B) C(B, C)
Thus, one obtains:
m(B, C)"m(A, C)!m(A, B) [44]
Similar relationships are obtained for systems
with more than three counterions. Examples of
a comparison between experimental results and equi-
libria predicted from binary data are given in
Figures 10}12.
For amphoteric sorbents the individual equilibrium
parameters are already determined during evaluation
of data. Therefore, no additional calculations are
required. Figure 13 shows the comparison between
experimentally determined oxide loadings and devel-
opments calculated by means of the set of equilibrium
parameters determined previously. The data are plot-
ted as relative loadings (q(i)/qmax) as a function of pH.
The system contained the four components H#, Cl\,
SO24\ and K#. Obviously, sulfate ions are preferred
over chloride species. The constant sulfate loadings
for pH values below 5 are due to the constant initial
sulfate concentration in the system. Only minor load-
ings with monovalent cations are observed. There is
excellent agreement with the predicted loadings.
Figure 14 shows the results from one experiment
with activated carbon and protons, carbonate, sul-
phate, calcium, magnesium and potassium as
counterions. The relative loadings with both carbon-
[40] ate and nitrate ions decrease with increasing pH of
the solution, while the loadings of calcium and potas-
[41] sium slightly increase. Nevertheless their uptake
[42]
Figure 12 Comparison of experimental data and predicted
equilibria for the ternary system H#/Ca2#/Cu2# on the chelating
resin LEWATIT TP 207. C0 (HCl)"10 mmol L\1; C0
(CuCl2)"10 mmol L\1. (From Horst and HoK ll (1997), copyright
Academic Press.)
1650 II / ION EXCHANGE / Surface Complexation Theory
Figure 13 Comparison of experimental data and predicted loadings of activated alumina for a multicomponent system. Activated
alumina: COMPALOX AN/V800.
remains small, indicating that there is only a small
capacity for the uptake of cations. Again, there is
good agreement with the predicted developments.
Applications
The principal advantage of the surface complexation
theory lies in the fact that it can be applied to any
normal or reaction-coupled ion exchange process for
which other theoretical descriptions of equilibria fail.
One Reld of application is the prediction of multi-
species exchange kinetics. This has been demon-
strated by Franzreb, who included the description
into the solution of the Nernst}Planck equations for
Figure 14 Comparison of experimental data and predicted loadings of activated carbon with inorganic species. Activated carbon:
NORIT ROW 0.8 Supra.
several ternary and quaternary cases. The develop-
ment of systems with strong and weak exchange
resins and with normal and reaction-coupled multi-
species processes could be predicted with excellent
accuracy.
The surface complexation theory can also be easily
applied to the simpliRed prediction of the perfor-
mance of Rlters in which the Rlter is divided into
a series of equilibrium stages. By means of the poros-
ity of the Rlter section the quantities of resin material
and liquid phase are known. Consequently, the ex-
change equilibrium in each stage can be calculated.
The efSuent composition of the Rlter is given by
the equilibrium in the last stage. By this means and
by an appropriate assumption of the number of

Figure 15 Comparison of experimental (䢇, 䉱, 䊏) and pre-
dicted (*) breakthrough curves during softening/dealkalization
of tap water. Resin: AMBERLITE IRC 50, regenerated by car-
bonic acid. Volume of resin, 1 L; throughput, 6.9 BV h\1;
feed: Ca2#"2.44 mmol L\1; Mg2#"0.48 mmol L\1; HCO\ 3 "
4.42 mmol L\1.
equilibrium stages a fairly good agreement between
experimental and predicted concentration histories
can be obtained. Figure 15 shows the comparison
between experimental data and predicted develop-
ments of concentrations for a system with a weakly
acidic resin which is applied for softening/dealkaliz-
ation of tap water. Clearly a very satisfactory agree-
ment is obtained.
Further Reading
Davis JA, James RO and Leckie JO (1978) Surface ioniz-
ation and complexation at the oxide/water interface.
Theory of Ion Exchange
R. Harjula, University of Helsinki, Helsinki,
Finland
Copyright ^ 2000 Academic Press
Proper theoretical tools are necessary at all stages in
the development of ion exchange applications. First-
ly, in the laboratory, when a new application is being
developed, theory is helpful in the design of new
materials and necessary for the interpretation and
quantiRcation of research results. Secondly, theory is
needed in the design of industrial and laboratory
processes. When the process is being operated, theory
is again useful to keep the process under optimal
control.
II / ION EXCHANGE / Theory of Ion Exchange 1651
Computation of electrical double layer properties in
simple electrolytes. Journal of Colloid and Interface
Science 63: 480}499.
Franzreb M, HoK ll WH and Sontheimer H (1993) Liquid-
phase mass transfer in multi-component ion exchange. I.
System without chemical reactions in the Rlm. Reactive
Polymers 21: 117}133.
Franzreb M, HoK ll WH and Eberle SH (1995) Liquid-phase
mass transfer in multi-component ion exchange. 2.
Systems with irreversible chemical reactions in the Rlm.
Industrial and Engineering Chemistry Research 34:
2670}2675.
HoK ll WH, Horst J and Wernet M (1991) Application of the
surface complex formation model to ion exchange equi-
libria. II. Chelating resins. Reactive Polymers 14:
251}261.
HoK ll WH, Horst J and Franzreb M (1993) Application of
the surface complex formation model to ion exchange
equilibria. III. Anion exchangers. Reactive Polymers 19:
123}136.
HoK ll WH, Horst J, Franzreb M and Eberle SH (1993)
In: Marinsky J and Marcus Y (eds) Ion Exchange and
Solvent Extraction, a Series of Advances, vol. 11, ch. 3,
p. 151. New York}Basel}Hong Kong: Marcel Dekker
Inc.
Horst J and HoK ll WH (1997) Application of the
surface complex formation model to ion exchange equi-
libria. IV. Amphoteric sorption onto -aluminium
oxide. Journal of Colloid and Interface Science 195:
250}260.
Horst J, HoK ll WH and Eberle SH (1990) Application of the
surface complex formation model to ion exchange equi-
libria. I. Reactive Polymers 13: 209.
Stumm W (1992) Chemistry of the Solid}Water Interface.
New York: Wiley and Sons.
Ion exchangers are used in very diverse applica-
tions in the laboratory and in industry. Considering
this, and the fact that a very wide range of differ-
ent substances can act as ion exchangers, it is not
surprising to see in the literature that various sets of
nomenclature, conventions and diverse theories exist,
reSecting the special features of the applications or
materials involved. For instance, one Rnds rather dif-
ferent deRnitions for several basic terms of ion ex-
change in the Reld of water puriRcation compared to
those used in ion exchange chromatography and even
diversity in the terminology can be seen within the
Reld of ion chromatography. It is inevitable that
various theories must exist for the functioning
of different types of ion exchangers, but the
1652 II / ION EXCHANGE / Theory of Ion Exchange

discrepancies in the most basic deRnitions only lead Selectivity coefVcient (selectivity quotient)
to confusion and misconception that deter the utiliz-
ation of the theory. CM ZABC ZB A
The theory of ion exchange is a vast subject. kA/B" ZA ZB [2]
CM B C A
This article presents and discusses most of the
essential theory and concepts in connection with the
where CM s are the concentrations of the ions in
most common applications. There is emphasis on
the exchanger and Cs those in solution. Various
discussing details that are important in interpreting
concentration units are commonly used (molar,
correctly research results and in predicting and opti-
molal, equivalent fraction, etc.). When zAOzB, the
mizing ion exchanger performance. For simplicity,
numerical value of kA/B depends on the choice of the
the basic equations are written for cation exchange
concentration units. The selectivity coefRcient
but they are applicable to anion exchange as well,
(kA/B) usually changes as a function of exchanger
with minor modiRcations. Theories for the prediction
composition (kA/B"f (CM A); see Figure 1) and also as
of cation exchange selectivity are also discussed
a function of the total concentration (or ionic
brieSy.
strength) of the external solution, especially in con-
centrated solutions.
Basic Concepts
Corrected selectivity coefVcient By making the
Ion Exchange Equilibria activity (nonideality) correction for the solution
phase, the so-called corrected selectivity coefRc-
A binary ion exchange reaction between ion A
ient is obtained:
(charge zA) and ion B (charge zB) may be written as:
CM ZABaZB A CM ZABCZB A ZB A ZB A
zBA #zABM
ZA ZB
 zAB #zBAM
ZB ZA
[1] k
A/B" ZA ZB" ZA ZB ZB"kA/B ZB [3]
CM B aA CM B CA A A

where superscript bars refer to the ions in a solid ion
exchanger. Various equilibrium quantities are used to
measure and estimate the efRciency of the ion
exchanger for a given separation task. The most com-
mon of these include the following.
where aA, aB are the activities of ions A, B in the
solution. This quantity is independent of the total
concentration of the external solution by deRnition
and thus reSects the pure exchanger}ion interactions
contributing to the selectivity. In dilute solutions,

Figure 1 Selectivity coefficient kNa/H (see eqn [2]) for Na#/H# exchange is sulfonated polystyrene/divinylbenzene (DVB) resins as
a function Na# equivalent fraction in resin at different degrees of cross-linking (nominal DVB content). Circles, DVB 5.5%; squares,
DVB 15%; triangles, DVB 25%. (Data from Helfferich FG, 1995.)

k
A/B+kA/B. However, the corrected selectivity
coefRcient usually varies with the exchanger
composition (CM A).
Thermodynamic equilibrium constant The thermo-
dynamic equilibrium constant KA/B :
aN ZABaZB A CM ZABCZB A N ZB A ZB A N ZB A
KA/B" ZA ZB" ZA ZB ZB ZB"k
A/B ZB
aN B aA CM B CA N A A N A
can be obtained by integrating the corrected selectiv-
ity coefRcient as a function of exchanger com-
position. According to Argesinger et al. and HoK gfeldt
et al.:

1
KA/B" ln KH dEM A
0 processing capacity) and in ion chromatography (es-
where EM A is the equivalent fraction of A in the ex-
changer (EM A"zACM A/(zACM A#zBCM B)) and KH is the cor-
rected selectivity coefRcient written with mole
fractions (XM ) as concentration units for the ions in the
exchanger:
XM ZABCZB A ZB A
KH" ZA ZB ZB
XM B CA A
Gaines and Thomas supplemented the abstract
thermodynamic treatment to include the contribu-
tions of salt imbibition and water activity changes,
which need to be considered when ions are exchanged
in concentrated solutions.
Distribution coefVcient (distribution constant,
distribution ratio) Various distribution constants
and coefRcients are used to measure the ion ex-
change equilibria. In general, the distribution coef-
Rcient kd of ion A is deRned as a concentration ratio in
the exchanger and solution:
CM A
kd"
CA
This quantity is only a constant under special con-
ditions. In general, kd depends on the ionic composi-
tion of the exchanger and the solution. For a binary
exchange, one obtains from eqns [2] and [7] that:
ZA


1
CM B ZB
kd"k ZB
A/B
CB
Under the special condition that A is present in the
solution and in exchanger at much lower concentra-
II / ION EXCHANGE / Theory of Ion Exchange 1653
tion than B (CACB, CM ACM B), kA/B and CM B are essen-
tially constant (CM B+Q, the ion exchange capacity)
and:
1 ZA
log kd" log (kA/BQZA)! log CB [9]
ZB ZB
Under these circumstances, kd depends only on the
[4]
concentration of ion B and, on a logarithmic scale,
the slope of kd equals !zA/zB, the ratio of cation
charges.
Experimentally determined graphs of log kd vs
log CB (eqn [9]) are frequently used in research to
study sorption mechanisms, the charges of the ex-
changing species and in the estimation of exchanger
[5] performance, e.g. in water puriRcation (estimation of
timation of retention volume). Great care should be
taken, however, in the interpretation of the data and
in making sure that the assumptions leading to
eqn [9] are valid. Because of the widespread use of
distribution coefRcients in ion exchange, it is use-
ful to emphasize this point by taking a binary uni-
univalent exchange (zA"zB"1) as an example here.
For this equilibrium, eqn [9] can be further manipu-
[6]
lated to give:
Q
kd" [10]
CB
#CA
kCs/Na
This equation now shows that, in fact, the
condition for linear dependence of kd
on CA (log kd"log (kA/BQ)!log CB) is that
CACB/kA/B. Thus, even if CACB, the dependence
may not be linear if the selectivity coefRcient is
very large. This feature of kd is shown as calculated
examples in Figure 2. It can be seen that, if the selec-
tivity coefRcient is low, kd falls linearly with the
concentration of the macro-ion B, on a logarithmic
[7] scale with a slope of !1, as eqn [9] implies. How-
ever, when the selectivity increases, the kd starts to
level off at lower concentrations of B and ulti-
mately becomes independent of CB when the selectiv-
ity is very high. In the studies of highly selective
exchangers (zeolites and some other inorganic mater-
ials, chelating resins) such independence of kd on the
macro-ion concentration is often observed and every
now and then the incorrect conclusion is made that
the uptake of trace ions is not ion exchange but some
[8]
sort of surface adsorption reaction. Figure 2 also
shows an interesting feature of the link between
kd and selectivity: in dilute solutions the kds tend
to a common value, which is determined by the ratio
1654 II / ION EXCHANGE / Theory of Ion Exchange
Figure 2 Calculated values (eqn [10]) for the distribution coefficient kd for trace ion A (CA"10\6 mol L\1) in binary univalent A#/B#
exchange as a function of macro-ion concentration CB at different values of the selectivity coefficient kA/B. Dotted line, kA/B"5 000 000;
dashed line, kA/B"50 000; continuous line, kA/B"500. Ion exchange capacity of the exchanger 2.0 mmol g\1.
of Q/CA. The value of selectivity thus becomes
unimportant in dilute solutions.
In general, eqn [9] is valid for several parallel trace
ion exchange reactions (A, C, D2) in the presence of
one common macro-ion B, since the ions that are
present at trace level will have a negligible effect
on other trace ion equilibria.
Separation factor Separation factor is usually used
in ion exchange chromatography to estimate the sep-
arability of two trace ions. Considering the separ-
ation of two trace ions A and C using macro ions B as
an eluent, one obtains for the separation factor (A/C):
1 ZA



CM ACC kd(A) kA/B ZB
CM B ZC
A/C" " "
CACM C kd(C) kC/B CB
ZA
"Constant/C ZC
B [11]
In the case that zAOzC, the separation factor in-
creases as the concentration of B is decreased.
Ion exchange isotherm An ion exchange isotherm is
a function that represents the ionic composition of
the exchanger (EM A) as a function of the ionic composi-
tion of the solution (EA), or vice versa, at constant
temperature (Figure 3). Traditionally, the selectivity
of the exchanger is estimated from the isotherm. If the
isotherm is concave towards the axis representing the
ion concentration in the exchanger, the ion exchanger
is considered to be selective for that ion (curves a and
b in Figure 3). If the isotherm lies on the diagonal of
the presentation (EM A"EA), the exchanger has no
preference for either ion A or B (curve c) and a bend-
ing of the isotherm towards the EA-axis indicates that
the exchanger is nonselective. The magnitude of the
selectivity coefRcient cannot be always deduced
from the isotherm because when zAOzB, the shape of
the isotherm depends strongly on the total ion con-
centration (CT) in the solution. This behaviour arises
from the difference between the cation charges,
which can be clearly seen if the equation for the
selectivity coefRcient is expressed in terms of
equivalent fractions and rearranged:
EM ZAB EZAB
" T \
kA/BC (ZB ZA)
[12]
EM ZB A EZB A
At a given point EM A on the isotherm (the left-hand
side of eqn [12] constant), the ratio EA/EB must de-
crease as CT is decreased when zA'zB. Thus, the
relative concentration of ion A must decrease with
decreasing CT. This feature, the increased preference
of an ion exchanger for the ion having a higher charge
with the dilution of the solution, is called electroselec-
tivity. It should be also noted that the ion exchanger
may prefer ion A strongly even though the value of
the selectivity coefRcient is equal to or less than
unity (see calculated examples in Figure 3).
Ion Exchange Kinetics
The rate of ion exchange is governed by the various
diffusion processes in the system. In general,
 II / ION EXCHANGE / Theory of Ion Exchange 1655

Figure 3 Calculated ion exchange isotherms for a hypothetical di-univalent (A2#/B#) exchange having a selectivity coefficient
kA/B"1. The isotherms have been generated using constant total ion concentrations: 1, 0.01 mol L\1; 2, 0.1 mol L\1; 3, 1 mol L\1; 4,
5 mol L\1. The isotherms show increasing preference of the ion exchanger for ion A with increasing dilution of the solution
(electroselectivity effect.)

diffusion can be described by Fick’s Rrst law. The
Sux of ion A(JA) is given by:
JA"!D grad CA
where D is the diffusion coefRcient. This
equation describes purely statistical diffusion
that is driven by the concentration gradient. In ion
exchangers it is usually necessary to consider also the
electric potential (
) and then the Sux of ion A (JA) is
given by the Nernst}Planck equation:
Nernst, often satisfactorily describes the diffusion
processes at solid}solution interfaces. The internal dif-
fusion of the ions in the exchanger phase is called
[13] particle diffusion. Most often, the particle is con-
sidered homogeneous, so that the different dif-
fusional processes within the particle (pore diffu-
sion, matrix diffusion) are represented by a single
particle diffusion coefRcient. Either particle
or Rlm diffusion may be the rate-determining step
for the exchange process or both may contribute to the
rate in intermediate cases. In general, Rlm diffu-
sion may dominate at early stages of exchange (A low
in exchanger, B low in solution) when the concentra-


F tion gradient in the particle is large (fast rate in particle).
JA"!D grad CA#zACA grad
[14]
RT
where F is the Faraday constant. Most commonly, the
kinetics of ion exchange reactions are interpreted in
the terms of external or internal diffusion. As the
external solution is usually agitated, there is essentially
no concentration gradient in the bulk of the external
solution. Gradients arise, however, within a thin layer
of solution adhering to the surface of the exchanger
particle. Diffusion across this layer is called Rlm
diffusion. This concept, developed Rrst by
However, as the exchange proceeds further, the concen-
tration gradients in the particle decrease and particle
diffusion may become the rate-determining step.
In the exchange process, ions A and B move in
opposite directions. Therefore, generally, the so-called
interdiffusion coefRcient (DA/B) must be used
in eqns [13] and [14]. For particle diffusion:
DM ADM B(z2ACM A#z2BCM B)
DM A/B" 2 [15]
zACM ADM A#z2BCM BDM B
1656 II / ION EXCHANGE / Theory of Ion Exchange
Normally, the interdiffusion coefRcient is
not constant, but changes with the ionic composition
(CM A) of the exchanger. To calculate ion exchange
rates in a given ion exchange system, the
Nernst}Planck equations must be solved simulta-
neously for each diffusing species under bound-
ary conditions speciRc to the system. In general, the
resulting equations are nonlinear differential
equations, which have analytical solutions only in
some special cases. Such a case, for instance, is iso-
topic ion exchange, for which a so-called self-dif-
fusion constant can be used. Assuming also that the
solution has indeRnite volume } the concentration of
ion in the solution remains essentially constant } it is
obtained for the half-time (t1/2) of the exchange reac-
tion. In the case of particle diffusion:
r20
t1/2"0.030 [16]
DM
where r0 is the radius of the exchanger particle and
DM is the particle diffusion coefRcient. For
Rlm diffusion:
r0Q
t1/2"0.23 [17]
DC
where  is the Rlm thickness, Q the ion exchange
capacity, D the Rlm diffusion coefRcient and
C is the ion concentration. For particle diffusion
(eqn [16]), the rate of exchange increases (t1/2 de-
creases) as the particle radius decreases, being pro-
portional to 1/r20. For Rlm diffusion (eqn [17]),
the rate increases less strongly as r0 is decreased (the
proportionality is to 1/r0). In Rlm diffusion
the exchange rate can also be increased by increasing
the efRciency of agitation, which will decrease the
Rlm thickness. In real applications of ion exchange
the exchange rates do not usually follow the simple
relationships of eqns [16] and [17] and the equations
are presented here just to give a simple view of the
factors that can affect the ion exchange rates.
The values of Rlm diffusion coefRcients
are of the same order of magnitude as the diffu-
sion coefRcients of ions in the external salt solu-
tion (D+10\5 cm2 s\1). The values of particle dif-
fusion coefRcients depend strongly on the charge
of the ion and on the structure and porosity of the
exchanger matrix. In sulfonated polystyrene resins
DM decreases with increasing degree of cross-linking,
being in the range of 10\5}10\7 cm2 s\1 for univalent
cations. For multivalent cations, the values are much
lower, falling in the range of 10\7}10\10 cm2 s\1. In
weakly acidic resins and in crystalline inorganic
ion exchangers, cation diffusion coefRcients
are typically 2}5 orders of magnitude lower.
Basic Ion Exchange Operations
Ion exchange reactions can be carried out as either
batch or column operations. Column operation is far
more common and efRcient than batch operation.
Batch operation is however used in research, because
the experiment is simple to carry out and a large
number of experiments can be carried out in parallel.
Batch Operation
In batch operation, a given amount (m) of ion ex-
changer is contacted with a given volume (V) of
solution. The mixture is agitated until equilibrium
has been attained. In typical binary batch process
used for ion exchange studies (e.g. determination of
kA/B as a function CM A, Figure 1), the exchanger is
initially in a homoionic form (e.g. the B form) and the
solution initially contains only the ion A. Considering
the simple uni-univalent exchange as an example, the
ratio of ion concentrations at equilibrium is given by:


CM A CA
"kA/B [18]
CM B CB
In general, to achieve a high conversion to the
A form in a single batch equilibration, the selectivity
coefRcient kA/B and the ratio CA/BB must be high.
In practice, the degree of conversion to the A form is
controlled by adjusting the initial concentration of
A (CA0) in the solution and the solution to solid ratio
(V/m), often also called the batch factor (BF):
V
CM A"(CA0!CA) [19]
m
At constant CA0, the degree of conversion increases
as V/m is increased (Figure 4). It is usually difR-
cult to obtain a high conversion in a single batch
equilibration, since the selectivity often decreases
with increasing conversion and there is always ion
B in the solution released from the exchanger. Re-
moving B from the solution can enhance conversion.
This can be achieved by equilibrating the ion ex-
changer successively with fresh portions of
solution A.
Column Operation
There are several types of column operation, classi-
Red according to the technical design of the apparatus

Figure 4 Batch ion exchange: calculated degree of conversion from the B# form to the A# form as a function of solution volume to
exchanger mass ratio (batch factor). Selectivity coefficient kA/B"1, ion exchange capacity 2 mmol g\1. The solid line represent the
conversion in single batch equilibration. The broken line represents successive batch equilibrations with a constant batch factor of
10 mL g\1.
(e.g. Rxed-bed or Soating-bed operation) or
according to the purpose of the application (e.g. col-
umn chromatography or column separation).
Column separation Column separation usually in-
volves elimination of undesirable ions from water
(deionization, softening, decontamination). Taking
the binary exchange discussed earlier as an example,
a solution containing harmful ions (A) is passed
through the column that contains an exchanger in the
B form. The A ions are then taken up by the ex-
changer and B ions are released into the solution.
Because B ions are constantly removed from the sys-
tem, the operation is much more efRcient than
batch exchange in removing A ions from the solution
(see eqn [18]). The column efSuent is Rrst free of
ion A, but when a given amount of solution has been
passed through, A starts to emerge in the efSuent
and its concentration increases gradually to that in
the inSuent solution (Figure 5). The graph of the
concentration of A in efSuent as a function of
efSuent volume is called the breakthrough curve.
The area above the breakthrough curve gives the
total volume of solution that has been freed from ion
A. Dividing this volume by bed mass or volume gives
the total processing capacity, or theoretical capacity
(QT), of the column (L kg\1 or L L\1). In this simple
example, QT"Q/CA (Q"ion exchange capacity in
mmol L\1 or mmol mL\1) since at equilibrium A ions
have taken up all of the ion exchange capacity. In
general, QT is equal to kd, which can be easily cal-
culated in binary systems from eqn [9] for trace ions
II / ION EXCHANGE / Theory of Ion Exchange 1657
to be separated, provided that selectivity coefR-
cients are known. However, because ion A is con-
sidered harmful, operation is not continued until total
processing capacity has been used, but the feed is
discontinued when the concentration of A in the
efSuent reaches a measurable or a regulated
value. The capacity at this point is called the process-
ing capacity, or breakthrough capacity (QB). The ra-
tio QB/QT is called the column utilization factor, FU.
For efRcient separation process FU should be
maximized.
Column chromatography In column chromatogra-
phy ions are separated from each other for analysis or
for chemical production purposes. Considering
a simple example in which ions A and C are separated
for analysis, a sample solution containing A and C is
passed into the column containing an exchanger in
the B form. The sample volume is so low that A and
C take up only a very small fraction of the column
capacity near the inlet. After sample injection, an
eluent solution containing ion B is passed through the
column. A and C in the exchanger are exchanged for
B and begin to move through the column at dif-
ferent velocities. At a given volume, the less preferred
ion A Rrst emerges in the eluent as a concentration
peak followed by ion C. The eluent volumes at which
A and C emerge, i.e. the volumes at the peak maxima,
are called the retention volumes (VR) and they can be
obtained from the relation:
VR"kdVS#VM [20]
1658 II / ION EXCHANGE / Theory of Ion Exchange

Figure 5 Examples of column breakthrough curves generated for different number of theoretical plates N (see eqns [22]}[27]). In this
example, the capacity of the exchanger (Q ) is 1 mmol mL\1 and the exchanger bed volume is 1 mL. The exchanger is initially in the
B# form and the feed contains only ion A# at a concentration of 0.001 mmol mL\1 (CA). The total processing capacity QT is thus
Q/CA"1000 mL mL\1 exchanger and the area above the breakthrough curves is 1000 mL for the 1 mL bed. The breakthrough
capacity QB depends on N, which is affected by the operating conditions. Continuous line, N"30; broken line, N"10.

where VS is the volume of ion exchanger bed and H0, HP, HF and HL are the contributions of particle
VM is the free solution volume in the bed. In analytical size, particle diffusion, Rlm diffusion and longit-
separations A and C are present at trace levels, so udinal diffusion to the effective height:
kd values are again easily calculated from eqn [9]. In
analytical work, efRcient operation requires that H0"1.64 r0 [23]
the concentration peaks of A and B are well separated
(the peaks are sharp). The retention volumes k
d 0.14 r20 u
HP" [24]
VR should not be too large, because this leads to (k
d#) 2
DM
a long analysis time and to broadening of the peaks.


2
k
d 0.266 r20 u
Theory of Column Exchange HF" [25]
k
d# D(1#70r0u)
Most models for ion exchange column operation are
based on the concept of effective plates or trans- D(2
fer units. Martin and Synge Rrst used this concept for HL" [26]
u
chromatography and the theory was reRned by
Glueckauf, who obtained for the material balance in
where u is the linear Sow rate. The number of theoret-
the ion exchange column, under linear equilibrium
ical plates (N) in the column is then obtained by
(kd constant):
dividing the column length L by H:




Ci Ci H 2Ci
#q(k
d#) ! "0 [21] L
z V V Z 2 z2 V N" [27]
H
where z is the longitudinal coordinate in the column,
q is the cross-sectional area of the bed, k
d is the In general, the column performance improves as
column distribution coefRcient in which the ex- N is increased. In column separations the increase in
changer phase concentration of species i is calculated N makes the breakthrough curve steeper (break-
per unit volume of the bed, i.e. k
d is obtained from the through capacity QT increases). In column
kd of eqn [7] as k
d"kd (1!), where  is the bed chromatography, increase in N makes the elution
void fraction. H is the effective height of the peaks sharper and so increases the separation of two
theoretical plate given by: peaks. For a column with a constant length Z, N can
be increased by decreasing the plate height H. The
H"H0#HP#HF#HL [22] easiest way of doing this is to decrease the particle

radius (see eqns [23]}[25]), but there is a practical
lower limit for r0, because the hydrodynamic pressure
increases with decrease in r0. Another way to increase
N is to decrease Sow rate (eqns [24] and [25]), but
a very low rate is not usually acceptable due to
the effect of longitudinal diffusion (eqn [26]),
which causes poor separation.
Equilibrium Theories
Much theoretical work has been carried out to ex-
plain the nonideality of the ion exchange systems, e.g.
for the calculation of ion exchange equilibria and to
understand the factors that give rise to ion exchange
selectivity. For ions in solution, sufRcient theories
exist to calculate the nonideality (activities) in the
liquid phase. For the ions in the exchanger phase, no
generally valid theories exist. For the calculation of
ion exchange equilibria, it is always possible to
measure the nonideality of the exchanger phase, e.g.
the corrected selectivity coefRcient k
A/B as a func-
tion of exchanger ion composition at a given total
solution concentration (CT) and then use the mea-
sured function for the calculation of equilibria at
other CT values from equation:
EM ZB A EZB A ZB A (ZA ZB)
k
A/B " CT \ [28]
EM ZAB EZAB ZAB
In general, k
A/B and A/B are not known simulta-
neously, so iteration must be used to solve eqn [28].
The same approach can be extended to systems con-
taining more than two counterions, e.g. for the ter-
nary system the corrected selectivity coefRcients
k
A/BC, k
B/CA and k
C/AB can be measured and used for
the calculation of the equilibria. For instance, k
A/BC is
deRned as:
EM 2Z
A
BZC
CZB AZCCZCAZB ZB AZCZCAZB
k
A/BC" ZAZC ZAZB 2ZBZC [29]
EM B EM C CA 2Z
A
BZC
for the exchange reaction:
2zBzCA#zAzCBM #zAzBCM  2zBzCAM #zAzCB#zAzBC
[30]
k
B/CA and k
C/AB are deRned accordingly. It is intrin-
sic to this method that it gives precise results provided
that the selectivity coefRcients are measured and
described precisely within the exchanger composition
range of interest. In practice this requires large num-
ber of measurements, which makes the method very
laborious. Less effort is associated with ap-
II / ION EXCHANGE / Theory of Ion Exchange 1659
proaches in which the solid-phase activity coefR-
cients in multicomponent systems are estimated from
the binary interaction parameters (), e.g. from the
Wilson equation it is obtained that:



M M
XM ki
k
ln N i"1!ln  XM iij !  [31]
j"1 k"1  M
XM jkj
j"1
Parameter  can be determined from the measure-
ments of the corrected selectivity coefRcients of
the binary equilibria i/j, k/i and k/j by curve Rtting.
Activity coefRcients obtained by eqn [31] are
then used in the binary equations of thermodynamic
equilibrium constant (eqn [4]) for the calculation of
ion exchange equilibria. This method has given accu-
rate results even in four-component systems (e.g.
Na/K/Ca/Mg in a strong acid cation resin). Several
other related approaches have been developed.
The theories above are based on the measurements
of nonideality and make no assumptions about the
interactions that give rise to the selectivity. Thus, they
do not allow the calculation of KA/B, k
A/B or N i from the
fundamental data or explain the changes of k
A/B or N i.
The Rrst theory to explain the nonideality of the
exchanger phase was developed by Kielland (the
graphical presentations of Figure 1 are often called
Kielland plots), who considered van der Waals-type
interactions and showed that for the solid-phase
activity coefRcients:
N AR"cXM 2AR [32]
N BR"cXM 2BR [33]
Here, AR and BR denote the salt forms of the
exchanger, R being the common anion. This theory
predicts that the function k
A/B"f (XM A) is linear,
which is in agreement with the observed behaviour in
many cases. Quite often, the k
A/B functions are not
linear, but slightly, or even strongly, curved (see Fig-
ure 1). Some of these nonlinear functions can be ex-
plained by assuming that the exchanger has several
types of exchange sites, each subsite having a charac-
teristic selectivity coefRcient k
i . The measured
overall k
A/B decreases with XM A as sites with higher
selectivity are Rlled Rrst. If these subsites behave
ideally, sigmoidal curves are obtained for
k
A/B"f (XM A). By assuming nonideal behaviour for the
subsites, the k
A/B functions exhibit a wide variety of
different forms. This theory has been found to be
consistent with the behaviour of several zeolite systems.
A related approach is to consider the different
states that a given counterion may assume depending
on the neighbouring counterions. In a polymer chain
1660 II / ION EXCHANGE / Theory of Ion Exchange
(as in an organic ion exchange resin), considering the
two nearest neighbours, each counterion can have
three different energetic levels. As a conse-
quence, in general k
A/B"f (XM A) is a second-order
polynomial function in XM A, which is often the ob-
served trend in organic resins (Figure 1). If two of the
three energy levels are close to each other, the selectiv-
ity function is linear. Theories of this type are helpful
in the calculation of ion exchange equilibria and in
presenting the equilibria in a mathematical form, but
they give no information about the magnitude of selec-
tivity. In organic resins, various osmotic theories have
been developed to estimate the relative magnitude of
selectivity. The base in these theories is that:

KA/B" (zB B!zA A) [34]
RT
where  is the osmotic pressure difference be-
tween the external solution and exchanger pore liquid
and A and B are the partial molar volumes of A and
B in the exchanger. The osmotic theory predicts the
selectivity trend (I):
Cs#'Rb#'K#'Na#'Li#
observed in strong-acid cation resins, i.e. ions with
smaller hydrated radius (smaller partial molar vol-
ume) are preferred, because replacing larger ions with
smaller ones will reduce the swelling pressure. The
same selectivity trend can also be predicted from
purely electrostatic calculations. The dielectric the-
ory:



!e2 zA zB 1 1
ln KA/B" ! ! [35]
8 kT rA rB Z S
where Z and S are the macro-permittivities of the
exchanger and solution phases, respectively, predicts
that in uni-univalent exchange, the selectivity de-
creases as the framework charge density increases for
selective exchange (KA/B'1). This trend is commonly
observed for zeolite ion exchange.
The selectivity sequence I for alkali metal ions,
shown above, is common in organic resins have a low
degree of cross-linking and in zeolites with low
framework charge density. Other selectivity se-
quences appear as the degree of cross-linking or
framework charge density increases:
II. Cs'K'Rb'Na'Li
III. K'Cs'Rb'Na'Li
IV. K'Cs'Na'Rb'Li
V. K'Na'Cs'Rb'Li
VI. K'Na'Rb'Cs'Li
VII. Na'K'Rb'Cs'Li
VIII. Na'K'Rb'Li'Cs
IX. Na'K'Li'Rb'Cs
X. Na'Li'K'Rb'Cs
XI. Li'Na'K'Rb'Cs
Most of these sequences have been observed in ion
exchangers and they can be predicted from Eisenman
theory, originally developed for selective glass elec-
trodes. The theory considers cation exchange site and
cation water (hydration) interaction energies. The
free energy of exchange is obtained from:
F 0AB"(FM elA!FM elB)!(FM hyd
A !FM B )
hyd
[36]
where F el is the coulombic interaction energy between
cation and the anionic exchange site and Fhyd is the
hydration energy of the cation. The coulombic inter-
action energy for a univalent cation can be calculated
for widely separated sites from:
FM el"!332/(r##r ) [37]
\
and for closely spaced sites:
FM el"!1.56H332/(r##r ) [38]
\
where r# is the cation radius and r is the radius of
\
the anionic exchange site. The anionic Reld strength
decreases as r increases. Selectivity pattern I is ex-
\
hibited by exchangers having a low Reld strength and
cations are exchanged in the hydrated state with
a preference for a smaller hydrated radius. As the Reld
strength is increased, the less hydrated cations be-
come desolvated and the selectivity patterns start to
change. At high Reld strength, pattern XI is exhibited
and cations are exchanged as bare cations.
Conclusions
Rather simple theoretical concepts are available to
describe ion exchange phenomena and applications in
a qualitative manner. In some cases these concepts
may give a good quantitative agreement, but gener-
ally more rigorous theories are required, considering
the speciRc details of given systems. The application
of even the simplest theories usually involves much
experimental and computational effort when
systems comprising more than two exchanging ions
are involved.
See also: II/Ion Exchange: Historical Development.
 II / MASS SPECTROMETRY / Spectrometry^Mass Spectrometry Ion Mobility
Further Reading
Dorfner K (1991) Ion Exchangers. Berlin: Walter de
Gruyter.
Franklin KR and Townsend RP (1988) Prediction of multi-
component ion exchange equilibria in zeolites: a com-
parison of procedures. Zeolites 8: 367.
Helfferich FG (1995) Ion Exchange. New York: Dover
Publications.
Lehto J and Harjula R (1996) Proceedings of the
Workshop on Uniform and Reliable Formulations.
MASS SPECTROMETRY
Spectrometry^Mass
Spectrometry Ion Mobility
H. R. Bollan, DERA Bridgwater, Bridgwater, UK
Copyright ^ 2000 Academic Press
Introduction
The principle of mass spectrometry (MS) is the separ-
ation of ions in a vacuum, using an electrical or
magnetic Reld or a combination of both. The ions
may be formed through a variety of processes, but it is
perhaps the fragmentation of the molecular ion that
produces much of the analytical power of the tech-
nique. Mass-to-charge ratios are recorded and the
structure of the parent ion may be determined from
the ion molecular mass and the pattern of the fragment
ions recorded. Experienced mass spectrometrists can
recognize typical fragment ion patterns, however, al-
though there are libraries available for the automated
identiRcation of mass spectra, careful judgement must
be used in the Rnal assignment of the compound’s
identity. The theory and uses of MS have been well
documented as an analytical technique both as a stand-
alone and a hyphenated technique, for example
coupled with gas chromatography (GC-MS).
Less is known about the chemistry within ion mo-
bility spectrometers, which are used in the Reld to
monitor for contraband substances such as explos-
ives, drugs, and on the battleReld to detect chemical
warfare agents. Originally referred to as plasma
chromatography, ion mobility spectrometry (IMS) is
a technique concerned with the formation of ion-
molecule clusters in air and their movement in an
electric Reld, at or close to atmospheric pressure. The
average ion velocity of an ion species in an electric
Reld, vd, is the product of that electric Reld, E, and
1661
Nomenclature and Experimentation for Ion Exchange.
Special Issue of Reactive and Functional Polymers
27: 93.
Marinsky JA and Marcus Y (1966}97) Ion Exchange,
A Series of Advances, vols 1}13. New York: Marcel
Dekker.
Mehablia MA, Shallcross DC and Stevens GW (1996) Ter-
nary and quaternary ion change equilibria. Solvent Ex-
traction and Ion Exchange 14: 309.
Recommendations on Ion Exchange Nomenclature (1972).
Pure and Applied Chemistry 29: 619.
a constant of proportionality, K, i.e. vd"KE. K is
called the mobility of the ions, and is characteristic of
a particular ion species in a speciRed drift gas. K may
be calculated indirectly from drift time, td, from the
equation td"ld/vd, where ld is the drift length. The
theory of ion mobility and reaction chemistry is
covered in two monographs listed in the Further
Reading section, and need not be reproduced here.
Notably, the Mason}Schamp equation for mobility
(an equation that attempts to reconcile fundamental
properties of ions with their mobility) includes a term
containing a collision integral, to which mobility is
inversely proportional. The value of the collision inte-
gral is determined by the cross-section. Therefore, the
mobility, and consequently the ion drift velocity, is
dependent upon mass, size, shape, and polarizability.
The mobilities observed for ions are weighted aver-
ages of the mobilities of all the cluster ions participat-
ing in a localized equilibrium between the ion swarm
and the neutral molecules they encounter as they
traverse the drift region. If the drift gas, electric Reld
gradient, temperature, pressure, and therefore the
molecular number density remain constant, mobility
depends only on ion charge, reduced mass, and colli-
sion cross-section. The collision processes undergone
by ions during their drift time are very complicated,
and are much too complicated to go into here. How-
ever, it must be noted that these processes are
affected by variations in temperature and pres-
sure in the drift region. Ion cluster formation and
fragmentation are also governed by temperature.
Therefore, to simplify the situation, and to allow easy
comparison between different systems, mobility
of an ion is normalized for temperature and pressure,
the corrected term being referred to as reduced mobil-
ity, K0 ( 0 in some texts).
The initial distribution of ions immediately follow-
ing ionization is modiRed by various chemical
1662 II / MASS SPECTROMETRY / Spectrometry^Mass Spectrometry Ion Mobility
reactions, forming more stable ions. In clean air, these
ions form what is called a reactant ion peak (RIP).
Positive ion chemistry can involve proton transfer,
nucleophilic attachment, hydride or hydroxide ex-
traction, and oxidation; negative ion chemistry in-
volves electron capture, charge transfer, dissociative
capture, proton abstraction, and electrophilic attach-
ment; both positive and negative chemistries can be
subject to complex rearrangements.
When a sample atmosphere enters the ion mobility
spectrometer, many competitive reactions occur and
to a Rrst approximation proton or electron afRn-
ities may deRne the reaction pathways. These compet-
ing species may be target or possible interference
compounds. Ion mobility spectrometers respond to
a broad range of compounds with various functional
groups. Therefore, complicated spectra are common
in ion-molecule systems based upon water chemistry,
due to the relatively low proton afRnity of the
water molecule. Selectivity may be improved with the
introduction of trace quantities of an appropriate
dopant chemical into the detector carrier gas, thereby
altering the degree of afRnity required for reac-
tion. This can have an effect on resolution, sensi-
tivity, response and recovery times.
Whilst ion mobility spectrometers respond to many
compounds, in the Reld the operator is only able to
identify the compound being detected, by an ion
mobility spectrometer, from its display. The efR-
cacy of the instrument display depends upon calib-
ration and software programming. However, as the
observed peaks represent cluster ions participating in
a localized equilibrium, even in the laboratory, with
instrumentation capable of displaying the mobility
spectra, the accurate identiRcation of species may be
difRcult.
Although identiRcation of unknowns by IMS alone
is problematic, the coupling together of IMS and MS
(IMS-MS) produces a powerful technique. The mass-
es of ion-molecule clusters forming the RIP and prod-
uct ion peaks are recorded either in positive or nega-
tive mode mass spectra, depending on the polarity of
the ions being studied. When tuned ion analysis is
performed on a speciRc mass in the mass spectrum,
the mobility of the ion mass can be determined, i.e. its
position in the mobility spectrum. With the technique
enhanced, further by coupling IMS to tandem MS,
the composition of ion-molecule clusters can be iden-
tiRed from the results of collision-induced dissocia-
tion (CID).
History
In the 1960s, Cohen (of the Franklin GNO Corpora-
tion) worked on the development of the ion mobility
spectrometer, resulting in a US patent in 1971. The
instruments were developed to generate information
concerning negative ions produced from speciRc com-
pounds in air under atmospheric pressure conditions.
This early instrumentation was to have wider applica-
tion for the analysis of ultra-trace quantities of many
organic molecules forming either positive or negative
ions. Cohen went on to form the company PCP, and
to produce commercially available IMS-MS instru-
mentation.
By the early 1970s, Karasek was already employing
IMS as part of a hyphenated technique, using IMS-
MS to determine the identity of species separated
through a GC. Even without the GC in-line, it was
becoming evident that IMS-MS was a powerful iden-
tiRcation technique.
In the late 1970s and the 1980s, IMS research was
directed from fundamental studies to application re-
search, with a view to solving speciRc analytical prob-
lems relating to the rapid detection of volatile organic
compounds in the Reld. SpeciRcally, IMS was the
subject of military research programmes, designed to
enhance the real-time detection of chemical warfare
agents. IMS-MS still played an important role in
understanding the ion-molecule chemistry, which
was necessary to progress the development and relia-
bility of the Reld deployable IMS instrumentation.
A study of the ion-molecule behaviour of selected
agents and interference compounds has been made by
IMS-MS. IMS-MS has also been used to support some
of the IMS programmes that have been applied to
more general, as well as speciRc, monitoring require-
ments. Industrial applications have been directed to-
wards monitoring for toxic chemicals, and chemicals
considered to be hazardous to man or the environ-
ment. These include acid and stack gases (e.g. hydro-
gen Suoride), aliphatic and aromatic amines, ketones,
isocyanates, halogens, solvents, ethers, anaesthetics,
fuels, nicotine, polychlorinated biphenyls, mixtures
of organic compounds, organophosphorus com-
pounds, certain hydrocarbons (e.g. benzene), per-
Suoroisobutene, and volatile organic compounds
used in the semiconductor industry. With social
pressures for a greener environment, further IMS
techniques are being developed, for example, for the
identiRcation of polymers, using laser ablation-IMS,
to assist with sorting plastics for recycling. The need
to detect pollutants in liquid media is becoming more
desirable, e.g. the detection of aniline in hexane, and
of aqueous ammonia in rivers, wastewaters and
drinking water treatment facilities. This problem re-
quires a means of separating the analyte from the
liquid medium, usually in the form of a selectively
permeable membrane. IMS analysis proceeds once
the analyte has been transferred from the liquid to the
 II / MASS SPECTROMETRY / Spectrometry^Mass Spectrometry Ion Mobility 1663

vapour phase. IMS has also been investigated for ture is inadmissible. Drugs detected include heroin,
the detection of bacteria in water and wastewater cocaine, barbiturates, amphetamines, and LSD. Pre-
sources, using pyrolysis before introduction. scription drugs such as benzodiazepines are also de-
Detection of explosives is a speciRc and very impor- tected by IMS. The pressure to be absolutely certain
tant area of contraband detection at trace levels: about the identity of target compounds emphasizes
RDX, TNT, PETN, NG, EGDN, HMX, EGDN, 2,4- the advantages of powerful analytical techniques such
DNT, ammonium nitrate, and tetryl detection by IMS as IMS-MS, which enhances the development and
have all been investigated. Detection of illicit drugs calibration of detection equipment.
by IMS is another important area, which has bene- A more speciRc application is of potential use to the
Rted from conRrmation of detected species by IMS- forestry industry, which involves the identiRcation
MS analysis. Due to legal requirements for forensic of different timbers before processing. Fast ther-
and law enforcement purposes, alleged criminals molysis-IMS has proved successful for certain wood
charged with the clandestine manufacture of illegal species. Wetwood, an abnormal condition of wood
drugs in the USA must be charged with the manufac- from both deciduous and coniferous trees caused by
ture of speciRc drugs in order for the case to go to bacterial infection, was detected in Northern Red
court; a blanket charge of clandestine drug manufac- Oaks using this technique.
Although IMS-MS is necessarily a laboratory-
based technique, it continues to play a very important
Table 1 History
part in the understanding of ion-molecule chemistry
1890s Studies of the ionization of air and the development of IMS equipment that, through
1897 Study of the velocity of positive rays in an electric field user requirements, is becoming miniaturized. Hill
(Thomson & Rutherford) continues to develop hyphenated IMS techniques,
1900s Langevin studies of ionized air and mathematical/
sometimes employing different ionization methods,
chemical models for ion mobilities
1907 First mass spectrometer including electrospray IMS-MS. The ion-molecule
1919 Determination of atomic weights using MS chemistry of different ionization techniques can be
1946 Time of flight mass analyser characterized readily by IMS-MS techniques. These
1953 Quadrupole MS fundamental studies have led the way for IMS-MS
1960s Drift tube mass spectrometers
research into biomolecular sciences because, until the
1965 First IMS instruments (plasma chromatographs) were
developed (Franklin GNO Corporation) application of electrospray ionization, the bio-
1966 Studies of chemical ionization molecules had been too large for successful ionization
1967 Tandem MS by more traditional methods employed for IMS.
1968 Electrospray ionization for MS Clemner and Jarrold were able to further the research
1970s Research led by Cohen and Karasek using both IMS
by determining the conformation of biomolecules by
and IMS-MS techniques
1971 First IMS patent issued in the USA (Cohen MJ, US IMS.
patent 3,621,239) A chronology of the history of IMS-MS is given in
1974 Atmospheric pressure ionization MS Table 1.
1977 IMS-MS demonstration of ion-molecule behaviour of
chemical warfare agents
1982 High field asymmetric waveform IMS-MS (FAIMS-MS) An Ion Mobility Spectrometer^
1983 IMS-MS studies of reactant ion distributions in an ion Mass Spectrometer
mobility spectrometer with a membrane inlet system
1984 GC-IMS An ion mobility spectrometer consists of an ion-mol-
Corona discharge IMS ecule reaction chamber, incorporating an ionization
1985 IMS-MS identification of structurally different ions of the
region, coupled to a drift region via a shutter grid.
same mass
1986 IMS-MS analysis of prescription and illicit drugs A schematic diagram is shown in Figure 1. The drift
1987 Laser desorption IMS-MS region contains a screen grid and an ion collector.
1990 IMS-MS used to study the site of protonation in anilines A typical cell consists of metal guard rings, separated
IMS-MS used in the semi-conductor industry to deter- by insulators, connected to a resistance network with
mine cleanliness
a high voltage attached to one end of the resistor
1992 IMS-MS analysis of marijuana vapours and cigarette
smoke chain, to produce a uniform electric Reld along the
1994 IMS-MS confirmation of ion-molecule clusters dis- cell, usually of the order of a few hundred V cm\1.
closed Clean carrier gas is ionized by irradiation, usually
1997 Development of ESI-high resolution IMS-MS (Wash- with beta particles from a 63Ni radioactive source, to
ington State University)
form positive and negative reactant ions and conse-
Development of ion-trap-IMS-TOFMS systems for the
elucidation of biomolecular structures quently RIPs. The ion-molecule chemistry can be al-
tered by the introduction of a dopant chemical at
1664 II / MASS SPECTROMETRY / Spectrometry^Mass Spectrometry Ion Mobility
Figure 1 Schematic of an ion mobility spectrometer.
a controlled rate. Samples introduced into the ion
mobility spectrometer may react to form product
ions, the equilibrium concentrations of which are
governed by proton afRnity or electron afRn-
ity. If introduced into an electric Reld they will
migrate according to their polarity and that of the
applied Reld as, between collisions, individual ions
have a component of acceleration in the direction of
the applied Reld. Ions pass from the reaction region to
the drift region via a shutter grid, which is pulsed
open to allow a Rnite number of ions to enter the drift
region. Operation of the shutter grid starts the timing
sequence, which measures drift time. A counter-Sow
of clean drift gas enters the drift region near the
collector electrode, which is shielded by a screen grid
in order to prevent induced charge, which would lead
to a distorted current peak. By monitoring the collec-
tor electrode from the instant the voltage pulse is
applied to the grid, a mobility spectrum (see Figure 2)
is generated. Mobility spectra can be generated con-
Figure 2 An ion mobility spectrum.
tinuously by repetitive pulsing of the grid. Typically,
25 ms is sufRcient time to allow all ions to drift
from the grid to the collector electrode. The signal-to-
noise ratio is relatively noisy because only small ion
currents are involved. The signal-to-noise ratio may
be improved by averaging the signal over several
scans.
In a mass spectrometer, molecules are ionized by
any one of a number of techniques. These ions are
then analysed using either magnetic or electric Relds
or a combination of both and are separated according
to their mass-to-charge ratio before being detected. In
mass spectrometers using magnetic Reld separation,
a repeller plate directs ions to a set of accelerator
plates, used to produce a beam of rapidly moving
ions, which are directed into a uniform beam by
focussing slits. Neutral molecules are drawn off
by vacuum pumps. In a quadrupole mass spectrom-
eter, an oscillating electrostatic Reld is set up between
four rods, two diagonally opposite rods having direct
current voltage applied and the other two rods having
radio frequency applied. Ions acquire an oscillation in
the electrostatic Reld set up according to the ratio of
the direct current to the radio frequency amplitude.
Ions of the correct m/z value undergo a stable oscilla-
tion of constant amplitude and pass through the
analyser to reach the detector. Other ions undergo
unstable oscillation and the amplitude of the oscilla-
tion increases until the ions strike one of the rods.
Current IMS-MS Applications
APCI-MS enables ion chemistry at pressures used in
typical IMS systems to be studied, but some issues
remain regarding cluster formation in the interface
region and this could inSuence the interfacing of IMS
 II / MASS SPECTROMETRY / Spectrometry^Mass Spectrometry Ion Mobility
Figure 3 An ion mobility spectrometer}mass spectrometer.
(which operates close to atmospheric pressure) to
MS. However, Spangler has recently published details
regarding a better understanding of the behaviour at
the IMS-MS interface.
An ion mobility spectrometer may be coupled to
a mass spectrometer (see Figure 3) with sample trans-
fer via a pinhole, typically 50
to 100
in diameter.
The mass spectrometer used in conjunction with an ion
mobility spectrometer enables m/z identiRcation of the
reactant and product ions. The mass spectrometer is
initially programmed to scan through the chosen mass
range with the IMS shutter grids continuously open. (If
the ion mobility spectrometer is used in the normal
pulsed mode it may take a very long time to obtain
a mass spectrum, which may then not be representa-
tive.) Thus, it is possible to record ions created in an
ion mobility spectrometer, and a mass spectrum of
IMS sample ions is shown in Figure 4.
The mass spectrometer is then programmed to
detect ions of one chosen mass. In this ‘tuned’
Figure 4 A mass spectrum of IMS sample ions.
1665
ion mode (with the IMS shutter grids operating nor-
mally), a drift spectrum of the selected ion species
is generated (see Figure 5). Thus, it is possible
to associate a particular ion mass with a particular
ion mobility peak. Hence, IMS-tuned MS enables the
reduced mobility for ions to be determined, but the
signals are weak and a signiRcant amount of data
averaging is required. Because the signals are very
low, mass spectrometers used in conjunction with
ion mobility spectrometers are set to pulse counting
mode. Sometimes the average of thousands of spectra
is necessary to produce a mobility peak. Using IMS-
quadrupole MS to determine the reduced mobilities
of all the ion-molecule clusters in a mobility spectrum
could take from several hours to days. IMS-TOF is
much faster, because it is able to scan at 50 to
60 Hz about 1000 scans per mobility spectrum. Ew-
ing and Stone have an IMS-tuned MS for investigat-
ing the kinetic thermodynamic relationship for the
ion reactions.
1666 II / MASS SPECTROMETRY / Spectrometry^Mass Spectrometry Ion Mobility
Figure 5 A drift spectrum of a selected ion species.
Figure 6 A mass spectrum of CID product ions.
IMS-MS-MS studies can be performed when IMS is
coupled to a triple quadrupole mass spectrometer. An
ion selected by the Rrst quadrupole can be injected
into a collision gas, for example argon, in a second
quadrupole at 2;10\5 Torr (subjecting the cluster
ions to CID), and then the product ions can be ana-
lysed in the third quadrupole. Figure 6 shows a mass
spectrum of CID product ions. Consequently, MS-
MS analysis is used extensively in assigning ion ident-
ities. IMS coupled to triple quadrupole MS enables
the composite identiRcation of the ion-molecules
found in a drift tube. However, the number of ions
reaching the detector in IMS-MS-MS is extremely
low and a large amount of averaging is required to
determine structures.
See Colour Plate 46.
See also: I/Mass Spectrometry. II/Chromatogra-
phy: Gas: Gas Chromatography-Mass Spectrometry.
III/Biomedical Applications: Gas Chromatography-
Mass Spectrometry; Drugs and Metabolites: Liquid
Chromatography-Mass Spectrometry; Explosives: Gas
Chromatography; Liquid Chromatography; Thin-Layer
(Planar) Chromatography; Forensic Toxicology: Thin-
Layer (Planar) Chromatography; Forensic Sciences:
Capillary Electrophoresis; Liquid Chromatography; Heroin:
Liquid Chromatography and Capillary Electrophoresis.
Further Reading
Carr TW (ed.) (1984) Plasma Chromatography. New York:
Plenum Publishing Corporation.
Cohen MJ and Karasek JW (1970) J. Chromatogr. Sci. 8:
330.
Eiceman GA and Karpas Z (1994) Ion Mobility Spectro-
metry. CRC Press Inc.
Hill HH Jr, Siems WF, St. Louis RH and McMinn DG
(1990) Ion mobility spectometry. Analytical Chemistry
62: 1201A}1209A.
Knighton WB and Grimsrud EP (1996) Advances in Gas
Phase Ion Chemistry. JAI Press Inc.
 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
MEMBRANE SEPARATIONS
Bipolar Membranes and
Membrane Processes
H. Strathmann, University of Twente, The Netherlands
Copyright ^ 2000 Academic Press
Bipolar membranes are gaining increasing attention as
an efRcient tool for the production of acids and
bases from the corresponding salts by electrodialytic
water dissociation. The process is economically very
attractive and has a multitude of interesting potential
applications. The large scale utilization of bipolar mem-
branes, however, is still limited today by unsatisfactory
membrane properties and by a lack of application
know-how. A bipolar membrane should have adequate
water dissociation capability, low electrical resistance,
high permselectivity and a long useful life under operat-
Figure 1 Schematic diagram illustrating the principle of electrodialytic water dissociation in bipolar membranes.
1667
ing conditions, which means that it must be stable in
highly concentrated acid or alkaline solutions. The
monopolar anion and cation exchange membranes
which are also needed in the process should have
good proton and hydroxide ion-blocking capability
in addition to stability in strong bases and acids.
Although today’s membranes do not meet all of
these required properties, they are used successfully in
a number of relevant applications.
The Principle of Electrodialytic
Water Dissociation
The process of electrodialytic water dissociation us-
ing a bipolar membrane is illustrated in Figure 1,
which shows a bipolar membrane consisting of cation
and anion exchange layers arranged in parallel be-
tween two electrodes. If an electrical potential dif-
ference is established between the electrodes, charged
species are removed from the interphase between the
1668 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
two ion exchange layers. When all salt ions are re-
moved from the interphase region, further transport
of electrical charges can only be accomplished by
protons and hydroxide ions, which are available in
a concentration of ca. 1;10\7 mol L\1. Bipolar mem-
branes resemble a laminate of a cation and an anion
exchange layer with a very thin (4}5 nm) transition
region in which the water dissociation occurs accord-
ing to the water dissociation equilibrium given by:
2H2O 8 H3O##OH\
The reversible energy required for the production of
acids and bases in a bipolar membrane at constant
temperature and pressure can be calculated by the
Nernst equation for a concentration chain of solutions
with different H# ion activities, i.e. pH values:

G"F )
U"2.3 RT
pH
Figure 2 Schematic drawing illustrating the principle of electrodialytic production of acids and bases from the corresponding salts
with bipolar membranes.
Here
G is the reversible Gibb’s free energy,
U
the electrical potential difference between the
two solutions, R is the gas constant, T is the absolute
temperature, F is the Faraday constant, and
pH is
the difference between the pH values of the two
solutions separated by the bipolar membrane. For
1 mol L\1 acid and base solutions in the two phases
separated by the membrane,
U is 0.8 V and
G is
0.02 kWh mol\ at 253C.
The potential drop across the bipolar membrane
[1] measured in a water dissociation experiment is al-
ways higher than the calculated theoretical value be-
cause of irreversible effects due to the electrical
resistance of the membrane and the solutions.
To utilize bipolar membranes for the production of
acids and bases from the corresponding salt solution
they must be combined with monopolar ion exchange
membranes, as illustrated in Figure 2. This schematic
[2] drawing shows bipolar and cation and anion
 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
exchange membranes arranged in parallel between
two electrodes to form individual compartments. The
electrodialysis cell arrangement consists of three indi-
vidual compartments and three membranes, i.e. the
bipolar and the cation and anion exchange mem-
brane. As in conventional electrodialysis, a large
number of the three-compartment units can be
stacked between electrodes. When a salt solution is
introduced in the middle compartment and an electri-
cal potential difference between the electrodes is
established, the cations in the salt solution migrate
towards the cathode. They permeate the cation ex-
change membrane and form a base with the hydrox-
ide ions generated in the bipolar membrane. On the
other side of the bipolar membrane protons, which
are generated simultaneously with the hydroxide
ions, form an acid with anions migrating from the salt
solution through the anion exchange membrane to-
wards the anode. The net result of the process is the
production of an acid and a base from the corre-
sponding salt solution.
The Mechanism of Water Dissociation
in Bipolar Membranes
The water dissociation rate in the bipolar membrane
determines the overall efRciency of the process. It
can easily be shown, however, that the dissociation
rate constant of pure water is much too low to ex-
plain the experimentally determined high acid and
base generation rate in bipolar membranes.
As indicated earlier, a bipolar membrane consists
of a laminate of cation and anion exchange layers.
The speciRc resistance  of a strong acid or base ion
exchange layer is in the order of 50}100  cm. As-
suming a thickness of 100
m each for the cation and
anion exchange layers, the total area resistance r of
the ion exchange layers of the bipolar membrane is in
the order of 1}2  cm2.
The electrical resistance of the interphase layer of
a bipolar membrane which is assumed to consist of
deionized water can be calculated by:
 two solutions outside the bipolar membrane of about
rin" [3]
 14. A further small increase in the voltage then leads
where rin is the area resistance,  the thickness, and
 is the speciRc conductivity of the interphase layer.
If the interphase layer contains only pure water, its
speciRc resistance is approximately 18;106  cm.
Thus, the area resistance of a 1 nm thick interphase is
approximately 1.8  cm2. The above argument how-
ever is only correct if the ion concentration in the
interphase is constant and all ions which are removed
1669
by the electric current across the bipolar membrane
are replenished by the water dissociation. This means
that the ion Suxes from the bipolar membrane into
the outer phases cannot exceed the rate of their gen-
eration in the interphase. Thus, the maximum Sux
of H# and OH\ ions of the bipolar membrane is
given by:
JH>"JOH\"kdCH2O  [4]
where J is the maximum ion Sux from the bipolar
membrane into the outer phases, kd is the water
dissociation rate constant, CH2O is the concentration
of water in the interphase and  is the thickness of the
interphases. The subscripts H#, OH\ and H2O refer
to H#, OH\ ions and water, respectively.
The water dissociation rate constant kd for
pure water at 253C is given in the literature as
2.5;10\5 s\1.
According to eqn [4], the maximum Suxes JH#
and JOH\ from a bipolar membrane that has a
1 nm thick interphase of pure water would be
1.4;10\13 mol cm\2 s\1.
The electrical current I through the bipolar mem-
brane is proportional to the sum of all ion Suxes and
is given by:
I"F( JH##JOH\) [5]
Thus, the maximum current density through a bi-
polar membrane is, according to eqns [4] and [5],
approximately 1.4;10\8 A cm\2. A current density
exceeding this value would lead to a depletion of ions
in the interphase and thus to a drastic increase in its
electrical resistance. In practice, however, bipolar
membranes can be operated at current densities in
excess of 0.1 A cm\2, as demonstrated in Figure 3A,
which shows the current through a bipolar membrane
as a function of the applied voltage. When an increas-
ing voltage difference across a bipolar membrane
is established, the current hardly increases until the
voltage drop reaches a value of about 0.8 V, corre-
sponding to the concentration potential calculated by
eqn [2] for a pH value difference between the
to a drastic increase in the current density to values in
excess of 0.2 A cm\2. Thus, the current}voltage
curves determined with bipolar membranes show two
plateau values that indicate a limitation in the current
with increasing voltage drop across the membrane, as
depicted in Figure 3B. The Rrst plateau value indi-
cates a limitation of the current density due to a lim-
itation of ions in the interphase. However, at 0.8 V
accelerated water dissociation begins and the current
1670 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
Figure 3 Schematic diagram of the current density as a function of the applied voltage (A) determined with a typical commercially
available bipolar membrane and (B) three distinct areas of operation with bipolar membranes.
is no longer limited by a lack of ions until the second
plateau value is reached at ca. 0.2 A cm\2. Water
dissociation is then limited by the supply of water to
the interphase.
Thus, there are three distinct regions in the opera-
tion of a bipolar membrane. In the Rrst region the
current is very low and mainly transported by salt
ions. In the second region, high water dissociation
occurs and the current is transported by protons and
hydroxide ions generated in the interphase. In the
third region the production rate of protons and hy-
droxide ions is limited by the water transport rate
into the interphase. Operation of bipolar membranes
at current densities that exceed the second plateau
value leads to destruction of the membrane.
The experimentally determined current densities
indicate that the simple model of a bipolar membrane
depicted in Figure 1 is incorrect. Either the water
dissociation rate is faster by several orders of magni-
tude in the bipolar membrane than in free solution or
the interphase is much thicker. A thick interphase,
however, would lead to a high area resistance of the
interphase, which is not the case. From scanning
electron microscope photographs and calculations
based on the Poisson and Bolzmann relation for the
space charge at an interphase between differently
charged ion exchange membranes, it can be con-
cluded that the thickness of the interphase is less than
5 nm. This means that, in bipolar membranes the
water dissociation is at least 106 times faster than in
free solution.
Various mechanisms have been suggested to ex-
plain the accelerated water dissociation in bipolar
membranes. One possible explanation, suggested by
Wien, is that at high electric Reld densities the ion
mobility as well as the degree of dissociation of weak-
ly dissociated electrolytes increases with increasing
Reld density. The increase in the dissociation constant
of weak electrolytes by the electric Reld effect
can be expressed by an increase in the water dissocia-
tion rate constant, assuming that the recombination
rate of H# and OH\ ions is unaffected.
Other theoretical considerations and experimental
evidence support a hypothesis that the accelerated
water dissociation is caused by a reversible proton
transfer reaction between charged groups and water.
This means that, in the presence of certain ionic
groups, the water dissociation rate constant may be
several orders of magnitude higher than in pure
water. In the case of the bipolar membrane the anion
exchange groups of the membrane polymer adjacent
to the interphase layer are assumed to react with the
water molecules at the membrane surface as follows:
k2
B#H2O 0 BH#OH\ and
k
\2
k3
BH##H2O 0 B#H3O# [6]
k 3
\
where B is a neutral base, e.g. a tertiary ammonium
group.
Both models can explain the acceleration of the
water dissociation in the interphases between the
anion and cation exchange layer of the bipolar mem-
brane and serve as theoretical basis for the develop-
ment of bipolar membranes.
The Preparation and Performance
of Bipolar Membranes
The properties required of bipolar membranes to be
useful in practical applications are low electrical
resistance at high current density, high water dis-
sociation rates, low co-ion transport rate, high ion
selectivities, good thermal and, most importantly,
good chemical stability since the cation-selective layer
of the bipolar membrane is in direct contact with an
 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
acid and the anion exchange layer with an alkaline
solution.
Low electrical resistance of the cation and anion
exchange layer of the bipolar membrane can be ob-
tained by using a strong acid, such as sulfonic acid
groups, and a strong base, such as quaternary am-
monium groups as Rxed charges in high concentrations
in the polymer matrix. To minimize the resistance of
the interphase between the cation- and the anion-selec-
tive layers the thickness of this interphase must be as
thin as possible, as indicated earlier. There are various
ways to prepare bipolar membranes with satisfactory
properties. Most commonly, membranes are prepared
as laminates with some kind of interphase which forms
a transition region where the actual water dissociation
takes place. In some membranes heavy metal hydrox-
ides are deposited in the interphase to catalyse the
water dissociation. However, tertiary ammonium Rxed-
charge groups at the surface of the anion exchange
membrane seem to have the same effect.
A bipolar membrane with satisfactory properties
can be prepared, e.g. as a laminate of highly perm-
selective anion and cation exchange layers which
have good alkaline and acid stability. An anion-selec-
tive layer with the required properties can be ob-
tained by reacting chloromethylated polysulfone dis-
solved in n-methyl pyrrolidone with the monoquater-
nary salt of 4,4-diazabicyclo-[2.2.2]-octane. The
cation-selective layer can be prepared by introducing
sulfonic acid groups as Rxed charges into a poly-
ether-ether-ketone matrix using chlorosulfonic acid.
The co-ion transport and the swelling behaviour can
be controlled in both layers by partial cross-linking.
The properties of ion exchange membranes prepared
following the above procedures are listed in Table 1.
The Performance of Bipolar Membranes
Bipolar membranes are usually characterized in terms
of their water dissociation capability, their resistance
Table 1 Electrochemical properties of the cation- and anion-
selective layers of a bipolar membrane prepared by the technique
described above
Anion exchange Cation exchange
layer layer
Ion exchange capacity 1.2 1.0
(mmol g\1)
Membrane thickness 60 60
(
m)
Area resistance 1.05 1.31
( cm2)
Permselectivity (%) 97.5 98.5
Swelling (%) 8 12.5
1671
and their long-term stability. The water dissociation
rate and electrical resistance of a membrane prepared
by the procedure described above is shown in Fig-
ure 3A. Here the current density is shown as a func-
tion of the potential drop across the membrane. The
test solutions in both compartments adjacent to the
bipolar membrane are 1 molar Na2SO4. The results
indicate that the current density is extremely low at
potential differences of less than ca. 0.8 V. Then
the current density increases up to 0.250 A cm\2 with
very little increase in voltage drop. When this value is
exceeded, the resistance of the membrane increases
drastically, due to limitations in the water transport
into the interphase region.
Problems in Practical Applications of Bipolar
Membranes
Electrodialytic dissociation of water with bipolar
membranes is economically very attractive for creating
acids and bases. There are, however, several severe
problems in practical applications, such as the con-
tamination of the products by salts and low current
efRciency at high acid and base concentrations.
Salt contamination of the products is related to the
properties of the bipolar membrane. The poor current
efRciency is the consequence of the proton and
hydroxide ion transport in monopolar membranes, as
indicated in Figure 4, which illustrates the conversion
of Na2SO4 into H2SO4 and NaOH by electrodialytic
water dissociation. Figure 4(A) shows the ion trans-
port in the bipolar membrane. What is desired is
a Sux of H# and OH\ ions from the interphase of the
bipolar membrane as the result of the water dissocia-
tion. However, in addition there is a Sux of Na# and
SO24\ ions through the bipolar membrane due to
incomplete permselectivity of the anion and cation
exchange layers. This leads to a contamination of the
base by SO24\ ions and the acid by Na# ions. Since
the permeability of the ion exchange layers to SO\ 4
and Na# increases with increasing acid and base
concentration, the contamination is also increasing
with increasing concentration, as shown in Figure 4B.
This Rgure shows the salt contamination in sulfuric
acid and sodium hydroxide produced by water dis-
sociation in bipolar membranes from a 1 mol L\1
Na2SO4 solution as a function of the concentration of
the acid and base produced.
The current efRciency in water dissociation
with bipolar membranes is mainly affected by
the properties of the anion exchange membrane
which has very poor retention of the protons, as
illustrated in Figure 4C. The transport mechanism of
protons is based on a tunnelling mechanism, with the
consequence that protons can permeate the anion
1672 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes

Figure 4 Schematic diagram illustrating (A) the contamination of an acid and a base by salt to incomplete permeability of the cation
and anion exchange layers of a bipolar membrane; (B) experimentally determined salt contamination as a function of the acid and base
concentrations; (C) the decrease in current efficiency during the production of acids and bases due to the poor acid-blocking capability
of the anion exchange membrane; (D) experimentally determined current efficiency as a function of the produced acid and base
concentration.

exchange membrane rather freely. The same is true
for the hydroxide ions which can permeate the cation
exchange membrane. The net result of the process is
that protons and hydroxide ions generated in the
bipolar membrane neutralize each other. The proton
and hydroxide Suxes and thus the current efR-
ciency depend on the concentration, as shown in
Figure 4D. With increasing acid and base concentra-
tion, the current efRciency decreases rapidly.
Application of Bipolar Membranes
One interesting application of bipolar membranes is
the production of caustic soda. Currently, caustic
soda is produced as a co-product of the products of
chlorine by electrolysis of salt. The worldwide de-
mand for polyvinyl chloride and other chlorinated
hydrocarbons has led to the development of a large
market for chlorine. Because of environmental prob-
lems caused by chlorinated hydrocarbons and the
disposal of polyvinyl chloride wastes, the demand for
chlorine is steadily decreasing, however, and it can be
expected that the demand for caustic soda will soon
exceed that produced in the chlorine alkaline electro-
lysis. Thus interest in alternative processes for obtain-
ing caustic soda is increasing. Electrodialytic water
dissociation with bipolar membranes is one of the
more promising techniques for the future large scale
economic production of caustic soda. However, to-
day’s bipolar membranes produce caustic soda with
some salt contamination. The production of NaOH
and H2SO4 from the corresponding salts has been
investigated in great detail. Test results obtained in
laboratory studies are shown in Figure 4. These tests
were carried out with a 1 mol L\1 solution Na2SO4
feed at room temperature and an applied current
density of 0.1 A cm\2. The test results indicate that
up to three normal acid and base solutions can be
achieved with a current utilization of 60}70%. How-
ever the produced acid and base are contaminated by
salt and the salt contamination increases with increas-
ing acid and base concentrations due to decreasing
 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
selectivity of the bipolar membrane with increasing
acid or base concentrations. Salt concentration can
reach values in excess of 0.03 mol L\1 at 4 molar
base or acid concentrations. To improve the overall
efRciency of the electrodialytic dissociation pro-
cesses and to obtain less salt contamination in the
acids and base produced, better proton-blocking
membranes have to be developed in addition to more
selective bipolar membranes.
Fortunately, there are a large number of other
potential applications of the electrodialytic water
dissociation where the purity of the product, i.e.
the produced acid or base, is not critical and traces
of salts can be tolerated. Typical applications of
bipolar membranes with large industrial relevance
are:
E Recovery of acids and bases such as sulfuric,
hydrochloric or hydroSuoric acid and sodium hy-
droxide from the salts generated in neutralization
reactions
E The recovery of organic acids such as formic,
acetic, citric, lactic and itaconic acid or certain
amino acids from fermentation broths
E Adjustment of pH values in fermentation or chem-
ical production processes without increasing the
ion potential
E Regeneration of H2SO4 and NaOH from Na2SO4
obtained in industrial efSuents, for example, in
the production of viscose or regenerated cellulose
E Regeneration of acids and bases from scrubbers
used to remove SO2, NOx from contaminated air
streams.
This list of potential applications of the elec-
trodialytic water dissociation with bipolar mem-
branes is not complete and as more efRcient bipo-
lar membranes become available, more applications
will certainly be identiRed. In this outline three typi-
cal examples for the use of bipolar membranes are
described in more detail.
Figure 5 Schematic diagram illustrating a conventional process for recovering an amine from a contaminated air stream using an
acid scrubber.
1673
Recycling H2SO4 and DimethylIsopropylamine
from an Acid Scrubber
Alkaline or acid scrubbers are often used to remove
components that are harmful to the environment,
such as NOx, SO2 or certain amines from waste air
streams. In these processes large amounts of acids or
bases are consumed and salts are produced. In gen-
eral, only dilute acids and bases are required in
scrubbers. This makes the use of electrodialytic water
dissociation with bipolar membranes a very suitable
process to recover the acids or bases from the corre-
sponding salts. The recovery of base from scrubbers
used to remove SO2 and NOx from coal-burning
power plants is described in detail in the literature.
Another similar application is the recovery of
dimethylisopropylamine removed from a waste air
stream by a sulfuric acid scrubber. This type of waste
air stream is generated when aluminium casting
moulds are made from a sand/epoxy resin mixture by
injecting dimethylisopropylamine in a mixture with
air as catalyst to cure the resin instantaneously. The
amine is not consumed in the process and is emitted in
a waste air stream containing ca. 0.5 g amine per m3
waste air. The amine can be recovered as amine
sulfate in an acid scrubber, as indicated in Figure 5.
The amine can then be regenerated by adding sodium
hydroxide and distilled. The net result of the process,
however, is the production of large amounts of so-
dium sulfate.
A complete recycling of the amine, the sulfuric acid
and water is achieved without the production of a salt
by combining the electrodialytic water dissociation
with distillation. The process is illustrated in
Figure 6. The waste air stream containing the amines
is fed into an acid scrubber where the free amine is
converted into amine sulfate. The efSuent of the
acid scrubber containing about 10% amine sulfate in
a mixture with sulfuric acid is then fed into
the electrodialytic water dissociation apparatus
1674 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes
Figure 6 Schematic diagram illustrating the recovery of dimethylisopropyl amine from a waste air stream by combination of acid
scrubber, diffusion dialysis and electrodialytic water dissociation using bipolar membranes and distillation.
containing bipolar membranes and anion and cation
exchange membranes in alternating series between
two electrodes. Here the amine sulfate is converted to
the free amine while the sulfate ions form sulfuric
acid which is recycled to the acid scrubber. The
amine}water mixture is distilled to recover the amine
and the water is recycled to the electrodialysis unit.
Thus, the process allows complete recovery of the
amine from a waste air stream by combination of an
acid scrubber and electrodialytic water dissociation.
Production of Itaconic Acid in a Continuous
Fermentation Process
One of the more promising applications of bipolar
membranes is the adjustment of the pH value of
fermentation solutions to recover the organic acids
from the spent medium. As an example the produc-
tion of itaconic acid is described below.
Conventionally, itaconic acid is produced by
a batch fermentation process. During fermentation
the pH value shifts towards lower values due to the
production of the acid. To avoid product inhibition
the pH is maintained at a high level by addition of
sodium or ammonium hydroxide which form soluble
salts with the produced itaconic acid. At the end of
the fermentation processes, the free acid is recovered
from the spent medium by lowering the pH value by
adding sulfuric acid. The adjustment of the pH values
in the fermenter as well as in the spent medium is not
only costly, but also creates salts mixed with the
desired product and thus further puriRcation steps are
required. By applying bipolar membrane technology
the production of salts can be eliminated and the
itaconic fermentation can be carried out more ef-
Rciently in a continuous process, as illustrated in the
production scheme depicted in Figure 7. The Sow
scheme shows a fermenter combined with an elec-
trodialysis unit Rtted with bipolar membranes. The
fermenter is continuously fed with substrate and its
constituents passed through an ulRltration unit. The
retained biomass is recycled to the reactor while the
product containing Rltrate is fed to the middle cell of
a three-compartment electrodialyser repeating unit. In
this cell the solution will be depleted of the ions. The
cations, i.e. sodium or ammonium ions, permeate the
cation exchange membrane and form, with the OH\
ions generated in the bipolar membrane, NaOH which
is concentrated and then fed back into the bioreactor
to adjust the pH value. The anions, i.e. the itaconate
ions, permeate the anion exchange membrane and
form, with protons generated at the bipolar mem-
brane, the itaconic acid which is then concentrated and
precipitated. Thus, the itaconic acid is produced in
a continuous process without the addition of acids or
bases, i.e. without the production of additional salts.
The Electrodialytic Production of Sodium
Methylate by Methanol Dissociation
Bipolar membranes may be used not only for the
electrodialytic dissociation of water. They can also be
applied for the dissociation of alcohol and thus for
the production of alcoholates, as illustrated in the
following example. Methanol, like water, is both
a weak base and a weak acid. Its dissociation con-
stant, however, is somewhat less than that of water.
Thus, sodium methanolate can be efRciently pro-
duced from methanol and sodium acetate in
nonaqueous media by the use of bipolar membranes
 II / MEMBRANE SEPARATIONS / Bipolar Membranes and Membrane Processes 1675

Figure 7 Schematic diagram illustrating a continuous fermentation process for the production of itaconic acid without further addition
of acids or bases using bipolar membranes.

according to the reaction scheme illustrated in
Figure 8, which shows a bipolar membrane elec-
trodialysis stack consisting of two compartment cell
systems in a repeating unit between electrodes.
Water-free methanol and sodium acetate are fed into
the cell formed by the bipolar and the cation ex-
change membrane which is directed towards the cath-
ode while water-free methanol is passed through the
other cell. Under the driving force of an electrical
potential gradient, methanol is split in the bipolar
membrane into protons and CH3O\ ions which react
with sodium ions migrating from the sodium acetate-
containing cell to form CH3ONa. The acetate ions
recombine on the other side of the bipolar membrane
with the protons which were produced simulta-
neously with the CH3O\ ions in the bipolar mem-
brane to form acetic acid. Thus, sodium acetate and
methanol are converted into sodium methanolate.
The current efRciency decreases with increasing
methylate concentration due to proton transfer from
the acetic acid-containing compartment through the
bipolar membrane to the sodium methanolate-
containing cell. But all in all the process seems to be
technically feasible.

Figure 8 Schematic diagram illustrating the electrodialytic production of sodium methanolate from methanol and sodium acetate in
bipolar membranes containing a two-compartment cell unit.
1676 II / MEMBRANE SEPARATIONS / Catalytic Membrane Reactors
Conclusions
The mechanism of water dissociation in bipolar mem-
branes can be rationalized by a hypothesis which
postulates a catalytic reaction between a weak base
and water. Based on this hypothesis, very stable
chemical and thermal bipolar membranes can be pre-
pared and operated efRciently at current densities
in excess of 0.1 A cm2. The process has many poten-
tial applications. There are, however, still a multitude
of problems to be solved. Some are related to the poor
selectivity of the bipolar membranes and poor acid-
blocking capability of the anion exchange mem-
branes; others are caused by the lack of application
know-how and practical experience.
Catalytic Membrane Reactors
M. E. Rezac, Georgia Institute of Technology, Atlanta,
GA, USA
Copyright ^ 2000 Academic Press
Introduction
The concept of completing both a reaction and separ-
ation in a single process unit has motivated research
into the development of catalytic membrane reactors.
For example, it has long been recognized that pallad-
ium metal has the capacity both to permeate hydro-
gen and to promote a variety of reactions. Thus,
harnessing both of these features in a single device
seemed a logical combination. In the mid 1960s,
Wood and co-workers demonstrated that the dehyd-
rogenation of cyclohexane to cyclohexene could be
increased if the hydrogen produced was removed
from the reaction vessel through semipermeable pal-
ladium walls. In this case, the palladium walls also
acted to catalyse the dehydrogenation reaction.
A membrane reactor of this type is illustrated in
Figure 1.
In Russia, Gryaznov conducted much of the re-
search that followed. Starting in the late 1970s, Gry-
aznov began publishing his results on the use of
palladium membrane reactors both to produce and to
recover hydrogen from a myriad of dehydrogenation
reactions. In the dehydrogenation reactions, hydro-
gen leaves the reactor by permeating through the
semipermeable membrane. However, reactors can
also be used in reactions where hydrogen or other
reaction products enter the reaction chamber by pen-
etration through the membrane. The commonest
classes of reactions that have been successfully in-
Suenced by the use of membrane reactor technology
Further Reading
Liu KJ, Chlanda FP and Nagasubramanian KJ (1977) Use
of bipolar membranes for generation of acid and base:
an engineering and economic analysis. Journal of Mem-
brane Science 2: 109}124.
Mani KN (1991) Electrodialysis water splitting technology.
Journal of Membrane Science 58: 117}138.
Simons R (1985) Water splitting in ion exchange mem-
branes. Electrochimica Acta 30: 275}282.
Strathmann H, Kroll JJ, Rapp JJ and Eigenberger G (1997)
Limiting current density and water dissociation in bi-
polar membranes. Journal of Membrane Science 125:
123}142.
Strathmann H, Bauer B and Rapp HJ (1993) Better bipolar
membranes. Chemtech June: 17}24.
are listed in Table 1. Details relating to the large
volume of research reported are provided in the Fur-
ther Reading section. None of these membrane reac-
tors are in commercial use. But some } the selective
oxidation of methane, for example } are the subject
of a very large industrial research effort. If suc-
cessfully developed, this process would change the
feedstock basis of a number of petrochemical
processes.
Most research on the development of membrane
reactors involves the use of these devices to shift
equilibrium-limited reactions (often dehydrogena-
tions). The thermodynamic equilibrium of the react-
ants and products at the temperature and pressure of
the reaction determine the conversion achievable in
any given reaction. For dehydrogenation reactions,
increasing temperature and decreasing pressure pro-
mote an enhanced reaction. Unfortunately, each of
these solutions has an associated cost. Increasing the
reaction temperature typically results in a reduced
Figure 1 Schematic of a membrane reactor using hydrogen-
permeable palladium membranes to shift the equilibrium of the
dehydrogenation reaction cyclohexane to cyclohexene.
 II / MEMBRANE SEPARATIONS / Catalytic Membrane Reactors 1677

Table 1 Reaction classes that may be amenable to membrane reactor technology

Reaction class Example Role of membrane

Hydrogenation C2H2#H2PC2H4 in presence of C2H4 Controlled addition of hydrogen

Hydrogenolysis Cyclopentadiene#H2P cyclopentene# Controlled addition of hydrogen

cyclopentane
Dehydrogenation CyclohexanePbenzene#3H2 Remove hydrogen to shift equilibrium limitation

Partial oxidation Butane#O2P maleic anhydride Recovery of intermediate product of control reactant at
addition rate to promote formation of intermediate
product
Esterifications R}OOH#CH3OHPR}O}O}CH3#H2O Selective water removal to shift equilibrium limit without
loss of reactant
Syn gas CH4#12 O2PCO#2H2 Selective oxidation of methane
Oxidative coupling 2CH4#O2PC2H4#2H2O Selective oxidation of methane

catalytic selectivity for the desired product. Reducing
pressure comes at the cost of adding a diluent to the
reactor, paying for the additional capital to handle
this component and paying the price of downstream
separation.
Figure 2 provides a schematic representation of the
behaviour of a conventional reactor and a theoretical
membrane reactor. The conventional data are for
a highly active butane dehydrogenation catalyst oper-
ating at 1 atm total pressure (pure normal butane
feed). In the conventional system, the selectivity of the
catalyst degrades rapidly at temperatures that are just
beginning to promote reaction. Thus, the catalytic
yield (deRned as the product of conversion and cata-
lytic selectivity) goes through a pronounced max-
imum. Incorporation of an appropriately designed
membrane into the reactor system results in the re-
moval of hydrogen from the system. The catalytic
selectivity does not appear to be inSuenced by this
process, but the conversion of butane to butene is
enhanced by the reduction in the hydrogen partial
pressure. Thus, the yield of the membrane reactor
system is markedly improved.
The ability to operate at acceptable conversions
while maintaining very high catalytic selectivity is
a strong driving force for the use of membrane reactor
technology. By operating in a high selectivity region,
the production of by-products that can act as catalyst
poisons is minimized. This results in a longer catalyst
life between regenerations and reduced waste produc-
tion.

Possible Membrane Con\gurations

Figure 2 Influence of product hydrogen removal on the dehyd-
rogenation of butane. Based on pure butane feed with 1.1 atm
total pressure. Continuous line, conventional reactor; dashed line,
membrane reactor.
Incorporation of a reaction and separation zone in
a single process unit allows for a variety of possible
conRgurations. The optimum design of the equipment
is closely tied to the reaction conditions and the
ability of the membrane material to serve as a cata-
lyst. Several of the more common conRgurations are
shown in Figure 3. For illustrative purposes, the de-
hydrogenation of a compound to form hydrogen will
be considered. The hydrogen is removed from the
reaction zone to increase the equilibrium conversion.
Similar conRgurations can be employed for the other
reactions listed in Table 1. As described below, these
conRgurations represent the most frequently em-
ployed designs, however, the list is not exhaustive and
new conRgurations are developed and patented
regularly.
1678 II / MEMBRANE SEPARATIONS / Catalytic Membrane Reactors
Figure 3 Common membrane reactor configurations. (A) Catalytic membrane; (B) membrane tube packed with catalyst;
(C) membrane-assisted batch reactor.
Catalytic Membrane
One of the earliest catalytic membrane conRgurations
employed was that of the reactive tube. In this conRg-
uration, the material used to construct the tube fulRls
both the roles of separation medium and catalyst.
Few materials have this special capability. Palladium
is one. Palladium has the ability to transport hydro-
gen through the matrix by a process of adsorption,
dissociation, diffusion, and then reassociation on
the low pressure side. Palladium is also a reasonable
catalyst for many of the reactions detailed in Table 1,
especially hydrogenation and dehydrogenation reac-
tions. Thus, using this material to achieve both func-
tions was an obvious consideration.
Transport can only be achieved through a chemical
potential driving force of the hydrogen from the reac-
tion zone to the separation zone. Such a driving force
has been established with the use of a sweep gas on
the permeate side to keep the hydrogen concentra-
tions low. The sweep gas may either be inert or
reactive with the hydrogen. Inert sweep gases of-
fer the advantage of being simple to employ. Unfor-
tunately, to achieve a partial pressure difference
across the membrane, the sweep gas rate must be high
and the permeated hydrogen is recovered as a dilute
component in the sweep gas.
Reactive sweep gases offer other engineering
possibilities and challenges. The use of air as a sweep
gas in catalytic dehydrogenation membrane systems
has been reported. At dehydrogenation temperatures
(300}6003C), oxygen can react with hydrogen to
form water. This reaction is highly exothermic. In
contrast, the dehydrogenation reaction is endother-
mic. Thus, thermal matching of the heat released by
the hydrogenation of oxygen and the heat consumed
by the dehydrogenation reaction would allow for an
isothermal system. Because the hydrogenation reac-
tion is rapid, the effective partial pressure on the
permeate side of the membrane can be maintained
near zero.
Catalytic membrane systems require that the mem-
brane material be stable for both reaction and
separation and that it operates well in both modes
simultaneously.
Packed Tube
Optimization of a single material for both catalytic
and separative functions is challenging. Few materials
have the ability to transport the desired component
and act as a catalyst for the desired reaction. Further-
more, even for materials that possess both character-
istics, precise matching of the rates of reaction and
 II / MEMBRANE SEPARATIONS / Catalytic Membrane Reactors
transport is difRcult. To overcome these limita-
tions, a packed membrane tube conRguration has
been employed. In such a conRguration, a catalyst is
packed in the bore of a tubular membrane. Reactants
are fed into the catalyst zone and products have the
potential to be transported through the membrane
walls and out of the reaction zone. This conRguration
offers tremendous Sexibility in the selection of
the catalyst and membrane to be used. Both homo-
geneous and heterogeneous catalysts have been em-
ployed in this conRguration.
Membrane-assisted Batch Reactor
The membrane-assisted batch reactor is most fre-
quently considered for implementation because it
requires the smallest process modiRcation from
traditional catalytic reactors. In this conRguration,
a membrane unit is added in the recycle line of a batch
reactor. In so doing, the membrane has the capacity
to selectively remove a product component or to
selectively add a reactant. It has the advantage of
allowing the pressure or temperature of the mem-
brane and reactor unit to be controlled indepen-
dently. Therefore, the properties of the membrane
can be varied to optimize the separation achieved.
Membrane-assisted conRgurations suffer from
the inability to remove product components com-
pletely as they are produced. This limits the conver-
sion to values that are lower than those that are
theoretically possible in the other conRgurations
considered.
Available Membranes
The development of catalytic membrane reactors is
limited by the availability of membranes capable of
controlling the reaction environment that are stable
at reaction conditions. A brief review of transport
through membranes is provided and then the addi-
tional membrane requirements are summarized.
Figure 4 provides a schematic of a membrane em-
ployed for the transport of a gaseous component.
Transport of a component through a solid is only
possible if there are differences between the chem-
ical potential of the component on the two faces of
the solid. For gas-phase systems that operate at mod-
erate pressures and can be considered to be ideal,
transport can be described by:
Fluxi"Pi(pHi!pLi)/l [1]
where Pi is the permeability of component i through
the membrane, pHi is the partial pressure of compon-
ent i on the high pressure side, pLi is the partial
1679
Figure 4 Schematic of transport through a solid.
pressure of component i on the low pressure side and
l is the membrane thickness.
Thus, the difference in the partial pressure of
the component to be transported controls the rate of
transport. A high partial pressure driving force can be
produced through:
E high total pressure on the feed side
E high concentration of the component of interest on
the feed side
E low total pressure on the permeate side (using a
vacuum)
E very low concentration of i on the permeate side
using a very high dilution ratio of sweep gas
The rate at which a component is transported
through a solid is deRned as the Sux of the compon-
ent, (eqn [1]). To facilitate the comparison of a
variety of materials, the properties of the material
(permeability) have been separated from the process
conditions (membrane thickness, pressures and
concentrations). For equivalent process conditions,
the material with the highest permeability will have
the fastest transport.
The selectivity of a membrane for any pair of gases
(A, B) is usually deRned by the term
A/B, equal to the
ratio of the gas permeabilities:

*A/B"PA/PB [2]
For membrane reactors, membranes with high
selectivities are required, so only the required com-
ponent enters or leaves the reactor.
Speci\c Membranes
Certain materials have the ability to transport a single
component with the complete exclusion of all others.
Transport of hydrogen and oxygen through two of
these materials is reviewed here.
Metals (hydrogen) Palladium and some of its alloys
have the ability to transport hydrogen while
1680 II / MEMBRANE SEPARATIONS / Catalytic Membrane Reactors
completely excluding all other compounds. The per-
meation of hydrogen through metals is a multistep
process involving:
E chemisorption of hydrogen on to the metal surface
E dissociation of the hydrogen
E dissolution of the atomic hydrogen from the sur-
face into the bulk of the metal
E diffusion across the metal layer
E the desorption from the bulk of the metal to the
surface
E reassociation
E desorption of molecular hydrogen from the metal
surface
The diffusion of atomic hydrogen across the
metal layer is typically the rate-limiting step. There-
fore, the transport of the hydrogen can be modelled
using a Fickian diffusion equation.
Alloys of palladium have proven effective for
the transport of hydrogen. Pure palladium undergoes
a phase transformation in the presence of hydrogen
at moderate temperatures, and a density change
ensues. Even though these changes are small, they are
sufRcient to produce a brittle, cracked and non-
selective material after only a few cycles. Therefore,
few of the membranes evaluated are pure palladium
and most commercial metallic hydrogen puriRers are
prepared from a palladium/silver alloy containing
23% silver.
A signiRcant limitation to the use of palladium
membranes is their strong susceptibility to poisoning
by sulfur compounds and CO compounds frequently
found in the hydrocarbon streams of interest. Re-
search into the development of more resistant mater-
ials is underway. However, the improved resistance
has so far been attained at the cost of permeability.
This issue must be resolved before these membranes
can be used in the chemical process industry.
Nonporous ceramics (oxygen) Ceramic membranes
have been developed which will selectively transport
hydrogen or oxygen. Hydrogen has been shown to be
transportable through nonporous silicon dioxide.
While the transport rates are extremely slow, the
selectivity to hydrogen transport is inRnite, just as in
defect-free palladium. Silicon dioxide has the advant-
age of being more resistant to the presence of sulfur
compounds that act as poisons for palladium. Never-
theless, the challenge of forming this material into
extremely thin layers has limited its use. Perhaps as
preparation techniques continue to improve, the use
of this material for the highly selective transport of
hydrogen will be re-examined.
Nonporous ceramics have also been used for the
transport of oxygen. Oxide-conducting materials, in-
cluding stabilized zirconia, have been used in mem-
brane reactors. These membranes can successfully
transfer oxygen while barring the transport of all
other compounds. Current limitations relating to the
temperatures required for operation are discussed in
subsequent sections.
Nonspeci\c Membranes
Commercial utilization of membranes has relied al-
most exclusively on the use of nonspeciRc membrane
materials. These materials have the ability to trans-
port one component of a gas mixture in preference to
a second. However, they are permeable to all compo-
nents to at least some degree. Therefore, unlike pal-
ladium, which can act as a perfect separator, these
membranes transport all stream components. The
properties of both porous ceramics and polymeric
membranes will be considered here.
Porous ceramics Porous ceramic membranes, with
pore sizes ranging from a few nanometers to several
microns, have been produced and are commercially
available. These membranes separate by size exclu-
sion. For the separation of gases and low molecular
weight liquids, Knudsen diffusion is typically
employed. For gases, Knudsen diffusion occurs for
pore sizes of about 4}100 nm. For systems operating
under the Knudsen diffusion regime, the separ-
ation of two molecules can be deRned as:

A/B"(MWB/MWA)0.5 [3]
Thus, high degrees of separation selectivity are only
possible if the molecular weight difference be-
tween the two components is large. These materials
have been employed for the separation of hydrogen
from hydrocarbon streams. The ideal separation sel-
ectivity for hydrogen over butane, for example, is 5.8.
Thus, a small but measurable separation can be
achieved.
The microporous nature of these membranes
allows them to have very high transport rates, as
compared to nonporous palladium or ceramic mater-
ials. Yet, the separation achievable is limited.
Polymers The Rnal category of membrane materials
to be considered is polymers. Polymeric membranes
are employed for the separation of gas streams, the
recovery of organic vapour from air, the separation of
mixtures of organic liquids and Rltration of particles
from aqueous streams. Nonporous polymeric mem-
branes have been considered for membrane reactor
applications. Nonporous polymeric membranes sep-
arate on the basis of sorption of the component into
 II / MEMBRANE SEPARATIONS / Catalytic Membrane Reactors
the polymeric matrix, diffusion across and desorp-
tion from the low pressure side.
Polymers offer several advantages when com-
pared with the porous ceramics. For many gas pairs,
the inherent selectivity of the polymeric membrane is
substantially higher than that of a Knudsen diffu-
sion-controlled ceramic. The polymeric membranes
are also easier to prepare in high surface per volume
modular conRgurations, resulting in a considerably
lower price. Recent estimates put the price of com-
mercial polymeric membranes at well under $10 per
square foot membrane area, installed. In contrast,
ceramics may cost 10 times as much.
Many applications of membrane reactors, includ-
ing dehydrogenation of hydrocarbons, hydrogenations
and partial oxidation reactions are high temperature
reactions. Conventional polymeric membranes do not
have either the chemical or thermal stability to be
successfully used in these reactions. However, recent
advances have provided materials that can be pro-
cessed using conventional solvent-based techniques
and are later cross-linked to provide the chemical and
thermal resistance necessary for membrane reactor
applications. New polyimide-based materials provide
such characteristics.
Polyimides that are thermally stable to 3003C for
extended periods have been reported. When incor-
porated in a membrane reactor for the dehydrogena-
tion of butane, the system performance increased
markedly. With no membrane, conversions of 22%
were achieved. Following addition of the membrane
to the integrated system, the conversion increased to
over 30%.
The development of chemically stable polymeric
materials provides an opportunity to inSuence liquid-
phase organic reactions.
Applications
While the use of membranes to inSuence catalytic
reactions has been explored in great detail, few systems
have been employed commercially. Some of the pos-
sible reasons for this slow adoption are listed below.
Nevertheless, a few materials are produced through
the use of membrane reactor technology. Gryaznov,
for example, has reported the production of vitamin
K using a single-step process utilizing a membrane
reactor. The membrane was employed to control the
hydrogenation of a mixture of quinone and acetic
anhydride to form vitamin K. The membrane reactor
process resulted in a 95% yield using an external
hydrogen pressure of 1 atm. The conventional process
required several processing steps and resulted in only
80% yield. This membrane reactor process is re-
ported to be employed commercially in Russia.
1681
Current Limitations
Some of the more important technical issues that have
limited the implementation of membrane reactor
technology are detailed below. In addition to tech-
nical problems, economic considerations are a con-
cern. For many of the systems considered (Table 1),
commercial production facilities using conventional
reactor technology are available. If membrane reac-
tors are to supplant these existing systems, the eco-
nomic beneRt must be substantial. Furthermore, the
cost of the membrane reactor system must be only
marginally higher than the conventional system. For
most reactions, this is not currently the case.
Need for Pressure Drop
When one considers the use of membrane reactor
technology for the selective removal of a product com-
ponent (such as in dehyhdrogenations), the design of
the membrane will be governed by the need for a par-
tial pressure difference of this component. To
increase the equilibrium conversion, these systems are
run at low pressure with a goal of complete removal of
the component as it is produced. Thus, the partial
pressure of the product component is nearly zero in the
reaction zone. For transport to occur, the partial pres-
sure on the permeate side must be lower. Several
techniques have been employed to attain partial
achievement of these goals. These include high volumes
of sweep gas on the permeate side; vacuum on the
permeate side; and transforming the system to a batch
reactor with continuous removal of the product com-
ponent. In the last case, the partial pressure on the feed
side is maintained at some Rnite level, and additional
conversion is achieved by long residence times.
Sweep gas The use of high sweep gas ratios (nitro-
gen or argon is commonly employed in the labora-
tory) is effective in reducing the partial pressure
of products in the reaction zone and enhancing con-
versions. For nonspeciRc membranes, it has been
shown that two processes reduce the partial pressure
of products: transport of the product from the reac-
tion zone to the separation zone, and transport of
sweep gas from the separation zone to the reaction
zone. This latter process can occur because the mem-
brane is nonselective and the partial pressure gradient
of the sweep gas drives the transport. As the sweep
gas permeates into the reaction zone, it acts as a dilu-
ent and provides a mechanism for an increase in the
percentage conversion in the reactor. Unfortunately,
the downstream separations required in these systems
are signiRcant and the economics are less favourable
than simply mixing the diluent with the reactants in
a conventional reactor.
1682 II / MEMBRANE SEPARATIONS / Concentration Polarization
Vacuum permeate An alternative method to pro-
duce a pressure drop is the use of a vacuum on the
membrane permeate. This has been shown to be
highly effective in laboratory settings. However,
the economics are not favourable for the large scale
production of inexpensive components. Nevertheless,
vacuum permeate systems may prove viable for small,
high value-added systems.
Batch versus continuous Continuous reactor sys-
tems are preferred; they require less down time and
have higher production rates than batch systems of
similar size. However, as previously detailed, if the role
of the membrane is to remove a product component,
the available partial pressure difference is limited
and the process will always be working with a very
limited pressure drop that will require very large mem-
brane areas. Batch and semi-batch processes allow the
system to develop some limited partial pressure dif-
ference before membrane separation is attempted.
Membrane Degradation
The stability of the membrane is another important
consideration. Ideally, for integrated systems, the
membrane should be stable in all possible reaction
environments: catalyst activation, normal reaction,
catalyst regeneration and any thermal cycling experi-
enced upon transitions. This presents speciRc chal-
lenges for each system and there are few materials
that can satisfy all of these requirements. Thus,
special engineering solutions are necessary. Even if
the membrane material can fulRl these speciRcations,
the many components needed to produce a mem-
brane reactor module may not.
Future Possibilities
Organic Separations
A great deal of research is currently focusing on the
development of membranes (either polymeric, inor-
Concentration Polarization
H. Wijmans, Membrane Technology and Research,
Inc., Menlo Park, CA, USA
Copyright ^ 2000 Academic Press
Introduction
All membrane separation processes are accompanied
by a phenomenon called ‘concentration polarization’
ganic, or hybrids of the two) for the selective separ-
ation of liquid organic mixtures. If this research is
successful, it will allow for incorporation into liquid-
phase membrane reactors.
Control of Reactant Addition for Intermediate
Product Recovery
A second area of immense current research activity is
the development of oxygen-permeable membranes to
inSuence the conversion of methane to either meth-
anol or syn gas. The goal in these processes is a
mechanism for the conversion of natural gas to
a transportable liquid that may be further converted
to high valued products. Current research has
shown that membranes can be developed and that the
appropriate catalysts are available for these conver-
sions. Many engineering challenges lie ahead. These
membrane reactor processes operate in excess of
7003C (sometimes much higher). Sealing these ce-
ramic membranes into a housing remains a limita-
tion. Further, the thermal stresses, which develop
when cycling from 25 to '7003C, may result in
membrane damage. While these are complex prob-
lems, the incentive to succeed is large and numerous
research efforts continue in this area.
Further Reading
Armour JN (1989) Catalysis with permselective inorganic
membranes. Applied Catalysis 49: 1.
Gokhale YV, Noble RD and Falconer JL (1995) Effects
of reactant loss and membrane selectivity on a dehyd-
rogenation reaction in a membrane-enclosed catalytic
reactor. Journal of Membrane Science 103: 235.
Govind R and Itoh N (eds) (1989) Membrane reactor
technology. AIChE Symposium Series 85: 268.
Saracco G, Versteeg GF and van Swaaij WPM (1994) Cur-
rent hurdles to the success of high-temperature mem-
brane reactors. Journal of Membrane Science 95: 105.
Shu J, Grandjean BPA, Van Neste A and Kaliaguine S (1991)
Catalytic palladium-based membrane reactors: a review.
Canadian Journal of Chemical Engineering 69: 1036.
in which the composition at the feed}membrane in-
terface differs from the composition in the bulk
of the feed mixture. This gradient in composition is
generated by the separation performed by the mem-
brane and, as such, cannot be avoided. However, it is
important to minimize the effects of concentra-
tion polarization because the gradient in composition
reduces the separation performance of the membrane
and increases the potential for membrane fouling.
1682 II / MEMBRANE SEPARATIONS / Concentration Polarization
Vacuum permeate An alternative method to pro-
duce a pressure drop is the use of a vacuum on the
membrane permeate. This has been shown to be
highly effective in laboratory settings. However,
the economics are not favourable for the large scale
production of inexpensive components. Nevertheless,
vacuum permeate systems may prove viable for small,
high value-added systems.
Batch versus continuous Continuous reactor sys-
tems are preferred; they require less down time and
have higher production rates than batch systems of
similar size. However, as previously detailed, if the role
of the membrane is to remove a product component,
the available partial pressure difference is limited
and the process will always be working with a very
limited pressure drop that will require very large mem-
brane areas. Batch and semi-batch processes allow the
system to develop some limited partial pressure dif-
ference before membrane separation is attempted.
Membrane Degradation
The stability of the membrane is another important
consideration. Ideally, for integrated systems, the
membrane should be stable in all possible reaction
environments: catalyst activation, normal reaction,
catalyst regeneration and any thermal cycling experi-
enced upon transitions. This presents speciRc chal-
lenges for each system and there are few materials
that can satisfy all of these requirements. Thus,
special engineering solutions are necessary. Even if
the membrane material can fulRl these speciRcations,
the many components needed to produce a mem-
brane reactor module may not.
Future Possibilities
Organic Separations
A great deal of research is currently focusing on the
development of membranes (either polymeric, inor-
Concentration Polarization
H. Wijmans, Membrane Technology and Research,
Inc., Menlo Park, CA, USA
Copyright ^ 2000 Academic Press
Introduction
All membrane separation processes are accompanied
by a phenomenon called ‘concentration polarization’
ganic, or hybrids of the two) for the selective separ-
ation of liquid organic mixtures. If this research is
successful, it will allow for incorporation into liquid-
phase membrane reactors.
Control of Reactant Addition for Intermediate
Product Recovery
A second area of immense current research activity is
the development of oxygen-permeable membranes to
inSuence the conversion of methane to either meth-
anol or syn gas. The goal in these processes is a
mechanism for the conversion of natural gas to
a transportable liquid that may be further converted
to high valued products. Current research has
shown that membranes can be developed and that the
appropriate catalysts are available for these conver-
sions. Many engineering challenges lie ahead. These
membrane reactor processes operate in excess of
7003C (sometimes much higher). Sealing these ce-
ramic membranes into a housing remains a limita-
tion. Further, the thermal stresses, which develop
when cycling from 25 to '7003C, may result in
membrane damage. While these are complex prob-
lems, the incentive to succeed is large and numerous
research efforts continue in this area.
Further Reading
Armour JN (1989) Catalysis with permselective inorganic
membranes. Applied Catalysis 49: 1.
Gokhale YV, Noble RD and Falconer JL (1995) Effects
of reactant loss and membrane selectivity on a dehyd-
rogenation reaction in a membrane-enclosed catalytic
reactor. Journal of Membrane Science 103: 235.
Govind R and Itoh N (eds) (1989) Membrane reactor
technology. AIChE Symposium Series 85: 268.
Saracco G, Versteeg GF and van Swaaij WPM (1994) Cur-
rent hurdles to the success of high-temperature mem-
brane reactors. Journal of Membrane Science 95: 105.
Shu J, Grandjean BPA, Van Neste A and Kaliaguine S (1991)
Catalytic palladium-based membrane reactors: a review.
Canadian Journal of Chemical Engineering 69: 1036.
in which the composition at the feed}membrane in-
terface differs from the composition in the bulk
of the feed mixture. This gradient in composition is
generated by the separation performed by the mem-
brane and, as such, cannot be avoided. However, it is
important to minimize the effects of concentra-
tion polarization because the gradient in composition
reduces the separation performance of the membrane
and increases the potential for membrane fouling.
 II / MEMBRANE SEPARATIONS / Concentration Polarization
Therefore, minimizing concentration polarization is
one of the most important objectives in designing and
engineering membrane separation systems.
Mathematical Description of
Concentration Polarization
The velocity proRle of a Suid Sowing in a channel
is not constant across the thickness of the channel,
because of friction at the Suid}channel surface
interface. The Suid velocity decreases as the
distance from the channel surface decreases. The
same phenomenon occurs in the channels of a mem-
brane module, and the resulting velocity gradient
adjacent to the feed side of the membrane is charac-
teristic of all membrane processes. To facilitate mass
transfer analysis, the velocity gradient is usually rep-
resented by a step function, and it is assumed that
a stagnant boundary layer exists adjacent to the mem-
brane. Any component permeating the membrane
must Rrst pass through the boundary layer as illus-
trated in Figure 1.
Although the boundary layer is stagnant in the
direction of the feed bulk Sow, the boundary layer is
subject to convective Sow perpendicular to the mem-
brane surface which is generated by the permeate
Sux. The convective transport of a component into
the boundary layer from the bulk solution is given by
the product vp ) cp , where vp (cm s\1) is the convective
velocity and cb (g cm\3) is the concentration in the
bulk of the feed. The rate at which the same com-
ponent leaves the boundary layer is vp ) cp, where
cp (g cm\3) is the permeate concentration. In general,
if separation is achieved, cp does not equal cb, and the
convective Sows into and out of the boundary layer,
generate a mass imbalance. This imbalance then
forms a concentration gradient in the boundary layer,
and the concentration gradient increases until dif-
fusion of the component down the concentration
Figure 1 Schematic of the boundary layer adjacent to the
membrane surface. If c p'c b: component is enriched in per-
meate. If c p(c b: component is depleted in permeate.
1683
gradient is sufRcient to restore mass balance in
the boundary layer.
At steady state, the sum of convective and dif-
fusive transport in the boundary layer equals the
amount permeated through the membrane. This
steady state is expressed for each component by the
equation:
vpci!Ddci/dx"Jwi [1]
where D (cm2 s\1) is the diffusion coefRc-
ient, x (cm) is the coordinate perpendicular to the
membrane surface and Jwi (g cm\2 s\1) is the mass
Sux of i permeating through the membrane.
In liquid-phase separations (including pervapora-
tion) concentrations are typically expressed as
a weight fraction, wi"ci/
where
(g cm\3) is the
density of the liquid. Assuming that the density of the
feed is constant in the boundary layer:
dwi
vp ) wi )
!D )
"Jwi [2]
dx
and assuming that the feed density is equal to the
density of the permeate:
Jwi"wp ) Jwtot"wp ) vp )
[3]
where wp (g g\1) is the weight fraction of i in the
permeate and J wtot (g cm\2 s\1) is the combined mass
Sux of all components permeating the membrane.
Combining eqns [2] and [3] and eliminating the den-
sity
gives:
dwi
vp ) wi!D "vp ) wp [4]
dx
which, integrated over the thickness  (cm) of the
boundary layer, yields the polarization equation:
wm!wp
"exp(vp ) /D)
wb!wp
"exp(vp/kbl)
"exp(Jwtot/
) kbl) [5]
where wm and wb are the weight fractions of i at the
membrane surface and in the bulk of the feed, respec-
tively, and kbl"D/ (cm s\1) is the mass-transfer
coefRcient in the boundary layer.
In gas-separation applications, concentrations are
typically expressed as mole fraction ni, which is equal
1684 II / MEMBRANE SEPARATIONS / Concentration Polarization

to the volume fraction, assuming the gas mixtures are Integrating eqn [10] in the same way as eqn [4] gives:
ideal. Starting again with eqn [1], the mole fraction
ni can be substituted for ci by using: nm!np
"exp(vp ) /D)
nb!np
ni"ci ) 22 400 ) T/(Mi ) pf ) 273) [6]
"exp(vp/kbl)
3 1
where 22 400 (cm (STP) mol\ ) is the molar volume
of an ideal gas, T (K) is the gas temperature, "exp(Jvtot ) T/pf ) 273 ) kbl) [11]
Mi (g mol\1) is the molecular weight of i, pf (bar) is
the feed gas pressure, and 273 K is the standard where nm and nb are the mole (or volume) fraction of
temperature. Also, the volume Sux J vi i at the membrane surface and in the bulk of the feed.
3 2 1
(cm (STP) cm\ s\ ) can be substituted for the mass Eqns [5] and [11] describe the concentration pro-
Sux Jwi using: Rles that develop in the boundary layer, as illustrated
in Figure 2. Any component enriched in the permeate
Jvi"Jwi ) 22 400/Mi [7] will be depleted in the boundary layer and any com-
ponent depleted in the permeate will be enriched in
Elimination of the term Mi/22 400 gives: the boundary layer.

dni Factors Determining the Extent

vp ) ni ) pf ) 273/T!D ) pf ) 273/T "Jvi [8]
dx of Concentration Polarization
The ratio of the concentration of a component at the
Since Jvi"np ) Jvtot and: membrane interface to the concentration in the bulk
of the feed is called the ‘concentration polarization
vp"Jvtot ) T/(pf ) 273) [9] modulus’ and is a measure of the inSuence of concen-
tration polarization on the separation process. The
elimination of the term pf ) 273/T gives: following expression for the modulus can be obtained
from eqn [5]:
dni
vp ) ni!D "vp ) np [10] wm exp(vp/kbl)
dx " [12]
wb 1#Eo[exp(vp/kbl)!1]

where Eo"wp/wm is the intrinsic enrichment

Figure 2 Schematic of the concentration polarization phenom-
enon. The concentration profiles in the boundary layer result from
the separation achieved by the membrane. The type of concen-
tration profile formed depends on the value of wp relative to wb (or
n p relative to nb).
achieved by the membrane (and equal to the actual
enrichment if concentration polarization were ab-
sent). An equation equivalent to eqn [12] but ex-
pressed in mole fractions can be derived from eqn
[11].
Eqn [12] allows the concentration polarization
modulus to be calculated as a function of vp/kbl for
different values of the intrinsic enrichment fac-
tors, Eo. The ratio vp/kbl is a Peclet number and is
a measure of the inSuence of convection relative to
the inSuence of diffusion in the boundary layer.
The results of this calculation are shown in the very
informative Figure 3, which conRrms that the con-
centration polarization modulus is smaller than 1
(boundary layer depletion) if the permeating com-
pound is enriched in the permeate and larger than 1
(boundary layer build-up) if the permeating com-
pound is depleted in the permeate. The concentration
polarization modulus increasingly deviates from
unity as the ratio vp/kbl increases, that is, as the Sux
through the membrane increases or as the turbulence
 II / MEMBRANE SEPARATIONS / Concentration Polarization 1685

higher concentrations. The primary requirement for

signiRcant concentration polarization effects is
a high value for the enrichment factor, E. However,
because E has an upper bound equal to 1/wb, a low
feed concentration is a secondary requirement for
severe concentration polarization effects. This
conRrms an empirical rule long held by membrane
separation practitioners.

Transport Equations Incorporating

Concentration Polarization
Figure 3 Concentration polarization modulus, wm/wb, as func- As pointed out in the previous sections, concentration
tion of vp/kbl for a range of values of the intrinsic enrichment factor polarization primarily affects membrane per-
Eo. Lines calculated through eqn [12]. This figure shows that meation by the change in composition at the mem-
compounds that are enriched by the membrane (Eo'1) are more brane interface relative to the bulk of the feed mix-
affected by concentration polarization than compounds that are
rejected by the membrane (Eo(1).
ture. To calculate the effect of concentration
polarization on Sux and separation, the transport
equation for the membrane can be combined with
eqn [5] or eqn [11] to arrive at a set of equations that
of the feed Suid decreases. At high values for the ratio
predict the permeate Sux and composition.
vp/kbl, the concentration polarization modulus,
UltraRltration, nanoRltration and reverse osmosis
wm/wb, approaches the limiting value 1/Eo. At this
are membrane processes in which a solute is separ-
point, the boundary layer completely negates the sep-
ated from a solvent using a solute-rejecting mem-
aration power of the membrane permeation step. The
brane. Typically the permeate is essentially pure
concentration polarization modulus also increasingly
solvent, free of the solute. A simple but very
deviates from unity as the intrinsic enrichment in-
effective transport equation developed for this
creasingly deviates from unity, that is, as the separ-
situation is given below.
ation power of the membrane increases.
The pure solvent Sux Jwsolvent (g cm\2 s\1) of the
A striking feature of Figure 3 is the asymmetry
membrane is given by:
with respect to enrichment and rejection. For
example, when the term vp/kbl has a value of 10\1,
concentration polarization is essentially nonexistent Jwsolvent"P/Rm [13]
for a component rejected by the membrane with an
intrinsic enrichment Eo of 10\4. On the other hand, where P (bar) is the pressure difference applied
concentration polarization is very severe for a com- across the membrane and Rm (bar cm2 s g\1) is the
ponent enriched by the membrane with an intrinsic membrane resistance to the solvent. When a solute is
enrichment Eo of 104. The reason for this asymmetry present, the driving force for permeation is reduced
is that the concentration polarization effect is by the osmotic pressure difference between the
generated by the difference in concentration be- feed at the membrane interface and the permeate,
tween the permeate and the feed, wp!wb" m (bar), therefore:
wb (E!1), where E"wp/wb is the actual enrichment
factor. It is clear that the absolute value of wp!wb is Jwsolvent"(P!m)/Rm [14]
signiRcantly larger if E'1 than if E(1.
A second feature of the calculations shown in
Eqn [14] is called the ‘osmotic pressure model’, in
Figure 3 is that the concentration polarization
which the osmotic pressure is a measure of the ther-
modulus values are independent of the bulk concen-
modynamic work required to produce solvent from
tration, wb. This means that at a constant
a solvent}solute mixture. Assuming that the permeate
enrichment factor, E, the inSuence of concentration
solute concentration is neglible:
polarization is the same, no matter whether the com-
ponent is present in the feed at a concentration of one
part per hundred, one part per million, or one part m"a ) wnm [15]
per billion. Thus, concentration polarization does not
necessarily affect components present at low where a is a constant and n is an exponent equal to
concentrations more than components present at approximately 1 for low-molecular-weight solutes,
1686 II / MEMBRANE SEPARATIONS / Concentration Polarization
Figure 4 Solvent flux as a function of applied pressure as
calculated from eqn [17]. The flux observed with solvent}solute
mixtures is always less than the pure solvent flux. The deviation
increases with increasing applied pressure, increasing solute con-
centration, and decreasing mass-transfer coefficient in the bound-
ary layer.
but equal to 2 or higher for macromolecular solutes.
Combining eqn [15] with eqn [5] and assuming
wp"0 gives:
m"a ) wnb ) exp(n ) Jwsolvent/
kbl) [16]
and:
Jwsolvent"(p!a ) wnb ) exp(n ) Jwsolvent/
kbl))/Rm [17]
From eqn [17] it is clear that an increase in the Sux
Jwsolvent leads to an exponential increase in the osmotic
pressure and that the Sux will increase less than
linearly with the applied pressure. This means that
any increase in driving force P will be negated at
least in part by the increase in osmotic pressure. The
general effect of pressure on Sux predicted by
eqn [17] is illustrated in Figure 4 and is in agreement
with the vast majority of experimental data. As can
be seen from Figure 4, the Sux observed with sol-
vent}solute mixtures is always less than the pure
solvent Sux, and the deviation increases with increas-
ing applied pressure, increasing solute concentration
and decreasing mass transfer coefRcient in the
boundary layer. Figure 4 also shows that at higher
applied pressures the Sux becomes essentially inde-
pendent of the applied pressure. This is often ob-
served in ultraRltration applications and is referred to
as the limiting Sux. Eqn [17] predicts that under
‘limiting Sux’ conditions the Sux is independent of
the membrane resistance, which also has been con-
Rrmed experimentally.
Gel Layer Formation
When the solute is a macromolecular compound such
as a protein or a polymer, there is the possibility that
the solute concentration at the membrane interface
exceeds the gel concentration, wg, at which concen-
tration the solution is no longer a Suid. A gel layer
thus forms at the membrane interface which creates
an additional resistance to the permeation Sux which
consequently decreases. The Sux continues to de-
crease until the solute concentration at the membrane
interface equals the gel concentration, at which point
steady state is reached. The Sux at that point can be
obtained from eqn [5]:
(wg!wp)
Jwlimit"
) kbl ) ln [18]
(wb!wp)
and because wp is typically close to zero:
Jwlimit"
) kbl ) ln(wg/wb) [19]
The steady-state Sux Jwlimit is called the ‘limiting Sux’
because any increase in applied pressure will just
result in a thicker gel layer and not in a higher Sux.
From eqn [19] it can be seen that the limiting Sux as
predicted by the gel layer model is independent of the
applied pressure as well as the membrane resistance.
Additionally, eqn [19] predicts a straight-line plot of
Jwlimit versus ln(wb) with a slope equal to !
) kbl. All
these predictions have been conRrmed in a vast num-
ber of ultraRltration experiments. Interestingly, the
osmotic pressure model also predicts a limiting Sux
with the same attributes.
Approaches to Minimize
Concentration Polarization
The primary method of reducing the negative inSu-
ence of concentration polarization is to maximize the
mass-transfer coefRcient in the boundary layer.
Usually the Rrst method used is to increase the feed
velocity. This has the drawbacks of a high feed-to-
residue pressure drop and the requirement of long,
thin modules, which have higher capital costs than
shorter, larger-diameter modules. A more efR-
cient approach is to choose optimized feed-spacer
materials and/or to create non-linear feed channels
which induce mass-transfer-enhancing vortices. More
complicated methods used for a feed mixture with
a high viscosity and/or a high membrane fouling
 II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
potential employ spinning membranes or vibrating
modules.
Further Reading
Belfort G, Davis RH and Zydney AL (1994) The behavior
of suspensions and macromolecular solutions in cross
Sow microRltration. Journal of Membrane Science 1: 96.
Dialysis in Medical Separations
W. R. Clark, Baxter Hemodialysis Research Lab.,
Wishard Hospital, Indianapolis, IN, USA
M. J. Lysaght, Brown University, Province,
RI, USA
Copyright ^ 2000 Academic Press
Introduction
Although haemodialysis (HD) as a therapy for ura-
emia (kidney failure) was Rrst described early in the
1900s, its widespread use did not occur until the
1950s. At this time, Travenol Laboratories (now Bax-
ter International) unveiled the ‘coil’ dialyser (‘artiR-
cial kidney’) in which tubes composed of cellophane
membranes were wound around a support structure
and immersed in a recirculated dialysis solution.
Relative to contemporary models, the mass transfer
efRciency of this type of dialyser was extremely
poor, due to high mass transfer resistances in all three
compartments (blood compartment, membrane, and
dialysate compartment). In the early 1960s, solution
mass transfer resistances were decreased with the
introduction of parallel Sow dialysers, in which sheet
membranes were formed in a stacked conRguration.
The improvement in dialysate-side mass transfer with
these dialysers was particularly large because the dialy-
sis solution contacted the membrane under Sow condi-
tions as opposed to the semi-batch operation of the coil
dialyser. In addition, the membranes used in these
devices were thinner in structure, providing less dif-
fusive resistance than earlier versions. Although the
earliest manufactured parallel Sow dialysers were not
disposable, design improvements permitted the pro-
duction of disposable units by the late 1960s.
The last truly major development in haemo-
dialysers occurred more than 30 years ago when
the hollow Rbre artiRcial kidney was developed.
Blood compartment mass transfer was reduced fur-
ther with this design due to the high shear rate that
could be achieved in the annular space of the hollow
Rbre. Additional beneRts of the hollow Rbre artiRcial
kidney included an enhanced ability to control trans-
1687
Brian PLT (1966) Mass transport in reverse osmosis. In:
Merten U (ed.) Desalination by Reverse Osmosis, p.
181. Cambridge, MA: MIT Press.
Cheryan M (1998) UltraTltration and MicroTltration
Handbook. Lancaster: Technomic Publishing.
Zeman LJ and Zydney AL (1996) MicroTltration and
UltraTltration. New York: Marcel Dekker.
membrane pressure (see below) and a lower extracor-
poreal blood volume. This type of dialyser is now
used in virtually all HD treatments.
On a global basis, approximately 800 000 patients
receive chronic haemodialysis therapy for the treat-
ment of end-stage renal disease (ESRD) and this
population is growing at a rate of 8}10% per annum.
This Rgure represents approximately 85% of the
ESRD population, with the remaining patients receiv-
ing peritoneal dialysis. Numerous dialysis membrane
and haemodialyser manufacturers are situated
around the world, with the vast majority based in the
three largest markets: United States, Western Europe
and Japan.
The Haemodialysis Procedure
In addition to the dialyser, the other fundamental
component of a HD system is a dialysis machine,
which serves a number of purposes. First, it is equip-
ped with a roller pump that delivers blood, usually at
a rate of 200}500 mL min\1, from the patient to the
dialyser and back to the patient. Second, the dialysis
machine prepares dialysate by mixing (‘proportion-
ing’) water and a concentrated bicarbonate solution
in such a ratio that the dialysis Suid produced is the
same as that prescribed by a physician to meet the
needs of an individual patient. The typical dialysate
Sow rate is 500}800 mL min\1 and its major con-
stituents are sodium, potassium, calcium and bicar-
bonate. The pathophysiology of uraemia is such that
during the period between dialysis treatments, potas-
sium levels in the plasma rise while calcium and
bicarbonate levels fall. Consequently, the concentra-
tion of potassium in the dialysate is typically lower
than that in the plasma at the beginning of the pro-
cedure while dialysate calcium and bicarbonate con-
centrations are typically higher. The third major func-
tion of the dialysis machine is to provide an accurate
measurement of transmembrane pressure (TMP) in
the dialyser, which is deRned as the difference
between the average pressure in the blood and
1688 II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
dialysate compartments. Fluid removal requirements
are quite patient-speciRc in this patient population
such that both the rate and total volume of plasma
water ultraRltration need to be controlled accurately.
Accurate control of ultraRltration is achieved by con-
tinuous monitoring of dialyser TMP, which essential-
ly is an ultraRltration surrogate for a membrane of
speciRc hydraulic permeability. Finally, monitoring
components of the dialysis machine safeguards
against potentially catastrophic events, such as air
embolism or a massive blood leak related to a mem-
brane defect.
Classi\cation of Uraemic Solutes
In the properly functioning human kidney, plasma
water and blood solutes are removed by ultraRltra-
tion and convection, respectively. Solutes of molecu-
lar mass less than approximately 40 000 Da have
essentially unrestrained passage through the
glomerulus, the kidney’s Rltration unit. As such, the
clearance of these solutes approximates to the
plasma water ultraRltration rate, which is about
120 mL min\1 for humans of normal size. By deRni-
tion, ESRD is associated with absent or minimal
native kidney function. As a result, blood solutes
normally removed by the above Rltration mechanism
are retained in the blood stream with a resultant
several-fold increase in their plasma concentrations.
The classiRcation of uraemic solutes is typically
based on molecular mass and three well accepted
classes currently exist (Table 1). The Rrst category,
simply called ‘small solutes’, is comprised of nitro-
genous compounds of molecular mass less than
200 Da. These solutes are by-products of protein
metabolism and include the compounds urea (mo-
lecular mass 60 Da) and creatinine (113 Da), which
are commonly measured in clinical medicine. The
second category, referred to as ‘middle molecules’,
consists of a diverse group of molecules in the 200 to
2000 Da range. Although this class has been widely
studied from an experimental perspective, a represen-
Table 1 Classification of ureamic solutes
Solute class Molecular mass Examples
range (Da)
Small solutes (200 Urea
Creatinine
Middle molecules 200}2000 Appetite suppressant
Osteoblast inhibitor
Peptides/proteins 2000}40 000 AGE-peptides

2-Microglobulin
Parathyroid hormone
Source: Vanholder R and De Smet R (1999).
tative solute, which is clinically measurable, has not
yet been identiRed. Low molecular mass peptides and
proteins (molecular masses 2000 to 40 000 Da) are
the most recently identiRed class of uraemic toxins.
The plasma concentrations of these compounds are
typically increased 50}100-fold in ESRD. Recently,
a speciRc toxin in this class,
2-microglobulin (
2M:
molecular mass 11 800 Da), has been identiRed as
a causative factor in the development of dialysis-
related amyloidosis, a deposition disorder speciRc to
the ESRD population.
Dialyser Speci\cations
Contemporary hollow Rbre dialysers have nominal
surface areas ranging from 1.0 to 2.2 m2, although
the trend in clinical practice is to use devices at the
upper end of this range. Both the length (approxim-
ately 23 cm) and inner diameter (i.d.: usually 200
m)
of hollow Rbres used for clinical HD are fairly stan-
dard. The i.d. parameter represents a compromise
between the desirable characteristics of a short dif-
fusive pathlength and high shear rate with a small
i.d., and a low axial pressure drop and hydraulic
resistance with a large i.d. Rbre. On the other hand,
the variation in wall thickness is considerable, with
values ranging from 6 to 55
m. (See below for an
expanded explanation.) Based on the surface area of
the dialyser, the total number of Rbres comprising the
dialyser ranges approximately from 7000 to 12 000.
Extracorporeal Therapy Modes Used
in ESRD Patients (Figure 1)
In a typical haemodialysis procedure, although trans-
membrane mass transfer occurs predominantly by dif-
fusion, a modest degree of convective mass transfer is
also achieved in association with the ultraRltered
plasma water. However, the recent recognition of

2M and other low molecular mass proteins as impor-
tant uraemic toxins has prompted interest in using
dialytic therapies with increased convective removal
capabilities for these poorly diffusible solutes. In
haemodialysis, the total ultraRltration volume and net
ultraRltration rate are determined by the degree to
which a patient’s plasma volume needs to be reduced
and the duration of the treatment. (The total ultraRl-
tration requirement is dictated by the amount of Suid
ingested by the patient in the period between dialysis
treatments.) The total volume of plasma water ultraRl-
tered is approximately 3}4 L, resulting in a typical net
ultraRltration rate of 15}20 mL min\1.
As a means to augment convective solute removal,
haemoRltration (HF) was developed by Henderson,
 II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
Figure 1 Extracorporeal therapy modes used in end-stage renal disease.
Lysaght and colleagues in the early 1970s. This is
a purely convective therapy in which no dialysate is
used but an ultraRltration rate that far exceeds the net
ultraRltration requirements of the patient is em-
ployed. As plasma water is typically ultraRltered at an
absolute rate of at least 100 mL min\1 (6 L h\1) in
HF, the much lower net ultraRltration rate required
for Suid removal from the patient is achieved by
‘replacing’ most of the ultraRltrate with a bicarbon-
ate-based solution. For the large volume of intra-
venous-quality ‘replacement Suid’ that is required,
the Rltrate produced by sequential ultraRltration of
dialysate is used. This ‘on-line’ mechanism, in which
the dialysate precursor of the replacement Suid is
produced by the same HD machine that performs the
HF treatment, allows very high volumes of ultraRl-
trate to be produced. In HF, only dialysers with very
high hydraulic permeability (see below) are used.
Although HF is a signiRcant improvement over HD
with respect to relatively large sized uraemic toxin
removal, the absence of diffusion renders it only
a marginal therapy with respect to small solute re-
moval. To overcome this deRciency of HF, Canaud
and colleagues approximately 15 years ago Rrst em-
ployed online haemodiaRltration (HDF). As its name
implies, this therapy is essentially a HD/HF hybrid in
which both dialysate Sow and high ultraRltration rates
are used. At present, HDF offers the broadest
solute removal spectrum of all dialytic therapies.
Permeability Classi\cation of
Dialysis Membranes
Although numerous classiRcation schemes have
been proposed, HD membranes are traditionally clas-
siRed according to water Sux. The clinical parameter
used to characterize the water permeability of a dialy-
1689
ser is the ultraRltration coefRcient (KUF:
mL h\1 mmHg). In fact, the only dialyser classiRca-
tion scheme recognized by the United States Food and
Drug Administration is based on water permeability,
with low and high permeability dialysers having
KUF values of (8 and 58 mL h\1 mmHg, respec-
tively. The water permeability of a dialyser is usually
derived from in vitro experiments in which bovine
blood is ultraRltered at varying transmembrane pres-
sure. Based on a commonly used model which as-
sumes that a membrane is composed of parallel cylin-
drical pores, the Sux of plasma water through each
pore is dependent on the fourth power of the radius
so that small changes in mean pore size have a very
large effect on water permeability.
A common misconception relating to dialyser per-
formance is the assumption that a membrane’s solute
removal capabilities are necessarily correlated with
its water permeability. Based on a model in which
a membrane has N (straight) cylindrical pores (per
unit surface area) of radius r, diffusive solute Sux
can be expressed as:
"DC/t [1]
where  is the solute partition coefRcient, D is
solute diffusivity,  is membrane porosity, C is
the transmembrane concentration gradient, and t is
membrane thickness. (While the partition coefRc-
ient is essentially unity for solutes such as urea and
creatinine, larger solutes with incomplete access to
the membrane pores have  values that are less than
one.) Membrane porosity is a function of both pore
size and number:
"Nr2 [2]
1690 II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
For all dialysis membranes, small solutes such as
urea and creatinine have free pore access ("1).
Therefore, small solute transport is highly dependent
on membrane porosity. As eqn [2] indicates, one
membrane with a large number of relatively small
pores and a second membrane with a small number of
relatively large pores can have equivalent porosities.
Although the small solute transport properties of
these two hypothetical membranes would be equiva-
lent, the Sux (water permeability) properties would
greatly differ. This difference is explained by
the strong dependence of ultraRltrate Sux on mem-
brane pore size, described above.
Polymeric Composition of Dialysis
Membranes
From a relatively simplistic perspective, dialysis mem-
branes can be divided into those comprised of cellu-
lose-based material and those comprised of synthetic
materials.
Cellulosic Dialysis Membranes
The monomeric subunit of cellulosic membranes is
cellobiose, a naturally occurring saccharide found in
plants. Chemically, cellobiose is a ringed structure
richly endowed with hydroxyl groups. The interac-
tion of complement cascade products with these hy-
droxyl groups is felt to be responsible, at least partly,
for the relatively pronounced complement activation
observed when unsubstituted cellulosic membranes
contact blood. For the past several years, a major
objective among manufacturers has been the develop-
ment of modiRed (substituted) cellulosic membranes
in which a certain fraction of these hydroxyl groups
are replaced with other moieties. The substitution
groups diminish the degree of complement activation
by at least three different mechanisms. One
mechanism is the replacement of a large percentage of
the hydroxyl groups with acetate groups. In the Rrst
substituted cellulosic membrane, cellulose (di)acet-
ate, approximately 70}80% of the hydroxyl groups
on the cellulosic backbone were replaced with an
acetate group. Most likely because this modiRcation
eliminates a large fraction of the active surface sites
for interaction with complement components, an at-
tenuation of the intense complement activation seen
with unmodiRed cellulosics was achieved. This mem-
brane modiRcation also resulted in a moderate in-
crease in pore size, yielding a slightly higher water
permeability and broader solute removal spectrum
for cellulose acetate in comparison to unsubstituted
cellulosic membranes of similar surface area. Extra-
polation of this process to total replacement of the
hydroxyl groups resulted in the cellulose triacetate
Rbre characterized by further attenuation of comp-
lement activation and higher water permeability.
A second cellulosic substitution mechanism is the
replacement of a relatively small percentage (less than
5%) of the hydroxyl groups with a bulky chemical
group, which sterically reduces the degree of interac-
tion between complement activation products and the
membrane. Examples for which this strategy is em-
ployed are Hemophan威 (tertiary amine substitution)
and synthetically modiRed cellulose (SMC; benzyl
substitution group).
The evolution in cellulosic membranes has resulted
in a wide spectrum of biocompatibility and Sux pro-
Rles. If complement activation and neutropenia are
used as the major biocompatibility criteria, regen-
erated cellulose is the least biocompatible while cellu-
lose triacetate is the most biocompatible, with the
other modiRed cellulosic membranes having inter-
mediate proRles. However, characterization of the
Sux properties of these membranes is not as straight-
forward. For dialysers of comparable surface area,
a simplistic approach is to report KUF values in the
following ascending order: regenerated cellulose(
Hemophan威, synthetically modiRed cellulose(cellu-
lose acetate(cellulose triacetate. In this simplistic
scheme, a 1.5 m2 dialyser having a regenerated cellu-
lose, Hemophan威, or SMC membrane generally falls
in the low Sux category (KUF(8 mL h\1 mmHg)
while comparably sized dialysers having cellulose
acetate and cellulose triacetate membranes fall in the
midSux (KUF 10}20 mL h\1 mmHg) and high Sux
(KUF'20 mLh\1 mmHg) categories, respectively.
However, this simplistic categorization scheme
breaks down in several respects. High Sux cellulose
acetate membranes have now been produced and
cellulose triacetate dialysers of low water permeabil-
ity (KUF 9.5 mL h\1 mmHg) are also available. Fi-
nally, the recent development of unmodiRed cellu-
losic and cellulose acetate membranes having rela-
tively low water permeability but solute removal ca-
pabilities that include
2M further confounds this
classiRcation scheme and provides additional exam-
ples of a dissociation between water and solute Sux.
Synthetic Dialysis Membranes
The monomeric subunits of the various synthetic
membranes individually vary and all differ sig-
niRcantly from cellobiose. The absence of surface
hydroxyl groups on synthetic membranes is one
factor responsible for the reported differences in
complement activation between synthetic membranes
and either unsubstituted cellulosic membranes or
modiRed cellulosic membranes of low permeability.
 II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
Subsequent to the introduction of the AN69威 (sul-
fonated polyacrylonitrile) membrane in the early
1970s, numerous additional synthetic membranes
have been introduced for clinical use. Similar to
AN69威, polysulfone and polyamide were brought to
the market for use in both high Sux HD and haemoRl-
tration. One obvious reason accounting for the use of
these membranes in a haemoRltration mode is their
signiRcantly larger pore size and hydraulic permeabil-
ity than regenerated cellulose membranes. The other
reason relates to the structural differences be-
tween the synthetic and unsubstituted cellulosic
membrane groups. Cellulosic membranes have rela-
tively thin walls (generally in the 6}15
m range)
which have a uniform (symmetric) composition
across their entire thickness. Although the relative
thinness of cellulosic membranes is desirable with
respect to diffusive solute transport, this same
characteristic renders many cellulosic membranes un-
able to withstand the high transmembrane pressures
required to perform convective therapies employing
high ultraRltration rates. The synthetic membranes
have thicker walls (20
m or more) which may be
structurally symmetric (e.g. AN69威, polymethyl-
methacrylate (PMMA)) or asymmetric (e.g. polysul-
fone, polyamide, polyethersulfone). In the latter
category, a very thin ‘skin’ (less than 5
m) contacting
the blood compartment lumen acts primarily as the
membrane’s separative element with regard to solute
removal while the remaining thickness (stroma) im-
parts mechanical strength. In turn, the composition of
the stroma layer is quite variable for the various
synthetic membranes. For the Fresenius polysulfone
membrane, the stroma is relatively homogeneous
with a sponge-like structure while the Gambro poly-
amide membrane has, adjacent to the skin, a sponge-
like stroma layer which has progressively larger pores
(‘macrovoids’ with a Rnger structure) in the radially
outward direction. Finally, a new synthetic (poly-
ethersulfone) membrane developed by Membrana
GmbH (formerly Akzo Nobel) has a novel conRgura-
tion consisting of a sponge-like stroma layer inter-
posed between skin layers on both the inner (blood-
side) and outer (dialysate-side) aspects.
In the production of synthetic membranes made of
primarily hydrophobic polymers (polysulfone, poly-
amide, polyethersulfone), a hydrophilic additive
(polyvinylpyrrolidone: PVP) acts as a polymer alloy.
PVP is used to impart sufRcient hydrophilicity to
the membrane to allow clinical use and, as a wetting
agent, modulates surface tension and viscosity within
the pore structure during membrane formulation.
This latter feature explains PVP’s importance in de-
termining the overall pore size distribution of syn-
thetic membranes.
1691
Although synthetic membranes are employed for
both haemoRltration and high Sux HD, it is in the
latter mode that these membranes have found their
widest application. Another synthetic membrane for-
mulation was reported in the late 1980s with the
introduction of low Sux versions. Low Sux polysul-
fone and PMMA have been used clinically for several
years and recently a low Sux version of a poly-
amide/polyethersulfone copolymer has been intro-
duced.
Effect of Membrane Composition
and Structure on Dialytic Solute
Removal
Small Solute Removal
Small solute removal during HD occurs almost ex-
clusively by diffusion. To quantify a particular
membrane’s diffusive capabilities, its mass trans-
fer resistance is frequently used:
RO"RB#RM#RD
In the above equation, the overall resistance to dif-
fusive mass transfer of a particular solute (RO) has
three components: blood compartment resistance
(RB), resistance due to the membrane itself (RM) and
dialysate compartment resistance (RD). Minimizing
the mass transfer resistance in the blood compart-
ment primarily requires the use of relatively high Sow
rates (i.e. shear rates) that decrease unstirred layers.
Dialysate-side mass transfer resistance is likewise de-
creased by increasing Sow rate but optimal dialysate
perfusion of Rbre bundles is also a consideration.
Although increasing dialysate Sow rate may itself
improve Rbre bundle perfusion (see below), another
mechanism by which this can be achieved is the inclu-
sion of spacer yarns. These devices are spacing Rla-
ments placed external to the Rbres and are designed to
facilitate dialysate distribution and reduce channel-
ling. The resistance related to the membrane itself
actually has two components:
RM"XM/DM
where XM is the effective diffusion path-
length for a solute and DM is the solute-speciRc mem-
brane diffusivity. This equation indicates that
a decrease in membrane resistance can be achieved
either by a decrease in membrane thickness or an
increase in membrane diffusivity, the latter of
which is inSuenced strongly by membrane porosity.
1692 II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
Middle Molecule Removal
Vitamin B12 (molecular mass 1350 Da) is commonly
used for in vitro characterizations of dialysers. How-
ever, due to its extensive binding to plasma proteins,
this compound is not useful in vivo. In fact, the
removal of uraemic solutes having molecular masses
which fall in the classic middle molecule category has
been difRcult to quantify due to the lack of an
easily measured in vivo surrogate molecule (cf. urea
and creatinine for the small solute category). Because
recent evidence suggests that uraemic appetite sup-
pression is mediated by the retention of solute(s) in
this size range, an understanding of removal mecha-
nisms for middle molecules is important. Based on
dialysis practices used in the 1960s and early 1970s
(i.e. relatively low Sow rates and thick, low perme-
ability cellulosic membranes), diffusive middle
molecule removal was so limited that any convective
removal contributed relatively substantially to total
removal. However, the situation is vastly differ-
ent in contemporary HD, in which higher Sow rates
and dialyser membranes of signiRcantly greater
diffusive permeability for middle molecules are
employed.
Low Molecular Mass Protein Removal
Recent interest in increasing the extracorporeal re-
moval of
2M has provided insight into the general
mechanisms mediating the removal of low molecular
mass proteins. A number of studies published in the
past 15 years support several general conclusions.
First,
2M removal by low Sux unsubstituted cellu-
losic membranes is usually negligible, although
certain exceptions do exist. Second, the primary
mechanism by which
2M is removed during high
Sux HD varies widely among membranes. For certain
membranes, such as AN69威 and particularly PMMA,
removal is achieved predominantly or solely by ad-
sorption. At the other end of the spectrum is the
cellulose triacetate membrane, for which adsorption
is minimal and removal occurs primarily by dif-
fusion. High Sux polysulfone and unsulfonated PAN
membranes have intermediate adsorptive character-
istics and achieve transmembrane
2M removal by
a combination of diffusion and convection.
Third, at least for the high Sux synthetic membranes,
use of convection-based therapies (HF and HDF) in-
creases
2M removal relative to standard (diffu-
sion-based) HD. Although many clinicians consider

2M to be surrogate for the low molecular mass
protein class of uraemic solutes, this assumption has
not been conclusively proved. Nevertheless, it is rea-
sonable to use the abundant transport data available
for
2M to provide insight into the transport charac-
teristics of other low molecular mass proteins, such as
complement activation products and cytokines.
Interaction Between Biocompatibility
and Flux
Measurement of complement pathway by-products is
one technique used to assess the inSammatory re-
sponse elicited by exposure of blood to a dialysis
membrane. However, numerous previous studies
have failed to account for the fact that the clinically
measured complement components (C3a and C5a)
are low molecular mass proteins. Therefore, the con-
centration of these inSammatory mediators repres-
ents the net result of the simultaneous processes of
generation and any dialytic removal that may occur.
In this regard, complement activation products are
similar to most uraemic solutes, for which the plasma
concentration is determined by both generation and
net removal. The corollary of this observation is that
the permeability properties and not just the polymeric
composition of a dialysis membrane must be con-
sidered when evaluating complement activation data.
Recent data indicate that the relatively low levels of
complement activation associated with high per-
meability synthetic membranes is at least partially
related to their ability to remove, either by adsorption
or transmembrane transport, the generated inSam-
matory mediators.
It is simplistic to limit the discussion about mem-
brane biocompatibility to complement activation as
a number of agents have been identiRed as potential
inSammatory mediators in chronic HD patients.
A list of these putative mediators appears in Table 2.
Some of these compounds, such as Lipid A and LPS
fragments, potentially have their origin in dialysate,
a nonsterile Suid. Due to their relatively low
molecular mass, these inSammatory mediators may
undergo transmembrane passage and induce cytokine
Table 2 Inflammatory mediators
Mediator Molecular mass
(kDa)
Lipid A 2}4
Lipopolysaccharide (LPS) fragments (8
C3a 8.9
Granulocyte inhibitory peptide (GIP) II 9.5
C5a 11
Interleukin-1 17
Tumour necrosis factor (monomeric) 17
Factor D 23
Granulocyte inhibitory peptide (GIP) I 28
Tumour necrosis factor (trimeric) 55
Lipopolysaccharide (LPS) '100
Source: Lonneman G (1993).
 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
production in the blood stream, either directly via an
effect on mononuclear cells or indirectly via an
effect on the alternative complement pathway.
Conversely, the majority of the mediators that are
potentially elicited in the blood, such as C3a and
IL-1, may be simultaneously eliminated during high
Sux therapies by an adsorptive or transmembrane
mechanism, as discussed above. Other investigations
have conRrmed that adsorption is also important in
the removal of other inSammatory mediators, such as
Factor D and cytokines.
Summary
Dialysers used in contemporary HD are equipped with
a wide variety of membranes and within both the
cellulosic and synthetic classes, water and solute Sux
properties vary widely. For small and middle-sized
solutes, abundant clinical data point to the importance
of membrane thickness in diffusive mass transfer.
The removal of low molecular mass proteins may
occur largely by adsorption for some high Sux mem-
branes, particularly those of hydrophobic synthetic
composition. Because many of the mediators of in-
Sammation in dialysis patients fall in this low molecu-
lar mass protein category, the biocompatibility of
a particular membrane must be interpreted in con-
junction with its permeability properties.
See Colour Plate 47.
See also: II /Membrane Separations: Membrane Bio-
separations. III/Membrane Preparation: Hollow Fibre
Membranes; Interfacial Composite Membranes.
Further Reading
Anderstam B, Mamoun A, Sodersten P and Bergstrom J
(1996) Middle-sized molecule fractions isolated from
uraemic ultraRltrate and normal urine inhibit ingestive
behavior in the rat. Journal of the American Society of
Nephrology 7: 2453}2460.
Cheung AK, Parker C, Wilcox L and Janatova J (1990)
Activation of complement by haemodialysis
membranes: polyacrylonitrile binds more C3a than cu-
prophan. Kidney International 37: 1055}1059.
Diffusion Dialysis
T. A. Davis, Annandale, NJ, USA
Copyright ^ 2000 Academic Press
Introduction
Diffusion dialysis is a separation process in which an
ion exchange membrane separates a source solution
1693
Clark WR, Macias WL, Molitoris A and Wang NHL (1995)
Plasma protein adsorption to highly permeable haemo-
dialysis membranes. Kidney International 48: 481}488.
Clark WR, Hamburger RJ and Lysaght MJ (in press) The
effect of membrane composition and structure on
solute removal and biocompatibility in haemodialysis.
Kidney International.
Colton C, Henderson L, Ford C and Lysaght M (1975)
Kinetics of hemodiaRltration. I. In vitro transport char-
acteristics of a hollow Rber blood ultraRlter. Journal of
Laboratory and Clinical Medicine 85: 355}371.
Deppisch R, Gohl H and Smeby L (1998) Microdomain
structure of polymeric surfaces } potential for improving
blood treatment procedures. Nephrology, Dialysis and
Transplantation 13: 1354}1359.
Henderson L (1996) In Jacobs C, Kjellstrand C, Koch K and
Winchester J (eds) Biophysics of UltraTltration and
HemoTltration, 4th edition, pp. 114}145. Dordrecht:
Kluwer Academic Publishers.
Henderson L, Colton C and Ford C (1975) Kinetics of
hemodiaRltration. II. Clinical characterization of a new
blood cleansing modality. Journal of Laboratory and
Clinical Medicine 85: 372}391.
Jindal KK, McDougall J, Woods B, Nowakowski L
and Goldstein MB (1989) A study of the basic
principles determining the performance of several high-
Sux dialyzers. American Journal of Kidney Disease 14:
507}511.
Ledebo I (1998) Principles and practice of hemoRltration
and hemodiaRltration. ArtiTcial Organs 22: 20}25.
Leypoldt JK, Cheung A, Agodoa L, Daugirdas J, Greene
T and Keshaviah P (1997) Hemodialyzer mass transfer-
area coefRcients for urea increase at high dialysate
Sow rates. Kidney International 51: 2013}2017.
Lipps B, Stewart R, Perkins H, Holmes G, McLain E, Rolfs
M and Oja P (1967) The hollow Rbre artiRcial kidney.
Transactions of the American Society of ArtiTcial Inter-
nal Organs 13: 200}207.
Lonneman G (1993) Dialysate bacteriological quality and
the permeability of dialyzer membranes to pyrogens.
Kidney International (Suppl. 41) S195}S200.
Lysaght MJ (1988) Haemodialysis membranes in
transition. Contributions to Nephrology 61: 1}17.
Vanholder R and DeSmet R (1999) Pathophysiologic ef-
fects of uremic retention solutes. Journal of the Ameri-
can Soceity of Nephrology 10: 1815}1823.
and a receiving solution, usually water. Anion ex-
change membranes are notoriously permeable to
acids, and diffusion dialysis exploits this property to
separate acids from salts. A common application
of diffusion dialysis is recovery of acids from
waste metal pickling solutions, the strong acid solu-
tions that are used to remove oxide coatings from
metal parts before they are painted, galvanized or
1694 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
electroplated. Cation exchange membranes are per-
meable to bases, and this is utilized to recover NaOH
from aluminium etching solutions.
Diffusion dialysis of acids through anion exchange
membranes was reported as early as 1964, and was
installed on an industrial scale by 1980. There have
been many laboratory studies on membrane proper-
ties and transport of acid through such membranes.
Therefore, the discussions that follow concerning the
theory and practice of diffusion dialysis will focus
primarily on acid transport through anion exchange
membranes. Base dialysis is relatively new, and there
is not a large body of knowledge about the mecha-
nism of transport, design criteria and performance of
that process. Until such information becomes avail-
able, it is reasonable to assume that the theory and
practice of base dialysis parallels that of acid dialysis.
Since ion exchange membranes have an ionically
charged polymeric structure, their discrimination be-
tween solutes is based on ionic charge. Anion ex-
change membranes are easily permeated by anions,
but cations are rejected, because the positive ionic
change of the membrane matrix repels the cations.
Unlike other cations, hydrogen ions are an integral
part of the water that pervades the membrane, and
hydrogen ions seem to permeate by a different mecha-
nism that avoids the rejection of the charged polymer
structure. Anion exchange membranes transport
acids while rejecting salts.
Figure 1 illustrates diffusion dialysis for recovery
of HNO3 from a solution also containing Fe(NO3)2.
The anion exchange membrane is quite permeable to
the NO\ 3 ions, but an equivalent amount of cations
must also pass through the membrane to maintain
electroneutrality. Because of their double positive
Figure 1 Diffusion dialysis to recover HNO3 from pickling solution.
charge, the Fe2# ions are strongly rejected by the
membrane, but the protons are transported rather
easily. Thus, a useful separation of acid and salt is
achieved.
Background and Theory
Transport in diffusion dialysis is described by Fick’s
law:
Flux"!U
C [1]
where
C is the concentration difference of the dif-
fusing solute (the driving force for diffusion) and U is
a mass transfer coefRcient, expressed in units of
length time\1. Since the concentrations can be mea-
sured only in the bulk solutions, the measured value
of
C is the driving force for diffusion through the
membrane and the solution boundary layers next to
the membrane. Therefore, an overall mass transfer
coefRcient Uo is needed to describe the observed Sux.
The reciprocal of the mass transfer coefRcient is the
diffusional resistance, and the diffusional resistances
of the membrane and the adjacent liquid boundary
layers are additive.
1/Uo"1/Um#1/Ul [2]
Values of U for a particular solute through a particu-
lar membrane are conveniently measured in a stirred
cell with the membrane separating the source solution
from the receiving solution, usually pure water. With
sufRcient stirring the resistance of the liquid can be
minimized so that the measured value of U is essen-
tially Um. Acids permeate some anion exchange
 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
membranes rapidly, with U values of about 10\4 to
10\3 cm s\1 while salts have U values of about
10\6 cm s\1. Therefore, there is sufRcient difference
in the diffusion rates to achieve useful separations of
acids from salts by diffusion dialysis.
Since solution velocities in commercial dialysers
are slow, Ul could be a signiRcant part of Uo. A rough
idea of the resistance in the boundary layer can be
estimated by examining the elements of the equation
for diffusive Sux through a Rlm of liquid:
Flux"!D
C/z [3]
where D is the diffusivity of the solute through the
solvent, typically about 10\5 cm2 s\1, and z is the
thickness of the Rlm of liquid through which diffusion
occurs. Spacing between membranes in a commercial
dialysis apparatus is somewhat less than 0.1 cm, so
liquid Rlm thickness z would probably be about
0.01 cm. Then D/z"10\3 cm s\1, which is a U value
for the liquid Rlm of the same order of magnitude as
the typical U values for dialysis membranes.
Consequently, both the membrane and the liquid
Rlms in contact with it are likely to contribute to the
resistance to diffusion in a real dialysis application,
even at rather high solution velocities.
Transport of solvent through dialysis membranes
can be great enough to inSuence diffusion dialysis
performance. Osmotic forces provide a driving force
to transport solvent from the dilute solution to the
concentrated solution. However, the diffusing solute
can drag along solvent, both in the solvation shells of
the ions and by convection, in the direction opposite
to that of normal osmosis. Further, osmotic pressures
are caused by the concentration difference of nondif-
fusing solutes across the membrane, the values of
which can be difRcult to determine. Consequently,
even the direction of solvent transport can be difRcult
to predict in certain circumstances, and prediction of
the rate of solvent transport is quite difRcult.
Mathematical analysis of dialysis is rather simple if
the assumptions are made that the overall value of
U is independent of C and that solvent transport is
negligible. There are two typical cases that are usually
encountered with dialysis in general or with diffusion
dialysis.
Case 1 is an experiment done in an apparatus used
to measure dialysis coefRcients, i.e. U values.
A sample of the membrane is placed between two
chambers of equal volume in a stirred cell, with a sur-
face area A exposed to both solutions. The source
solution Rlls one chamber, and an equal volume V of
pure water, the receiving solution, Rlls the other
chamber. Because the volumes of the two solutions
are equal, the concentration of the diffusing solute
1695
decreases in the source solution at the same rate as it
rises in the receiving solution on the other side of the
membrane. The rate of concentration change, dC/dt,
is related to the Sux, volume and area of the mem-
brane by the equations below. On the side containing
the receiving solution:
dC/dt"Sux;A/V [4]
and on the side containing the source solution:
dC/dt"!Sux;A/V [5]
To integrate this equation, an expression is needed
for Sux in terms of concentrations on one side of the
membrane. The appropriate expression can be ob-
tained by material balance. Let Cs represent the con-
centration of the diffusing solute on the side with the
source solution. Then the concentration of the diffus-
ing solute in the receiving solution would be
Cr"C0!Cs, where C0 is the initial concentration of
the source solution. Now an equation for solute Sux
can be written as follows:
Flux"!U;
C"!U;[Cs!(C0!Cs)]
"!U;(2Cs!C0) [6]
The differential equation can be integrated to yield:
Cs"C0(1#e\2tUA/V)/2 [7]
on the side of the source solution and:
Cr"C0(1!e\2tUA/V)/2 [8]
on the side of the receiving solution.
The experiment described in case 1 is a useful
technique for measuring values of U for a membrane.
SufRcient stirring can reduce solution Rlm resistance
to negligible levels, and even volume changes are
insigniRcant in short experiments. Volume changes
and analytical inaccuracy can cause substantial errors
if source solution concentrations are used in this de-
termination, so determination of the U value should
be based on the measured concentrations in the re-
ceiving solution.
Although the apparatus described in case 1 is useful
for determining membrane properties, it is of limited
commercial value as a separation process because no
more than half of the diffusing solute can be removed
when equal volumes are used on both sides of the
membrane. A high degree of removal would require
a volume of the receiving stream much larger than
that of the feed, but that has limited commercial
1696 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
appeal. A commercially useful separation can be
achieved with countercurrent Sow of the solutions
through the dialyser.
Case 2 is an example of countercurrent Sow of the
solutions on opposite sides of the membrane. The
system operates at steady state so that concentrations
do not change with time, but they do change with
position along the Sow path of the solutions. To
simplify the equations it will be assumed that there is
no solvent transport through the membrane and that
the source and receiving solutions have the same Sow
rate, F. When pure water is used for the receiving
stream, the material balance is simply:
Cf"Cd#Cr [9]
where the subscripts represent the feed, depleted and
recovered streams respectively. The amount of solute
transferred across the membrane is equal to the
amount of solute appearing in the recovered stream:
UA
C"FCr [10]
Because the Sow rates are equal, the concentration
change within a solution compartment is linear with
respect to distance along the Sow path, so the concen-
tration difference across the membrane is equal to the
arithmetic mean concentration difference:

C"(Cf#Cd!Cr)/2 [11]
Combining this with the material balance equation
yields:

C"Cf!Cr [12]
which can be combined with the transfer equation:
UA(Cf!Cr)"FCr [13]
and rearranged to show the fraction of solute re-
covered:
Cr/Cf"U/(U#F/A) [14]
In practice, the values for U are often expressed in
the same units as the Sow rate per unit area of
membrane, L h\1 m\2. For the diffusion of HCl from
pickle liquor through Neosepta AFN anion exchange
membrane, reported values of U are 8.6 L h\1 m\2
for HCl and 0.17 L h\1 m\2 for Fe, and a typical
value for F/A might be 1 L h\1 m\2. With these
values the HCl recovery would be 8.6/(8.6#1)"
0.9, and the Fe leakage would be 0.17/(0.17#1)"
0.15. Thus, 90% of the HCl is recovered, and 15% of
the Fe appears in the recovered acid. Leakage of Fe
could be reduced to 8% by doubling the Sow rates,
but HCl recovery would drop to 81%.
The simpliRed equation developed above for
countercurrent dialysis is only applicable when Sow
rates of both streams in the dialyser are equal. For
those more general situations with unequal Sow
rates, the log-mean concentration difference would
be used as the driving force in Fick’s law.
Deviations from Simple Modelling
Osmotic forces play a key role in the water balance,
and water transport through the membrane can in-
validate the simple mathematical models described
above. The following discussion is based on the re-
covery of acid from a steel pickling solution, which is
a signiRcant industrial application of diffusion dialy-
sis. In the recovery of acid from a mixture with
a metal salt the major driving force for osmosis is the
difference in concentration of salt across the mem-
brane. The osmotic Sow of water can cause the vol-
ume of the receiving stream to decrease as much as
20% as it passes through a typical industrial dialyser.
Therefore, a good mathematical model of diffusion
dialysis should account for water transport through
the membrane.
The presence of salt in the source solution can
substantially affect the concentration of acid in the
recovered stream. In diffusion dialysis of metal pickle
liquors the source solution has two important compo-
nents } the free acid that can diffuse through the
anion exchange membrane rather easily and the metal
salt that is rejected by the membrane because of
repulsion of the metal cations by the Rxed positive
charge on the membrane matrix. There are numerous
reports of countercurrent diffusion dialysis in which
the acid concentration in the recovered stream is
higher than the free acid concentration in the feed.
Some writers have attempted to explain these obser-
vations in terms of osmotic removal of water from the
receiving stream, but it seems more plausible that
a concentration difference of the common anion pro-
duces a driving force for transport of protons through
the membrane. That driving force is the Donnan
potential (discussed in ‘Membrane Separations: Don-
nan Dialysis’) generated by the difference between the
activity of anions in the two solutions. That potential
difference provides a driving force for proton trans-
port in addition to the driving force provided by the
difference in concentration of the free acid.
Fick’s law describes acid Sux due to simple diffu-
sion as the product of the driving force
CA and
the mass transfer coefRcient for diffusion, UA. A
similar expression can be used to describe the acid
 II / MEMBRANE SEPARATIONS / Diffusion Dialysis 1697

Figure 2 Diffusion dialysis of pure 4 mol L\1 HNO3 feed stream (continuous line) with a pure water-receiving stream (dotted line).

transport due to the concentration difference of the
salt with a common anion, with
CS as the driving
force and US as the mass transfer coefRcient. The total
acid Sux can be expressed as the sum of the Sux due
to individual driving forces as follows:
Total acid Sux"UA
CA#US
CS
This empirical equation was tested with published
data by Edwards, who measured HNO3 concentra-
tions on both sides of a Tokuyama AFN membrane in
a stirred cell. Graphs of data for three different start-
ing compositions are shown in Figures 2}4. In each
graph the data at the top connected by the continuous
line show the reduction in concentration as acid dif-
fuses from the source stream, and the data at the
bottom show the increase in acid in the receiving
stream. Figure 2 shows simple diffusion of HNO3,
which is well described by the dotted line that was
calculated by Fick’s law with UA"12.8 L h\1 m\2.
But when NaNO3 was added to the source solution,
Fick’s law with UA"12.8 L h\1 m\2 (shown in Fig-
ure 3 by the dotted line) predicts a much slower
[15] appearance of acid than the data indicated. The equa-
tion for total acid Sux with values of UA"
12.8 L h\1 m\2 and US"0.45 L h\1 m\2 shown by
the dashed line gave a good correlation of the data. It
is interesting to note in Figure 3 that the inSuence of
the excess nitrate forced so much HNO3 out of the
source stream that its concentration fell below that in
the receiving stream. The situation was reversed when
NH4NO3 was placed in the receiving solution.
Figure 4 shows that the total acid Sux equation
with values of UA"12.8 L h\1 m\2 and US"

Figure 3 Diffusion dialysis of 0.1 mol L\1 HNO3 and 5 mol L\1 NaNO3 feed stream (continuous line) with a pure water-receiving
stream (dashed and dotted lines). Lines calculated as described in the text.
1698 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
Figure 4 Diffusion dialysis of a 2.1 mol L\1 HNO3 feed stream (continuous line) against a 4 mol L\1 NH4NO3-receiving solution
(dashed and dotted lines).
0.45 L h\1 m\2 again described the data better than
Fick’s law and correctly showed that the appearance
of acid in the receiving solution was retarded by the
presence of nitrate in that solution. It should be em-
phasized that the total acid Sux equation is empirical,
but it does seem to be a useful way of accounting for
the effects of salts on acid Sux and would thus be
useful for mathematical modelling of diffusion dialy-
sis of acids.
Competing Processes
Diffusion dialysis, like most other membrane pro-
cesses, must compete with other processes that can
achieve the desired separation. Lime neutralization,
sorption on ion exchange resins and bipolar mem-
branes are competing processes for the treatment of
waste acids from metal pickling. When disposal and
replacement acid costs are low, lime neutralization is
the most economical alternative. When the recovered
acid can be used in a diluted form, sorption on ion
exchange resins is attractive. When it is necessary to
minimize discharge, bipolar membranes, though ex-
pensive, can be the preferred process. Diffusion dialy-
sis offers the important advantages of very low oper-
ating costs and long membrane life that can exceed
5 years if clean, nonfouling feeds are used. Therefore,
if diffusion dialysis can achieve the desired separation
and if capital costs are tolerable, diffusion dialysis can
be the process of choice for recovering waste acids.
Membranes for Diffusion Dialysis
Modern membranes for acid dialysis are made of
anion exchange polymers which have an afRnity for
acids and reject cations other than protons. The anion
exchange membranes that are most permeable to
acids seem to be those with a very high water content.
The water content is important because the high
electromobility of protons through water is attribu-
table to a transfer mechanism that is not available to
other cations. A proton can transfer from a hy-
dronium ion to an adjacent water molecule by
a mechanism which was Rrst suggested by Grotthus
in 1806.
The major suppliers of diffusion dialysis mem-
branes and devices are Asahi Glass, who make
Selemion DMV, and Tokuyama Corporation who
make Neosepta威 AFN and AFX membranes. Since
the membrane devices for diffusion dialysis are sim-
ilar to those used in electrodialysis, any supplier of
electrodialysis equipment is capable of supplying dif-
fusion dialysis equipment as well. The supplier with
the largest number of installations in the USA is Pure
Cycle Environmental Technologies in Palmer, Mass-
achusetts, and the largest supplier in Europe is Euro-
dia in Paris, France.
Membranes for base dialysis were parchmentized
paper when the process was Rrst developed in the
1930s. The early membranes had little selectivity
between caustic and salts, but they effectively re-
tained hemicellulose from rayon and viscose pro-
cesses. Modern membranes for base dialysis are made
of cation exchange polymer that rejects anions other
than hydroxyl, presumably because of the Grotthus
mechanism of hydroxyl transport. The typical cation
exchange membrane made for electrodialysis has low
Sux for base diffusion, but Tokuyama Corporation
has made a special cation exchange membrane,
Neosepta威 CMX-SB, with acceptably high Sux.
 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
Design of Processes and Equipment
Industrial diffusion dialysis usually operates
with countercurrent Sow of the solutions on opposite
sides of the membrane. Countercurrent Sow produces
the maximum concentration difference over the en-
tire length of the membrane and allows recovery of
a substantial portion of the most highly diffusive
solute while minimizing the transport of the less dif-
fusive solutes. Since Suxes in diffusion dialysis are
relatively low compared to other membrane pro-
cesses, the solution velocity across the membrane
surface must also be slow in order to have enough
residence time for adequate removal of the solute.
Typical solution velocities in diffusion dialysis are
about 1 cm min\1. Convective effects due to density
changes in the solutions can be important with such
low velocities.
Dialysers for industrial applications must be ro-
bust, cleanable, efRcient and economical. Industrial
diffusion dialysers usually have Sat sheet membranes
with some type of spacer to keep the membranes
apart and to form solution compartments. The mem-
brane arrangement (without showing the spacers)
and the directions of solution Sows in a typical diffu-
sion dialyser are shown in Figure 1. A single dialyser
can contain hundreds of identical, vertically oriented
membranes. Membranes and spacers have holes that
are aligned to form manifolds, and each spacer has
entry ports that connect the solution compartment to
the appropriate manifold. This manifolding, which is
also typical in electrodialysis, distributes the solution
equally to the parallel compartments. The feed solu-
tion usually enters the bottom of the dialyser and
the solvent usually enters at the top, as shown in
Figure 1.
Solutions Sowing through the industrial diffusion
dialyser should be free of particulate matter, because
the solution velocities are too slow to sweep out
particles. The dialyser can be expected to perform
maintenance-free for several years if the feed solution
is clean and no precipitation occurs within the dialy-
ser. A single Rlter on each supply line should sufRce if
the feed solution is inherently clean. However, pri-
mary and secondary Rltration is recommended if par-
ticles are expected to be present in the feed because of
the possibility of contamination during the cleaning
or replacement of primary Rlter elements.
The low solution velocities that characterize diffu-
sion dialysis cause extremely low pressure drops
through dialysers, usually just a few kPa. Many dialy-
sers can be fed in parallel from a single header tank
positioned just above the dialysers. The solutions
exiting the dialysers also enter a header tank with
adjustable overSow levels. The header tanks should
1699
have covers and Rltered vents to avoid the entrance of
dust. Transparent tanks or sight glasses positioned
close together allow the operator to monitor visually
the pressure drop through the dialyser. Density differ-
ences of the solutions should be considered in deter-
mining the actual pressure head.
In metal Rnishing plants the dialysis process is
normally set up to run continuously to treat a small
stream of the metal-laden acid in the pickling tank
and return the recovered acid to the tank. This allows
the dialysis to run as a steady-state process. The waste
stream, which typically contains 10% of the acid and
90% of the metals from the feed, is usually neu-
tralized to precipitate the metal as hydroxides for
disposal or recovery.
It is important that the solution Sow is uniformly
distributed to all solution compartments that are fed
in parallel. Density changes caused by solute transfer
across the membranes are utilized to achieve uniform
Sow distribution. The feed solution, which has the
highest density, enters the bottom of the dialyser and
decreases in density as acid is removed. Osmotic
water transport into this concentrated solution also
contributes to the decrease in density. The receiving
solution increases in density as it Sows downward.
The uniform gradation in density allows the solutions
to approach plug-Sow conditions in each solution
compartment.
Entrapped gas can cause Sow disruptions in the
receiving solution compartments of acid dialysers.
Water entering the top of the receiving solution com-
partments contains some dissolved gases (O2, N2,
CO2) that can form bubbles in the downward-Sowing
solution. Even if the water is not initially super-
saturated with dissolved gases, the addition of solute
diffusing across the membrane can lead to super-
saturation within the receiving stream. The slow
downward Sow of solution is not sufRcient to force
the bubbles out the bottom of the dialyser, but it
could hinder their rise in the compartments, Bubbles
eventually reach such a large size that buoyancy for-
ces exceed the forces of surface tension. Then the
large bubbles rise and collect in the top of the com-
partment and eventually in the entry ports where they
block off the Sow of water into some of the receiving
solution compartment. As more compartments be-
come blocked, the solution velocity in the remaining
compartments increases. But that increase in Sow rate
means that the residence time in those open compart-
ments is shorter, so the performance of the dialyser
deteriorates.
Removal of the dissolved gases from the water
before it enters the dialyser is beneRcial. Gases can be
removed by heating the water in an open or vented
tank, by application of a vacuum or by passing the
1700 II / MEMBRANE SEPARATIONS / Diffusion Dialysis
water through a nonwetting microRltration device
with a vacuum applied to the opposite side of
the microporous membrane. Another remedy is peri-
odically to reverse the Sow of the receiving stream
and force the bubbles out of the top of the dialyser
into a vent tank. Flow reversal can be accomplished
easily with a centrifugal pump situated in the line of
the receiving solution at the entrance or exit to the
dialyser. The header tanks must have sufRcient surge
capacity to accommodate the volume of the Sow
reversal. Flow reversal for a few seconds is sufRcient
} just long enough to displace any gas that has
accumulated in the top of the receiving compartments
and entry ports. These two remedies are often used
together.
The heat of dilution of the acid can also cause
problems of overheating in diffusion dialysis. Because
the dialyser acts like a countercurrent heat exchanger,
the heat released in the dialyser tends to become
trapped inside. When the concentration of acid in the
feed is high, the peak temperature, which occurs
about halfway through the Sow path, can be high
enough to damage the membrane.
Limitations of Diffusion Dialysis
A necessary condition of dialysis is that the solute
concentration in the recovery stream must be lower
than in the feed stream in order to provide a driving
force for diffusion. This is not a real limitation in
applications where the diffusing solute is a waste that
can be easily discarded. But this condition can be
a limitation when the diffusing solute is the desired
product, because the product is often recovered at
a low concentration. Fortunately, the acid from steel
pickling solution can be recycled to the pickling bath
at the concentration at which it was recovered. An-
other limitation is that the nondiffusing solutes are
left in the original solution in a slightly diluted state,
which means that the waste volume can be consider-
able.
The selectivity of diffusion dialysis membranes
for rejecting metal ions is inSuenced by the ionic
charge on the metal ion. Metal ions with multiple
positive charge are rejected more efRciently than
ions with a single charge. However, zinc and some
other metal ions form complexes with the anions of
the acid. In HCl solutions, zinc forms ZnCl\
and ZnCl\ 4 complexes that behave as anions in the
anion exchange membrane. These complex ions do
not diffuse through the membrane as readily as Cl\
ions do, but they diffuse much faster than Zn2# ions.
However, zinc does not form a complex in H2SO4
solution, so zinc is rejected quite well in the sulfate
system.
All of the halogens form complexes with some
metals. Chloride complexes of Cu, Ga, Fe (ferric
forms a much stronger chloride complex than fer-
rous), V and Zn have been reported. The existence of
a chloride complex does not necessarily mean that
HCl cannot be recovered from the metal salt by
diffusion dialysis. Since the chloride complexes are
rather bulky, they do not pass through the anion
exchange membranes as easily as chloride ions do.
Applications
The Rrst important industrial application for dialysis
seems to have been for recovery of caustic from vis-
cose, hemicellulose, wood-pulping solutions and
textile-processing solutions. In the 1930s there
were many patents describing dialysers that utilized
diaphragms of parchmentized paper and regenerated
cellulose. Publications of that era described dialysis
as a method for separating crystalloids (sub-
stances that form true solutions and are capable of
being crystallized) from colloids (small particles in
suspension).
By far the most important application for dialysis
was begun during the 1940s when Dr. Willem Kolff
discovered that treatment of blood by dialysis re-
moved urea and other metabolic wastes, and he pro-
ceeded to develop the artiRcial kidney. The artiRcial
kidney and other conventional dialysis processes are
described in detail in ‘Membrane Separations: Dialy-
sis in Medical Separations’.
Conclusion
Diffusion dialysis utilizes membranes that contain
ion exchange groups, and those were not available
until the 1950s. Diffusion dialysis plants have been
recovering and recycling acids in Japan since 1980,
and many are being installed in the USA, particularly
in metal-Rnishing facilities. Acids that have been re-
covered include HCl, HF, HNO3, H2SO4 and
methanesulfonic. The recovered acid is sufRciently
concentrated to be returned to the pickling tank, and
the acid-free solution of metal salts requires consider-
ably less base to precipitate the metal hydroxides.
Recovery of mixed HF and HNO3 from the pickling
of stainless steel is important because these
3 cids are expensive and cause severe pollution prob-
lems if they are discarded. Diffusion dialysis has
been applied to the recovery of H2SO4 from
aluminium anodizing baths where the trivalent alu-
minium cation is well rejected by the anion exchange
membrane.
Base dialysis membranes have been used commer-
cially for the recovery of NaOH from the waste
 II / MEMBRANE SEPARATIONS / Donnan Dialysis
generated by the chemical milling of aluminium air-
craft parts. Chemical milling is used to remove metal
from aluminium parts, such as curved sections of
wing or fuselage that are difRcult to machine with
mechanical devices. The part is dip-coated with a Rlm
of rubber, and then a selected portion of the rubber is
stripped away to expose the metal surface. Then the
part is immersed in boiling NaOH that rapidly and
uniformly dissolves the metal from the exposed sur-
face. The dissolved aluminium accumulates in the
etch tank as NaAlO2, which must be discarded event-
ually. When the NaOH is removed from the solution
by dialysis, the NaAlO2 hydrolyses to Al(OH)3 and
NaOH. The Al(OH)3 is recovered by Rltration and
sold as a pure product, and the released NaOH is
returned to the etch tank along with the dialysed
NaOH. Dialysis allows recovery of essentially all of
the NaOH and completely eliminates the need for
disposal of the waste etchant. An industrial installa-
tion of base dialysis has been operating success-
Donnan Dialysis
T. A. Davis, Annandale, NJ, USA
Copyright ^ 2000 Academic Press
Introduction
Donnan dialysis is a separation process that utilizes
counterdiffusion of two or more ions through an
ion-exchange membrane to achieve a separation.
It can also be viewed as a continuous deionization
process. For example, water softening can be done
with a cation}exchange membrane. Hard water Sows
on one side of the membrane, and NaCl brine Sows
on the other side. Na# ions from the brine
diffuse across the membrane and cause the Ca2#
and Mg2#ions to diffuse in the opposite direction.
Donnan dialysis is usually performed as a continuous,
countercurrent process so that a substantial portion
of a cation from a dilute solution could be concen-
trated into a small volume. Differences in the
volumes and concentrations of the two solutions can
be exploited to achieve some interesting and useful
separations.
Donnan dialysis can be used for changing composi-
tions of process or analytical solutions, pollution con-
trol, and even deionization of a process stream. The
deionization process, called ‘neutralization dialysis’,
1701
fully in a chemical milling plant in California since
1991.
See Colour Plate 48.
Further Reading
Bailey DE (1993) Acid recycling system. US Patent 5 264
123.
Davis TA (1991) Recovery of sodium hydroxide and alumi-
nium hydroxide from etching waste. US Patent 5 049
233.
Marshall RD and Storrow JA (1951) Dialysis of caustic
soda solutions. Industrial and Engineering Chemistry.
Engineering and Process Development 42: 2934}2943.
Saddington AW and Julien AP (1938) Dialysis of aqueous
caustic solution. US Patent 2 138 357.
Shigekuni N and Motomura K (1979) Diffusion dialysis
method. Japanese Patent 54 136 580.
Zender J (1946) Process and apparatus for dialyzing solu-
tions. US Patent 2 411 238.
combines Donnan dialysis through both cation-ex-
change and anion-exchange membranes in one appar-
atus with H# and OH\ ions exchanging for the
cation and anion of a salt.
In the discussions that follow, the fundamental
principles of Donnan dialysis will be presented,
and some of its applications and capabilities
will be described. The type of equipment and
membrane arrangements appropriate for both
Donnan dialysis and neutralization dialysis will be
presented.
Background
The Donnan dialysis process is named after F. G.
Donnan who in 1924 described the equilibrium that
resulted when a semipermeable membrane separated
two solutions of electrolytes. NaA on one side and
KA on the other. The membrane he used was pre-
pared by Rlling the pores of parchment paper with
a gel of copper ferrocyanide, and he used ferrocyan-
ide as the common anion A of the two salts. When the
initial volumes and concentrations of the two salt
solutions were the same, counterdiffusion of equal
amounts of Na# and K# through the membrane led
to an equilibrium condition where the two solutions
had equal concentrations of NaA and KA. But when
1702 II / MEMBRANE SEPARATIONS / Donnan Dialysis
initial concentrations were different, counterdiffusion
of equal amounts of Na# and K# through the mem-
brane produced solutions with equal ratios of
Na#/K# on both sides of the membrane at
equilibrium. This relationship of concentrations of
the ions in the solutions on opposite sides of the
membrane is called the ‘Donnan equilibrium’.
Before proceeding with the theory, it is useful to
deRne some terms that are often used in discussions of
modern polymeric ion-exchange membranes. First,
an ‘ion-exchange membrane’ is a plastic Rlm with
Rxed ionically charged groups dispersed more or less
uniformly within the Rlm. Associated with the Rxed
charges are mobile charges of opposite sign called
‘counterions’. The counterions are free to exchange
with other counterions in the adjacent solutions.
When the membrane has Rxed negative charges, e.g.,
sulfonic acid groups, it is called a ‘cation-exchange
membrane’, and the counterions are cations. In the
external solutions, mobile anions are associated with
the cations, but in the membrane the charge balance
is satisRed by the Rxed negative charges. Therefore,
anions tend to be excluded from the interior of the
cation-exchange membrane. The ions with the same
charge as the Rxed charge in the membrane are called
‘co-ions’.
Theory
The Donnan equilibrium relationship is derived from
thermodynamics. Under conditions of equilibrium
the chemical potential
i of any dissolved species i is
the same in every phase present:

i"
3i #RT ln ai [1]
Here,
3i is the chemical potential of species i in the
standard state, R is the gas law constant, T is the
absolute temperature, and ai is the activity of the
particular chemical species i being considered. How-
ever, electrical potentials must also be considered
when the chemical species are ionic, so the electro-
chemical potential i must be used to describe the
equilibrium:
i"
3i #RT ln ai#ziF [2]
where zi is the ionic charge, F is Faraday’s constant,
and  is the electrical potential. When the two liquids,
phase 1 and phase 2, are at equilibrium with the
membrane, there is also equilibrium between the two
liquid phases, and the electrochemical potential of any
mobile species i in the two phases can be equated.
i1"i2, or RT ln ai1#ziF1"RT ln ai2#ziF2 [3]
It should be noted that the
3i terms cancel because the
same standard state exists in both liquid phases. But
the co-ions are not free to move through the mem-
brane that separates the two liquid phases, so there is
no opportunity for their concentrations to change.
Whenever salt concentrations on opposite sides of the
membrane differ, there will be a potential difference
across the membrane caused by the concentration
difference. This potential difference, called the ‘Don-
nan potential’, EDon, is described by rearrangement of
eqn [3]:
EDon"2!1"RT/ziF (ln ai1!ln ai2)
"RT/F ln(ai1/ai2)1/zi [4]
Since the Donnan potential acts on all mobile ionic
species, the value of (ai1/ai2)1/zi is the same for all of the
counterions in the system. In other words, the con-
centration difference of the co-ions causes an electri-
cal potential that acts on the counterions.
As Donnan pointed out in his seminal description
of the theory, a precise treatment of the equilibria
would require the use of activities rather than concen-
trations of ions in the solutions. But the use of molar
concentrations greatly simpliRes the presentation of
the theory, so that is the approach taken here. For the
experiment described by Donnan where zi"#1 for
both Na# and K# ions, the equilibrium written with
concentrations would be:
[Na#]1/[Na#]2"[K#]1/[K#]2 or [Na#]1/[K#]1
"[Na#]2/[K#]2 [5]
Figure 1 illustrates the Sow of ions in the approach to
Donnan equilibrium. Two dilute salt solutions NaCl
and KCl are separated a cation-exchange membrane,
labelled C, which is permeable to the cations Na#
and K# but impermeable to the common anion Cl\.
The concentration difference of Na# ions across the
membrane provides a driving force for their diffusion
through the membrane. There is no net Sow of elec-
tric current through the membrane, so any net trans-
fer of Na# to the right must be balanced by transfer
of an equivalent amount of K# to the left. Those
diffusive processes will occur until an equilibrium is
established.
The equilibrium concentrations can be expressed in
terms of the initial molar concentrations c1 of NaCl
on the left and c2 of KCl on the right, x moles
transported through the membrane (still the same for
both cations) and the volumes V1 and V2 of the
 II / MEMBRANE SEPARATIONS / Donnan Dialysis
Figure 1 Donnan potential forces K> ions to higher concen-
tration.
solutions, expressed in litres:
(c1!x/V1)/(x/V1)"(x/V2)/(c2!x/V2) [6]
Solving this equation for x yields:
x"c1c2/(c1/V2#c2/V1) [7]
Donnan dialysis is particularly effective for recov-
ery or removal of multivalent ions. The Donnan equi-
librium for a divalent Ca2# ion and a univalent K# is
described by the equation:
([Ca2#]1/[Ca2#]2)1/2"[K#]1/[K#]2 [8]
For maintenance of electroneutrality in the system,
the transport of x moles of Ca2# ions through the
membrane requires the transport of 2x moles of
K# ions in the opposite direction. Thus the equilib-
rium is described by:
[(c1!x/V1)/(x/V1)]1/2"(2x/V2)/(c2!2x/V2) [9]
For V1"10, V2"1 and initial concentrations of
c1"0.01 and c2"1, the value of x"0.095 is cal-
culated by eqn [9], which means that more than 95%
of the calcium would be driven through the mem-
brane. The effect of valence is even more dramatic
1703
when the c1/c2 ratio is much lower than the value used
in this example.
Comparison of Donnan Dialysis with
Conventional Ion Exchange
In many respects Donnan dialysis can be viewed as
a continuous ion-exchange process. For instance,
both processes can be used for the softening of water.
In conventional ion-exchange softening, the ion-
exchange resin beads are initially in the Na# form.
When hard water Sows through a column of the Na#
form resin beads, Ca2# ions in the water exchange
with Na# ions on the resin. When the supply of Na#
ions on the resin approaches exhaustion, the Sow of
hard water is stopped, the resins are regenerated by
passing NaCl brine through the column, and then the
cycle is repeated. In contrast, water softening with
Donnan dialysis is a continuous process. The Na#
ions for regeneration are always available in the solu-
tion on one side of the membrane, and the Ca2# ions
are continuously removed from the feed stream. Both
processes are effective for water softening. Donnan
dialysis works best in a situation where a continuous
Sow is expected. Softening with a bed of resin beads
can easily accommodate variable Sow rates or even
on/off operation. Since water softening is most often
done in situations where Sow rates Suctuate widely,
the use of resin beads rather than membranes is the
accepted practice.
If a particular exchange of ions can be accomp-
lished by the cyclic process with ion-exchange resin
beads or the continuous process with ion-exchange
membranes, what factors might inSuence the selec-
tion of the membrane process over the conventional
process? Because the membrane process is continu-
ous, it operates at steady-state conditions everywhere
in the dialyser. In cases where acid or base is used to
drive the process, conditions of pH extremes can be
avoided more easily with Donnan dialysis than with
conventional ion exchange. Compared to mem-
branes, ion-exchange beads offers a large surface area
in a small volume, so the rates of mass transfer are
faster with beads. But that large surface area
can be a problem if components of the solution are
subject to denaturation at surfaces where pH ex-
tremes exist. Donnan dialysis can be a much gentler
process.
A major limitation to Donnan dialysis is that, like
ion exchange, there is a need to add chemicals to
recover or remove chemicals. Essentially the stripper
solution in Donnan dialysis serves the same function
as the regenerant solution in ion exchange. Donnan
dialysis has the added limitation that osmotic water
transport makes the recovered electrolyte more dilute
1704 II / MEMBRANE SEPARATIONS / Donnan Dialysis
than it would be in the spent regenerant of an ion-
exchange process. However, Donnan dialysis has the
advantages that it is continuous and does not require
the rinse step after regeneration. If the application is
one in which the stripper solution is already available
and needs to have the transported ion in it, then the
Donnan dialysis can be an advantageous process.
Neutralization Dialysis
Anion- and cation-exchange membranes can be com-
bined into a single dialyser where both acid and base
drive the transport of the cation and anion of a salt as
illustrated in Figure 2. Since the hydrogen and hy-
droxyl ions that drive the transport are neutralized in
the feed solution, the driving force for desalting can
be maintained until almost all of the acid and base are
consumed. Neutralization dialysis has been used ef-
fectively for desalination of aqueous solutions of or-
ganic compounds. Research has demonstrated that
the Sux of salt ions increases with acid and base
concentration up to about 0.1 M. Above those levels
there was little beneRt in raising the acid and base
concentrations.
It should be noted from Figure 2 that the mem-
brane arrangement in neutralization dialysis is differ-
ent from that in electrodialysis. Here there are four
membranes in a repeating sequence, two anion-
exchange membranes and then two cation-exchange
Figure 2 Neutralization dialysis for desalination. C"cation;
A"anion.
membranes. It is a fortunate circumstance in neutral-
ization dialysis that the acid only contacts cation-
exchange membranes, because anion-exchange mem-
branes are notoriously leaky to acids. Likewise, the
two anion-exchange membranes that bound the base
compartments resist leakage of base.
Neutralization dialysis competes with deionization
with mixed bed ion-exchange resins which has lower
capital cost. But neutralization dialysis has the ad-
vantage of being a continuous process that can be
controlled by the Sow rates and concentrations of the
acid and base. Moreover, the feed solution is not
exposed to such severe pH extremes as it would be in
deionization with ion-exchange resins.
Membranes for Donnan Dialysis
Many ion-exchange membranes are manufactured
for electrodialysis, and most of them would be poten-
tially useful for Donnan dialysis. In addition, the
NaRon Suoropolymer cation-exchange membranes
made by DuPont are uniquely suited to Donnan di-
alysis because of their ability to withstand severe
thermal and chemical attack. Fluoropolymer mem-
branes have been used to recover chromic acid,
a strong oxidizing agent that attacks hydrocarbon-
based membranes. NaRon has been made as small
tubules and fabricated into shell-and-tube dialysers.
The preparation of hollow Rbres for use in analyti-
cal devices was reported in 1981. Low-density poly-
ethylene was extruded into 300
M i.d., 380
M o.d.
hollow Rbres that were sulfonated with 10% chloro-
sulfonic acid in methylene chloride. A bundle of eight
hollow Rbres was inserted into a coiled tube and
sealed with silicone rubber caulk.
Although many types of ion-exchange membranes
have been used for Donnan dialysis, there seem to be
no published studies comparing the performance of
commercial membranes for Donnan dialysis or estab-
lishing criteria for desirable membranes. One would
expect diffusional transport properties of the mem-
branes to parallel those of electrodialysis. In Donnan
dialysis, membranes with low electrical resistance
would be expected to have low resistance to ion diffu-
sion, and those with low electroosmotic water trans-
port would be expected to have low osmotic water
transport. Of course the membranes should be stable
at the anticipated operating temperatures of the pro-
cess and resist chemical attack by solution compo-
nents. Experimental studies have established that, un-
der typical hydrodynamic conditions, the major resist-
ance to transport is in the boundary layer of the dilute
solution rather than in the membrane; therefore, even
some of the thicker commericial ion-exchange mem-
branes could be candidates for use in Donnan dialysis.
 II / MEMBRANE SEPARATIONS / Donnan Dialysis
Japanese researchers have made cation exchange
membranes, speciRcally selective to the transport of
uranyl ions, by forming a copolymer of 2,3-epithio-
propyl methacrylate and 2-acrylamide-2-methyl-
propanesulfonic acid. They observed that the propor-
tions of the monomers had a great effect on the ability
of the membranes to transport uranyl ions. High
transport was achieved when the 2-acrylamide-2-
methylpropanesulfonic acid content was at least
34%, but very low transport occurred when that
monomer comprised less than 21% of the membrane.
By contrast, they found that the Selemion DLE mem-
brane did not transport the uranyl ions under the
same experimental conditions. The same researchers
also made anion-exchange membranes of 2,3-
epithiopropyl methacrylate}dodecyl methacrylate}
methylacrylamide propyltrimethylammonium chlor-
ide terpolymer that selectively transported ferric
oxalate complex anions from acidic ferric sulfate
solution to a receiving solution containing sodium
oxalate. These studies suggest that the development
of ion-speciRc membranes could lead to selective sep-
arations by Donnan dialysis that cannot be economi-
cally achieved by other methods.
The amount of membrane area required for a par-
ticular Donnan dialysis application depends on the
mass transfer rates that can be achieved. However,
the user should be cautious about relying upon mass
transfer data obtained in stirred cells to design indus-
trial systems, because the high shear rates obtained by
stirring might not be achieved in a commerical dialy-
ser. Results reported for Donnan dialysis of uranyl
ions in a countercurrent dialyzer should offer realistic
mass transfer coefRcients for design purposes. AM-
Fion C-103 cation-exchange membranes separated
solutions of 0.01 M UO2(NO3)2 and 2 M HNO3. The
solution compartments between the membranes were
formed with 0.38-mm-thick woven screens of stain-
less steel. Uranyl ion Suxes were about
9;10\10 mol cm\2 s\1. The treatment rate was
about 3 L h\1 m\2, and 96% of the UO2(NO3)2 was
recovered.
Apparatus
The equipment used for Donnan dialysis is similar to
that used for dialysis and diffusion dialysis. Mem-
brane shapes include sheets, tubes, and hollow Rbres.
Sheet membranes have been fabricated into spiral-
wound devices, but the dominant conRguration for
Sat membranes is the plate-and-frame arrangement
used in electrodialysis. Thus any supplier of elec-
trodialysis equipment is a potential supplier of stacks
for Donnan dialysis. Since the major resistance to
transport usually lies in the boundary layer of the
1705
dilute solution, equipment designers strive to achieve
high shear in the feed solution. There is a major
advantage to countercurrent Sow of the feed and
stripper solutions, because that conRguration leads to
maximum driving forces and maximum total transfer
of the desired ions.
In a commercial dialyser, the simultaneous achieve-
ment of both high shear and countercurrent Sow is
difRcult. High shear can be achieved by rapid stirring
or by high Sow rates, but the beneRts of countercur-
rent Sow can only be achieved if the residence time in
the device is sufRcient for a large fraction of the
desired ion to be transferred through the membrane.
As discussed below, studies of the effects of solution
velocity on mass transfer rates show correlation with
Vn where the exponent n is always less than unity.
This means that increasing the solution velocity
through a dialyser of a given length always results in
a smaller fraction of the solute being transferred
through the membrane. Therefore, an increase in
solution velocity requires an increase in path length to
achieve the same fraction of transfer. (This Vn rela-
tionship also applies to dialysis, diffusion dialysis,
and electrodialysis.)
Process Design and Control
Most often the objective of Donnan dialysis is to both
recover a target ionic species from a feed solution and
raise its concentration. An increase in the concentra-
tion is achieved by the use of a small volume of
a stripper solution with a diffusing ionic species that
is higher in concentration than that of the target ionic
species in the feed. A strip solution of about 1 M
might be used with a typical 0.001 M feed solution.
One might expect that a more concentrated strip
solution would recover the target ionic species at
a higher concentration, but stripper concentrations
higher than 1 M do not seem to be beneRcial. More
concentrated strip solutions lead to osmotic dehydra-
tion of the gel structure of the membrane which
reduces membrane permeability. Moreover, co-ion
transport due to reduced Donnan exclusion at the
higher concentrations allows loss of solutes from the
receiving solution across the membrane, and osmotic
transport of water through the membrane dilutes the
strip solution.
The effects of hydrodynamics on mass transfer
rates in Donnan dialysis have been studied by several
researchers who correlated the data using the stan-
dard equation Sh"kScmRen. The Sherwood number
Sh"KD/h, Schmidt number Sc"
/D, and
Reynolds number Re"hV/
are dimensionless
parameters where K is the mass transfer coefRcient,
D is the diffusivity of the diffusing ionic species, h is
1706 II / MEMBRANE SEPARATIONS / Donnan Dialysis

Table 1 Dimensionless parameters for correlation of mass
transfer rates in Donnan dialysis through Nafion
Cation Strip acid k m n
Cu2# H2SO4 0.48 0.33 0.28
Ni2# H2SO4 0.166 0.33 0.475
Na#, K# HNO3 0.201 0.4 0.62
the characteristic thickness or diameter of the con-
duit,
is the solution viscosity,  is speciRc gravity,
and V is solution velocity. The coefRcient k, and the
exponents m and n are used to correlate the data. The
results of three studies are shown in Table 1, and
some comments about the experiments follow.
Studies with CuSO4 and NiSO4 were done with the
feed Sowing inside a Suoropolymer membrane tube
a few millimetres in diameter and with H2SO4 stripper
Sowing outside the tube. The correlation parameters
apply to Re(1000. The following observations were
made in the studies with CuSO4 and NiSO4. Increasing
the stripper concentration from 1 to 5 M did not
improve mass transfer. Increasing the solution velocity
of the stripper solution produced minor increases in
mass transfer. The mass transfer rate was propor-
tional to the metal ion concentration in the feed for
concentrations below 2000 mg L\1 and Re(1000.
Studies of NaNO3 and KNO3 feeds with HNO3 strip-
ping were done with a Sat sheet Suoropolymer mem-
brane in a stirred cell with 15'Re'400.
Since the Sow rates in Donnan dialysis are normally
rather low, one must be concerned with convective
Sow in the dialyser caused by density changes in the
solutions as their compositions change. The use of
excessively high acid concentrations in the stripper
exacerbates the problem of changing densities. Osmo-
tic transport of water into the acid stripper solution at
the top of the solution compartment can cause sub-
stantial dilution of the acid, and the resulting density
decrease allows the denser incoming stripper to stream
downward through the diluted acid. The problem can
be alleviated by mild pulsation of the solution at the
top of the stripper side to improve mixing.
ery from rinse water in a Watts nickel electroplating
process with 1 M H2SO4 as the stripper. With typical
feed concentrations of up to 1 g L\1, Ni2# Suxes
exceeded 20 g h\1 m\2 and the recovered Ni concen-
tration exceeded 30 g L\1.
An early application of Donnan dialysis with poly-
meric ion-exchange membranes was described in
1967. Multiple AMFion C-103C cation-exchange
membranes were assembled in a plate-and-frame di-
alyser with solution compartments arranged so that
the feed and strip streams could Sow in a pattern that
was countercurrent overall. This apparatus was used
to recover UO2(NO3)2 from a 0.01 M feed to a Rnal
concentration of 0.28 M with 2 M HNO3 as the strip-
per and to 0.46 M with 2 M H2SO4 as the stripper.
He also used a Donnan dialyser with AMFion A-
104B anion-exchange membranes to remove acid
from the UO2(NO3)2 feed to improve the driving force
for cation transport. Complexing agents, EDTA and
DPTA, were used as strippers to separate Ag# and
Cu2# ions. The Cu2# ions formed such strong com-
plexes that the free Cu2# ion concentration in the
stripper was low enough to maintain a driving force
until virtually all of the Cu2# ions were transported
across the membrane.
The use of Donnan dialysis for water softening was
reported in 1970. The process is illustrated in
Figure 3. A brine of NaCl and hard feed water Sow
countercurrent. The diffusion of Na# ions from the

Applications
Although many industrial applications have been
demonstrated in the laboratory and at the pilot scale,
there appear to have been few industrial-scale ap-
plications of Donnan dialysis. Industrial applications
have been mainly in recovering heavy metals from
rinse waters of metal-Rnishing operations. One of the
few descriptions of an industrial installation reports
results of an experimental evaluation of a DuPont- Figure 3 Donnan dialysis removes Mg2# ions from hard water.
made NaRon hollow-Rbre dialyser for metal recov- C"cation.
 II / MEMBRANE SEPARATIONS / Electrodialysis
brine causes a driving force for transport of Mg2#
ions from the feed into the brine. Countercurrent
operation allows the large driving force to be main-
tained over the length of the membrane.
A hollow-Rbre device has been used as a suppressor
for ion chromatography. Compared to an ion-ex-
change column that is normally used to suppress
conductance in the eluant, Donnan dialysis allowed
more control over the conductivity and eliminated
the need to regenerate the resins. Moreover, the
resolution was improved by the use of Donnan
dialysis.
The use of neutralization dialysis for desalination
of cheese whey has been described.
Conclusion
Donnan dialysis can be used in many applications
where ion-exchange beads are currently applied. The
process might Rnd a niche in water softening in a ca-
pacity range between that of home water softeners
and the large-scale lime-soda softeners. Capital costs
are higher for membranes than for resin beads, but
the use of membranes offers the advantages of steady-
state operation without the need for rinse-down,
which produces large volumes of water that must be
discarded. The capital cost of small-scale Donnan
dialysis could be reduced by the availability of more
ion-exchange membranes in hollow-Rbre form that
could be assembled into compact modules.
Further Reading
Bleha M and Tishchenko GA (1992) Neutralization dialysis
for desalination. Journal of Membrane Science 73:
305}311.
Electrodialysis
H. Strathmann, University of Twente, The Netherlands
Copyright ^ 2000 Academic Press
Electrodialysis is a process in which ion exchange
membranes in combination with an electrical poten-
tial difference are used to remove ionic species
from an aqueous solution. The large scale industrial
utilization of the process began about 30 years ago
with the development of highly selective ion exchange
membranes of low electric resistance arranged in
a multicell stack.
Until the mid 1970s electrodialysis stacks were
operated in a unidirectional mode, that is, the polar-
1707
Davis TA, Wu JS and Baker BL (1971) Use of the Donnan
equilibrium principle to concentrate uranyl ions by an
ion-exchange membrane process. AIChE Journal 17(4):
1006}1008.
Donnan FG (1924) The theory of membrane equilibria.
Chemical Reviews 1(1): 73}90.
Grot WG (1986) Ion Exchange Process and Apparatus.
U.S. Patent 4,591,439.
Ng PK and Snyder DD (1981) Mass transport character-
istics of Donnan dialysis: the nickel sulfate system.
Journal of the Electrochemical Society 128(8):
1714}1719.
Ng PK and Snyder DD (1983) Uranyl nitrate and nitric
acid. Combined electrodialysis and dialysis for regenera-
tion of chromic acid etching solution. Journal of Mem-
brane Science 13: 327}336.
Nonaka T, Ogawa H, Morikawa M and Egawa H (1992)
Uphill and selective transport of uranyl ions through
2,3-epithiopropyl methacrylate-2-acrylamide-2-methyl
propanesulfonic acid copolymer membranes. Journal of
Applied Polymer Science 45: 285}292.
Nonaka T and Fujita K (1998) Transport of ferric ions
through 2,3-epithiopropyl methacrylate}dodecyl meth-
acrylate}methylacrylamide propyltrimethylammonium
chloride terpolymer membranes. Journal of Membrane
Science 144: 187}195.
Roach ET (1985) Evaluation of Donnan Dialysis for
Treatment of Nickel Plating Rinse Water. Final Report
to U.S. Environmental Protection Agency, NTIS PB85-
200046.
Stevens TS and Davis JC (1981) Hollow Rber ion-exchange
suppressor for ion chromatography. Analytical Chem-
istry 53: 1488}1492.
Wallace RM (1967) Concentration and separation of ions
by Donnan membrane equilibrium. Industrial and En-
gineering Chemistry, Process Design and Development
6(4): 423}431.
ity of the electrodes was Rxed. A signiRcant step
towards the efRcient application of electrodialy-
sis was the introduction of a new mode of operation
referred to as electrodialysis reversal. In this operat-
ing mode the Sow streams and the polarity in an
electrodialysis stack are periodically reversed, which
reduces membrane fouling and scaling. Costly and
time-consuming membrane cleaning procedures are
then unnecessary.
The main application of electrodialysis is the de-
salination of brackish water for domestic and indus-
trial use. In Japan electrodialysis is also used on
a large scale to concentrate sodium chloride from sea
water for the production of table salt. More recently
1708 II / MEMBRANE SEPARATIONS / Electrodialysis

exchange membranes, while those carrying positively

charged groups are referred to as anion exchange
membranes.
In a cation exchange membrane, the Rxed negative
charges are in electrical equilibrium with mobile ca-
tions in the interstices of the polymer. Figure 2 shows
a cation exchange membrane with Rxed anions and
mobile cations; the latter are referred to as counter
ions. The mobile anions, called co-ions, are more or
less completely excluded from the polymer matrix
because their electrical charge is identical to that of
Figure 1 Schematic diagram illustrating the principle of elec-
trodialysis.
the Rxed ions. Because of the exclusion of the co-ions,
cation exchange membranes are preferentially per-
meable for cations. Anion exchange membranes
utilization of electrodialysis in the food and chemical which carry positive Rxed charges and exclude ca-
industry and to treat certain industrial efSuent tions are preferentially permeable to anions. The ex-
streams has become important. tent to which co-ions are excluded from an ion
exchange membrane depends on the membrane as
well as on the solution properties.
The Principle of Electrodialysis The most desirable properties for ion exchange
The principle of electrodialysis is illustrated in membranes are:
Figure 1. A typical electrodialysis cell arrangement
E High permselectivity } the membrane should be
consists of a series of anion and cation exchange
permeable to counter-ions only
membranes arranged in an alternating pattern be-
E Low electrical resistance } the membrane should
tween an anode and a cathode to form individual
have high counter ion permeability
cells. If an electrolyte solution is passed through these
E Good mechanical and form stability } the
cells and an electrical potential is established between
membrane should be mechanically strong and
the electrodes, the positively charged cations migrate
should have a low degree of swelling in diluate
towards the cathode and the negatively charged an-
solutions
ions towards the anode. The positively charged
E High chemical stability } the membrane should be
cations can easily permeate the negatively charged
stable over the entire pH range and in the presence
cation exchange membrane but are retained by the
of oxidizing agents and organic solvents.
positively charged anion exchange membrane. Like-
wise, negatively charged anions permeate the anion The properties of ion exchange membranes are
exchange membrane but are retained by the cation determined by the base polymer and the type and
exchange membrane. The overall result is an increase concentration of the Rxed charges. The base polymer
in the ion concentration in alternate compartments, determines the mechanical, chemical and thermal
while the other compartments simultaneously
become depleted. The depleted solution is referred to
as diluate and the concentrated solution as brine or
concentrate. The driving force for the ion transport in
the electrodialysis process is the applied electrical
potential. The total space occupied by the diluate and
the concentrated solutions and the contiguous anion
and cation exchange membranes make up a cell pair.
The cell pair is the repeating unit in an electrodialysis
stack.

The Ion Exchange Membranes

Ion exchange membranes are the key components
in electrodialysis. They consist of highly swollen
gel-type polymer structures carrying Rxed positive
or negative charges. Polymer structures carrying Figure 2 Schematic drawing illustrating the structure of a
negatively charged groups are referred to as cation cation exchange membrane.
 II / MEMBRANE SEPARATIONS / Electrodialysis 1709

stability of the membrane. The type and concentra- The concentration of individual ions is related to
tion of the Rxed ions determine the permselectivity that of the salt by the stoichiometric coefRcient
and the electrical resistance. The moieties often  which determines into how many ions a salt will
used as Rxed charges are }SO\ 3 and }COO\ groups dissociate in the solution. Thus:
in cation exchange membranes and }R3N# and
}R2NH# groups in anion exchange membranes. The Ci"iCs [3]
sulfonic acid group }SO\ 3 is completely dissociated
over the entire pH range, while the carboxylic acid Another assumption in electrodialysis is that elec-
group }COO\ is virtually undissociated in the pH trical charges are transported exclusively by ions.
range (3. The quaternary ammonium group Thus:
}R3N# again is completely dissociated over the entire
pH range, while the tertiary ammonium group I" ziFJi [4]
}R2NH# is only weakly dissociated. Accordingly, i
ion exchange membranes are referred to as weakly or
strongly acidic or basic in character depending on the Here Ji is the Sux of the individual ions and I is the
charged groups they contain. total electrical current.

Transport and Transference Numbers

Mass Transfer in Electrodialysis The transport number Ti and the transference number
Mass transfer in electrolyte solutions is determined by ti of an ion i are given by:
the driving forces acting on the individual ions of the
solution and by the friction of the ions with other zi Ji Ti Ji
Ti" and ti" " [5]
components in the solution. The driving forces can be  zi Ji zi  zi Ji
i i
expressed by gradients in the electrochemical poten-
tial of individual components. The friction that has to
be overcome by the driving force can be expressed by Here Ti indicates the fraction of the total current
the ion mobility or diffusivity. carried by the ion i, and ti determines the number of
To describe the mass transport in a system, thermo- moles of the ion i transported per mole of electrons,
dynamic and kinetic parameters must be mathemat- i.e. per Faraday.
ically related. Several relations are described in the The transference number is directly related to the
literature. The one most frequently used is the ion concentration and their mobility and its sum is 1.
Nernst}Planck equation which describes transport of Thus:
ions under isobaric and isothermal conditions in an
ion exchange membrane as follows: Cimi
ti" and ti"0 [6]
 ziCimi i


i
d
i d
di
Ji"miCi "miCi ziF #
dz dz dz
The transference numbers of different salt ions


d
d ln ai
"miCi ziF #RT [1]
dz dz
For the deRnition of symbols in this and all other
equations, see Table 1. Ji is the Sux in the direc-
tion perpendicular to the membrane surface and z
refers to the number of charges carried by an ion
and indicates whether these charges are positive or
negative.
A boundary condition for describing the mass
transport in electrolyte solutions is the electroneutral-
ity requirement which postulates that on a macro-
scopic scale there is no excess in positive or negative
charges. Thus:
in solution are not very different. In an ion ex-
change membrane, however, there are the Rxed ions
of the membrane in addition to the mobile ions of the
electrolyte. The Rxed ions do not contribute to the
transport of electrical charges. Their transference
number is therefore 0. Furthermore, the concentra-
tion of the counterions is much lower than that of the
co-ions. Their concentration in the membrane deter-
mines the permselectivity of a membrane.
Membrane Permselectivity
The permselectivity of cation and anion exchange
membranes is deRned by:

c !tc
tmc a !ta
tma
ziCi"0 (i"1, 2, 32 n) [2] mc" and ma" [7]
i ta tc
1710 II / MEMBRANE SEPARATIONS / Electrodialysis

Table 1 Definition of symbols used in mathematical equations Table 1 Continued

Symbol Definition Symbol Definition

Electrochemical potential bd Diluate solution in the bulk

 Chemical potential diff Ion diffusion

Electrical potential l Current leakage through the manifold
 Stoichiometric coefficient which determines into m Membrane
how many ions a salt will dissociate in the solution ma Anion membrane
 Permselectivity of a membrane mc Cation membrane
 Current utilization md Diluate solution at the membrane surface
 Thickness of a cell mig Ion migration
 Resistance of ion exchange membranes ms Membrane selectivity

Potential difference between solutions separated by o Feed
a membrane assuming electrochemical equilibrium p Product solution
C Difference in salinity of feed and product water s Solution

Don Donnan potential sc Concentrate solution
G Energy required for production of the dilute sd Diluate concentration
m
Average activity coefficients of salt in the mem- se Electrode rinse solution
!
brane sf Feed solution
n s Amount of salt in moles transferred from a feed to sp Product solution
a concentrate solution w Water transport through the membranes
p Hydrodynamic pressure loss through the stack
Subscripts
s Equivalent conductivity
s a Anion
Average activity coefficients of salt in the electrolyte
! c Cation
solution
cou Counterion
z Thickness of the boundary layer
i Ion
a Activity
s Salt
A Effective area
w Water
A min Minimum membrane area required for a certain
plant capacity
C Concentration differences between feed and diluate
Cmco Co-ion concentration in the membrane
The permselectivity of an ion exchange membrane
Cm f Concentration fixed ions in the membrane
C sco Co-ion concentration in the electrolyte solution relates the transport of electric charges by counter-
D Diffusion coefficient ions to the total transport of electric charges through
dz Directional coordinate perpendicular to the mem- the membrane. An ideal permselective cation ex-
brane surface change membrane would transmit positively charged
E des Energy required for desalination
c "1,  "1. The permselectiv-
ions only, i.e. for tmc mc
Ep Pumping energy
F Faraday constant ity approaches zero when the transference number
I Total electrical current within the membrane is identical to that in the elec-
i Current density trolyte solution. For the anion exchange membrane
i lim Limiting current density the corresponding relationship holds.
Ji Flux of component i (individual ions)
k Constant referring to efficiency of pumps
Diffusion Potential, Donnan Equilibrium and
m Mobility of ions in the membrane
n Number of cell pairs in a stack Ion Exclusion
Q Volumetric flow of the product The diffusion potential can be derived by integ-
R Gas constant
R Electric resistance ration of eqn [1] and is given for a monovalent elec-
T Temperature trolyte when the ion activity is expressed by the salt
t Transference number concentration by:
t Time
Ti Transport number of component i (ions)


u Linear flow velocity of solution in electrodialysis RT ma!mc Cs

"
!
" ln [8]
cells F ma#mc Cs
U Voltage drop across the electrodialysis stack
V Volume
z Electrochemical valence An electrical potential difference is not only
Superscripts
established between two solutions of different
and Two phases separated by the membrane concentrations separated by a membrane but also
b Boundary layer between a membrane and the adjacent electrolyte
ba Concentrate bulk solution solution if the ion concentration in the membrane
bc Concentrate bulk solution is different from that in the adjacent solution,
 II / MEMBRANE SEPARATIONS / Electrodialysis
which is generally the case with ion exchange mem-
branes. This potential difference is referred to as
the Donnan potential.
The Donnan potential cannot be measured directly.
It can, however, be calculated from the electrochemi-
cal equilibrium of ions between the membrane and
the adjacent solution. By introducing the proper rela-
tions for the electrochemical potential, the Donnan
potential } the electrical potential difference be-
tween an ion exchange membrane and a solution of
a monovalent salt } can be calculated to a Rrst ap-
proximation by:
RT am
i

Don"
m!
s" ln s [9]
F ai
The numerical value of 
Don is negative for the
cation exchange membrane and positive for the anion
exchange membrane in equilibrium with a dilute
electrolyte solution.
The Donnan potential is also the basis for
calculating the Donnan exclusion, which determines
the co-ion concentration in a membrane. For a mono-
valent electrolyte, i.e. zi"1, and assuming a
high Rxed ion concentration in the membrane com-
pared to the electrolyte concentration, the co-ion
concentration in the membrane is given to a Rrst
approximation by:


s s 2
C
co"
Cm
co
m
!
m [10]
C f
! more detail later.
Energy Requirements in
Electrodialysis
The energy required in an electrodialysis process
is the sum of two terms: Rrstly, the electrical
energy needed to transfer the ionic components from
a feed solution through the membranes into the con-
centrate solution, and secondly, the energy required
to pump the solutions through the electrodialysis
unit. Energy consumption due to electrode reactions
can generally be neglected because of the large num-
ber of cell pairs usually stacked between the two
electrodes.
Minimum Energy Required for Transfer of Ions
from a Feed to a Concentrate Solution
In electrodialysis the minimum energy required for
the transport of salt from a feed to a concentrate
solution can be expressed by the Gibbs free energy of
mixing. Taking into account the electrolyte concen-
trations in the feed, diluate and concentrate, the min-
1711
imum desalting energy is given by:
CsfsCsfs
 
ln ln
CsdsCsc
s
G"zRTns sf ! [11]
Cs Csfs
!1
Csc
s s !1
Csd
Practical Energy Requirements in Electrodialysis
As discussed previously, the minimum energy re-
quired for desalting a given feed solution refers to
a reversible process. In electrodialysis there are also
irreversible energy losses and the total electric energy
required for the transfer of ions from a feed solution
to a concentrate, i.e. the actual energy used for desali-
nation is much larger than the theoretical minimum
value. This is given by:
Edes"UIt"RI2t [12]
The electric current required for the desalination
of a feed solution is directly proportional to the
concentration difference between the feed and
the diluate solution. It is given by:
V p za zc vsF(Csfs!Csp
s )
I" [13]
t
The current utilization is the fraction of the total
current passing through the electrodialysis stack that
is used for the transfer of ions. It will be discussed in
The electrical resistance of an electrodialysis stack
is determined by the resistances of the membrane and
diluate and concentrate solutions and is given to
a Rrst approximation by:

 
U n 2 1 1
R" " sd# sc # #
ma mc
[14]
I A s Cs Cs
The electrical resistance of the solutions is inversely
proportional to their salt concentrations, which are
changing while passing through the stack from the
feed to the product concentration. The concentration
in the diluate cell is decreasing and that in the concen-
trate cell increasing. An electrical resistance of a stack
can thus be calculated as a function of the cell thick-
ness, i.e. the distance between two membranes. Gen-
erally, the resistances of the ion exchange membranes
and the average resistance of the concentrate solution
are much lower than the average resistance of the
diluate and can therefore be neglected.
The electrical resistance of a stack can be cal-
culated to a Rrst approximation from the integral
1712 II / MEMBRANE SEPARATIONS / Electrodialysis

average of the diluate concentration and is given by: current through the stack manifold and water trans-
fer across the membranes due to osmosis and

 
Csfs electroosmosis.
 ln
n Csp
s The ratio of the actual current needed for salt
R" [15] transport from a feed to a concentrate stream to that
A s(Csf
s !C sp
s )
calculated theoretically is referred to as the current
The superscripts sf and sp refer to the feed and the efRciency, which for one cell pair is given by:
diluate at the cell outlet, which is the required prod-
uct.
"
ms
w
l [19]
A combination of eqns [13] and [15] gives the
energy required to remove a certain amount of salt The efRciency term
ms is a function of the
from a feed solution. For the desalination of a mono- membrane permselectivities.
w is caused by convec-
valent salt, i.e. where za, zc and vs are all unity, the tive Sow due to the hydrostatic pressure differ-
electrical energy is given by: ence between the diluate and concentrate cells, by
transfer of water in the hydration shell of ions and by
osmosis.
l is determined by parallel current through

 
Csfs
 ln the stack manifold.
n V F (C !C )
p2 2 sf
s
sp
s Csp
s
Edes" [16] The overall current utilization  can be deRned as
A  s
a function of the number of cell pairs, membrane
selectivity, water transfer and manifold current Sow.
For a given plant capacity, salt solution and cell It is given by:
design, the equivalent conductivity and the number
and area of cells are constant. Thus, the energy re- "n(mcta#matc) (1![tmc
w #tw ])
ma
quired for the desalination of a monovalent salt solu-
tion can be expressed to a Rrst approximation by the
;0.018(Csc
s !Cs )

sd l
[20]
constant factor k, by the current utilization and by the
feed and the product concentration.
For relatively dilute solutions, Cs(0.1 mol L\1,
k Csfs the efRciency loss due to water transfer is quite
Edes" (Csfs!Csp
s ) log [17] low. However, for higher feed solution salt con-
 Csp
s
centrations the water transfer may affect the
efRciency of electrodialysis quite signiRcantly.
Pumping Energy Requirements The current leakage through the manifold system
can, in a well-designed stack, be neglected, i.e.
l+1.
The operation of an electrodialysis unit requires one
or more pumps to circulate the diluate, the concen- Concentration Polarization and Limiting
trate and the electrode rinse solution through the Current Density
stack. The energy required for pumping these solu-
Concentration polarization occurs in all mass separ-
tions is determined by the volumes to be circulated
ation processes and is the result of changes in mass
and the pressure drop. It can be expressed by:
transport properties at an interface. In electrodialysis,
separation of ions is the result of differences in
Ep"k(Vpsd#Vscpsc#Vsepse) [18]
their transport numbers in solution and in the mem-
branes. At the surface of an ion exchange membrane
facing the diluate the concentration of counterions is
Processes Affecting the reduced and at the surface facing the concentrate the
Ef\ciency of Electrodialysis concentration of counterions is increased because of
the lower transport number of the counterions in the
In practical application electrodialysis is effected solution than that in the membrane. Because of
by concentration polarization and by incomplete cur- the electroneutrality requirement, the co-ions in the
rent utilization. Both phenomena inSuence the ef- boundary layers are transported in the opposite direc-
Rciency of the process. tion. Thus, salt concentration gradients are estab-
lished in the boundary layers at membrane surfaces,
Current Utilization
which leads to an additional mass transport towards
Current utilization in an electrodialysis stack is im- the membrane surface in the diluate and away from it
paired by incomplete membrane selectivity, parallel in the concentrate solution. Due to turbulent mixing

Figure 3 Schematic drawing illustrating concentration profiles of a salt in the boundary layer on both sides of an ion exchange
membrane and the fluxes of cations and anions in the boundary layer and the membrane surface. For abbreviations, see Table 1.
of the bulk solutions, the concentration gradients are
limited to a relatively thin laminar boundary layer at
the membrane surfaces, as indicated in Figure 3,
which shows the salt concentration proRles in the
solutions near the surface of an anion exchange
membrane.
The concentration proRles at the membrane surface
can be determined by a mass balance based on the
so-called Nernst Rlm model, which assumes static
boundary layers at the membrane surfaces, where con-
centration and electrical potential gradients perpen-
dicular to the membrane surfaces are the only driving
forces for the mass transport. The bulk solution be-
tween the laminar boundary layer is well mixed and
has a uniform concentration. It can be assumed that
the transport of ions through an ion exchange mem-
brane is the result of migration caused only by an
electrical potential gradient, while in the solution ions
are transported by both migration and diffusion.
In a steady state the ion Sux through the membrane is
identical to that through the boundary layer. For
a strictly permselective membrane it is given by:
m i i dCs
m mig
Jcou "!bJmig
cou # Jcou"tcou "tcou !Ds
b diff b
[21]
F F dz
The current density can be obtained from eqn [21]
by integration over the boundary layer. For the
boundary layer at the membrane surfaces adjacent to
the diluate the current density is:
FDs bd
Cs!mdCs
i"! m
(tcou!tcou) z
II / MEMBRANE SEPARATIONS / Electrodialysis 1713
When the hydrodynamic Sow conditions are kept
constant the boundary layer thickness, z, will be
constant and the current will reach a maximum value
independent of the electrical potential gradient if the
salt concentration at the membrane surface, mdCs,
becomes 0. This maximum current is referred to as
the limiting current density, ilim, which is given by:
FDs bd
Cs
ilim"! m [23]
(tcou!tcou) z
Exceeding the limiting current density in practical
applications of electrodialysis can affect the ef-
Rciency of the process severely by increasing the elec-
trical resistance of the solution and by causing water
splitting, which leads to changes in the pH values of
the solutions, causing precipitation of metal hydrox-
ides on the membrane surface.
Electrodialysis Process and
Equipment Design
The performance of electrodialysis in practical ap-
plications is not only a function of membrane proper-
ties but is also determined by the membrane stack and
the overall process design.
Electrodialysis Stack Design
Two different stack designs are used in large
scale applications today. One is the so-called sheet
Sow and the other is the tortuous path Sow design.
[22]
A typical sheet Sow electrodialysis stack is shown in
1714 II / MEMBRANE SEPARATIONS / Electrodialysis

Figure 4 Exploded view of an electrodialysis stack arrangement, indicating the individual cells and the sheet flow-type spacer, also
containing the manifolds for distribution of different flow streams.

Figure 4. The membranes are stacked between
electrodes in such a way that the Sow streams are
kept separate. The gaskets that separate the mem-
branes contain manifolds to distribute the process
Suids to the different compartments. The supply
ducts for the diluate and the concentrate are formed
by matching holes in the gaskets, in the membranes,
and in the electrode cells. To minimize the resis-
tance of the aqueous solution, the distance between
the membrane sheets is made as small as possible and
is normally between 0.5 and 2 mm in industrial
electrodialysis stacks. In an industrial electro-
dialysis system, 200}1000 cation and anion
exchange membranes are installed in parallel to form
an electrodialysis stack with 100}500 cell pairs.
Spacers between the individual membrane sheets sup-
port the membranes and provide mixing of the Sow
streams.
A proper electrodialysis stack design provides
the maximum effective membrane area per unit
stack volume and ensures uniform Sow distribution
through each compartment. The spacer screen should
provide a maximum of mixing of the solutions at the
membrane surfaces to reduce concentration polariza-
tion, but the pressure loss must be small.
Process Design and Modes of Operation
The efRciency of electrodialysis in a given ap-
plication depends greatly on the process, design
and mode of operation. Two different operating
modes are currently used: the Rrst is referred to as
unidirectional electrodialysis and the second as elec-
trodialysis reversal.
A Sow diagram of a typical unidirectionally oper-
ated electrodialysis plant is shown in Figure 5. Feed
solution pumped into the stack is converted to
a diluate and a concentrate which are collected in
storage tanks when the desired degree of concentra-
tion or depletion is achieved. To prevent the forma-
tion of free chlorine by anodic oxidation, the
electrode cells are rinsed with a separate solution that

Figure 5 Flow scheme of the unidirectional electrodialysis operating mode.

 II / MEMBRANE SEPARATIONS / Electrodialysis
Figure 6 Flow scheme of the electrodialysis reversal operating mode.
does not contain chloride ions. Unidirectionally oper-
ated electrodialysis plants are rather sensitive to
membrane fouling and scaling and often require care-
ful feed solution pretreatment and stack-cleaning
procedures.
Membrane fouling and scaling can be greatly re-
duced by operating in the electrodialysis reversal
mode. In this operating mode, the polarity of the
current is changed periodically every few minutes to
a few hours. Simultaneously, the hydraulic Sow
streams are reversed, as shown in Figure 6. The ad-
vantage of the electrodialysis reversal operating mode
is that precipitates that are formed the concentrate
cells are redissolved when the Sow is reversed and
these cells become the diluate cells. In the elec-
trodialysis reversal operating mode there is a brief
period when the concentration of the desalted prod-
uct does not meet the product quality speciRcation.
Thus, a certain amount of the product will be lost to
the waste stream.
Electrodialysis Process Costs
The economics of an electrodialysis process are usu-
ally expressed as cost per unit product. These costs
are the sum of Rxed charges associated with amorti-
zation of the investment and operating costs such as
energy and labour.
Investment costs include items such as the
electrodialysis stacks, pumps, electrical equip-
ment and membranes and are proportional to the
required membrane area. The minimum required
membrane area for a certain plant capacity is
1715
given by:
zFQCn
A" [24]
i
The required membrane area for a given capacity
electrodialysis plant is proportional to the amount of
ions removed from a given feed solution and inversely
proportional to the applied current density.
As indicated earlier, the applied current density
should not exceed a certain limiting value. According
to eqn [23] this value is proportional to the diluate
concentration and the mass transfer in the boundary
layers at the membrane surfaces. The mass transfer
depends on the boundary layer thickness, which is
a function of Sow velocity. For given stack and feed
solution properties, the limiting current density is
given by:
ilim"bdC a ub [25]

Here a and b are constants, the values of which are
determined by a series of parameters such as the cell
and spacer geometry, the solution viscosity and the
transference numbers of ions in the membrane and
the solution.
Eqn [25] shows that the limiting current density is
proportional to the diluate concentration. However,
the diluate concentration is changing during the de-
salting process from the concentration of the feed to
that of the product. Thus, the limiting current density
is decreasing along the Sow path through the stack.
The average limiting current density is proportional
to the average concentration in the diluate cell and
1716 II / MEMBRANE SEPARATIONS / Electrodialysis
Figure 7 Membrane, energy and total costs of the actual
desalination process in electrodialysis.
given by:
Cos!bdCps
bd
iM lim"bdCM s a ub"a ub bd o [26]
C
log bd p
Cs
Combining eqns [24] and [26] leads to:
bd o
Cs
log bd p
Cs
Amin"a bd o bd p [27]
Cs ! Cs
The constant a is determined by the feed Sow vel-
ocity, the stack design, etc.
The operating costs of an electrodialysis plant are
mainly determined by the energy consumption which
is the sum of the electrical energy required for the ion
transfer and the energy necessary for pumping the
solution through the stack, as indicated in eqns [17]
and [18]. The energy required for the desalting pro-
cess is a function of the feed solution concentration.
The pumping energy depends on the Sow velocities in
the stack.
It should be noted that, according to eqn [12],
the energy costs increase with increasing current
density, while the required membrane area decreases
with increasing current density. Thus the total desali-
nation cost } which is the sum of capital, energy
and operating costs } will reach a minimum at a cer-
tain current density, as illustrated in Figure 7, where
the total cost of an electrodialysis process is shown
schematically as a function of the applied current
density.
Applications of Electrodialysis
Electrodialysis is mainly used to desalinate saline
solutions such as brackish water. But other applica-
tions such as the treatment of industrial efSuents,
demineralization of whey and deacidiRcation of fruit
juices are becoming increasingly important. In Japan,
electrodialysis is also used in the production of table
salt from sea water.
Desalination of Brackish Water by Electrodialysis
In terms of installed plant capacity, the most impor-
tant application of electrodialysis is the production
of potable water from brackish water. Here, elec-
trodialysis competes directly with reverse osmosis
and multistage Sash evaporation. For water with
a relatively low salt concentration (less than
5000 p.p.m.), electrodialysis is generally considered
to be the cheapest process. Another signiRcant feature
of electrodialysis is that salts can be concentrated to
comparatively high values without affecting the
economics of the process severely.
Production of Table Salt
In the production of table salt from sea water,
electrodialysis is used to concentrate sodium
chloride up to 200 g L\1 prior to evaporation.
This application is developed and used nearly exclus-
ively in Japan. Key to the success of electrodialysis
in this application has been the development of mem-
branes with a preferred permeability of monovalent
ions.
Electrodialysis in Waste Water Treatment
Treatment of metal ion-contaminated rinse waters
produced in electroplating operations is an important
application of electrodialysis. Complete recycling of
the water and the metal ions can be achieved in
favourable cases. A disadvantage, however, is that in
electrodialysis only ions can be removed from a feed
stream. Uncharged components that are also present
in the rinse waters cannot be recovered.
Dump leach waters containing heavy metal ions
have also been successfully treated by electrodialysis.
The removal of nitrate from drinking water by elec-
trodialysis is an application that seems to be competitive
to processes such as ion exchange or reverse osmosis.
While in most of these applications the average
plant capacity is considerably lower than that in
brackish water desalination or table salt production,
there is also a signiRcant number of large plants
installed for the treatment of reRnery efSuents
and cooling-tower waste streams.
Concentration of Reverse Osmosis Brines
Often the disposal of large volumes of brine obtained
in reverse osmosis plants is difRcult and further
concentration is desirable. Because of the osmotic
pressure, the brine concentration cannot exceed a
certain value in reverse osmosis. A further

concentration, however, may be achieved at reason-
able cost by electrodialysis.
Electrodialysis in the Food and Chemical
Industry
Several applications of electrodialysis in the food in-
dustries, such as the demineralization of cheese whey,
have considerable economic signiRcance and are well
established today. Other applications, such as the
deashing of molasses or de-acidiRcation of fruit
juices, are still in an experimental stage. In the chem-
ical industry electrodialysis is used for the desalina-
tion of protein, dextran or sugar solutions. Here,
electrodialysis is often in competition with other sep-
aration procedures such as dialysis and solvent
extraction. The separation of organic acids is an ap-
plication of electrodialysis that is of interest to the
pharmaceutical industry.
Production of Ultra Pure Water
Electrodialysis is now being used for the production
of ultra pure water for the semiconductor industry.
By combining electrodialysis with mixed-bed ion ex-
change resins, deionized water is obtained without
a chemical regeneration of the ion exchange resin.
The process has been commercialized recently.
Conclusions
Electrodialysis has a long and proven history in
the desalination of brackish waters. However, new
applications in waste water treatment as well as in the
Filtration
R. Sahai, EuTech Scientific Services,
Morganville, NJ, USA
Copyright ^ 2000 Academic Press
Introduction
Filtration is a key processing operation in the pharma-
ceutical, chemical and cosmetic industries. For
example, Rltration may be necessary to clear process
solutions before analysis or as process step in manufac-
turing or in the sterilization of process solutions. Ana-
lytical testing requires only laboratory-scale Rltration
and is usually performed by a variety of membrane
types depending upon the application. Filtration in
manufacturing requires large-scale Rltering in engin-
eered devices called membrane modules or cartridges.
II / MEMBRANE SEPARATIONS / Filtration 1717
food and the chemical industry are becoming more
and more important. There are still a multitude of
problems to be solved. Some are related to the proper-
ties of the membranes and the process design, while
others are caused by the lack of application know-
how and practical experience.
Further Reading
Bergsma F and Kruissink ChA (1961) Ion-exchange mem-
branes. Fortschrifte in der Hochpolymer-Forschung 21:
307}362.
Helfferich F (1962) Ion-Exchange. London: McGraw-
Hill.
Katz WE (1979) The electrodialysis reversal (EDR) process.
Desalination 28: 31}40.
Lacey RE (1972) Basis of electromembrane processes. In:
Lacey RE and Loeb S (eds) Industrial Processing with
Membranes. New York: John Wiley.
Nishiwaki T (1972) Concentration of electrolytes prior to
evaporation with an electromembrane process. In: Lacey
RE and Loeb S (eds) Industrial Processing with Mem-
branes. New York: John Wiley.
Schaffer LH and Mintz MS (1966) Electrodialysis. In:
Spiegler KS (ed.) Principles of Desalination, pp. 3}20.
New York: Academic Press.
Spiegler KS (1956) Electrochemical operations. In: Nachod
FC and Schubert J (eds) Ion Exchange Technology,
pp. 118}181. New York: Academic Press.
Strathmann H (1995) Electrodialysis and related processes.
In: Nobel RD and Stern SA (eds) Membrane Separation
Technology, pp. 213}281. Amsterdam: Elsevier.
Wilson JR (1960) Design and operation of electrodialysis
plants. In: Wilson JR (ed.) Demineralization by Elec-
trodialysis. London: Butterworth.
Filtration Mechanism
Filtration is a mechanical phenomenon, which is
sometimes aided by chemical manipulations of the
Rltration medium to make it more efRcient. In
any case, a driving force across the Rlter media is
required. The following methods can be used to gen-
erate this driving force:
E Vacuum
E Pressure difference
E Centrifugal force
E Gravity pull
E Concentration difference
E Electrical potential difference
E Temperature difference
E A speciRc chemical attraction}repulsion
1718 II / MEMBRANE SEPARATIONS / Filtration
Filtration is either through a membrane or bed of
Rlter media. The chemical composition of the Rlter
media and physical conditions to perform the Rltra-
tion constitute a large number of Rltration choices
available today.
Membrane Filtration
Membrane Rltration through a very thin Rlter
medium is also known as ‘surface Rltration’.
The solid particles to be separated are usually large
compared to the pore size characteristic of
the membrane. The pores on the surface are of
irregular shapes. The rejection of particles is
dependent on several factors affecting the trans-
port through these pores into the tortuous channels.
The separation is based on exclusion discrimination
by physical size, charge or afRnity or a combina-
tion of these properties. Large particles are rejected
on the surface and do not accumulate on the surface
and do not get a chance to enter into the interior of
the Rlter.
Other types of membrane Rlters are screen Rlters
and here the pores do not lead into tortuous capillary
paths. The pore size is uniform but the distribution of
the pores is random on the Rlter surface. The Rlter is
made by bombarding a thin polycarbonate Rlm with
neutrons in a reactor. The Rlm is then placed in a bath
of etching solution which preferentially attacks the
polymer along the track of the neutrons. The pore size
is regulated by selecting the appropriate reagent, ex-
posure time and temperature.
Membrane Rltration can be dead end or cross-Sow.
In dead-end Rltration all the solution is forced
through the membrane. Retained particles collect on
the membrane surface and in the Rlter greatly reduc-
ing Sow. A current application of dead-end Rltration
is in bacterial testing where the liquid to be tested is
passed through the Rlter retaining all bacteria on the
surface. Most chromatographic Rltration applica-
tions are of this type. In cross-Sow membrane Rltra-
tion, the feed liquid Sows tangentially to the mem-
brane surface, which prevents the build up of cake on
the membrane. Both types of Rltration use similar
membranes.
By convention, membrane Rltration or microRltra-
tion is limited to membranes used to remove particles
larger than 0.1
m in diameter. Membranes able to
remove smaller particles are called ‘ultraRltration
membranes’ and microsolutes can be removed by
reverse osmosis. UltraRltration and reverse osmosis
are discussed elsewhere. This article is limited to the
process of microRltration.
The Rltration thresholds of common membrane-
Rltration processes are shown in Table 1.
Table 1 Filtration threshold of common membrane-filtration
processes
Type of filtration Impermeability of membrane
Reverse osmosis (0.001
m
Ultrafiltration 0.001}0.1
m
Microfiltration 0.1}10
m
Micro\ltration
MicroRltration is used to separate suspended solids or
colloidal particles between 0.1 and 10
m in diameter
from solution. Most of the chromatography applica-
tions are microRltration based. The same type of
membrane with different pore size is used for
these applications. The membrane acts like a physical
sieve. The Suid passes through tortuous channels
while the particles are rejected on the surface of the
Rlter. It can be easily understood as a mechanical
sieve with pores leading into a capillary forming
a tortuous path; within this tortuous path, there could
be mechanical entrapment and adsorption (Figure 1).
MicroRltration membranes can be subjected to har-
sher conditions compared to ultraRltration mem-
branes. Membranes of different polymers in
varying pore sizes are available. Even nominally the
same pore-size membranes of a polymer may dif-
fer from each other in Rltration characteristics
because they may have different pore-size distri-
butions, i.e. varying pore size all across the mem-
brane. To aid in wetting, many membranes have
some surfactant pretreatment and their effective
pore size may be different from the real pore size.
Often a membrane Rlter becomes more efRcient
as small particles are entrapped within the pores. The
large particles captured on the Rlter can also alter the
effective particle-size rejection in subsequent Rl-
tration. Filter capacity may vary depending on the
solute particle-size variation in the feed. Uniform size
particles result in faster clogging of Rlters.
Depth Filter
In depth Rltration, the Rlter medium has larger pores
than the particles it is meant to remove. The process
starts out at the surface of the Rlter and proceeds
in the cake portion of the membrane. The
medium traps the particles in the interstices of the
internal structure. Particles enter into the Rlter me-
dium and separate by gravity settling, diffusion,
and attachment to the media owing to electrostatic
forces. These Rlters usually have a pressure drop
across the Rlter caused by pressure, vacuum, or cen-
trifugation. These Rlters usually have a long life, but

Figure 1 Tortuous path of micro- and ultrafiltration.
eventually a cake is formed over the medium stopping
the Sow through the Rltration device. When the Rlter
bed is full of solids and the pressure drop is very high,
the entrapped solids can be back-washed. Usually less
than 0.1% solids concentration is Rltered through
this type of Rlter to avoid pressure build up. There is
always some liquid left behind in depth Rlters and
some solid material still makes it through the Rlter
medium depending upon the efRciency of the
system. To use the system effectively, variations
of this procedure using moving-bed Rlters, radial Sow
Rlters, or travelling back-wash Rlters can be em-
ployed. The commercial products available for this
kind of Rltration are application speciRc.
Filtration Matrices
A variety of polymers are used to manufacture Rlter
media. Each type has speciRc attributes and could be
best for certain applications but could be a complete
failure for other applications. The same Rlter material
from different manufacturers can differ in
physical properties and in Rltration characteristics.
However, the chemical compatibility of the material
is almost the same irrespective of the manufacturer
(Table 2).
Incompatible chemicals can cause shedding, af-
fecting pore size, as well as adding extractables. Hy-
drophobic membranes have to be wet before starting
the Rltration. In some cases, it is desirable to convert
the hydrophobic membranes into hydrophilic mem-
branes by modifying the surface. Surface reactions
can also be used for changing the surface charge. The
II / MEMBRANE SEPARATIONS / Filtration 1719
polyvinylidene Suoride (PVDF) membranes can be
treated to render them hydrophilic and can be reacted
to modify the surface charge. The use of aggressive
solvents should be avoided with these Rlters to min-
imize deterioration of the surface. Membranes used in
Rltration usually have surface-active agents incorpor-
ated in them to make the pre-wetting easier.
Filtration Devices
The Rltration application dictates the type of device to
be used. Industrial applications demand a high surface
area, ease of cleaning and low clogging. For enhanced
yield and capacity, open-channel tangential Sow sys-
tems, which require two pumps for recirculation and
permeation, are available from several manufacturers.
Different designs are used to overcome gel forma-
tion and to continuously sweep away the contamina-
tion in the Rltration process. The details of these sys-
tems are beyond the scope of this article.
The most common disposable Rltration devices
used in laboratories are syringe Rlters. The membrane
is held in a polypropylene housing with an adapter for
syringe attachment on one end. This adapter can be
a luer lock or friction Rtting. The other end is de-
signed for easy extrusion of permeates. The Rlter
membrane can be housed alone or with pre-Rlters.
The membrane is bonded to the housing ultrasoni-
cally, without the use of any chemical adhesive, to
avoid unwanted extractables in the Rltering process.
Syringe Rlters are disposable and used in very
high volume, particularly in the pharmaceutical
industry. Because the continuous use of syringe Rlters
1720 II / MEMBRANE SEPARATIONS / Filtration

Table 2 Chemical compatibility of common filter membranes with widely used solventsa

Chemical Nylon PTFE PVDF PS Polypropylene Regen. Cellulose Cellulose Cellulose

cellulose nitrate acetate triacetate

Acids
Glacial acetic acid LC C C C C C NC NC NC
Hydrochloric acid NC C C C C C NC NC NC
Sulfuric acid NC C NC NC C NC NC NC NC
Nitric acid NC C C NC C NC NC NC NC
Phosphoric acid (25%) NC C C LC LC C C
Bases
Ammonium hydroxide (25%) C C LC C C LC C C C
Sodium hydroxide 3 mol L\1 C C C C C LC NC NC NC
Common solvents
Acetone C C NC NC LC C NC NC NC
Benzene C C C NC C C C C C
Benzyl alcohol C C C ND C C LC LC LC
Butanol C C C C C C C C C
Carbon tetrachloride C C C NC LC C C LC LC
Chloroform C LC C NC LC C C NC NC
Dichloromethane C C C NC LC C C NC NC
Dimethylformamide LC C NC NC C LC NC NC NC
DMSO C C NC NC C C NC NC NC
Ethanol/methanol C C C C C C LC C C
Ethyl acetate C C C NC LC C NC NC NC
Ethyl ether C C LC NC C C NC NC NC
Glycerol C C C C C C C C C
Hexane C C C NC C C C C C
Isopropanol C C C C C C LC C C
Methyl ethyl ketone C C LC NC ND C LC LC LC
Tetrahydrofuran C C LC NC C C NC NC NC
Application MF MF MF, UF MF, UF, RO MF MF, UF MF MF, UF, RO MF, UF, RO

a
PTFE, polytetrafluoroethylene; PVDF, polyvinylidene fluoride; PS, polysulfone. C, compatible; LC, limited compatibility; NC, non-compatible;
ND, not done; MF, microfiltration; UF, ultrafiltration; RO, reversed osmosis.

can be tiring, a mechanical device is now available
which is helpful when repeated Rltration is
required.
Another common type of laboratory Rltration is
with centrifuge Rlters. These devices are the method
of choice for molecular weight cut-off Rltration
and for the Rltration of viscous materials. The driving
force here is centrifugal force. The Rlter is manufac-
tured to Rt in the rotors of laboratory centrifuges. In
these rotors, several Rltrations can be carried out
simultaneously.
Filtration Applications
In every type of Rltration process, the result is always
a retentate (restricted to pass through the Rlter media)
and permeate (down stream collection). Retentate or
permeate can be the desired product of the process.
Selective Filtration
Selective Rltration is used to retain only a particular
type of solute. Usually in these cases membranes are
modiRed for the desired afRnity. There are several
applications and products available based on ionic
attraction.
Purifying Water
The constantly increasing demand for drinking water
requires the sea or other sources to be converted into
potable water. Most of the potable water plants use
reverse-osmosis treatment. Water used in injectables,
buffers and chromatography generally has very
deRned speciRcations. For most laboratory applica-
tions, water with an electrical resistance of not less
than 18 M
is required to be pyrogen- and bacteria-
free. The water used in chromatography should be
free of UV/vis-absorbing and ionic impurities.
Chromatographic Applications
Filtration is required in chromatography for prepar-
ing a sample for injection. The preparation may
include concentration and/or puriRcation. Sample Rl-
tration helps in trouble-free operation of chromato-
graphy instruments and columns. The use of Rltration

for processing samples and solvent is an essential part
of instrument preventive maintenance programmes.
The Rltration of the mobile phase also results in
degassing, which is essential for long pump life in
high pressure liquid chromatography. The most com-
mon devices used for sample preparations are syringe
Rlters.
Biological Applications
The use of Rltration as a sterilizing technique is be-
coming increasingly popular. Other sterilization tech-
niques such as autoclaving, radioactive exposure or
ethylene oxide treatment can be detrimental for the
product. A dead-end Rltration using 0.22
m pore-
sized membrane is considered good for sterilizing by
Rltration. Viruses can permeate through the mem-
brane of 0.22
m Rlter. A 0.1
m pore-size Rlter is
used to prepare a virus-free solution.
Filtration is also used for desalting or buffer
exchange of proteins and nucleic acids, deproteiniz-
ing samples, screening natural products and combina-
torial products, and separation of oligonucleotide
primers from nucleic acid preparations.
Selecting the Right Filtration System
Each application requires a speciRc Rltration charac-
teristic. Choosing the right Rltration device and media
are necessary when selecting the correct Rltration
system. The following considerations help when de-
ciding which Rlter device is to be used:
E Objective of Rltration.
E Sample size.
E Filter parameters required: permeability, capacity
and Sow rate.
E Physical conditions to which the Rltration is re-
quired to be subjected.
E Tangential Sow Rlter or dead-end Rlter.
The choice of Rlter material, pore size and physical
conditions depends on the following factors:
E The chemical and physical condition of the feed.
E Size and shape of molecules.
E Zeta potential and isoelectric point. Filtration car-
ried out at a pH close to the isoelectric point results
in reduced electrostatic interactions.
E Hydrophobicity or hydrophillicity.
E Solvent in which solute is dissolved.
E Properties of the Rlter feed, pH, viscosity, surface
tension, ionic strength, osmolarity and chemical
functionality.
E Intended use after Rltration of the sample
II / MEMBRANE SEPARATIONS / Filtration 1721
Choice is always application speciRc. For example,
in the bacterial examination of water, the purpose is
to retain all the particles on the Rlter surface. A dead-
end Rltration is used on a 0.2
m Rlter. The cross-Sow
Rlter is useful for concentrating particles with the
removal of solvent. In selecting the Rltration system,
it is necessary to always consider yield, simplicity,
technical reasons and cost.
In some applications, using a combination of dif-
ferent Rltration techniques in a certain order is the
most efRcient method. Sometimes it helps to pre-
treat the solution to be Rltered. The pretreatment
could include coagulation and Socculation, magnetic
treatment, pH adjustments, and an electric Reld.
A proper washing procedure is usually employed to
have the most efRcient Rltration.
Two Rlters supplied by Pall Corporation are shown
in Figures 2 and 3.
Filtration Validation
The validation of Rltration processes includes all
the equipment, physical conditions and material re-
quirements of the process. Usually the Rlter manufac-
turer performs the basic testing to ensure the type,
pore size and integrity of the Rlter. The Rlter material
characteristics are covered in this article. Some of the
most common tests used for this purpose are shown
below.
Bubble Point
Bubble point is a function of pore size, Rlter medium
wettability, surface tension and angle of contact. The
Rlter membrane is wetted and a gradual increasing
gas pressure is applied. The bubble starts forming
from the largest pore Rrst. The gas pressure at this
time is the bubble point for the membrane. This is an
indirect measurement of the size of the largest pore on
the Rlter. It does not indicate the variability of pore
sizes or irregularity of the membrane.
Water Breakthrough
The water-breakthrough test is used for hydrophobic
membranes. It is similar to bubble point as this test
also give information about the largest pore of the
Rlter membrane. In this test, the minimum pressure
required to permeate water from a Rlter membrane is
measured. The water-breakthrough number is depen-
dent on pre-wetting, temperature and pore size of the
Rlter medium. Water is Rrst permeated from the lar-
gest pore. This also ascertains Rlter usability as an
aqueous barrier.
1722 II / MEMBRANE SEPARATIONS / Filtration
Figure 2 (See Colour Plate 49). Pall Ultipor威 VF2+ Grade DV50 virus filters for high protein-transmissive virus filtration. (Photo
courtesy of Pall Corporation, East Hills, NY.)
Extractables
The Rlter devices and materials can be a source of
contamination in the Rltration process. The source of
impurities could be additives, stabilizers, surface
modiRers, detergents and monomers in the Rlter ma-
terial. Some contaminants occur in small quantities
but some detergents can make up as much as 2}3% of
the dry weight of the Rlter. This large amount of
detergent helps in efRcient Rltration, lower pres-
sure requirements and permits autoclaving for steril-
ization. The additives and monomers can be
entrapped within the body of the Rlter. Sometimes the
source of impurities is not from the Rltering material
but from the housing or support of the Rlter. This
housing material is usually plastic, and the manufac-
turer tests that the plastic used in containing the Rlter
material is not going to leach out impurities under
experimental conditions. Although aggressive sol-
vents or physical conditions may be very compatible
with the Rlter membrane, they may affect the
Rlter-containment system.
In some analyses, even a small amount of con-
taminant is enough to cause problems. Commercial
Rlter manufacturers now certify for speciRc applica-
tions. For many biological applications, the manufac-
turer certiRes the Rltration material to be pyrogen-free.
For chromatographic applications the Rlter material is
certiRed not to add impurities to the process. The safest
way to use Rlters for chromatography is to wash them
with the same solvent used during Rltration and to
discard an initial volume of the Rltrate. The Rlters used
in ion analysis should be completely free of any ionic
impurities. The standard operating procedure of the
Rltration step should clearly deRne the conditions and
if possible include the limits of the procedure.

Figure 3 (See Colour Plate 50). Pall Ultipleat威 high flow filters,
providing efficient and economical high-flow filtration with reduced
waste disposal costs. (Photo courtesy of Pall Corporation, East
Hills, NY.)
Flow Rate
The Sow rate is determined by using water or alcohol
to determine the permeability to Sow before any
extra pressure drop produced by the Rlter cake. Flow
rate is dependent on the hydrophobicity of the Rlter
material, temperature of the procedure, physical
thickness and pore-size distribution of the Rlter ma-
terial. It is expressed as millilitres per minute per
square centimeter. An optimum Sow rate is needed
for the expected life of a Rlter.
Capacity
Capacity of Rltration is the ability to maintain an
acceptable permeability. The capacity of a Rlter is
measured until an increase of about three times in
differential pressure or &60% decrease of in-
itial Sow. It is expressed as time, volume of liquid, or
by quantity of retained particles.
Pore Size
Pore size is probably the most misunderstood prop-
erty of the Rlter membrane. The estimation of pore
size depends upon the method employed to determine
the porosity. The usual methods are all indirect. For
II / MEMBRANE SEPARATIONS / Filtration 1723
a nominal rating, a range of neutral polymers of
different sizes is challenged individually on the
membrane and the percentage of a particular size
retained on the surface rates it for that size. It could
be anywhere from 60 to 98% for a given size rating
by the manufacturer. The variability of pore sizes is
also polymer dependent. The pore sizes are irregular
in membranes manufactured by solvent casting.
The pore size is averaged to give a mean pore size
assuming all pores are circular. The importance of
this point is that the efRciency of the Rlter should
be measured above this point. In actual practice,
pore size is used only as a guide; retained particle size
data are closer to reality in the Rltration process.
Most Rlter manufacturers give particle size retained
data traceable to standards from the National Insti-
tute of Standards and Testing. In membranes manu-
factured by neutron bombardment, the pores are cir-
cular and same-size pores are randomly distributed
along the surface of the membrane. The pore size
given is the actual pore size of the membrane. In
many Rlter membranes, detergents are used for en-
hancing Rlter characteristics; the effective pore
size in these membranes is usually larger than the
actual pore size.
Microbial Challenge Test
The absolute rating of a membrane is determined by
challenging with test organisms (Table 3). The vol-
ume of the feed is such that it averages out to one
organism per pore on the membrane surface. The
absolute rated membrane is accepted if no more than
one organism is present in the permeate.
A membrane with a pre-rating of 0.22
m is ac-
ceptable for liquid Rltration sterilization. The ability
of a membrane to remove bacteria is dependent on
the size of the pores and the thickness of the mem-
brane. There is a Rnite number of speciRc bacteria,
which can be retained by the membrane before it
becomes effectively clogged.
Filtration Challenges
Despite the fact that a great deal of improvement in
the Rltration process and material has taken place,
Table 3 Microfiltration rating by test organisms
Microfiltration rating Test organism
1
m Candida albicans
0.8
m Lactobacillus
0.45
m Serratia marcescens
0.2
m Pseudomonas diminuta
0.1
m Acholeplasma laidlawii
1724 II / MEMBRANE SEPARATIONS / Filtration
there are still some areas where any advancement will
make Rltration a friendlier process.
Scaling Up for Manufacturing
The Rltration process development remains a chal-
lenge because the efRcient separation at small
volume level is not always transferable to pilot or
production scale with the same efRciency and
chemistry. Several manufacturers claim new scalable
technologies providing similar results in large scale as
applications using tangential Sow with the same Rbre
material used throughout the development of the Rl-
tration process. Special Rltration scale-up software is
available commercially.
Membrane Fouling, Gel and Cake Formation
The Rltration membranes may start fouling during
use. This means that particles start attaching on the
surface and in the internal porous structure of the
membrane. Large suspended or colloidal particles
usually are the cause of fouling. Fouling is a result of
van der Waals forces, electrostatic attraction, or hy-
drogen bonding. The fouling of Rlter media results in
a reduction in membrane permeability and uncontrol-
led solute removal efRciency. The pretreatment of
the feed can be helpful in delaying or completely
avoiding fouling. Gel formation and cake formation
on the surface can be reversible and Rlter media can
be reused. Macromolecules and some interacting
small organic molecules can result in gel formation on
the Rltration surface.
Cleaning the Filter Media
It is not cost effective to clean the Rlter in labor-
atory-scale Rltration. For large-scale Rltration, usu-
ally cleaning and validation protocols are used. The
cleaning process could involve cleaning with deter-
gents or other strong chemicals. It could also involve
treating with proteolytic enzymes to break down pro-
tein impurities trapped in the Rlter medium and
EDTA to arrest activity of bacterial enzymes. Devel-
opment of cleaning procedures and validation of Rlter
media is very application speciRc and requires experi-
enced people to design and implement.
Extractables
The extractables in the Rlter medium can create
a problem in the subsequent use of the permeate. This
remains a problem in some Rlter media where addi-
tives are used for improved performance. The origin
of extractables is either in the processing or the hous-
ing device of the Rlter. Various kinds of extract-
ables are found, including metals, oligomers, loose
polymers, plasticizers, wetting agents, antioxidants,
resins, Rllers and mould-release chemicals. The usual
practice is to wash off the Rlter material immedi-
ately before use. The type and amount of impurity in
Rlter media is not consistent. Each type of impurity
has its own rate of extraction from the medium.
Hence there is no universal Rlter-treatment procedure
which can ensure a contamination-free permeate. The
washing procedure could be under- or overdone in
certain applications. The challenge exists to manufac-
ture consistent contamination-free Rlter media.
Conclusion
Tremendous developments have taken place in both
laboratory and large scale Rltration techniques in
recent years. Various new types of matrices have been
exploited for Rltration applications. The heavy use of
Rltration in industry has clearly identiRed the chal-
lenges that remain to be solved. Research continues
on selective Rltration as a cost-effective way of
separation for various applications. In the next few
years, we will witness improvement in both the chem-
ical and mechanical properties of Rltration equip-
ment.
See Colour Plates 49, 50.
See also: II/Membrane Separations: Microfiltration;
Ultrafiltration.
Further Reading
Chenoweth MB (ed.) (1986) Synthetic Membranes. New
York: Harwood.
Cheryan M (1986) UltraTltration Handbook. Lancaster:
Technomic.
Cooper AR (ed.) (1980) UltraTltration Membranes and
Applications. New York: Plenum Press.
Crespo JG and Boddekar KW (ed.) (1994) Membrane Pro-
cess in Separation and PuriTcation. Dordrecht: Kluwer.
Gutman RG (1987) Membrane Filtration. Bristol: IOP Pub-
lishing.
Levy RV and Leahy TJ (1991) Disinfection, Sterilization
and Preservation. Philadelphia: Lea and Febiger.
Lombardi R (1998) Membrane Filtration in Chromatogra-
phy } A Trivial Pursuit. LC}GC supplement, S47.
Matteson MJ and Orr C (ed.) (1987) Filtration: Principles
and Practices. New York: Marcel Dekker.
Murkes J and Carlsson CG (1988) Cross Filtration. New
York: John Wiley.
Nachinkin OI (1991) Polymeric MicroTlters. New York:
Harwood.
Osada Y and Nakagawa T (ed.) (1992) Membrane Science
and Technology. New York: Marcel Dekker.
Shoemaker W (ed.) (1977) What the Filterman Needs to
Know about Filtration, p. 171. New York: American
Institute of Chemical Engineers.
 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
Gas Separations with Polymer Membranes
D. V. Laciak and M. Langsam, Corporate
Science and Technology Center, Air Products and
Chemicals Inc., Allentown, PA, USA
Copyright ^ 2000 Academic Press
Introduction
In 1996 the worldwide industrial gas market was in
excess of $29 billion (US). It continues to grow at an
average rate of 4}5% per annum. Industrial gases
account for some of the largest production volume
chemicals (1998 US): nitrogen (843 bcf (billion cubic
feet)), oxygen (698 bcf) and ammonia (19 700 mil-
lion tons). Oxygen and nitrogen are separated from
puriRed air. Ammonia is produced by the reaction of
nitrogen and hydrogen. Certainly, the vast majority
of industrial gases are puriRed using cryogenic distil-
lation or adsorption technology. However, in the last
20 years there has been a growing interest in and an
intense effort on the part of major gas producers to
evaluate and develop membrane technology to pro-
duce or purify gases. By 1999 sales of gas separation
membrane technology exceeded $100 million per
year. This article will describe basic concepts along
with various practical aspects of polymeric gas separ-
ation membranes including permeability measure-
ment, membrane formation, module fabrication and
applications.
A polymeric membrane is deRned as a thin,
semipermeable barrier between two gaseous phases.
Gases will permeate the membrane if a difference in
their chemical potential exists between the two gas-
eous phases. The chemical potential difference is
most often a result of pressure differences across the
membrane. Thus, gases will solubilize into the mem-
brane at the high pressure interface, diffuse across the
membrane in a concentration gradient to the low
pressure interface and evolve into the low pressure
gas phase (Figure 1). If a mixture of gases comprised
of components i and j is brought into contact with the
membrane, the permeate stream will be enriched in
the more permeable gas i, leaving the retentate en-
riched in gas j.
The realization that gases permeate through poly-
mers is not new. Every child knows that a balloon
Rlled with air or helium deSates over time. Indeed,
this phenomenon was observed by Mitchell in 1831.
Balloons made of natural rubber Rlled at different
rates depending on the gaseous atmosphere they were
placed into. Carbon dioxide Rlled the balloon fastest,
air slowest. Thirty-Rve years later Graham expanded
1725
on Mitchell’s experiments and quantitatively mea-
sured the permeation rates of gases through natural
rubber. He found that the permeation rate was not
related to the known gaseous diffusion coefRcients
and so concluded that permeation does not proceed
through microscopic pores in the rubber but must
occur within the rubber itself. He also demonstrated
that natural rubber could be used to produce from air
a permeate which was enriched in oxygen to 46%.
A mathematical description of the permeation pro-
cess was proposed by Fick. The relationship between
permeation rate J, gas pressure P, membrane area
A and membrane thickness l, known as Fick’s Rrst
law, is governed by eqn [1], where
P is the pressure
difference across the membrane:
J"Po ) A )
P/l [1]
The proportionality constant, Po , is termed the per-
meability:
J)l volume gas ) thickness
Po" " [2]
A )
P area ) time )
P
The customary unit of permeability is the barrer
where:
cm3 ) cm
1 barrer"10\10 [3]
cm2 ) s ) cmHg
Permeability can also be written as the product of
the gas solubility times its diffusivity, the so-called
solution}diffusion mechanism (eqn [4]). The per-
mselectivity () for two gases i and j is deRned as the
ratio of the permeabilities:
Po"D ) S [4]
ij"Poi/Poj [5]
Figure 1 Schematic representation of membrane permeation.
1726 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes

Permeation is an activated process. The effect of tem-

perature on permeation is given by eqn [6], where
Ep is the activation energy of permeation, R is the gas
constant and T is temperature:

Po"Po ) exp(!Ep/RT) [6]

In typically encountered cases Ep is postive and the

permeability increases exponentially with temper-
ature. Additionally, Ep is related to penetrant size and
therefore selectivity usually decreases with increasing
temperature. This treatment is not true when dealing
with gases below their critical temperature. The Figure 3 Dual-mode sorption isotherm.
reader is referred to the Further Reading section for
these special cases. production of oxygen-enriched air for medical ap-
The above equations give a mathematical, phe- plications or the recovery of C4 hydrocarbons, low-
nomenological description of gas permeation through selectivity membranes and hence rubbery polymers
polymers but imply nothing of the molecular-level have found limited commercial utility in the puriRca-
processes giving rise to permeation. While we speak tion of industrial gases.
of gas-separating polymers as being dense Rlms, on
a molecular level one must consider that the mem-
brane is not ‘solid’; that is, there are molecular-size Dual-Mode Permeation in Glassy
gaps between the polymer chains. These gaps arise Polymers
from packing defects in the solid state and also arise
Polymers in the glassy state possess ‘free volume’ as
from the thermal motions of the polymer chains
shown in Figure 2 } that is, a polymer quenched
themselves. It is through pemanent and transient gaps
below its Tg to a nonequilibrium state in which its
that gas transport is believed to occur. Solution-diffu-
molar volume is higher than the equilibrium value.
sion behaviour has proven adequate to describe per-
This free volume can be visualized as long-lived mo-
meation through rubbery polymers } those whose
lecular-level gaps between the polymer chains. One
glass transition temperature, Tg, is below the experi-
aspect of free volume is that glassy polymers exhibit
mental temperature. As a family, rubbery materials
excess sorption capacity. Sorption in glassy polymers
are highly permeable but unselective for the same
can be described by eqn [7]:
molecular-level rationalization. In the rubbery state
polymer chains are highly mobile, generating a high
frequency of transient gaps which the penetrant gases CHbP
S"kDP# [7]
can easily diffuse through. However, these gaps are 1#bP
not very selective. From a practical perspective, the
purity of the product is related to the membrane
permselectivity. With some exceptions, such as the

Figure 4 Coupling of Langmuir and Henry’s mode diffusion

Figure 2 Schematic of free volume. coefficients.
 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes 1727

Figure 5 Permeability test cells. A, supporting legs; B, lower plate; C, upper plate; D, adapter; E, vacuum valve.

where kD is the Henry’s law solubility constant, b is behavior. A typical dual-mode sorption isotherm is
the Langmuir equilibrium constant and CH is the shown in Figure 3. At low pressures, sorption is dom-
Langmuir capacity and can be related to the free inated by the Langmuir element, while at high pres-
volume. Such sorption is often termed ‘dual-mode’ sure sorption is described by Henry’s law.
1728 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes

Table 1 Air separation characteristics of some common

polymers

Polymer Po O2 (barrer)  O2/N2

Polyacrylonitrile 0.0002 '10

Polyvinylidene chloride 0.0053 5.6
Polyethylene terephthalate 0.059 4.5
Cellulose acetate 0.78 2.8
Polystyrene 2.63 3.3
Poly(4-methyl-1-pentene) 32.3 4.1
Silicone rubber 610 2.0
Poly(trimethylsilylpropylene) 8000 1.4

Figure 6 Time lag diffusion experiment.

Dual-mode permeation can be viewed as the simul-
taneous diffusion of gas molecules within and between
the dense polymer phase and the Langmuir gaps or
‘holes’. Dual-mode permeation then is described by
the sum of permeation in these phases (eqn [8]). The
two diffusion coefRcients DD and DH appear to be
strongly correlated to each other (Figure 4):
2. Condensible vapours such as water and higher
hydrocarbons are strongly adsorbed in the high
enthalpy free volume sites and deleteriously affect
permeation rates and permselectivity.
3. Strongly absorbing gases such as CO2 can swell
the membrane at high pressure.
4. The free volume, a manifestation of the non-
equilibrium state of the polymer, can be decreased
by annealing or upon aging.

Po"Po (dense)#Po (hole)"SDDD#SHDH

Methods of Measuring Permeability
CHbDH Conceptually, measuring the gas permeability of
Po"kDDD# [8] polymeric membranes is simple although fraught
1#bP
with experimental pitfalls related to establishing
Some consequences of dual-mode transport are: steady state Sow. The experimental parameters are
given in eqn [2]. Several methods and apparatus have
1. The permeability of a dual-mode system decreases been developed to conduct permeability measure-
as the pressure is increased. ments. For obtaining the permeation pure gases one

Figure 7 Upper bound representation of oxygen/nitrogen permselectivity } 1990.

 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
Figure 8 Relationship between the upper bound slope (n) and
kinetic diameter difference of gas pairs ij.
measurers either the rate of permeability pressure rise
(usually from a vacuum) in a constant volume/tem-
perature receiver or the volume of permeate gas at
a Rxed pressure (Figure 5). Commercially available
test cells are known as the volumetric or ‘Linde cell’
and the manometric or ‘Dow’ cell. The American
Society of Testing and Materials has published
a method for measuring gas permeability (ASTM
Method D1434-82). It is possible to obtain the
Figure 9 Upper bound representation of oxygen/nitrogen permselectivity } 1993.
1729
Table 2 Upper bound parameters
Pi"kijn
Gas pair (ij) k (barrer) n
O2/N2 389 224 !5.800
H2/CH4 18 500 !1.2112
CO2/CH4 1 073 700 !2.6264
He/N2 12 500 !1.0242
permeability coefRcient Po and the diffusion coefRc-
ient D through a time lag measurement as shown in
Figure 6; subsequent calculation of the solubility
term though an independent measurement of S is
advised.
Measuring the permeability of gas mixtures is more
complex. Usually a Sow of a pre-blended source is
passed over the feed side of the membrane. The
steady state permeate Sux can be measured by em-
ploying an inert gas such as helium to sweep the
permeating components into a gas chromatograph,
for example, for compositional analysis. Preferably
the experiment is conducted such that back-diffusion
of the helium sweep is insigniRcant and where the
feed gas composition is not altered by permeation.
Additionally, nuances in the Sow patterns within test
cells not speciRcally designed for mixed gas experi-
ments can lead to erroneous results.
Optimization of Polymer Permeation
Properties
Table 1 illustrates the range of oxygen permeability
and oxygen/nitrogen permselectivity for some
1730 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes

Table 3 Structure}property relationships in polymer membranes

Polymer Structure Density d-spacing P (O2) O2/N2

(g cm\3) (A> ) (barrer)

PMDA-ODA 1.40 4.6 0.61 6.1

PMDA-mDA 1.35 4.9 0.98 4.9

PMDA-IPA 1.28 5.5 7.1 4.7

6FDA-ODA 1.43 5.6 3.9 5.34

6FDA-mDA 1.40 5.6 4.6 5.70

6FDA-IPA 1.35 5.7 7.5 5.60

PSF 1.24 5.0 1.4 5.60

DMPSF 1.21 5.0 0.64 7.00

TMPSF 1.15 5.5 5.6 5.28

 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes 1731

common polymers. Permeability spans a wide range:

seven orders of magnitude. Further, as the permeabil-
ity of polymers increases, their ability to differentiate
between gases, the permselectivity, decreases. This
correlation is valid for nearly all polymeric mem-
branes and has been the subject of research. While
long recognized, this observation was Rrst formalized
by Robeson in 1991. Working with a database of
over 200 polymers, the selectivity of several gas pairs,
plotted on a log}log scale against the permeability of
the faster gas, exhibits a characteristic upper bound
deRning the combinations of permeability and perm-
selectivity simultaneously achievable with polymeric
materials (Figure 7). Upper bound performance can
be described by eqn [9]:

Pi"knij or ij"k\1/nPi1/n [9]

where the values of k and n, tabulated in Table 2,

are calculated from the upper bound relationship for
speciRc gas pairs. The parameter n is related to the
difference in kinetic diameters of the penetrant gas
pair
dij as shown in Figure 8. This empirical treat-
ment implies that the upper bound is a natural result
of the sieving characteristic of stiff chain glassy poly-
mers. A fundamental theory was later developed by
Freeman in which the constants could be related to
gas size and gas condensability and invoked just
a single adjustable parameter f:

ln ij"!ij ln Di#ln (Si/Sj)!ij(b!f(1!a)/RT)

[10]

where ij"[(di/dj)2!1] and di is the kinetic

diameter; Si/Sj"N(i/k!j/k), where k is the
Boltzmann constant and  is the potential energy well
depth in the Lennard}Jones potential energy func-
tion. The constants a and b are derived from linear
free energy relationships and are independent of gas
type. Moreover a is independent of polymer and
has a universal value of 0.64; b has a value of 11.5 for
glassy polymers and 9.2 for rubbery polymers.
The solubility selectivity among polymers is largely
Figure 10 Structure}property relationship: (A) FFV model
constant; consequently diffusivity considerations (Paul). (B) Robeson model.
dominate upper bound permselectivity.
The optimization of polymer structure to obtain
upper bound properties comprises the lion’s share of The solubility selectivity (Si/Sj) is nearly constant
polymer membrane research in the last 20 years and across a wide variety of polymers and for O2/N2 is
the reader is referred to the Further Reading section. about 2. Selectivity in glassy polymers is therefore
Researchers have heuristically developed the under- dominated by the diffusive selectivity which in turn
standing gained via the upper bound analysis. The results from the sieving properties of the imperfectly
permselectivity for gases i and j is given by eqns [4] packed polymer chains. The best trade-off in per-
and [5] as: meability properties within a polymer family is ob-
tained when both main chain mobility is limited
ij"Poi/Poj"(Si/Sj) ) (Di/Dj) [11] and intersegmental packing of polymer chains is
1732 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes

Figure 11 Structure units with imparting superior permselectivity.

inhibited. This behaviour is illustrated (Table 3)

for a family of pyromellitic dianhydride (PMDA) and
hexaSuoro-isopropylidene dianhydride (6FDA)-
based polyimides. Changes in the functionality lead
to different packing arrangements as measured by
density and X-ray d-spacings (average distance be-
tween polymer chains). Very small changes in chain
packing result in signiRcant changes in both perme-
ability and selectivity. Further, groups such as 6FDA
are particularly desired because they can increase
permeability without a large loss of selectivity. A fur-
ther example is that of ortho, di- versus tetra-substitu-
tion patterns on aromatic polymers. It is widely recog-
nized that incorporation of bulky substituents leads to
an increase in permeability, usually at a loss of selectiv-
ity. However, it is also noted that ortho di-substitution
patterns result in lower permeability and higher selec-
tivity than the symmetrically tetra-substituted ana-
logue. Using this intuitive approach a great many new
polymers were synthesized and characterized between
1990 and 1993 and as a result the empirically deter-
mined upper bound has been shifted upperwards and
its slope has changed (Figure 9), with many polymers
lying at or near the upper bound.

Predicting Polymer Permeability

The beneRt of being able to predict a priori the rela-
tionship between polymer structure or physical prop-
erties and permeability is obvious. One method is to
correlate the permeability with the reciprocal of the
fractional free volume (FFV), deRned as:

V!V
Figure 12 Cross-section of an asymmetric membrane. FFV"  [12]
V
 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
Figure 13 Phase diagram for a ternary dope system. PP, polymer-poor; PR, polymer-rich; CP, critical point.
where V is the speciRc volume of the polymer and Vo
is the volume occupied by the polymer chains. The
speciRc volume is obtained from experimental density
measurement. Direct measurement of Vo is not poss-
ible and therefore various computational approaches
have been developed. The most widely used is that
developed by Bondi relating the zero point volume
(the volume occupied at 0 K) to the van der Waals
volume of the molecule and later modiRed by Park to
account for the fact that different gases have access to
different FFV depending on the speciRc gas}polymer
interaction. This group contribution approach has
yielded good correlations (Figure 10A) but it is not
intuitively obvious how to relate polymer structure
directly to free volume.
The group contribution approach by Robeson as-
serts that the overall polymer permeability can be
represented by the sum of structural subunits that
comprise the polymer in proportion to their volume
fraction:
ln Po"
i ) Pi [13]
where i is the volume fraction of subunit i and Pi its
permeability contribution. Volume fractions are cal-
culated using molecular mechanics computer model-
ling and the Po’s are experimentally determined.
The Pi’s are found by regressing the set of simulta-
neous linear equations from eqn [13]. In addition to
adequately representing the experimental data
(Figure 10B), this method also identiRes those
subunits which exhibit upper bound properties
1733
(Figure 11). Desirable moieties include the 6FDA,
di-ortho substituted phenyl ethers and the spirobi-
indane fragment. A superior polymer would be tetra-
bromo-poly(phenylene oxide); however, no one has
as yet succeeded in its synthesis.
Membrane Formation
Most research on polymeric membrane materials is
conducted on thin, solvent-cast, Rlms. The practical
limit to such membranes is about 25 m for free-
standing Rlms and perhaps 1 m if the polymer solu-
tion is cast on a microporous substrate. However, at
these thicknesses the Sux through the membrane is
too low to be of practical value. That is to say, the
economics of a membrane process utilizing these
thick Rlms would not effectively compete with estab-
lished technologies such as cryogenic distillation. The
enabling development for fabricating ultrathin, high
Sux membranes was the integrally skinned asymmet-
ric membrane of Loeb and Sourirajan. This type of
membrane is produced by inducing phase separation
Figure 14 Series resistance model for defect repair.
1734 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes

Figure 15 Flat-sheet membrane fabrication.

in a thermodynamically stable solution, usually by ating a water-miscible solvent into a water coagula-
changing its composition through the introduction of tion bath. Such membranes possess an ultrathin dense
a nonsolvent. These membranes are prepared by cast- skin that gradually opens into a microporous sub-
ing a concentrated polymer solution (dope) incorpor- structure (Figure 12). The skin layer provides the

Figure 16 Spiral-wound membrane element.

 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
separation while the thicker integral substructure
provides robustness but very little resistance to gas
Sux.
Membrane formation processes can be grouped
into three methods:
1. Dry phase inversion processes involve thermally
quenching a dope or solvent evaporation in multi-
component solvent system dopes.
2. Wet phase inversion processes involve quenching
a cast dope in a coagulation bath.
3. Dry-wet phase inversion processes combine sol-
vent evaporation and a quench medium.
Most commercial gas separation membranes are pro-
duced by the dry}wet process although temperature-
induced phase inversion (TIPS) is also employed.
Phase diagrams illustrate the phase inversion pro-
cess (Figure 13). The binodal deRnes the boundary of
the two-phase region and is divided into two parts at
the critical point (CP). A second envelope, the
spinodal line, also emerges at the critical point. The
phase inversion process involves bringing a dope
solution into the two-phase envelope. The initially
stable solution (at a composition designated by * in
Figure 13) decomposes into a polymer-rich phase and
a polymer-lean phase, the compositions of which are
Figure 17 Hollow-fibre membrane spinning apparatus and cross-section of spinnerette.
1735
deRned by tie lines. If the decomposition takes place
in the region between the binodal and spinodal a nu-
cleation and growth mechanism of the polymer-rich
and polymer-lean phases dominates, leading to struc-
tures undesirable for gas separation. Decomposition
below the critical point leads to a dispersion of poly-
mer nodules within the polymer-lean phase; de-
composition above the critical point leads to a closed
cellular structure of an encapsulated polymer-lean
phase. The preferred inversion path is to quickly
bring the dope within spinodal envelope, thus
generating an interpenetrating network of polymer-
lean and polymer-rich phases which vitrify into a Rne-
ly porous substructure.
The thermodynamic framework for a ternary dope
system lies in Flory}Huggins theory and the Gibbs
free energy of mixing for a ternary system is given
by:

Gmix"n1 1#n2 2#n3 3#12n1 2
#13n1 3#23n2 3 [14]
where the subscripts refer to the nonsolvent (1), the
solvent (2) and the polymer (3). The notations ni and
i are the number of moles and the volume of
1736 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes

component i and ij is the interaction parameter ac-

counting for the nonideality of the mixture. At equi-
librium the chemical potentials (u) of the components
in all the phases are equal and are given by eqn [15],
which deRnes the binodal envelope. The composition
for the spinodal line is given by the solution to
eqn [16], where vi is the pure component molar vol-
ume of species i:

ui 
Gmix
" [15]
RT ni RT

(1/ 1# 1/ 2 2 !212)(1/ 1 # 1/ 3 !213)

3

;(1/ 1#23 1/ 2!12!13)"0 [16]

Composite Membranes
The above discussion also applies to the formation of
composite or multilayer membranes. Composite
membranes can be categorized as either a dense, iso-
tropic or asymmetric coating of a high performance
separating layer on a microporous substrate. Com-
posite membranes are fabricated in two operations,
substrate formation followed by dip coating, allow-
ing for the independent optimization of coating and
substrate properties. Gas transport through com-
posite membranes is described by the series resistance
model by analogy to an electrical circuit. For a two-
layer composite consisting of a thin layer of polymer Figure 18 Hollow-fibre membrane element.
A on a substrate of polymer B, the Sux of gas i though
the membrane is given by eqn [17], where l is the
thickness of the respective layers: circuit. The total resistance is given by the sum resist-
ances in the coating layer (Rc); the parallel resistance
Ji"
P(lB/PoB#lA/PoA) [17] in the defective skin layer; (Rsk,d); and the resistance of
the substrate Rsub:
Regardless of their method of formation a critical
element of gas separation membranes is that the skin Rtot"Rc#Rsk,d#Rsub
layer must be as thin as possible in order to produce
a high Sux membrane. The practical limit to the skin Rsk;Rd
layer thickness is thought to be in the range of "Rc# #Rsub [18]
Rsk#Rd
500}1000 A> . Concurrently, the skin layer must be
free of manufacturing defects or pinholes. A defective
surface area fraction as low as 10\5% can lower After repair, the resistance of the defect, Rd is greater
selectivity to an extent that the membrane is not than Rd and also, Rsk'Rd , so that the effective
suitable for gas separation. Conventional manufac- permeability of the composite approaches the intrin-
turing processes are not capable of achieving this level sic permeability of a defect-free membrane.
of reliability. The second enabling development in
commercializing gas separation membrane tech-
nology was the demonstration of a poly(dimethyl-
Membrane Devices
siloxane) defect repair coating to effectively ‘plug’ Gas separation modules can be prepared from Sat
manufacturing pinholes and eliminate bulk Sow sheets as plate and frame assemblies and spiral-
through the defects (Figure 14). Gas permeation wound elements. Hollow Rbres can be Rne Rbres
through this multicomponent system is described by ((1000 m OD) or tubular. In general only
the series resistance model analogous to an electrical Rne Rbres and spiral-wound elements combine
 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
Figure 19 Membrane technology: breakdown by application.
performance and cost parameters providing eco-
nomic viability. The beneRts of the spiral-wound
conRguration include ease of fabrication and low
pressure drop but manufacturing costs are high and
the membrane surface area to module volume ratio is
low (+700 m2 m\3), leading to larger, heavier sys-
tems. Hollow Rbre modules can achieve very high
surface area/volume ratio ('5000 m2 m\3) and pro-
vide a higher degree of countercurrent Sow but are
more difRcult to fabricate and can have high pressure
drop.
Flat Sheets and Spiral-Wound Elements
A schematic representation of a Sat-sheet asymmetric
casting device is shown in Figure 15. Flat-sheet
membranes are typically produced on a nonwoven
cloth which provides support for the nascent mem-
brane. These nonwoven cloths are commercially
available on large rolls of 24}48 inches. A thin
liquid Rlm of dope solution is metered by a doctor
blade onto the nonwoven cloth while the fabric ro-
Table 4 Selected membrane applications
Category Range of operation
Hydrogen recovery 95% H2
(5 MMSCFD
Nitrogen production 95}97% N2
0.1}2.0 T/D
Oxygen 40% O2
Carbon dioxide removal 95#% CH4
(40 MMSCFD
Dehydration Dewpoint to !403C
Misc.
1737
tates around a stainless steel roller. The nascent
membrane is quenched in the gel tank, washed and
collected on a take-up roller. A generic spiral-
wound element is shown in Figure 16. The simplest
conRguration consists of a central collection pipe
around which is wound and glued an envelope of
Sat-sheet membrane. The envelope contains the
membrane and feed and permeate spacer channels.
The spacer material is an extremely porous, inert
material. Feed gas Sows parallel to the permeate pipe;
permeating gas Sows into the permeate spacer and is
collected in the pipe. Higher membrane areas can be
achieved using the multileaved method in which two
to four sheets of membrane are wrapped simulta-
neously.
Hollow Fibres
A second asymmetric membrane geometry is that of
a hollow Rbre. The ultrathin skin is present on the
external surface of the Rbre. Like the Sat sheet, the
wall of the Rbre is microporous and offers low resist-
ance to gas Sow. The permeating gases collect in the
bore or lumen of the Rbre. A generic hollow-Rbre
spinning device is shown in Figure 17. The apparatus
contains a dope reservoir, a bore Suid reservoir,
pumps, coagulation baths, wash baths and a take-up
winder. The dope solution and bore Suid are co-
extruded through a die, resulting in a sheath of poly-
mer around the bore Suid (typically water). This
nascent Rbre is then coagulated, washed and dried.
Hollow-Rbre devices contain thousands, even hun-
dreds of thousands, of individual Rbres within a single
pressure housing (Figure 18). The feed gas is usually
fed to the outside of the Rbre (shell side) although
some processes route the feed gas through the Rbre
Application Gases separated
Ammonia synthesis purge gas H2/N2, Ar
Syngas ratio adjustment H2/CO
Hydrotreater off gas
Inerting: fuel tanks; food transportation N2/O2, CO2
Gas and oil drilling N2/O2, CO2, H2O
Oxygen-enriched air O2/N2
Natural gas sweetening CO2/CH4
Enhanced oil recovery
Landfill gas
Instrument air H2O/air or N2
Natural gas dehydration H2O/CH4
Semiconductor process gas Perfluorocarbon from N2
1738 II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
Figure 20 Triple-orifice die for co-extrusion. 1, bore fluid; 2,
substrate polymer dope; 3, coating polymer dope.
bore, especially when well-deRned Sow character-
istics are required.
Applications
The Rrst successful commercial activity in gas separ-
ation by membranes began in 1977 with the introduc-
tion of the Prism] membrane system by Monsanto
(now Air Products and Chemicals, Inc.), to recover
H2 from ammonia synthesis plant purge gas. Installed
membrane capacity in 1977 was only 5 MMSCFD
(million standard cubic feet per day) and grew to over
3500 MMSCFD by 1996. Approximately two-thirds
of current installed membrane capacity is used for
H2 separation, which includes ammonia purge gas,
reRnery and petrochemical applications (Figure 19).
SigniRcant growth is expected for N2 and CO2 separ-
ations. While the speciRcs will vary according to
application, a generic membrane system will include
a compressor if the source gas is not available at
pressure and pretreatment to remove condensible,
corrosive or reactive components. The partial listing
of applications (Table 4) represents established uses
for gas separations. Expanding the slate continues to
be a focus of membrane manufacturers and research
institutes.
Future Trends
The concurrent need for high performance polymers
and low cost membranes has resulted in research into
hybrid polymer systems and improved manufacturing
processes. Four areas currently being explored are:
1. Polymer blends: A hybrid polymer system in
which expensive, high performance polymers
form a continuous phase in a multipolymer blend.
Examples include Matrimid/Ultem威 1000 and
aromatic polyimides/polysulfone.
2. Coextruded composite hollow Tbres: An im-
proved hollow-Rbre spinning process in which
a forming asymmetric substrate is simultaneously
coated with a thin Rlm of separating polymer by
employing a die similar to that shown in Fig-
ure 20. This is a paradigm shift away from separ-
ate substrate fabrication and subsequent coating
practices. The co-extrusion processes should
lower manufacturing costs and relax price con-
straints on new high performance polymers.
3. Organic/inorganic mixed matrix membranes:
A nanocomposite composite is a hybrid in which
very small particles of a material with high dif-
fusive selectivity such as a microporous carbon is
dispersed with an organic polymer matrix. Such
materials combine the high selectivity of the inor-
ganic with the processability of the organic poly-
mer matrix.
4. Computational methods: Molecular modelling of
gas transport through rubbery polymers has al-
ready proven successful. Its application to glassy
polymers is signiRcantly more difRcult, primarily
because of the poorly deRned nature of the glassy
state, but signiRcant progress has been made in
the last several years. As our understanding
of gas}polymer and polymer}polymer inter-
actions improves and merges with advancing
computer technology, widespread use of molecu-
lar dynamics should provide signiRcant insight
into gas separation with polymer membranes.
Further Reading
Ho WS and Sircar KK (1992) Membrane Handbook. New
York: Chapman & Hall.
Hwang ST and Kammermeyer K (1975) Membranes in
Separations. New York: Wiley.
Kesting RE (1985) Synthetic Polymeric Membranes:
A Structural Perspective, 2nd edn. New York: Wiley.
Koros WJ and Fleming GK (1993) Membrane-based gas
separations. Journal of Membrane Science 83: 1}80.
Mulder M (1991) Basic Principles of Membrane Tech-
nology. Dordrecht: Kluwer.
Paul DR and Yampolski YP (eds) (1994) Polymeric Gas
Separation Membranes. Boca Raton, FL: CRC Press.
Robeson LM (1991) Correlation of separation factor vs.
permeability for polymeric membranes. Journal of
Membrane Science 62: 165.
Stern SA (1994) Polymers for gas separations: the next
decade. J ournal of Membrane Science 94: 1}65.
 II / MEMBRANE SEPARATIONS / Liquid Membranes
Haemodialysis
See II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
Kidney Dialysis
See II / MEMBRANE SEPARATIONS / Dialysis in Medical Separations
Liquid Membranes
L. Boyadzhiev, Institute of Chemical
Engineering, Bulgarian Academy of Sciences,
Sofia, Bulgaria
Copyright ^ 2000 Academic Press
Introduction
The separation of solutes by means of liquid mem-
branes is based on a simple and well-established idea:
two completely miscible liquid phases, separated by
a third liquid, immiscible with either of them, can
exchange solutes, provided there is a difference be-
tween their chemical potentials in the two phases and
provided the intermediate liquid is able to transport
them. In most cases the two miscible liquids, denoted
hereafter as donor and acceptor phases, are aqueous
solutions and the third (membrane) phase is an or-
ganic liquid. The conRguration involving two organic
solutions separated by an aqueous membrane is less
popular.
The growing interest in the recovery and separation
of solutes by means of liquid membranes may be
related to the advantages of this separation method
over the related separation operations } solid mem-
branes and solvent extraction } as well as to the
recent development of efRcient liquid membrane
techniques and contactors.
The main advantage of liquid membranes over
polymer ones is the higher Sux, owing to the very
much higher diffusion coefRcients of solutes in liquids
than in solids. Moreover, some liquid membrane
techniques allow a convective diffusion regime in-
stead of a molecular one, which also increases Suxes.
Another advantage of liquid membranes is the avail-
ability of a great number of substances which, when
added to the liquid membrane phase, increase selec-
tivity.
1739
A liquid}membrane process can be regarded as
a combination of extraction and a stripping process,
which take place simultaneously in the same device.
In solvent extraction, both the extractant amount
and the distribution coefRcient of the solute play
essential roles for process efRciency, whereas in liquid
membrane separation the selectivity is controlled
by the kinetics of the transport process. In
contrast to solvent extraction, in liquid membrane
separation the amount of transferred solute is not
proportional to the amount of the solvent used, in this
case the membrane liquid. The relatively small
amount of the latter permits the use of various highly
efRcient and selective } even expensive } carriers.
Mechanisms of Solute Transfer
Like some of the solid}membrane separation
methods, the difference between the chemical poten-
tials of the solute in the donor and acceptor solutions
controls the transport of the species. In other words,
the concentration difference is the driving force.
There are various mechanisms for the selective
transfer of solutes in the considered three-liquid-
phase system. They can be divided into two groups:
nonfacilitated and facilitated, or carrier-meditated,
transfer mechanisms.
In nonfacilitated processes, the membrane phase is
the solvent and carrier of the solute. In facilitated
processes, the membrane phase is a neutral medium,
dissolving a carrier, which reacts with some molecu-
les or ions and selectively transfers them to the accep-
tor phase. The carrier reacts reversibly with the solute
by binding it in the donor solution or at the interface
between this solution and the membrane phase; it
transports it across the bulk of the membrane, and
releases it at the other interface. When the transfer of
1740 II / MEMBRANE SEPARATIONS / Liquid Membranes
Figure 1 Basic transport mechanisms. (A) Simple nonfacilitated transport; (B) Simple uphill nonfacilitated transport; (C) facilitated
uphill transport; (D) facilitated (coupled) co-transport; (E) facilitated (coupled) countertransport. See text for details.
a solute is accompanied by an equivalent transfer of
one or more other solutes, it is designated as coupled
transport. Depending on the direction of the accom-
panying transfer, the mechanisms are called co-trans-
port and countertransport. Figure 1 illustrates the
Rve most popular transport mechanisms: (A) and (B)
refer to nonfacilitated mechanisms, while (C)}(E) re-
fer to facilitated mechanisms.
Nonfacilitated Mechanisms
Figure 1A shows the nonfacilitated transport of sol-
ute A from the donor solution to the membrane liquid
as a result of its solubility and the low concentration
in the latter. From this phase, it is transferred to the
acceptor phase again for the same reasons. This pro-
cess continues until the chemical potentials of the
solute, i.e. its concentrations in the donor and accep-
tor solution, are equal. The selectivity of separation
of solutes present in the donor solution mainly de-
pends on the difference between their transfer rates,
which in turn are related to their solubility in the
membrane and, to a lesser extent, on the difference
between their diffusion coefRcients. This rather
simple mechanism is of no practical interest. An
example is the separation of aromatic and aliphatic
hydrocarbons using water as the membrane phase.
Figure 1B shows a second example of non-
facilitated transport. The process differs from Fig-
ure 1A in that the acceptor solution has a component
B which is insoluble in the membrane; it reacts irre-
versibly with solute A that permeates through the
membrane. The reaction product AB is insoluble in
the membrane and cannot diffuse back to the donor
solution. In some cases an enzyme plays the role of
the reagent B, transforming transported solute into
products which are insoluble in the membrane. The
continuous consumption of A in the acceptor solution
maintains its concentration in this phase at a low
level, creating a sufRcient driving force to transfer the
whole amount of A from the donor solution. The
solute A in the form of the product AB can reach very
high concentrations in the acceptor solution, which is
generally of a smaller volume than the donor solu-
tion. This transfer, apparently against the concentra-
tion gradient, is known as a simple uphill transport
and it has a real practical value. A typical example is
the transfer of a phenol as a neutral solute which is
soluble and thus permeable through the organic
membrane phase. The acceptor phase is an alkaline
solution that converts the phenol to an ionized salt
which is not soluble or permeable through the mem-
brane phase.
Facilitated Transport Mechanisms
In facilitated transport mechanisms the neutral mem-
brane liquid contains an active substance C, which
selectively and reversibly reacts with the permeating
solute, forming a complex AC (Figure 1C). This com-
plex is formed at the donor interface of the membrane
phase and then, due to its concentration gradient,
moves to the acceptor solution membrane-phase in-
terface. The complex AC then reacts with a reagent B.
As a result of this reaction, A is irreversibly bound by
B, while the carrier C is restored and goes back across
 II / MEMBRANE SEPARATIONS / Liquid Membranes
the membrane to the feed}membrane interface to
bind a new portion of the solute A. Because of this
shuttle mechanism, small amounts of the carrier
C can transfer large amounts of the solute in the
acceptor phase. An example is the recovery of nitric
acid from dilute solution using a small amount of
the carriers tributylphosphate or trioctylphosphine
oxide. The adducts formed are unstable in strongly
alkaline media (the acceptor solution), where the acid
is neutralized and irreversibly converted into nitrate.
In transport processes shown in Figure 1D, some-
times called facilitated co-transport, the carrier C re-
versibly forms an intermediate complex not only with
the solute A but also with other (one or more) con-
stituents of the donor solution. The complex ABC so
formed is transported to the acceptor solution, where
it reacts with another additive, D, by forming a more
stable compound. The latter, like the reagent D, is
insoluble in the membrane liquid. An example of this
mechanism is the transport of silver which is selec-
tively recovered from complex polymetallic nitrate
solutions. The complex, transferred across the mem-
brane, is formed by a silver cation, a nitrate anion and
two molecules of the extractant triisobutylphosphine
sulRde, selective for silver. In the acceptor solution,
the complex is destroyed by ammonia. The chemical
reaction in the acceptor phase yields ammonium ni-
trate, the stable silver}ammonia complex and the
regenerated carrier.
Figure 1E illustrates the third, probably most often
used, facilitated transport mechanism, sometimes
called facilitated countertransport. In this case, ions,
initially present in the donor solution, are substituted
Figure 2 Bulk liquid membrane contactors for laboratory use. (A) U-tube contactor (Schulmann bridge); (B) beaker-in-beaker
contactor; (C) and (D) two cells separated by supported liquid membrane.
1741
by other ions of the same type, present in a sufRcient
amount in the acceptor solution. This is actually an
ion exchange process in which the ion-exchanging
agent, the carrier C, transports in one direction one
type of ion and in the opposite an equivalent amount
of substituent. A typical example for this transfer is
the recovery of some divalent metal cations, e.g.
Cu2#, from neutral or slightly acidic aqueous solu-
tions by means of oleophilic chelating oximes. The
latter transfers the metal ions to the strongly acidic
acceptor solution and returns protons according to
the scheme:
Me2##2HR 0 MeR2#2H#
The equilibrium conditions at the two interfaces, con-
trolled by the pH values of the aqueous phases, are
chosen so that the metal}organic complex is the
stable species at the donor}membrane interface,
while free cations exist at the membrane}acceptor
interface. This type of process is of great signiRcance
for hydrometallurgy and for the removal of heavy
metals from industrial efSuents.
These Rve types of transfer mechanisms do not
exhaust all possible schemes for the selective recovery
and separation of solutes by means of liquid mem-
branes. Liquid membrane processes have been de-
veloped during the last few years on the basis of
various transport and reaction mechanisms, including
redox reactions. For example, the selective transfer of
metals may result from the different solubility of their
ions at various oxidation states. Of equal interest are
some enzymatic reactions, both in the bulk of the
1742 II / MEMBRANE SEPARATIONS / Liquid Membranes

tactors shown in Figure 2C and D Rnd a broader

application. In these devices, the membrane liquid
permeates a porous membrane, which separates the
donor and the acceptor solutions. In modiRcation
(D), a cylinder with an attached porous barrier
rotates and stirs the donor and acceptor phases,
reducing or eliminating the mass transfer resistances
in these two phases. The type of device depends on
whether the membrane liquid is heavier or lighter
than the other two solutions.

Supported liquid membranes The laboratory con-

tactor shown in Figure 2C is the prototype of sup-
Figure 3 Spirally wound supported liquid membrane module. ported liquid membrane contactors. In these devices
R, Acceptor solution; F, donor solution. the membrane liquid Rlls the pores of a 25}100
m
thick porous membrane containing pores 0.01}10
m
in diameter. The membrane is usually made of
membrane and in the bulk of the acceptor solution. polypropylene, polysulfone or another oleophilic
The reader may Rnd further information in the Fur- polymer.
ther Reading section. Although the membrane is quite thin, the Suxes
across it are very low as a result of the total immobil-
ization of the membrane liquid in the pores, reduced
Liquid Membrane Techniques free section and pore tortuosity. This is overcome by
The main reason for the limited large scale appli- the use of large surface area modules such as the
cation of liquid membrane processes is the lack spirally wound (Figure 3) or containing bundles of
of efRcient equipment providing simultaneously large tiny porous hollow Rbres, as shown in Figure 4. Hol-
contact areas and high Suxes between the phases low Rbre membrane modules containing Rbres with
without deterioration of the membrane over time diameter of 0.2}1 mm can achieve interface areas of
causing intermixing of the donor and acceptor
phases. The realization of stable membranes is an
extremely difRcult task.
In general, liquid membrane techniques can be
divided into two groups: techniques in which there is
no dispersion of phases and techniques with at least
one dispersed phase. The Rrst group includes bulk
liquid membranes and the supported liquid mem-
branes, as well as some recent techniques combining
elements from both techniques. The second group is
mainly represented by the emulsion liquid membrane
technique.

Methods Without Phase Dispersion

Simple bulk liquid membranes Several simple con-

tact devices designed for studies of liquid}membrane
processes are shown in Figure 2. In all, there is a com-
mon compartment for the membrane liquid. The
other contactor space is divided into two compart-
ments, one for the donor solution and the other for
the acceptor solution. The interface between the
membrane liquid and the other two solutions is free
(A, B) or immobilized (C, D) by a solid porous mem-
brane. The Rrst device (A) is known as the Schulmann
bridge. Devices of the type shown in Figure 1A and Figure 4 Hollow fibre supported liquid membrane module.
B are limited to laboratory experiments, but the con- R, Acceptor solution; F, donor solution; S, membrane liquid.
 II / MEMBRANE SEPARATIONS / Liquid Membranes 1743

2000}10 000 m2/m3. In such modules, one of the
aqueous phases Sows in the lumen of the hollow
Rbres, while the other Sows outside the Rbres and the
pores of the Rbre walls are Rlled with the membrane
liquid.
The insigniRcant amount of membrane liquid re-
quired in these modules (10 cm3 per 1 m2 interface),
often pointed out as a major advantage, is actually
the chief drawback of supported liquid membrane
contactors, causing their operational instability and
short life. The life of the expensive modules is
shortened by the inevitable solubility of the mem-
brane liquid in the donor and acceptor phases, by its
washing out or by emulsiRcation caused by the pres-
sure difference on both sides of the membrane, the
lateral shear force, and the change of support wetta-
bility with time. In spite of numerous design improve-
ments, e.g. periodic or continuous reimpregnation of
the membrane and partial or total gelation of the
membrane liquid, this technique has not been used in
industrial applications.
This instability forced researchers as early as in the
1980s to look for other solutions. The combination of
this technique with stable bulk liquid membranes
yielded the bulk-supported liquid membranes.
permits free elongation of the Rbre package caused by
the swelled membrane liquid.
The latter two membrane techniques provide sig-
niRcantly longer life of the contactors, as the inevi-
table losses of membrane Suid are compensated by
the larger liquid volume. However, Suxes are lower
because of higher mass transfer resistance due to the
second porous support Rlled with immobilized liquid
and the two additional diffusion boundary layers in
the same phase. This drawback is, however, off-
set by the longer membrane life.
A further modiRcation of the contained liquid
membrane technique is the separation of the two
hollow Rbre packages in two modules } one where
the donor liquid exchanges solutes with the
membrane liquid and a second where the membrane

Flowing liquid membranes and contained liquid

membranes In these two variants of the bulk-
supported liquid membrane group, as well as in
numerous subsequent modiRcations, the membrane
liquid not only Rlls the pores of two closely spaced
porous supports separating the donor and acceptor
phase, but also the space between them, as shown in
Figures 5 and 6. Figure 5 shows a device introduced
by Teramoto et al., called the Sowing liquid
membrane: the spirally wound module contains one
additional layer and one additional porous barrier
(Figure 5) in comparison with the analogous sup-
ported liquid membrane module, shown in Figure 3.
Between the two, separated by porous support
spacers, Sows the membrane liquid, which also Rlls
the pores of the support which are preferentially
wetted by it.
In contained liquid membranes, a technique pro-
posed by Sirkar et al. in the late 1980s, the donor
phase Sows in the lumen of a part of the capillaries,
while the acceptor phase Sows in the lumen of the rest
of them. As Figure 6 shows, the membrane liquid
Rlling the space outside the hollow Rbres can also be
set in motion. When the hollow Rbre material is
wetted by the membrane liquid, the pores are Rlled
with it. In the reverse case, they are Rlled with the
other two phases. The module shown in Figure 6, in
which the inlets and the outlets of the feed and accep- Figure 5 Spiral-type flowing liquid membrane module.
tor phases are located in one end of the module case, R, Acceptor solution; F, donor solution; S, membrane liquid.
1744 II / MEMBRANE SEPARATIONS / Liquid Membranes

Figure 6 Contained liquid membrane contactor. R, Acceptor solution; F, donor solution; S, membrane liquid.

Figure 7 Liquid film pertractors: (A) falling film pertractor; (B) rotating film pertractor. R, Acceptor solution; F, donor solution;
S, membrane liquid.
 II / MEMBRANE SEPARATIONS / Liquid Membranes 1745

liquid contacts with the acceptor liquid. The mem-
brane liquid circulates between the two devices. This
technique, bearing the name two-module hollow Rbre
supported liquid membranes, differs little from the
arrangement in a conventional extraction-stripping
unit operation.
Liquid Vlm pertraction The technique known as
liquid Rlm pertraction attempts to combine the ad-
vantages of bulk liquid membrane and supported
liquid membrane. In the process all three liquids are
in motion and the interfaces between the phase pairs
are not immobilized, so that the transport rate in all
stages of the transfer process is controlled by convec-
tive transport instead of the much slower molecular
diffusion.
Two devices utilizing this technique are schemati-
cally presented in Figure 7. In the Rrst one, called the
falling Rlm pertraction, shown in Figure 7A, the do-
nor and acceptor solutions Sow down the surface of
alternating vertical supports. The spaces between the
opposite supports, covered by Rlms of donor and
acceptor liquids, respectively, are Rlled with the mem-
brane phase, Sowing countercurrent to the other two.
By independent control of the Sow rates of the feed
and acceptor phases, a signiRcant solute accumula-
tion in the acceptor solution can be achieved.
The second technique, rotating Rlm pertraction
uses a package of rotating horizontal discs wetted
only by the feed and receptor phases. This rotation
generates an intensive transfer regime in all three
liquids. As Figure 7B shows, the discs, alternately
mounted on a shaft, are partially immersed in the
corresponding wetting solutions and on rotation
form mobile Rlms which directly contact with the
membrane liquid Rlling the spaces between the discs.

Figure 8 Separation by emulsion liquid membranes. 1, Emulsion preparation (step 1); 2, feed treatment with the emulsion (step 2);
3, break-up of enriched emulsion (step 3). R, Acceptor solution; F, donor solution; S, membrane liquid.
1746 II / MEMBRANE SEPARATIONS / Liquid Membranes

The advantages of these two techniques consist in the

considerably larger Suxes per unit interface and in
their practically unlimited life. However, the rather
low ratio between the contact interface and the bulk
of the solution neutralizes, the Rrst advantage.

Methods with Phase Dispersion

Emulsion (surfactant) liquid membranes Emulsion
liquid membranes were Rrst described in 1971 by Li
in a paper dealing with the separation of aromatic
and aliphatic hydrocarbons by stabilized dispersion
of three liquids: the above-mentioned mixture, water
and an inert hydrocarbon as a recipient phase. This
technique, known as emulsion (or surfactant) liquid
membranes, was the Rrst pertraction technique de-
veloped to industrial scale.
As the name implies, the three-phase system is
stabilized by an emulsiRer, added to the membrane
liquid, in some cases its concentration in the mem-
brane liquid reaches 5% or more. The acceptor solu-
tion is dispersed as Rne (2}20
m) droplets in the
membrane phase. The thick emulsion, stabilized by
the emulsiRer, is dispersed in its turn in the donor
solution as globules of 1}2 mm diameter and the
resulting dispersion is intensely stirred for several
minutes. During this contact time, the solutes, which
are more soluble in the membrane phase, are transfer-
red from the donor phase to the intermediate phase
and from there to the encapsulated acceptor solution.
This transfer is very fast due to the large contact
areas. After termination of the second (main) process
step and dispersion settling, the enriched emulsion is
separated and subjected to chemical, thermal or, most
often, high voltage electrocoagulation, which breaks Figure 9 Two-compartment pulsating column. Applied pulsa-
the emulsion into two phases. The separated mem- tions, D, exchange the membrane liquid S between central and
brane liquid phase is fed back for a new cycle of the annular compartments across the porous wall. R, Acceptor solu-
process and the enriched acceptor solution phase is tion; F, donor solution; S, membrane liquid.
subjected to further treatment. The scheme in Fig-
ure 8 illustrates this three-stage batch process which
in some modiRcations is carried out as a continuous Other techniques with phase dispersion In addition
process. In this process, the recovery efRciency and to the disadvantages listed above, the added emulsi-
the separation selectivity are controlled by the trans- Rer contaminates both the donor and acceptor
fer kinetics, i.e. by the difference in the transfer rates phases, as in some cases its solubility in these phases
of the individual solutes. As these rates depend on exceeds that of the membrane liquid itself or of the
a great number of factors, for each case there is carrier added. To avoid using surface active sub-
a speciRc optimum contact time, which can only be stances, other techniques with phase dispersion
determined experimentally. A shorter than optimum were recently proposed, two of which are illustrated
contact time results in a lower solute recovery, while in Figures 9 and 10.
a longer contact time reduces selectivity and recovery Co-axially placed in the vertical tube is a second
efRciency. If the emulsion is too stable, this causes tube of porous hydrophobic material, e.g. porous
problems related to its break-up in the third process polypropylene. As shown in Figure 9, both internal
step. Irrespective of these drawbacks, the emulsion and annular spaces are Rlled with the membrane
liquid membrane technique is most often investigated liquid which, under laterally applied pulsations, par-
and practically applied. tially goes from one space to the other and back. The
 II / MEMBRANE SEPARATIONS / Liquid Membranes
Figure 10 Combination of hollow fibre supported liquid membranes with the emulsion technique in which a nonstabilized phase
R dispersed in phase S emulsion flows inside the hollow fibres.
two aqueous feed and acceptor phases are fed into the
top of the central and annular space, respectively, as
droplets of about 1 mm diameter. The porous Rlter
tube does not allow intermixing of the droplets of the
two aqueous phases. The aqueous droplets should
be small enough to guarantee sufRcient residence time
of the corresponding phase in the contractor,
but not too small that it penetrates into the other
compartment.
The second arrangement avoiding the use of sur-
face active substances is shown in Figure 10. The
technique is a combination of hollow Rbre and emul-
sion liquid membrane techniques without using an
emulsiRer for dispersion stabilization. The accep-
tor/membrane-phase emulsion Sows in the lumen of
porous capillaries wetted by the membrane liquid,
Rlling their pores. Evidently, no intense mass transfer
is possible with this technique, irrespective of the
continuous wash-out of the membrane liquid by the
acceptor solution dispersed in it. This drawback is,
however, again compensated for by the great number
of hollow Rbres used and by the recirculation of the
intracapillary dispersion.
Application Areas
The liquid membrane processes described above are
in principle highly efRcient chemical pumps selective-
ly separating and concentrating valuable solutes.
These processes have potential applications in a num-
1747
ber of industrial areas, e.g. hydrometallurgy, elec-
troplating and galvanic technologies, chemical and
pharmaceutical industries. One of the most promising
applications is in biotechnology, where pertraction,
can be integrated with the basic bioprocess in order to
increase process efRciency.
A very attractive feature of pertraction processes is
their low investment, and in particular, their opera-
tional costs. Being a membrane operation, the separ-
ation does not involve phase transitions and therefore
power consumption is very low. However, unlike
solid membrane separations, the costs of lost mem-
brane liquid and the puriRcation of treated solutions
sometimes required additionally contribute to the
process costs.
The Further Reading section lists titles containing
more information on various pertraction systems
studied in the last 25 years.
See also: I/Membrane Separations. II / Flotation: Flota-
tion Cell Design: Application of Fundamental Principles.
Further Reading
Araki T and Tsukube H (eds) (1991) Liquid Membranes:
Chemical Applications. Boca Raton, FL: CRC Press.
Bartsch RA and Douglas Way J (eds) (1996). Chemical
Separations with Liquid Membranes. New York:
American Chemica.
1748 II / MEMBRANE SEPARATIONS / Membrane Bioseparations
Boyadzhiev L (1990) Liquid pertraction or liquid mem-
branes } state of the art. Separation Science and Techno-
logy 25: 187.
Boyadzhiev L and Lazarova Z (1994) Liquid membranes
(liquid pertraction). In: Noble RD and Stren SA (eds)
Membrane Separation Technology. Principles and
Applications, pp. 283}352. Amsterdam: Elsevier.
Drioli E and Nakagaki M (1986) Membranes and Mem-
brane Processes. New York: Plenum.
Ho WSW and Sirkar KK (eds) (1993) Membrane Hand-
book. New York: Van Nostrand Reinhold.
Membrane Bioseparations
A. L. Zydney, University of Delaware, Newark, DE, USA
Copyright ^ 2000 Academic Press
Membrane processes are particularly well suited to
the separation and puriRcation of biological mol-
ecules since they operate at relatively low tempera-
tures and pressures and involve no phase changes or
chemical additives. Thus, these processes cause mini-
mal denaturation, deactivation and/or degradation of
highly labile biological cells or macromolecules. Al-
though essentially all membrane processes (Figure 1)
have been used for bioseparations, the greatest inter-
est has been in the application of the pressure-driven
processes of ultraRltration (UF) and microRltration
(MF). UltraRltration membranes have pore sizes be-
tween 1 and 50 nm and are used for protein concen-
tration, buffer exchange, desalting, clariRcation of
antibiotics and virus clearance. There is also growing
interest in the use of ultraRltration for protein puriR-
cation using high performance tangential Sow Rltra-
tion (HPTFF). MicroRltration membranes have a
pore size between 0.05 and 10
m and are thus used
Figure 1 Classification of pressure-driven membrane processes showing typical bioprocessing applications.
Li NN (1971) Permeation through liquid surfactant
membranes. American Institute of Chemical Engineers
Journal 17: 459.
Noble RD and Douglas Way J (eds) (1996) Liquid Mem-
branes. Theory and Application. American Chemical
Society Symposium Series no 347. Washington, DC:
American Chemistry Society .
Zhang R (ed.) (1984) Separation Techniques by
Liquid Membranes (in Chinese). Nanchang: Jiangxi
Renmin.
for initial clariRcation of protein solutions, cell
harvesting and sterile Rltration. In addition, ultraRl-
tration and microRltration of blood are used for the
treatment of a variety of metabolic and immunolo-
gical disorders.
The development of membrane processes for bio-
separations is very similar to the design of membrane
systems for nonbiological applications. However,
there are some important differences including:
1. increased concerns about deactivation or de-
naturation of biological molecules and cells
2. very high value (on a per unit mass basis) of most
biological products (particularly recombinant
therapeutic proteins)
3. tendency of biological macromolecules and cells
to cause signiRcant fouling of both ultraRltration
and microRltration membranes
4. critical importance of validation and integrity test-
ing in bioprocessing applications
This article provides a brief review of the historical
development of membrane systems for biosepara-
tions. This is followed by a general discussion of the
 II / MEMBRANE SEPARATIONS / Membrane Bioseparations
underlying principles governing the design of ultraRl-
tration and microRltration systems, with particular
emphasis on those factors that are most signiRcant for
bioseparations. The reader is referred to the Encyclo-
pedia articles on Membrane Separations } MicroRl-
tration and Membrane Separations } UltraRltration
for more detailed discussions of these membrane
technologies.
Historical Development
The Rrst mention of the process now known as ultra-
Rltration appears to have been in an 1856 study by
Schmidt on the Rltration of protein and gum arabic
through animal membranes. Thus, the idea of using
ultraRltration for bioseparations dates back well over
100 years. Bechhold coined the term ultraRltration in
1906 during a systematic study of the behaviour of
different pore size collodion membranes made by
impregnating Rlter paper with acetic acid and cellu-
lose nitrate. Zsigmondy obtained one of the Rrst
patents in membrane technology in 1922 for the prep-
aration of Sat collodion membranes from ether}
alcohol solutions. The Rrst efforts to develop micro-
porous membranes in the USA were motivated by the
need for rapid detection and analysis of biological
warfare agents. This technology was subsequently
transferred to the Lovell Chemical Company, which
ultimately led to the establishment of Millipore Cor-
poration.
The early historical development of ultraRltration
and microRltration is described in an excellent review
article by Ferry in 1936. The primary applications of
membrane technology in the early 1900s were for
a variety of biological, analytical and bacteriological
assays. Ferry also described the use of membranes for
enzyme concentration, analysis of bacteriophages
and viruses, blood ultraRltration to prepare cell- and
protein-free ultraRltrates, sterile Rltration of bio-
logical solutions and the partial separation of al-
bumin from globulins in blood serum. All of these
bioseparations remain areas of active commercial in-
terest even today.
Although many of the potential uses of membrane
systems in bioprocessing were identiRed well over 60
years ago, the collodion (cellulose nitrate) mem-
branes available at that time had inadequate chem-
ical, mechanical and mass transport properties for the
effective use of ultraRltration on an industrial scale.
The key breakthrough was the development of the
asymmetric cellulose acetate reverse osmosis mem-
brane by Loeb and Sourirajan in the early 1960s and
the subsequent extension of this technique to produce
asymmetric ultraRltration membranes. These asym-
metric membranes have a very thin skin (approxim-
1749
ately 0.5
m thick), which provides the membrane
with its selectivity, and a more macroporous sub-
structure, which provides the required mechanical
and structural integrity. The thin skin results in much
higher permeation rates than are obtainable with
homogeneous membranes, signiRcantly reducing the
required membrane area and/or process time.
UltraRltration is now used throughout the down-
stream separation process for the puriRcation of
therapeutic recombinant proteins, blood compo-
nents, natural protein products and industrial
enzymes. SpeciRc applications include protein con-
centration (i.e. volume reduction), desalting and buf-
fer exchange, all of which are used to condition the
product prior to, or immediately after, other separ-
ation processes or as part of the Rnal product formu-
lation. In addition, ultraRltration is used extensively
for the clariRcation of antibiotics, amino acids and
other small biological molecules. Recent work has
demonstrated that ultraRltration membranes are also
capable of effecting protein}protein separations using
a process known as HPTFF. MicroRltration mem-
branes are used for cell harvesting, initial clariRcation
of cell culture media and fermentation broths, and for
sterile Rltration of products that are directly added to
pre-sterilized containers. Sterile Rlters are also used to
remove bacteria and particles from feedstock solu-
tions and to reduce the overall bioburden in processes
where the product will be subjected to a terminal
sterilization step. Virus removal membranes are used
as part of the overall viral clearance required for the
production of therapeutic proteins and blood prod-
ucts. Virus Rlters can also provide a protective barrier
for bioreactors through the Rltration of media and
buffer solutions.
Ultra\ltration and Micro\ltration
Principles
Membrane Selection
Membrane selection should start with the choice of
a high quality vendor since robustness, reliability and
reproducibility of manufacturing operations are of
paramount importance in most bioprocessing ap-
plications. Consistent membrane and device charac-
teristics can be as important to product quality, yield
and economics as the inherent differences between
various membranes and devices. Cellulosic mem-
branes are attractive for many bioprocessing applica-
tions because of their low protein adsorption and low
fouling characteristics. Synthetic polymers (e.g. poly-
sulfone and polyvinylidene Suoride) are also attract-
ive due to their greater chemical and mechanical
stability. These polymers are often surface-treated to
1750 II / MEMBRANE SEPARATIONS / Membrane Bioseparations
render them more hydrophilic to reduce protein ad-
sorption and fouling. Membranes used for sterile Rl-
tration must be steam-sterilizable, have minimal par-
ticle shedding, low extractables and must pass United
States Pharmacopoeia (USP) Class VI toxicity testing.
Most manufacturers rate ultraRltration membranes
by their nominal molecular weight cutoff, which is
deRned as the molecular weight of a solute with
a particular retention coefRcient:
R"1!Cfiltrate/Cfeed [1]
where Cfiltrate and Cfeed are the solute concentrations in
the Rltrate solution and feed stream, respectively.
Data are typically obtained with a range of model
proteins or with polydisperse dextrans. Unfortun-
ately, the procedures used for assigning molecular
weight cutoffs, including the choice of solutes, the
speciRc buffer and Sow conditions, and the chosen
retention value (usually R"0.9) vary widely
throughout the industry. In addition, ultraRltration
systems used in bioprocessing generally require pro-
tein retention of at least 99%, and often as high as
99.9%, to minimize loss of high value products
through the membrane. Data obtained with solutes
having R"0.9 are often of little value in determining
whether a given membrane can provide these high
levels of protein retention due to differences in the
details of the pore size distributions.
MicroRltration membranes are typically rated by
their pore size or their particle retention character-
istics using the log reduction value (LRV), deRned as
the logarithm (base 10) of the ratio of the particle, cell
or virus concentration in the feed to that in the Rltrate
solution. Sterilizing-grade (0.2
m pore size) Rlters
are currently deRned by the Health Industry Manu-
facturing Association (HIMA) as a Rlter which pro-
duces a sterile Rltrate when challenged by 107 colony-
forming units of Brevundimonas diminuta (formerly
classiRed as Pseudomonas diminuta) per cm2 of mem-
brane area. This challenge uses the smallest bacteria
at a concentration that exposes essentially every pore
to the microorganisms. Sterile Rlters are often
thought of as operating via a purely size-based (siev-
Figure 2 Comparison of (A) dead-end and (B) cross-flow configurations.
ing) mechanism, although bacteria can also be re-
moved by adsorption on to the membrane surface.
The chemical compatibility of the membrane needs
to be veriRed with the feed, regeneration chemicals
and storage solutions. Sodium hypochlorite (NaOCl)
is used most extensively for chemical disinfection of
membrane systems in bioprocessing applications.
Many membrane systems are designed for steam-in-
place (SIP) sterilization, with the entire unit exposed
to Sowing steam as part of the completely assembled
Rltration system. Minimum requirements for an
effective steam sterilization are 15 min exposure to
steam at 1213C and 1 atm pressure. Polysulfone
membranes tend to have broader chemical and ther-
mal stability than cellulosic membranes but also re-
quire harsher chemical treatment for regeneration
due to their greater fouling characteristics. Inorganic
(ceramic) membranes have the greatest chemical
compatibility, but they are much more expensive than
polymeric membranes. The mechanical strength of
the membrane is important since reverse-pressure
spikes can cause membrane delamination and cata-
strophic yield loss.
Module Design
Dead-end, or normal-Sow, Rltration (Figure 2A) is
used primarily for laboratory-scale separations and
for systems in which the retained species are present
at very low concentration. For example, dead-end
microRltration cartridges are used extensively for
sterile Rltration since the retained bacteria are present
at very low concentration. Similar modules can be
employed for virus removal applications. Almost all
large scale commercial ultraRltration devices use tan-
gential Sow Rltration, also referred to as a cross-Sow
conRguration, in which the feed Sow is parallel to the
membrane and thus perpendicular to the Rltrate Sow
(Figure 2B). This allows retained species to be swept
along the membrane surface and out of the device
exit, signiRcantly increasing the process Sux com-
pared to that obtained with dead-end operation.
A number of tangential Sow modules have been
developed for ultraRltration and microRltration
 II / MEMBRANE SEPARATIONS / Membrane Bioseparations 1751

Table 1 Comparison of different module configurations

Module configuration Channel spacing Packing density Energy costs Particulate plugging Ease of cleaning
(cm) (m2 m\3) (pumping)

Flat sheet 0.03}0.1 300 Moderate Moderate Good

Hollow fibre 0.02}0.25 1200 Low High Fair
Tubular 1.0}2.5 60 High Low Excellent
Spiral wound 0.03}0.1 600 Low Very high Poor to fair

processes, differing primarily in the size and shape of
the feed and Rltrate Sow channels. Table 1 provides
a general summary of the physical characteristics of
the most common modules. Detailed descriptions of
these modules are available elsewhere.
The small channel spacing in Sat-sheet, hollow-
Rbre and spiral-wound modules provides high mem-
brane-packing densities. In addition, these modules
have low hold-up volumes, which facilitates the re-
covery of high value products. The screens used to
deRne the Sow path in spiral-wound modules and
many Sat-sheet cassettes are susceptible to particle
plugging and this may make cleaning more difRcult.
Hollow-Rbre membranes are self-supporting, so they
can often be cleaned by simple backSushing. The
large-bore tubular membranes can be cleaned by both
physical and chemical methods. However, these mod-
ules operate in the turbulent Sow regime which can
cause cell lysis, protein denaturation or aggregation.
A variety of enhanced mass transfer modules which
exploit Sow instabilities have also been developed for
bioprocessing applications. Rotating cylinder mod-
ules which induce Taylor vortices have very high
mass transfer rates, although there are concerns
about the moving parts. Another attractive approach
is to use helically coiled hollow Rbres wrapped
around a central core to induce Dean vortices.
The fed-batch conRguration utilizes an additional
tank to feed into the recycle tank (Figure 3). Fed-
batch conRgurations are commonly used to obtain
high concentration factors and to provide well-
mixed, low-hold-up, retentate reservoirs. These sys-
tems also provide Sexibility for use in multiple pro-
cesses. The disadvantages of the fed-batch system
include greater process time and greater number of
passes of the retentate through the pumps and valves
in the recycle line. The latter can lead to excessive cell
lysis, protein denaturation or aggregation.
DiaRltration is commonly used for buffer exchange
(for products in the retentate) and to enhance yield
(for products in the Rltrate). The diaRltration system
looks similar to the fed-batch conRguration shown in
Figure 3 except that the feed tank contains a buffer
solution which is added to the recycle tank. The most
common approach is constant retentate volume dia-
Rltration in which the buffer is added at the same rate
as Rltrate removed.
The yield and puriRcation obtained in ultraRltra-
tion and microRltration processes can be evaluated
from simple mass balances on the product and im-
purity assuming constant rejection coefRcients. The
Rnal product concentration (CF) at the end of a batch
concentration process is given as:



CF V0 1
\S
" [2]
Process Con\gurations C0 VF
Protein concentration can be carried out using either
batch or fed-batch operation (Figure 3). In a batch where VF is the Rnal retentate volume, V0 is the initial
process, the entire feed volume is contained within retentate volume and S is the product sieving coefRc-
the recycle tank. Tank design is critically important to ient (equal to one minus the rejection coefRcient). The
ensure adequate mixing while avoiding air entrain- analogous expression for a fed-batch process is:
ment and excessive foaming. Batch operation uses

 

a minimum of hardware and allows simple manual or CF 1 1 V0
automatic control. The Sux rates are also higher in " # 1! exp !S !1 [3]
C0 S S VF
batch processes since the bulk concentration follows
a more dilute path in going from initial to Rnal con-
The Rnal concentration after a constant retentate
centration. Disadvantages of the batch conRguration
volume diaRltration is:
include less Sexibility in using the same system for
multiple processes, greater difRculty in designing
a well-mixed system, and difRculties in obtaining CF
"exp(!SN) [4]
high concentration factors. C0
1752 II / MEMBRANE SEPARATIONS / Membrane Bioseparations
Figure 3 Comparison of (A) batch and (B) fed-batch processes for protein concentration.
where the number of diavolumes (N) is given by:
N"VD/V
where VD is the diaRltration buffer volume. Even very
small sieving coefRcients may result in substantial
product loss when a large number of diavolumes are
required in diaRltration processes. For example,
a diaRltration process with a product sieving coefRc-
ient of S"0.01 will result in a 10% product loss
during a 10 diavolume buffer exchange.
Concentration Polarization
One of the critical factors determining the overall
performance of tangential Sow Rltration devices is the
rate of solute/particle transport in the bulk solution
adjacent to the membrane. The Rltrate Sow causes an
accumulation of partially (or completely) retained
components at the upstream surface of the mem-
brane, a phenomenon referred to as concentration
polarization. The concentration thus varies from its
maximum value at the membrane surface (Cw) to its
[5] bulk value (Cb) over the thickness of the concentra-
tion boundary layer (
). Most analyses of concentra-
tion polarization have employed the simple stagnant
Rlm model in which:


Cw!Cf
J"k ln [6]
Cb!Cf
where J is the Rltrate Sux (typically in L m\2 h\1)
and k is the solute mass transfer coefRcient in the
particular membrane device. The accumulation of
particles/solutes at the membrane surface increases
the overall resistance to Rltrate Sow through the
formation of a particle cake or gel layer and it can
reduce the effective pressure driving force through the
osmotic pressure of the retained solutes. At high
transmembrane pressures, the wall concentration ap-
proaches a maximum value determined by the close-
packed concentration of the particles or cells, the
 II / MEMBRANE SEPARATIONS / Membrane Bioseparations
protein solubility limit or the concentration at which
the osmotic pressure of the retained solutes is essen-
tially equal to the applied transmembrane pressure.
The net result is that the Sux attains a nearly constant
pressure-independent value that increases with de-
creasing bulk concentration and increasing feed Sow
rate. The dependence on feed Sow rate is determined
by the module characteristics: approximately 1/3
power for laminar Sow in hollow Rbres and open
channels, 1/2 power for screened channels, and 0.8
power for turbulent Sow in tubular modules. The
dependence on feed Sow rate for cellular suspensions is
typically greater than that for protein solutions due to
shear-induced particle diffusion and inertial lift effects.
Process Control
UltraRltration and microRltration processes have tra-
ditionally been performed at constant transmem-
brane pressure. Constant-pressure processes are very
simple to control. The feed rate is ramped up to the
set point and the retentate valve is then partially
closed to obtain the desired transmembrane pressure.
The transmembrane pressure should be gradually in-
creased to minimize fouling. In some applications it
may not be possible to maintain constant transmem-
brane pressure without severe reductions in Rltrate
Sux over the course of the process due to membrane
fouling. This is particularly true for cell microRltra-
tion where the high initial Sux leads to very rapid
deposition of cells and cell debris on the membrane
surface. Several studies have shown that higher over-
all throughput can often be obtained in these applica-
tions by operating at constant Rltrate Sux. The Sux is
controlled by regulating the retentate pressure control
valve or by using a pump on the Rltrate line.
A third method of process control that is very
attractive for bioprocessing applications is to vary the
Rltrate Sux so that the wall concentration of retained
species (evaluated from eqn [6]) remains constant
during the process. Control is performed using a pro-
portional-integral-derivative (PID) loop that mea-
sures Sux and controls the transmembrane pressure
or Rltrate Sow rate to maintain a constant wall con-
centration throughout the process. The beneRts of
constant Cw control are that product yield is maxi-
mized, product quality is ensured, membrane area is
minimized and process time is consistent and inde-
pendent of variations in membrane permeability.
High Performance Tangential Flow
Filtration
UltraRltration and microRltration have traditionally
been limited to separating species that differ in size by
1753
at least 10-fold. In contrast, HPTFF enables the sep-
aration of solutes without limit to their relative size.
HPTFF is able to obtain the high selectivity required
for effective protein puriRcation by exploiting several
recent developments. Firstly, HPTFF is operated in
the pressure-dependent regime, with the Rltrate Sux
and device Suid mechanics chosen to minimize foul-
ing and exploit the effects of concentration polariza-
tion. Since optimal separation in HPTFF is obtained
at a speciRc Rltrate Sux, the membrane module
should be designed to maintain a nearly uniform Sux
and transmembrane pressure throughout the module.
This can be done using a co-current Rltrate Sow to
balance the feed-side pressure drop through the mod-
ule. Secondly, the buffer pH and ionic strength are
adjusted to maximize differences in the effective vol-
ume of the different species. The effective volume of
a charged protein (as determined by size exclusion
chromatography) accounts for the presence of the
diffuse electrical double layer surrounding the pro-
tein. Protein transmission through the membrane can
be reduced by increasing the effective protein volume,
e.g. by increasing the net protein charge (by adjusting
the pH) or by increasing the double-layer thickness
(by reducing the solution ionic strength). Thirdly, the
electrical charge of the membrane is chosen to in-
crease the electrostatic exclusion of all species with
like charge. Thus, a positively charged membrane will
provide much greater rejection of a positively charged
protein than will a negatively charged membrane of
the same pore size. Fourthly, protein separations in
HPTFF are accomplished using a diaRltration mode
to wash the impurity (or product) out of the retentate.
The diaRltration maintains an appropriate protein
concentration in the retentate throughout the separ-
ation, and it allows one to obtain puriRcation factors
for products collected in the retentate that are much
greater than the membrane selectivity due to the con-
tinual removal of impurities in the Rltrate.
Although HPTFF is still a new membrane techno-
logy, a number of recent studies have clearly demon-
strated the potential of this separation technique.
Several of these results are summarized in Table 2.
PuriRcation factors for the separation of bovine
serum albumin (BSA) from an antigen-binding frag-
ment (Fab) were greater than 800-fold with either
protein collected in the retentate depending upon the
choice of solution pH and membrane surface charge.
BSA and haemoglobin have essentially identical mo-
lecular weight but different surface charge character-
istics. In this case, operation at pH 7 caused a strong
electrostatic exclusion of the negatively charged BSA
from the negatively charged membrane. The separ-
ation of BSA monomer and dimer occurs primarily
because of the difference in protein size, with the
1754 II / MEMBRANE SEPARATIONS / Membrane Bioseparations
Table 2 Purification factors and yields for HPTFF processesa
Product (MW) Impurity (MW) Purification Yield
factor
BSA (68 000) Fab (45 000) 990 94%
Fab (45 000) BSA (68 000) 830 69%
BSA (68 000) Hb (67 000) 100 68%
IgG (155 000) BSA (69 000) 30 84%
BSA (68 000) BSA dimer (136 000) 9 86%
a
BSA, Bovine serum albumin; Fab, antigen-binding fragment from
recombinant DNA antibody; Hb, bovine haemoglobin; IgG, human
immunoglobin.
smaller monomer collected in the Rltrate. However,
electrostatic interactions are also important in this
system due to the combined effects of size and charge
on protein transmission and to possible differences in
the charge}pH proRles for the monomer and dimer.
Validation and Integrity Testing
Membrane systems used in bioprocessing applica-
tions need to be validated to demonstrate consistent
puriRcation and yield with minimal alteration in the
properties of the product. Food and Drug Adminis-
tration regulations provide speciRc guidelines for vali-
dation of sterile Rlters and virus removal membranes.
Validation should always be performed at the same
pH, ionic strength and chemical environment as used
in the actual process to ensure equivalent physical
and chemical characteristics of the product and impu-
rities. Viral clearance studies are typically performed
by spiking high titre infectious viruses (with different
physical characteristics) into scaled-down production
systems.
Integrity testing is critical for all sterile and viral
Rlters to ensure that the system operates at the re-
quired level of performance. Integrity tests should be
performed both prior to, and immediately after, the
actual process wherever possible. Integrity tests per-
formed prior to Rltration must not affect the sterility
of the connections downstream of the Rlter. The real
test for the sterile Rlter would be to challenge with B.
diminuta, but the Rlter could not be used after this
test. Thus, a number of surrogate nondestructive in-
tegrity tests have been developed. The industry stan-
dards are forward Sow, pressure decay and bubble
point. Each of these tests is based on the displacement
of a Suid from the pores by a second Suid (or gas),
with the rate of displacement providing a measure of
the membrane pore size characteristics. The gas or
intrusion liquid expels the wetting liquid out of the
pore when the feed pressure exceeds the capillary
force within the pore. The bubble point is deRned as
the pressure at which the pore is Rrst intruded by the
gas. The bubble point for sterilizing grade Rlters can
be correlated to the LRV of B. diminuta. Filters with
water bubble points of 55 psi or greater typically
yield the necessary LRV to be qualiRed as sterilizing-
grade Rlters. In the forward Sow test, one measures
the total gas Sow rate through the wetted membrane
at a Rxed pressure. High Sow rates indicate the pres-
ence of pressure-driven Sow through gas-intruded
(large) pores. The pressure decay test is performed in
a similar fashion, with the gas Sow calculated from
the rate of pressure decay. A variety of automated
integrity test instruments have been developed by the
different membrane manufacturers.
Bubble point tests with water-wetted membranes
cannot be used to verify virus Rlter performance since
the bubble points for these small pore size membranes
would exceed the membrane pressure limits. Air dif-
fusion and bubble point tests can be performed on
these membranes using wetting Suids having lower
surface tension (e.g. isopropyl alcohol). Liquid intru-
sion tests using two immiscible Suids (e.g. solutions
of a sulfate salt and polyethyleneglycol) have been
developed as integrity tests for virus Rlters and
HPTFF membranes.
Summary
Membrane processes should continue to be of critical
importance in bioprocessing applications, facilitating
the cost-effective production of a wide range of biolo-
gical products. UltraRltration has become the primary
means for protein concentration and buffer exchange
in the production of therapeutic proteins and indus-
trial enzymes. Sterile Rltration is used throughout the
bioprocessing industry, and viral Rltration is of grow-
ing importance in the production of blood products
and therapeutic recombinant proteins.
The future is likely to see the continued develop-
ment of high performance tangential Sow Rltration as
a viable alternative to existing separation technolo-
gies for protein puriRcation. There is also growing
interest in the development of membrane systems for
the preparation of enantiomerically enriched anti-
biotics, nutraceuticals and pharmaceuticals. These
membrane systems use chiral ligands to separate ra-
cemic mixtures or they employ immobilized enzymes
for direct production of single enantiomers in mem-
brane reactors. AfRnity membrane systems are also
being actively pursued as alternatives to standard
chromatographic resins for a range of adsorptive bio-
separations. In this case, the membrane provides an
attractive high surface area support with minimal
diffusional mass transfer resistance. New advances
in membrane materials, modules and processes
 II / MEMBRANE SEPARATIONS / Membrane Preparation 1755

should lead to continued development of membrane
systems for bioseparations.
See also: II / Membrane Separations: Microfiltration;
Ultrafiltration.
Further Reading
Belfort G, Davis RH and Zydney AL (1994) The behavior
of suspensions and macromolecular solutions in cross-
Sow microRltration. Journal of Membrane Science 96: 1.
Blatt WF, Dravid A, Michaels AS and Nelsen L (1970)
Solute polarization and cake formation in membrane
ultraRltration. Causes, consequences, and control tech-
niques. In: Flinn JE (ed.) Membrane Science and Techno-
logy, pp. 47}97. New York: Plenum Press.
Cheryan M (1997) UltraTltration and MicroTltration
Handbook. Lancaster, PA: Technomic.
Ferry JD (1936) UltraRlter membranes and ultraRltration.
Chemical Reviews 18: 373.
Ho WSW and Sirkar KK (eds) (1992) Membrane Hand-
book. New York: Chapman & Hall.
Lonsdale HK (1982) The growth of membrane technology.
Journal of Membrane Science 10: 81.
McGregor WC (ed.) (1986) Membrane Separations in
Biotechnology. New York: Marcel Dekker.
van Reis R and Zydney AL (1999) Protein ultraRltration.
In: Flickinger MC and Drew SW (eds) Encyclopedia of
Bioprocess Technology: Fermentation, Biocatalysis, and
Bioseparation, pp. 2197}2214. New York: John Wiley.
Zeman LJ and Zydney AL (1996) MicroTltration and
UltraTltration: Principles and Applications. New York:
Marcel Dekker.

Membrane Preparation
I. Pinnau, Membrane Technology and Research,
Inc., Menlo Park, CA, USA
Copyright ^ 2000 Academic Press
Background
A membrane (Latin, membrana, skin) is a thin barrier
that permits selective mass transport. Between 1850
and 1900, membranes were used to derive basic phys-
ical principles for gas and liquid transport across
a barrier material (see the work of Mitchell, Fick and
Graham). In these early studies it was already recog-
nized that membranes could be used to separate Suid
mixtures. Membranes used at that time included
dense Rlms of nitrocellulose, natural rubber, and pal-
ladium. The Rrst commercial synthetic membranes
were developed by Bachmann and Zsigmondy in the
early 1920s in Germany. These microporous nitrocel-
lulose membranes were used for laboratory purposes
as well as for the fast detection of bacteria in drinking
water. However, until the early 1960s, membranes
were not used in any industrial separation process.
The major event that ultimately resulted in the wide-
spread use of membranes for separations was the
development of integrally-skinned, asymmetric cellu-
lose acetate membranes for water desalination, by
Loeb and Sourirajan at UCLA from 1958 to 1960.
During a time span of only 10 years, a wide variety of
membranes was developed for reverse osmosis, ultra-
Rltration and microRltration applications based on
modiRcations of the original membrane preparation
method employed by Loeb and Sourirajan. Further-

Table 1 Major milestones in the development of membranes for industrial separations

Period of years Advances

1900}1920 Development of first ultrafiltration and microfiltration membranes made from nitrocellulose
(Bechhold, Zsigmondy, Bachmann).
1920}1940 Empirical studies on formation of phase inversion membranes (Bjerrum, Manegold, Elford).
Development of cellulose acetate ultrafiltration membranes (Dobry, Duclaux).
1940}1960 Development of integrally-skinned asymmetric cellulose acetate membranes for water desalination
by reverse osmosis (Loeb and Sourirajan).
1960}1970 Commercialization of reverse osmosis, ultrafiltration, microfiltration, and dialysis membranes.
1970}1980 Development of thin-film composite membranes made by interfacial polymerization (Cadotte, Riley).
Cellulose acetate gas separation membranes (Schell).
1980}1990 Commercialization of gas separation and pervaporation membranes (Henis and Tripodi, Tusel,
BruK schke).
1990}2000 Development of inorganic membranes for gas separation and pervaporation.
The next millennium Commercialization of inorganic membranes.
1756 II / MEMBRANE SEPARATIONS / Membrane Preparation
more, methods of efRciently packaging membranes
into modules (plate-and-frame, spiral-wound, and
hollow-Rbre) were developed during this period.
Around 1980, the use of membranes for separations
was established as a unit operation process in the
chemical process industry. Further optimization of
membrane preparation and modiRcation methods
from 1980 to 1990 made membrane separations
competitive in gas separation and liquid separation
applications. The most important production
methods that resulted in the commercial use of syn-
thetic membranes are listed in Table 1. Recent atten-
tion has been directed towards the development of
inorganic membranes. Optimized inorganic mem-
branes can have signiRcantly better separation
properties compared to state-of-the-art polymeric
membranes. However, currently the main limitations
for large-scale commercialization of inorganic mem-
branes are their poor mechanical strength (brittle-
ness) and extremely high manufacturing costs.
Membrane Types
Membranes can be distinguished based on their (i)
geometry, (ii) bulk structure, (iii) production method,
(iv) separation regime, and (v) application, as shown
in Figure 1. Most commonly, membranes are pro-
duced in Sat-sheet or tubular (hollow-Rbre) ge-
ometry. Flat-sheet membranes are either packaged in
plate-and-frame or spiral-wound modules, whereas
tubular membranes are packaged in hollow-Rbre
modules. The choice of the optimum membrane and
module type depends on a wide variety of process
Figure 1 Classification scheme of synthetic membranes based on their geometry, bulk structure, production method, separation
regime, and application.
speciRc conditions. Although hollow-Rbre modules
offer the highest membrane area per module volume
ratio, plate-and-frame and spiral-wound modules are
also commonly used for large-scale separations be-
cause of their better control of Suid dynamics.
Membranes either have a symmetric (isotropic) or
an asymmetric (anisotropic) structure. The structure
of a symmetric membrane is uniform throughout its
entire thickness, whereas asymmetric membranes
have a gradient in structure. The Sux of a Suid
through a symmetric membrane is typically relatively
low, as the entire membrane thickness contributes
a resistance to mass transport. Asymmetric mem-
branes consist of two structural elements, that is,
a thin, selective layer and a microporous substruc-
ture. Typically, the bulk structure (99#%) of an
asymmetric membrane is highly porous and provides
only mechanical strength. Separation of a Suid mix-
ture in an asymmetric membrane is performed in
a very thin surface layer, which is typically of the
order of 0.1}0.5
m thick. The most common sym-
metric and asymmetric membrane types are shown in
Figure 2.
Ideal Membranes for Separations
Membranes can be fabricated from a wide variety of
organic (e.g. polymers, liquids) or inorganic (e.g. car-
bons, zeolites, etc.) materials. Currently, most com-
mercial membranes are made from polymers. The
properties of a membrane are controlled by the mem-
brane material and the membrane structure. To be
useful in an industrial separation process, a
 II / MEMBRANE SEPARATIONS / Membrane Preparation
Figure 2 Schematic representation of symmetric and asymmetric membrane structures.
membrane must exhibit at least the following charac-
teristics: (i) high Sux, (ii) high selectivity (rejection),
(iii) mechanical stability, (iv) tolerance to all feed
stream components (fouling resistance), (v) tolerance
to temperature variations, (vi) manufacturing repro-
ducibility, (vii) low manufacturing cost, and (viii)
ability to be packaged into high surface area modules.
Of the above requirements, Sux and selectivity (rejec-
tion) determine the selective mass transport proper-
ties of a membrane. The higher the Sux of a mem-
brane at a given driving force, the lower is the mem-
brane area required for a given feed Sow rate, and,
therefore, the lower are the capital costs of a mem-
brane system. The selectivity determines the extent of
separation, and, therefore, the purity of the desired
product.
Typically, porous membranes are used in dialysis,
ultraRltration, and microRltration applications. Ideal
porous membranes have high porosity and a narrow
pore size distribution. Membranes having a dense,
selective layer are applied in reverse osmosis, per-
vaporation, and gas separation processes. Permeation
through dense membranes occurs by a solution/diffu-
sion mechanism. Ideal dense membranes should have
a very thin selective layer, because Sux is inversely
proportional to the membrane thickness. In addition,
the thin separating layer must be pinhole-free, because
even a very small area fraction of defects in the mem-
brane can cause a signiRcant decrease in selectivity.
1757
Polymeric Membranes
Currently, most commercial membranes are made
from polymers. Polymeric membranes can be fab-
ricated by a wide variety of methods and fulRll most
of the requirements of an ideal membrane listed
above. Membranes are made from amorphous as well
as semi-crystalline polymers by solution- or melt-
processes. A list of commonly used polymers for com-
merical membrane separation processes is given in
Table 2.
Symmetric Membranes
Dense Symmetric Membranes
Dense symmetric membranes with thicknesses greater
than 10
m can be made by melt extrusion or solution
casting and subsequent solvent evaporation. Because
the Suxes of Suids through dense polymer Rlms are
very low, this membrane type is rarely used for large-
scale separations. Dense symmetric, ion-exchange
membranes are used in electrodialysis applications
for production of potable water from brackish water.
Porous Symmetric Membranes
Typically, symmetric porous membranes have cylin-
drical, sponge-, web- or slit-like structures, and
can be made by a variety of techniques. The most
1758 II / MEMBRANE SEPARATIONS / Membrane Preparation

Table 2 Polymers used for production of commercial mem- aligned in the direction of orientation. In the second
branes step, slit-like pores, about 200}2500 A> wide, are for-
med between the stacked lamellae by stretching the
Membrane material
membrane in the machine direction. The pore size of
Cellulose regenerated D, UF, MF these membranes is determined by the rate and extent
Cellulose nitrate MF of stretching during the second process step. Com-
Cellulose acetate GS, RO, D, UF, MF mercial membranes made by the extrusion/stretching
Polyamide RO, NF, D, UF, MF
process are available from Hoechst}Celanese (Cel-
Polysulfone G, UF, MF
Poly(ether sulfone) UF, MF gard威) and W.L. Gore (Gore-Tex威).
Polycarbonate GS, D, UF, MF Template leaching is another method of producing
Poly(ether imide) UF, MF symmetric microporous membranes by melt-process-
Poly(2,6-dimethyl-1,4-phenylene oxide) GS ing of a semi-crystalline polymer. In this process,
Polyimide GS
a leachable component, such as a high-boiling par-
Poly(vinylidene fluoride) UF, MF
Polytetrafluoroethylene MF afRn, is uniformly dispersed in a polymer melt. After
Polypropylene MF extrusion and formation of a polymer Rlm, the leach-
Polyacrylonitrile D, UF, MF able component is extracted using a suitable solvent,
Poly(methyl methacrylate) D, UF and a sponge-like, microporous membrane is formed.
Poly(vinyl alcohol) PV
Symmetric porous membranes can also be made by
Polydimethylsiloxane PV, GS
a thermally-induced phase separation process (TIPS
MF"microfiltration; UF"ultrafiltration; NF"nanofiltration; process). In the TIPS process, the membrane structure
D"dialysis; PV"pervaporation; GS"gas separation. is formed by bringing an initially thermodynamically
stable polymer solution to an unstable state by
lowering the process temperature. The change in
important methods for the production of symmetric temperature causes phase separation of the initially
porous membranes are: (i) irradiation, (ii) stretching stable solution into two phases with different com-
of a melt-processed semi-crystalline polymer Rlm, positions. The membrane structure depends primarily
(iii) template leaching, (iv) temperature-induced on the initial polymer concentration and the kinetics
phase separation, and (v) vapour-induced phase of the phase separation process and the local distribu-
separation. tion of the polymer-rich phase at the point of solidiR-
Symmetric membranes with a cylindrical pore cation. A schematic phase diagram for a solution
structure can be produced by an irradiation-etching containing a polymer and a solvent is shown in Fig-
process. These membranes are often referred to as ure 3. The phase diagram is divided into three distinct
nucleation track membranes. In the Rrst step of this regions: (A) stable polymer solution region, (B) meta-
process, a dense polymer Rlm, such as polycarbonate stable or binodal region, and (C) spinodal region.
or polyester, is irradiated with charged particles, Phase separation can occur by two different mecha-
which cause chain scission and leave behind a sensi- nisms, that is, (i) nucleation and growth or (ii)
tized track of damaged polymer molecules. These spinodal decomposition. The quench paths of three
tracks are more susceptible to attack by chemical different polymer-solvent solutions from temperature
agents than the undamaged, base polymer Rlm. In the T1 to temperature T2 are illustrated in Figure 3. After
second step, the Rlm is passed through an etching lowering the initial solution temperature T1 to T2,
medium, typically a sodium hydroxide solution. Dur- solutions A and B will be in the meta-stable region of
ing this process, pores are formed by etching the the phase envelope and phase separation will occur
partially degraded polymer along the nucleation by nucleation and growth. Solution A forms nuclei
tracks. Membranes made by this method have a very with composition p@ (polymer-rich), whereas solu-
uniform pore size. The porosity and pore size of tion B will form nuclei with composition p? (solvent-
nucleation track membranes can be controlled by the rich). At equilibrium, both solutions phase-separate
irradiation time and etching time, respectively. into two phases composed of p? and @p . However,
Membranes with a symmetric slit-like pore struc- the ratio of
-phase to -phase is signiRcantly differ-
ture can be made from semi-crystalline polymers, ent after phase separation of both solutions. Solution
such as polyethylene, polypropylene or polytetra- A will consist of a very small volume fraction of
Suoroethylene, using a melt extrusion/stretching pro- polymer-rich-phase ( @p ) dispersed in a large volume
cess. In the Rrst process step, a highly oriented Rlm is fraction of solvent-rich phase ( p?). As a result of the
formed by melt-extrusion of a semi-crystalline poly- low volume fraction of polymer, solution A will form
mer and re-crystallization under high stress. The crys- a Rne powder of precipitated polymer. On the other
tallites in the semi-crystalline polymer Rlm are then hand, solution B will consist of a small volume
 II / MEMBRANE SEPARATIONS / Membrane Preparation
Figure 3 Schematic diagram of a binary polymer-solvent sys-
tem with an upper critical solution temperature (UCST).
fraction of solvent-rich phase dispersed in a large
volume fraction of polymer-rich phase. The resulting
morphology is a sponge- or foam-like porous struc-
ture. A thermal quench of solution C passes directly
into the unstable region of the phase diagram; there-
fore, phase separation occurs by spinodal decomposi-
tion. Typically, phase separation by spinodal de-
composition leads to an interconnected, porous struc-
ture. The Rnal membrane structure depends not only
on liquid}liquid phase separation phenomena but
also on the kinetics of the thermal quench process and
the distribution of the polymer-rich phase at the point
of solidiRcation. Typically, a rapid quench rate results
in a large fraction of small pores, whereas a slow
quench rate produces fewer, but larger pores.
Symmetric membranes with sponge- or web-like
pore structures can also be made by a vapour-precipi-
tation/evaporation technique. Membranes made by
this method are highly porous and are typically used
in microRltration applications. In its simplest form of
the method, a solution containing polymer, solvents
and non-solvents is cast onto a substrate and is then
exposed to a water-vapour-saturated air stream. The
water vapour induces phase separation in the initially
stable polymer solution. After phase separation, the
solvent and non-solvent components are evaporated
by blowing a hot air stream across the membrane.
The porosity and pore size of this membrane type can
be controlled by: (i) the polymer concentration in
the casting solution and (ii) the composition of
the vapour atmosphere. Low polymer concentration,
high humidity, and the addition of solvent-vapour
to the casting atmosphere lead to membranes with
high porosity and large pore size. Because mem-
branes made by the vapour-precipitation/evaporation
method have an essentially constant polymer concen-
tration proRle throughout the entire membrane thick-
1759
ness at the onset of phase separation, the resulting
membranes are porous and have a fairly symmetric
structure.
Asymmetric Membranes
The most commonly used asymmetric membranes
are: (i) integral-asymmetric with a porous skin layer,
(ii) integral-asymmetric with a dense skin layer, and
(iii) thin-Rlm composite membranes. Integrally-skin-
ned asymmetric membranes are typically made by
a non-solvent induced phase separation process (im-
mersion precipitation) and consist of a thin, selective
layer and a porous substructure. Both skin layer and
substructure are formed simultaneously during the
immersion precipitation process. Porous integral-
asymmetric membranes are applied in dialysis, ultra-
Rltration, and microRltration applications, whereas
integral-asymmetric membranes with a dense skin
layer are used in reverse osmosis and gas separation
applications. Thin-Rlm composite membranes consist
of a thin, selective polymer layer atop a porous sup-
port. In this membrane type, the separation and
mechanical functions are assigned to different layers
in the membrane. This membrane type was originally
developed for reverse osmosis applications; however,
nowadays thin-Rlm composite membranes are also
used in nanoRltration, gas separation, and pervapora-
tion applications.
Integrally-Skinned Asymmetric Membranes
The Rrst integrally-skinned asymmetric membranes
were developed by Loeb and Sourirajan in the early
1960s for seawater desalination by reverse osmosis.
In the original Loeb}Sourirajan technique, thin-skin-
ned (&0.2
m) cellulose acetate membranes were
made by a four-step process: (i) casting of a multi-
component polymer solution, (ii) partial evaporation
of a volatile solvent, (iii) immersion of the nascent
polymer Rlm into a non-solvent (water), and (iv)
thermal annealing of the water-wet membrane. Mem-
branes made by this method had water Suxes orders
of magnitude higher than those of thick, isotropic
cellulose acetate Rlms while maintaining high sodium
chloride rejection ('90%). The Loeb}Sourirajan
method has been modiRed and applied to a wide
variety of polymers other than cellulose acetate. In
fact, the Loeb}Sourirajan process is by far the most
important method for production of commercial
membranes for separations.
In the simplest case, integrally-skinned asymmetric
membranes are made from a binary solution contain-
ing a polymer and a solvent. Upon immersion of the
cast solution into a liquid (typically water), which is
1760 II / MEMBRANE SEPARATIONS / Membrane Preparation
a non-solvent for the polymer but miscible with the
solvent, an asymmetric structure with either a porous
or non-porous skin layer is formed. The structural
gradient in an integrally-skinned asymmetric mem-
brane results from a very steep polymer concentration
Figure 4 (A) Porous bulk structure and (B) skin layer of an
integrally-skinned asymmetric polysulfone membrane made by
the immersion precipitation process.
gradient in the nascent membrane at the onset of
phase separation. The structure of a typical mem-
brane made by immersion precipitation having
a highly porous substructure and a thin skin layer is
shown in Figure 4(A) and 4(B). In the immersion
precipitation process, phase separation can be in-
duced by: (i) solvent evaporation and/or (ii) sol-
vent/non-solvent exchange during the quench step.
Typically, the formation of membranes made by the
immersion precipitation method occurs over a very
short time scale, typically less than a few seconds.
Most commercial membranes made by the immersion
precipitation method are made from multi-compon-
ent solutions containing polymer, solvent(s), and
non-solvent(s) or additives. The porosity, pore size,
and skin layer thickness can be modiRed by the addi-
tion of non-solvents to the casting solution (e.g. alco-
hols, carboxylic acids, surfactants, etc.), inorganic
salts (e.g. LiNO3 or ZnCl2, etc.) or polymers (e.g.
polyvinylpyrrolidone, polyethylene glycol, etc.). Even
very small amounts of these solution additives can
have a signiRcant effect on the membrane structure,
and hence, its separation properties. The structure of
membranes made by immersion precipitation can
also be altered by using multi-component quench
media. For example, the addition of a solvent to the
quench medium results in an increase in the surface
porosity and pore size of the membrane. The
formation of membranes made by the immersion pre-
cipitation process depends on a large number of
material- and process-speciRc parameters including:
E choice of the polymer (molecular weight, molecu-
lar weight distribution)
E choice of the solvent(s) and additives
E composition of the casting solution
E choice of the quench medium
E composition and temperature of the casting atmo-
sphere
E temperature of the casting solution and quench
medium
E evaporation conditions
E casting thickness
E casting or spinning speed
E membrane support material (type of woven or non-
woven)
E drying conditions.
Thin-Film Composite Membranes
Composite membranes consist of at least two struc-
tural elements made from different materials, as
shown in Figure 5. A single-layer composite mem-
brane (5A) consists of a thin, selective layer atop
a microporous support. The support provides only
mechanical strength, whereas the separation is
 II / MEMBRANE SEPARATIONS / Membrane Preparation
Figure 5 Schematic diagram of (A) single-layer and (B) multi-layer thin-film composite membranes.
performed by the thin top-layer. A multi-layer com-
posite membrane (5B) consists of a porous support and
several layers of different materials, each performing
a speciRc function. Thin-Rlm composite membranes
are applied in nanoRltration, reverse osmosis, gas sep-
aration, and pervaporation applications. The selective
layer can be applied by lamination, solution coating,
interfacial polymerization, or plasma polymerization
methods. Compared to integrally-skinned asymmetric
membranes, composite membranes offer several signif-
icant advantages: (i) independent selection of materials
from which the separating layer and the porous sup-
port are formed, (ii) independent preparation of the
separating layer and the porous support membrane,
thereby making it possible to optimize each structural
element, and (iii) very expensive membrane materials
('1000 $/lb) can be used because only a very small
amount of polymer is required for the formation of the
thin separation layer (&1 g polymer/m2 of membrane
for a 1-
m-thick selective layer).
In most cases, porous, ultraRltration-type mem-
branes made by the immersion precipitation process
are used as mechanical support for thin-Rlm com-
posite membranes. Optimum porous supports for
thin-Rlm composite membranes should have the fol-
lowing properties: (i) porous support must be chemic-
ally resistant against the solvent or solvent mixture
from which the thin separating layer is formed and
(ii) the porous support should have a high surface
porosity and small pore size. High surface porosity is
1761
important because the support should not provide
any signiRcant resistance to mass transport in a com-
posite membrane. A small pore size is required for the
deposition of ultrathin, defect-free coatings.
The two most important methods for the com-
mercial production of thin-Rlm composite are based
on interfacial polymerization and solution coating
methods. The Rrst interfacially polymerized thin-Rlm
composite membranes were developed by Cadotte at
the North Star Research Institute and represented
a breakthrough in membrane performance for reverse
osmosis applications. The original interfacial polym-
erization process involved soaking a microporous
polysulfone support in an aqueous solution of a poly-
meric amine and then immersing the amine-impreg-
nated membrane into a solution of a di-isocyanate in
hexane. The membrane was then cross-linked by
heat-treatment at 1103C. The resulting polyurea
membrane had better salt rejection than that of an
integrally-skinned asymmetric cellulose acetate mem-
brane and high water Sux. ModiRcations in the chem-
istry of the original interfacial polymerization
reaction scheme resulted in further improvement in
performance of thin-Rlm composite membranes for
reverse osmosis applications.
The solution-coating method involves deposition
of a dilute polymer solution onto the surface of a por-
ous membrane and subsequent drying of the thin
liquid Rlm. The simplicity of this process is very
attractive for the production of membranes on a
1762 II / MEMBRANE SEPARATIONS / Membrane Preparation
commercial scale. However, it is generally very difR-
cult to produce defect-free thin-Rlm composite mem-
branes with a thickness of less than 1
m by the
solution coating process. These defects are caused by
incomplete coverage of surface pores in the support
membrane after complete evaporation of the solvent.
The difRculty in completely covering surface pores
results from penetration of the coating solution into
the porous support membrane structure. Because cap-
illary forces in the porous membrane tend to pull the
thin liquid polymer solution into the bulk support
membrane, the coating layer can be disrupted easily.
Several methods have been proposed to overcome
problems with the formation of the thin, selective
layer by the solution-coating process. One method is
to use ultrahigh molecular weight polymers for the
formation of the selective layer. An alternative ap-
proach for eliminating defects in the thin selective
coating layer is to fabricate multi-layer composite
membranes. These membranes, shown schematically
in Figure 5B, consist of: (i) a porous support, (ii)
a sealing layer, and (iii) an ultrathin, selective coating.
The function of the sealing layer is to plug the pores in
the support membrane and to provide a smooth sur-
face onto which the thin coating layer can be applied.
In addition, the sealing layer helps in channeling the
permeating components to the surface pores, thereby
rendering the entire surface area available for mass
transport. The sealing layer should not provide a sig-
niRcant mass transport resistance in a multi-layer
composite membrane. Hence, the sealing layer mater-
ial should be signiRcantly more permeable than the
thin, selective top-layer.
Membrane Modi\cation Methods
The development of high-performance polymeric
membranes involves the selection of a suitable mem-
brane material and the formation of this material into
a desired membrane structure. However, it is often
necessary to modify the membrane material or the
structure to enhance the overall performance of the
membrane. Generally, the objectives for modiRcation
of pre-formed membranes are: (i) increasing Sux
and/or selectivity and (ii) increasing chemical resistance
(solvent resistance, swelling, or fouling resistance).
The Rrst reported membrane modiRcation method
involved annealing of porous membranes by heat-
treatment. Zsigmondy and Bachmann demonstrated
in the early 1920s that the pore size of pre-formed
nitrocellulose membranes could be decreased with
a hot water or steam treatment. Loeb and Sourirajan
used the same method to improve the salt rejection of
integrally-skinned asymmetric cellulose acetate re-
verse osmosis membranes.
During the development of integrally-skinned
asymmetric cellulose acetate gas separation mem-
branes it was found that water-wet membranes col-
lapse and form an essentially dense Rlm upon drying.
This collapse occurs because of the strong capillary
forces within the Rnely porous structure during the
drying process. This phenomenon can be described
by the well-known YoungdLaplace relationship
((p"2/r) in the case of perfect wetting of the
liquid in the pores). Hence, the capillary pressure is
directly proportional to the surface tension of
a liquid, but inversely proportional to the pore radius.
If the modulus of the membrane material (in the
swollen state) is lower than the capillary force of the
liquid in the pore space, the pores will collapse and
form a dense polymer Rlm. Because water has a very
high surface tension, it is often difRcult to dry water-
wet membranes without collapsing the membrane
structure. An exchange of water with liquids having
lower surface tension, such as alcohols or aliphatic
hydrocarbons, results in maintaining the original
membrane structure upon drying. Typical solvent-
exchange methods involve replacing water Rrst with
iso-propanol and then with n-hexane. Other methods
of eliminating the collapse of Rnely porous membrane
structures include freeze-drying and the addition of
surfactants to the water prior to drying of the wet
membranes.
In the 1970s, commercialization of gas separation
membranes was severely limited by the very poor
reproducibility of making ultrathin, defect-free mem-
branes on a large scale. Methods for production of
thin-Rlm composite membranes as well as integrally-
skinned asymmetric membranes with separating layer
thicknesses of less than 0.2
m were known. How-
ever, production of these membranes without defects
was not possible. Defects as small as 20 A> over an
area fraction of less than 10\4% can severely reduce
the selectivity of gas separation membranes. How-
ever, a thin coating of a highly permeable polymer,
such as polydimethylsiloxane, can render defective
membranes suitable for gas separations. ModiRcation
methods developed by Browall for thin-Rlm com-
posite membranes and, in particular, Henis and Tri-
podi for integrally-skinned asymmetric membranes
resulted in rapid commercialization of gas separation
membranes. Surface coatings are also applicable in
improving the fouling resistance of membranes for
ultraRltration or nanoRltration applications. Chem-
ical surface modiRcation methods of gas separation
membranes include treatment with Suorine, chlorine,
bromine, or ozone. Typically, these treatments result
in an increase in membrane selectivity coupled with
a decrease in Sux. Cross-linking of polymers is often
applied to improve the chemical stability (swelling
 II / MEMBRANE SEPARATIONS / Membrane Preparation
resistance) and selectivity of membranes for elec-
trodialysis, reverse osmosis, pervaporation, and gas
separation applications.
Inorganic Membranes
Ceramic Membranes
Microporous ceramic membranes for ultraRltration
and microRltration applications can be formed from
a variety of metal oxides. SpeciRcally, aluminium and
titanium oxides are preferred precursors for the
production of ceramic membranes. Because ceramic
membranes are chemically inert and can be operated
at high temperatures, these membranes offer some
signiRcant advantages over polymeric membranes.
Pore diameters in ceramic membranes for ultraRltra-
tion and microRltration are in the 0.01 to 10
m
range and are typically made by a slip coating-sinter-
ing process. Other techniques, such as the sol-gel
method, produce ceramic membranes with pores in
the range of 10 to 100 A> . In the slip coating-sintering
process, a porous ceramic tube is made by pouring
a dispersion of a coarse ceramic material and a binder
into a mould. This mixture is then sintered at high
temperature. The resulting porous tube is then coated
with a mixture containing very small metal oxide
particles and a binder; this mixture is called a slip
suspension. Again, the mixture is sintered at high
temperature to form a more Rnely porous layer. The
slip-coating-sintering method can be used to make
membranes with pore diameters between 100 to
200 A> . More Rnely porous membranes can be fab-
ricated by the sol-gel technique. First, the metal ox-
ide, dissolved in alcohol, is hydrolyzed by addition of
excess water. Then, the colloidal polymeric or inor-
ganic hydroxide solution is cooled and coated onto
a preformed microporous support made by the slip
coating-sintering process. The coating must be dried
very carefully to avoid cracking of the thin ceramic
layer. The Rnal step of the sol-gel method involves
sintering of the coating at elevated temperature, typi-
cally between 500 and 8003C. In principle, mem-
branes made by this process can be used in a variety
of applications which require membranes that are
stable in harsh environment and at elevated temper-
ature. However, reproducibility of the membrane
formation process on a large commercial scale is
rather poor and the membrane costs are too high for
these membranes to be used in any industrial separ-
ation process.
Metal Membranes
Metal membranes have been considered for a long
time for gas separation applications, speciRcally hy-
1763
drogen separation. Certain noble metals, for example
palladium or palladium}silver and palladium}gold
alloys, are permeable to hydrogen but essentially im-
permeable to all other gases. In the 1950s and 1960s,
Union Carbide installed a pilot membrane system
containing 25-
m-thick, isotropic palladium mem-
branes. Because the hydrogen Sux through these thick
palladium membranes is quite low, the membranes
had to be operated at about 4003C. Although the
plant generated 99.9% hydrogen, commercialization
of this process was economically not feasible because
of the extremely high cost of the metal membrane
(&$5000/m2). Furthermore, contaminants in the
feed stream, such as hydrogen sulRde, poison the
metal which results in a dramatic decline in hydrogen
Sux.
Anodic Membranes
Symmetric and asymmetric microporous membranes
with a conical pore shape can be made from alumi-
nium using an anodic oxidation process. Symmetric
aluminium oxide membranes having a porosity of
65% and a pore size of about 200 nm can be made.
The surface pores of asymmetric aluminium oxide
membranes are about 25 nm. To prepare these mem-
branes, a thin aluminium foil is anodically oxidized in
an acid electrolyte, such as sulfuric or chromic acid,
thereby forming an aluminium oxide. The unaffected
fraction of the metal foil is subsequently removed
using a strong acid. The pore size of membranes made
by anodic oxidation is determined by the voltage and
the acid type.
Carbon Membranes
Microporous carbon membranes can be made by
compressing ultraRne carbon particles or by pyrolys-
ing polymeric precursors. Degradation of the base
polymer upon heating leads to carbonization. The
pore size and porosity of the pyrolysed membranes
depend primarily on the pyrolysis temperature and
the pyrolysis atmosphere. Molecular sieve mem-
branes made from pyrolysed polyacrylonitrile and
polyimide as well as selective surface Sow membranes
made from polyvinylidene chloride-acrylate ter-
polymer can have signiRcantly better separation per-
formance than polymeric membranes in gas separ-
ation applications. The pore sizes of microporous
carbon membranes are typically in the 5 to 20 A>
range.
Glass Membranes
Isotropic glass membranes with a sponge structure
can be made by thermal phase separation of an ini-
tially homogenous metal oxide mixture. Microporous
1764 II / MEMBRANE SEPARATIONS / Micro\ltration
glass membranes were produced by Corning
(Vycor威), Schott, and PPG. Glass membranes are
typically made as discs, tubes or hollow-Rbres. To
produce microporous glass membranes, a homogene-
ous melt consisting of 70 wt% SiO2, 23 wt% B2O3
and 7 wt% Na2O is formed between 1300 to 15003C.
Phase separation of the initially homogeneous glass
melt occurs by lowering the temperature to about
8003C. One phase consists primarily of insoluble sili-
con dioxide. The other phase, rich in alkali borate,
can be leached from the heterogeneous glass by treat-
ment with a mineral acid. After removal of the alkali
borate phase, a microporous silica membrane is
formed.
Future Developments
During the past forty years membranes have gained
signiRcant importance in a wide variety of industrial
separations. Currently, polymeric membranes are
most commonly used for commercial applications.
However, recent developments on inorganic mem-
branes are very promising and such membranes may
broaden the separation spectrum of membranes for
separations. The wide-spread use of inorganic mem-
branes in industrial applications is currently limited
by their poor mechanical stability and very high pro-
duction costs. If these problems can be solved in
future work, inorganic membranes will present a new
generation of high-performance membranes for the
next millennium.
Micro\ltration
I. H. Huisman, AMKM, TNO Voeding, AJ Zeist,
Holland
Copyright ^ 2000 Academic Press
Introduction
MicroRltration is a separation technique for remov-
ing micron-sized particles, like bacteria, yeast cells,
colloids, and smoke particles, from suspensions or
gases. The process uses membrane Rlters with pores
in the approximate size range 0.1 to 10
m, which are
permeable to the Suid, but retain the particles, thus
causing separation. Examples of particles with sizes
in the microRltration range are presented in Figure 1.
MicroRltration membranes were Rrst commercial-
ized in the 1920s, and were at that time mainly used
for the bacteriological analysis of water. After 1960
the number of successful microRltration applications
Further Reading
Baker RW, Cussler EL, Eykamp W et al. (1991) Membrane
Separation Systems } Recent Developments and Future
Directions. Park Ridge, NJ: Noyes Data Corporation.
Bhave RR (1991) Inorganic Membranes. New York:
Van Nostrand Reinhold.
Burggraaf AJ and Cot L (1996) Fundamentals of Inorganic
Membrane Science and Technology. Amsterdam:
Elsevier.
Cabasso I (1987) In Encyclopedia of Polymer Science and
Engineering, Vol. 9, pp. 509}579. New York:
John Wiley and Sons.
Kesting RE (1971) Synthetic Polymeric Membranes.
New York: McGraw-Hill Book Company.
Kesting RE and Fritzsche AK (1993) Polymeric Gas Separ-
ation Membranes. New York: John Wiley and Sons, Inc.
Koros WJ and Pinnau I (1994) In: Paul DR and Yampolskii
YP (eds) Polymeric Gas Separation Membranes,
pp. 209}271. Boca Raton: CRC Press.
Lloyd DR (1985) Materials Science of Synthetic Mem-
branes. ACS Symp. Ser. 269. Washington DC: ACS.
Mulder M (1996) Basic Principles of Membrane Technology,
2nd edn, Boston, MA: Kluwer Academic Publishers.
Petersen RJ and Cadotte JE (1990) In: Porter MC (ed)
Handbook of Industrial Membrane Technology,
pp. 307}348. Park Ridge, NJ: Noyes Publications.
Pinnau I (1994) Polym. Adv. Techn., 5, 733.
Strathmann H (1979) Trennung von molekularen Mischun-
gen mit Hilfe synthetischer Membranen. Darmstadt:
Dr. Dietrich Steinkopff Verlag.
Strathmann H (1990) In: Porter MC (ed) Handbook of
Industrial Membrane Technology, pp. 1}60. Park
Ridge, NJ: Noyes Publications.
grew rapidly, and nowadays microRltration processes
are operated in such different Relds as the biotech-
nological, automobile, electronics, and food industry.
Examples of applications are the harvesting of bacter-
ial and yeast cells, the recovery of latex pigments
from paints, and the puriRcation of water for the
electronics industry. In the food industry, microRltra-
tion is used in the clariRcation of fruit juices, wine,
and beer, in fat removal from whey and in removal of
bacteria from milk.
MicroRltration is the largest industrial market
within the membrane Reld, responsible for about
40% of total sales, both in Europe and in the USA. In
1997, the US microRltration membrane market
amassed revenues worth about $400 million, growing
at an average annual growth rate of 6.6%. MicroRl-
tration can be carried out in two different operation
modes: dead-end (in line) Rltration and cross-Sow

Figure 1 Particles in microfiltration size range.
(tangential Sow) Rltration (Figure 2). In dead-end Tl-
tration the main Sow direction is perpendicular to the
membrane. The suspended particles are continuously
dragged towards the membrane and deposit on the
surface or inside the membrane pores. The deposition
of particles leads to a continuously increasing resist-
ance to Sow and thus to a continuously decreasing
permeate Sux rate. To reduce this deposition process,
microRltration is often carried out in the cross-Uow
mode (tangential Sow) in which the main Sow direc-
tion is tangential to the membrane. The Sow ‘scours’
Figure 2 (A) Dead-end filtration and (B) cross-flow microfiltra-
tion using a tubular membrane.
II / MEMBRANE SEPARATIONS / Micro\ltration 1765
away particles from the membrane surface, and thus
limits particle deposition.
Micro\ltration Membranes
Two main types of membrane Rlters exist: screen
Tlters and depth Tlters. Screen Rlters contain capil-
lary-type pores; particles are retained on the mem-
brane surface primarily by a sieving mechanism.
Depth Rlters contain a random, tortuous porous
structure; particles are retained through adsorption
and mechanical entrapment within the bulk of the
Rlter. Screen Rlters are absolute: particles larger than
the pore size are retained, whereas particles smaller
than the pore size can pass relatively easily through
the membrane. Particle retention of depth Rlters is not
that clearly deRned: retention values increase slowly
over a broad particle size range and only reach 100%
for very large particles. Depth Rlters are often used
for dead-end Rltration, as they can retain a high
particle load.
Membrane Materials and Membrane Preparation
MicroRltration membranes are available in a wide
variety of materials and methods of manufacture.
Many membranes are made of polymers, such as
cellulose acetate, polysulfone, and polyvinylidene Su-
oride (PVDF). Most of these membranes are solvent
cast, through a phase inversion process. Other prep-
aration techniques are stretching (polytetraSuoro-
ethylene, PTFE, membranes) and track-etching (poly-
carbonate membranes). The track-etching process
results in cylindrical pores with a very narrow size
distribution.
Other microRltration membranes available are
made from glass, from ceramics, such as alumina,
titania, and zirconia, and from metals, such as silver
and stainless steel. Advantages of these inorganic
materials are their higher stability towards extreme
process conditions, such as high temperature, ex-
treme pH values, and solvents different than water.
Most metal and some ceramic membranes are pro-
duced by a sintering process, whereas other ceramic
membranes are produced by sol-gel processing or by
anodic oxidation. Some novel membranes are pre-
pared by lithographic techniques.
In Table 1, a number of different commercial
membranes and some of their key properties are pre-
sented, and in Figure 3 SEM (scanning electron
microscopy) and AFM (atomic force microscopy) im-
ages of some membranes are shown. Note that the
membranes shown here are only a fraction of the total
number of membrane materials and membrane
manufacturers available.
1766 II / MEMBRANE SEPARATIONS / Micro\ltration

Table 1 Various microfiltration membranes and their water fluxes

Manufacturer Trade name Material b Preparation method Pore size a (
m) Water permeability at
203C (L m\2 h\1 bar)

US Filter/SCT Membralox ] -Al2O3 Sintering 0.2 2 000

Anotec Anopore ] -Al2O3 Anodic oxidation 0.2 3 600
Carbon Lorraine Carbon Pyrolysis 0.2 1 500
Tech Sep Carbosep ] ZrO2 Sintering 0.14 400
Millipore Durapore ] PVDF Phase inversion 0.22 5 900
Fluoropore ] PTFE Stretching 0.22 12 000
MF-Millipore ] Mixed cellulose Phase inversion 0.22 14 400
esters
Osmonics PCTE Polycarbonate Track-etching 0.2 14 600
PES Polyethersulfone Phase inversion 0.2 20 500
MCS Mixed cellulose Phase inversion 0.22 15 400
esters
Whatman Cyclopore] Polycarbonate Track-etching 0.2 16 000
Aquamarijn MicrosieveTM Silicon nitride Photolithography 0.2 87 000

a
All these membranes are available with pore sizes in large ranges. The pore sizes closest to 0.22
m are mentioned here to compare
water fluxes of the different membranes. bPVDF, polyvinylidene fluoride; PTFE, polytetrafluoroethylene.

Membrane Characterization the case for sterile Rltration in the pharmaceutical

Originally the main goal in characterization of
porous membranes was to determine the pore-size
distribution. It has however been realized more re-
cently that membrane surface properties, such as hy-
drophobicity, zeta potential and surface roughness,
play an important factor in fouling and retention
properties of membrane processes. Characterization
is therefore nowadays performed by various tech-
niques, measuring different structural and physico-
chemical parameters. The relatively novel technique
of AFM microscopy has been shown to provide in-
formation on many membrane properties of interest:
pore size distribution, surface roughness, and ad-
hesion behaviour. In Table 2, various measurement
techniques are summarized.
Dead-end Micro\ltration
In dead-end Rltration, the Suid is forced perpendicu-
larly through the membrane, while all or most of the
particles are retained (Figure 2a). If screen Rlters are
used, these particles build a cake layer on the surface,
which causes an additional resistance to Sow. If depth
Rlters are used, these particles Rll the voids within the
membrane bulk, and in this way cause an increased
resistance. For both types of Rlters, the increased
resistance causes a continuous decline in Sux if a con-
stant transmembrane pressure is used (Figure 4).
After some time, the Sux has been reduced to unac-
ceptably low levels, and the membrane has to be
cleaned or replaced.
Dead-end Rltration is preferred over cross-Sow Rl-
tration in situations where the concentration of par-
ticles to be removed from the Suid is very low, as is
industry, for gas cleaning, and for guard-Rlters
positioned as last step in a high-purity water unit.
Dead-end Rltration is also used in situations where
backSush techniques and gas sparging are so
effective that the use of a cross-Sow is not necessary,
as found in some wastewater-treatment plants.
Fluid Flow through Membrane Pores
The capacity of a microRltration process is expressed
as Sux, J, which is the volume of permeate passing
through the membrane of area Am and per unit time:
1 dV
J" [1]
Am dt
where V is the volume of permeate, and t is time:
Most commercial liquid microRltration processes
operate at Suxes of typically about 10\4 m s\1
(360 L m\2 h\1).
The driving force for this Sux is the transmembrane
pressure (most commonly written as P), the pressure
difference between feed side and permeate side,
which results from applying either suction to the
permeate side or pressure to the feed side, or both.
Transmembrane pressures in liquid microRltration
are typically 5}100 kPa (0.05}1 bar). It was found
phenomenologically that the Sux increases linearly
with the transmembrane pressure (Darcy’s law):
P
J" [2]
Rm )
0
where
0 is the permeate viscosity, and Rm is the
hydraulic resistance of the membrane against
 II / MEMBRANE SEPARATIONS / Micro\ltration 1767

permeate Sow. The permeability of the membrane is where l is the membrane thickness, m is the
deRned as the inverse of its resistance (1/Rm). membrane porosity, and rp is the pore radius. For
MicroRltration membranes have permeabilities of membranes comprised of sintered spheres, the
typically 10\11 m. Kozeny} Carman equation may give a better approxi-
The permeability is related to the pore size. The mation:
exact relation between permeability and pore size
depends on the geometry. For straight cylindrical 45(1!m)2l
Rm" [4]
pores, the Hagen}Poiseuille equation yields: 3ma2m

8l where am is the radius of the particles that constitute

Rm" 2 [3]
mrp the membrane.

Figure 3 SEM (A}C, E}H) and AFM (D) of surfaces and cross-section of different membranes. (A) Durapore membrane, rating
0.22
m, PVDF solvent cast membrane (Millipore). (B) Fluoropore membrane, 0.1 rating, stretched PTFE (Millipore). (C) Polycarbonate
track-etched membrane (Osmonics). (D) Anopore membrane, 0.1
m rating, anodically oxidated Al2O3 (Anotec). (E) Microsieve,
photolithography, silicon nitride (Aquamarijn). (F) Silver membrane (Millipore). (G) AP15 glass fibre depth filter (Millipore). (H)
Cross-section of a P series membrane, solvent cast polyethersulfone (Osmonics). (A), (B), (F) and (G) were kindly supplied by the
Millipore corporation. (C) and (H) were kindly supplied by Osmonics; (D) was kindly supplied by the Group of Membrane Science and
Technology, University of Valladolid, Spain; (E) was kindly supplied by Aquamarijn.
1768 II / MEMBRANE SEPARATIONS / Micro\ltration

Figure 3 Continued

Figure 3 Continued
Screen Filters: Cake-layer Build-up
The Sux calculated by eqns [2]}[4] is the so-called
‘pure water Sux’. During Rltration, fouling and cake-
layer build-up continuously decrease the Sux to values
much lower than the pure water Sux. Darcy’s law
(eqn [2]) can be written for a fouled membrane as:
P
J" [5]
Rtot )
0
II / MEMBRANE SEPARATIONS / Micro\ltration 1769
where Rtot is the total hydraulic resistance. It can be
divided into the membrane resistance (Rm), the resist-
ance caused by fouling (Rf), and the resistance caused
by the cake layer (Rcake):
Rtot"Rm#Rf#Rcake [6]
Fouling can be caused by processes such as the ad-
sorption of macromolecules or bacteria. It is difRcult
1770 II / MEMBRANE SEPARATIONS / Micro\ltration

Figure 3 Continued


to predict the extent of fouling quantitatively; a dc dc
#J "(1!)
b [7]
qualitative description is given in a later section. dt dt
The cake resistance can be calculated by the Kozeny}
Carman equation, eqn [4], where the cake’s void where b is volume fraction of particles in the bulk.
fraction , the cake-layer thickness c, and the If a constant P is applied, the Sux is given by:
particle radius a are to be inserted for m, l and
am respectively.


P 2RK c bP ) t \1/2
The void fraction of the cake layer, , may depend J(t)" 1# [8]

0Rm (1!! b)
0R2m
on various parameters, such as transmembrane
pressure, particle size distribution, shape, and
compressibility, and the effect of particle}particle in- where RK c is the speciRc cake resistance ("Rc/c). Care
teractions. Often a value between 0.3 and 0.4 is must be taken when using eqn [8] as  and RK c often
found for . depend on time (cake compaction).
A model for the time dependent dead-end
Depth Filters
Rltration Sux is obtained by combining eqns [4]}[6]
with a mass balance describing cake layer build- Particle retention in depth Rlters is based on various
up: mechanisms. In the case of gas cleaning, the two
 II / MEMBRANE SEPARATIONS / Micro\ltration 1771

Figure 3 Continued

most important mechanisms are particle capture by line, which at some point passes close to the Rlter
interception and particle capture by diffusion. Inter- surface at a distance less than the particle radius, thus
ception occurs when a particle follows a Suid stream- causing contact between the particle and the Rlter.
1772 II / MEMBRANE SEPARATIONS / Micro\ltration

Table 2 Membrane characterization methods

Method a Parameters obtained Description/remarks

Microscopic techniques ‘Direct’ observation by microscopes.

SEM, TEM, FESEM
(#image analysis)
AFM (#image analysis)
Liquid penetration methods
Bubble point method
Extended bubble point method
Mercury porometry
Permporometry
Solute retention
Contact angle measurements
Sessile drop, Wilhelmy plate
Electrokinetic methods
Electroviscous method
Streaming potential,
electroosmosis
a
SEM, scanning electron microscopy; TEM, transmission electron microscopy; FESEM, field effect scanning electron microscopy;
AFM, atomic force microscopy.
Pore-size distribution, pore
morphology (surface roughness)
Pore-size distribution, morphology,
surface roughness, particle}
membrane interactions
Largest pore available
Pore-size distribution
Pore-size distribution
Pore-size distribution
Pore-size distribution (‘cut-off’)
Contact angle, surface tension,
hydrophobicity
Zeta potential, surface charge density
Zeta potential, surface charge density
Although these methods have many advanced possibili-
ties, they are mostly used for determining the pore-size
distribution in the membrane surface. Preparation of
sample necessary
See SEM (no preparation technique is necessary)
Liquids will fill larger pores at low pressures. To fill smaller
pores, however, higher pressures are needed
Simple method, contact angle of membrane}liquid needs
to be known
See bubble point method
High pressures are necessary that may damage the mem-
brane structure
Vapour condensation in pores is measured. Rather com-
plicated method
Membranes with smaller pores retain solutes of smaller
sizes. Simple method. More often used for ultrafiltration
than for microfiltration
Direct methods that measure the contact angle
liquid}air}membrane. Give qualitative infomation on hy-
drophobicity
Experimentally simple: measurement of water flux at vari-
ous ionic strengths. Interpretation of results more difficult
Direct measurements of electrokinetic effects. Interpreta-
tion of results sometimes complicated

Capture by diffusion occurs when the Brownian rapidly. Under these conditions sieving (entrapment)
motion of the particle results in contact between the is the only effective particle capture mechanism, mak-
particle and the Rlter matrix. ing the membrane permeable for all particles smaller
Interception is the dominant capture mechanism for than the pore size.
large particles; Brownian diffusion is the dominant
capture mechanism for smaller particles. Capture is
therefore least effective for intermediate size particles,
leading to the existence of a ‘most penetrating particle
size’ (Figure 5). The exact value of this most penetrat-
ing particle size depends on the membrane pore dia-
meter and the Sow velocity. It has been found, how-
ever, that capture of particles from gas streams by
membranes of pore diameters of about 0.2
m is so
effective that essentially all particles are retained.
For depth Rltration of liquids, the situation is dif-
ferent, as physicochemical (charge) effects alter the
relative magnitudes of the capture mechanisms de-
scribed above. If physicochemical conditions are fa-
vourable, capture efRciencies in liquids can be similar
to those in gases. If conditions are less favourable, Figure 4 Flux versus time for the dead-end microfiltration of
capture efRciencies for the smaller particles decrease a silica particle suspension.

Figure 5 Schematic representation of efficiency of capture by
interception and capture by diffusion versus particle size. The
most penetrating particle size is obtained by combining both
mechanisms.
Cross-]ow Micro\ltration
Dead-end microRltration, as stated, may suffer from
dramatic Sux loss because of deposition of particles
on the membrane surface and fouling phenomena.
Therefore microRltration is often carried out in the
cross-Uow mode (Figure 2b). The tangential Sow
(cross-Sow) ‘scours’ away particles from the mem-
brane surface, and thus limits cake-layer build-up and
fouling. Another advantage of cross-Sow Rltration is
the possibility for continuous operation. Cross-Sow
Rltration is used in most industrial large-scale micro-
Rltration plants. For cross-Sow microRltration, screen
Rlters are mainly used.
Cake-layer Build-up and Fouling
During a cross-Sow microRltration process, a Sux
behaviour is often observed as shown in Figure 6. The
Sux declines at Rrst rapidly with time; then the speed
of Sux decline decreases, and Rnally a steady state is
reached where the Sux does not decrease anymore.
The decrease in Sux is commonly ascribed to two
phenomena: cake-layer build-up and fouling.
When Rltering a suspension, the membrane retains
suspended particles. The particle concentration near
the membrane will therefore gradually increase.
Cake-layer build-up will occur when the particle con-
centration near the membrane surface reaches the
maximum packing density (0.6}0.7). Cake-layer
build-up is thus caused by the particles that are
retained by the membrane based on their size,
independent of any speciRc interaction between these
II / MEMBRANE SEPARATIONS / Micro\ltration 1773
particles and the membrane. Cake-layer build-up in
microRltration is a phenomenon similar to concentra-
tion polarization in ultraRltration.
Fouling, on the other hand, is based on a direct
contact between solutes and the membrane surface.
The term ‘fouling’ includes many processes, such as
adsorption and deposition of macromolecules, bac-
teria, or small organic molecules on the membrane
surface or within the pores. Fouling increases the hy-
draulic resistance against permeate Sow, and thus re-
duces the capacity of the microRltration process. More-
over, fouling in general increases the observed retention
of the membrane as it reduces the effective pore size.
If one plots the steady-state Sux for crossSow micro-
Rltration versus transmembrane pressure (P) often
a curve as given in Figure 7 is obtained. Three regimes
can be observed. For low values of P, the Sux in-
creases linearly with P and often equals the pure
water Sux. For higher values of P, the Sux curve
bends, because of cake-layer build-up, and Suxes be-
come less than the pure water Sux. The point where
the deviation from the straight line starts is often
referred to as the critical Uux. For even higher P, the
Sux is independent of the pressure. This pressure inde-
pendent Sux value is referred to as the limiting Uux.
Factors In]uencing Membrane Fouling and
Cake-layer Build-up
The extent of membrane fouling and cake-layer
build-up depends on many parameters, which can be
grouped in three main contributors:
E properties of the membrane,
E properties of the suspension, and
E properties of the process (hydrodynamics).
Membrane properties of importance are hydro-
phobicity, surface charge (zeta potential), surface
Figure 6 Flux versus time for the cross-flow microfiltration of
a silica particle suspension.
1774 II / MEMBRANE SEPARATIONS / Micro\ltration
Figure 7 Steady-state flux versus P for the cross-flow micro-
filtration of a silica particle suspension. **, water; - - - 䊐 - - -,
silica particle suspension.
roughness, and pore-size distribution. In general,
macromolecular adsorption is more severe for hydro-
phobic than for hydrophilic membranes. Fouling by
negatively charged colloids is less for negatively
charged membranes than for uncharged or positively
charged membranes. As most colloids in practical
suspensions acquire a negative charge, negatively
charged membranes are preferred in general. Mem-
brane fouling is further reduced by choosing
membranes with smooth surfaces, small pore sizes,
and narrow pore size distributions.
Feed suspension properties of importance are par-
ticle concentration, particle charge (zeta potential),
ionic strength, and overall composition. The amount
of cake-layer build-up increases with particle concen-
tration. Charge effects can reduce fouling by mem-
brane-particle repulsion, and can reduce cake-layer
build-up by particle}particle repulsion. Such charge
effects are less pronounced at high ionic strength, as
the ions present in solution ‘shield’ the charge of
membrane and particles. Overall composition of the
feed suspension is of great importance for the fouling
behaviour. Fouling may be caused not only by the
main particles retained, but also by macromolecules
and small organic molecules, which ‘geometrically’
should pass through the pores easily.
Process properties of importance are the transmem-
brane pressure and the cross-Sow velocity. Low foul-
ing normally occurs at low transmembrane pressures
and high cross-Uow velocities. More detailed in-
formation is given in a later section.
Calculating the Limiting Flux
To calculate the limiting steady-state Sux, local mass
balances near the membrane surface are used. It is
then assumed that the limiting Sux is reached when
the amount of particles transported towards the
membrane by the permeate Sux (convection) equals
the amount of particles transported away from the
membrane by the cross-Sow. The cross-Sow can
cause back-transport by at least four different mecha-
nisms:
E Brownian diffusion,
E shear-induced diffusion,
E inertial lift, and
E surface transport.
In the following, these mechanisms will be ex-
plained. It is assumed throughout this section that the
particles are spherical and monodisperse, and that
long-term fouling and physicochemical interactions
are negligible.
Brownian diffusion If back-transport is caused by
Brownian diffusion the standard concentration polar-
ization theory can be used, employing the Brownian
diffusion coefRcient for spherical particles:
kT
D" [9]
6
0a
where k is the Boltzmann constant, T is temperature,
and a is the particle radius. By numerical calculations,
using a suspension viscosity that depends on the par-
ticle concentration, it can be shown that the Sux is
given by:


wk2T2 1/3
1/3
J"0.0769 b [10]

30a2L
where w is the wall shear stress and L is the mem-
brane length.
Eqn [10] predicts Suxes of the right order of mag-
nitude for suspensions of small particles (up to about
10 nm). It under-predicts Suxes by one or two orders
of magnitude if applied to suspensions of larger par-
ticles. This discrepancy is called the ‘Sux paradox’.
This paradox is explained by assuming that there are
other mechanisms for back transport, apart from
Brownian diffusion.
Shear-induced diffusion When a shear Reld is ap-
plied to a layer of particles, the particles will tumble
over one another, leading to a more loosely packed
layer. Obviously the particles must move perpendicu-
lar to the applied shear stress to achieve this. The
resulting particle migration can be described by
employing an effective diffusion coefRcient, and is
called shear-induced diffusion.

It can be calculated, using an empirical value of
the shear-induced diffusion coefRcient, that the limit-
ing Sux is given by:


w a4(1!3.8 b) 1/3
J"0.060 [11]

0 bL
valid for b(0.2, i.e. for all practical applications.
Eqn [11] has been shown to give good Sux predic-
tions for suspensions of hard spherical particles, and
reasonable Sux predictions for complex bioSuids
such as milk. Although eqn [11] is derived for local
viscous Sow, Sux calculations have also been re-
ported to be accurate for many turbulent Sow
processes.
Inertial lift If a diluted suspension of particles Sows
through a duct, particles present close to the wall will
migrate towards the centre, perpendicular to the
streamlines. This migration, caused by complex hy-
drodynamic interactions, is called inertial lift. In
cross-Sow microRltration, inertial lift may be able to
prevent particles from depositing onto the membrane.
To model this phenomenon, it is assumed that a cake
layer builds up during microRltration until the con-
vective velocity towards the membrane (the Sux J)
equals the lift velocity, vL, away from the membrane:
0a32w
J"vL"0.036 [12]

30
The inertial lift theory neglects the inSuence of a par-
ticle on the motion of another particle, resulting in
a Sux equation which does not depend on the particle
concentration. The inertial lift model is therefore only
valid for very low particle concentrations. As the Sux
predicted by eqn [12] increases with the cube of the
particle size and the square of the wall shear stress,
inertial effects are expected to be important only for
large particles ('5
m) and high cross-Sow vel-
ocities (w'10 N m\2).
Surface transport A particle on top of a Rlter cake is
subject to different forces, as shown in Figure 8. The
horizontal drag force caused by the cross-Sow
F exerts a clockwise torque on the particle, and the
O
vertical drag force caused by the permeate Sux
FJ exerts a counterclockwise torque. If the torque
caused by the cross-Sow is larger than the torque
caused by the permeate Sux, the particle can roll over
the cake layer to the outlet of the membrane. This
mechanism of transport is called surface transport.
Equating the clockwise torque with the anticlock-
wise torque, an equation for the limiting Sux is
II / MEMBRANE SEPARATIONS / Micro\ltration 1775
Figure 8 Torque balance for the surface transport model.
F "horizontal drag force caused by the cross flow; FJ"vertical
O
drag force caused by the permeate flux J ; Fi"particleIparticle
interaction force; "angle of repose.
obtained:
2.36aw
J" [13]

0 tan (a2RK c)2/5
where  is the angle of repose (see Figure 8).
Just as for the inertial lift model, the present model
neglects the inSuence of a particle on the motion of
another particle, resulting in a Sux equation which
does not depend on the particle concentration.
Eqn [13] overpredicts Suxes for typical microRltra-
tion conditions by an order of magnitude or more.
Two of the models described above, the Brownian
diffusion model and the shear-induced diffu-
sion model, use a continuum approach. The other
two, the inertial lift model and the surface transport
model, are based on a single-particle approach. The
single-particle approach is only valid for low particle
concentrations and large particles.
In Figure 9, the Suxes predicted by the two con-
tinuum models are given as a function of particle size
for typical cross-Sow microRltration conditions. The
Sux predicted by the inertial lift model is plotted in
the same graph to indicate the order of magnitude of
inertial effects. For small particle sizes, Brownian
effects dominate and the Sux decreases with
particle size. For intermediate particle sizes, shear-
induced diffusion dominates and the Sux increases
with particle size. For large particle sizes ('5
m)
inertial effects dominate causing the Sux to increase
even faster with particle size.
The combined effect of Brownian and shear-in-
duced diffusion can be described by:
JBo#SI"(J2Bo#J2SI [14]
1776 II / MEMBRANE SEPARATIONS / Micro\ltration

Figure 10 Flux and amount of matter deposited on the mem-

Figure 9 Flux calculated according to different models as
a function of particle diameter. Calculations were performed for
w"32 N m\2, b"10\3, and L"1.2 m, using eqns [10]}[12]
for the Brownian, shear-induced, and inertial-lift models, and
using eqn [14] for combining Brownian and shear-induced diffu-
sion, - - - -, Brownian diffusion model; ) ) ) ) ) ) , shear-induced
diffusion model; *, Brownian and shear-induced diffusion model;
22, inertial lift model.
brane as a function of time for the filtration of a suspension of
0.48-
m silica particles. Circles represent experimental values for
a particle concentration of 1.7 kg m\3, a transmembrane pres-
sure of 0.42 bar and a cross flow velocity of 1 m s\1; lines repres-
ent model calculations.
a steady-state cross-Sow Rltration model. This ap-
proach is illustrated in Figure 10, modelling the tran-
sient Sux and cake-layer build-up in the cross-Sow
microRltration of a suspension of silica particles.

where JBo is the Sux according to Brownian theory,

eqn [10], and JSI is the Sux according to shear-
Process Considerations
induced theory, eqn [11]. Predictions according to Cross-Sow microRltration is usually carried out in the
eqn [14] are also given in Figure 9. feed-and-bleed mode, shown in Figure 11. The use of
a retentate recycle makes it possible to work at high
cross-Sow velocities (high Qc) while having low
Calculating the Transient Behaviour of
retentate Sows (i.e. high volumetric concentration
Cross-]ow Micro\ltration
factors Vc"1#QP/QR). When high concentration
The time dependence of the Sux can be predicted factors are desired, several recirculation loops may be
using an approach as outlined in the section on dead- placed in series, or in even more complicated
end Rltration, yet allowing for back-transport accord- schemes, with loops both in parallel and in series
ing to the particle transport mechanisms described (Christmas tree design).
above. Such descriptions are rather complicated, and In many cross-Sow microRltration systems and in
will not be treated here. some dead-end systems, backUushing is applied to
A simple but effective approach to model the tran- remove the fouling layer from the membrane. Back-
sient behaviour of the permeate Sux is the use of Sushing is achieved by forcing the permeate period-
a combination of transient dead-end Rltration theory ically back through the membranes. Effective back-
and a cross-Sow Rltration model for the steady-state Sushing is obtained by using high counterpressures
(limiting) Sux. While the cake is initially developing,
the effect of the cross-Sow is small and can be neglect-
ed, so that cross-Sow Rltration theory can be approxi-
mated by dead-end Rltration theory. Upon approach-
ing the steady state, the cross-Sow begins to arrest the
cake growth and dead-end Rltration theory is no
longer accurate. However, near the steady state the
Sux shows only minor time dependence, and the Sux
can be approximated by its steady-state value.
The procedure to predict the total transient behav-
iour of the permeate Sux is thus to use dead-end
Rltration theory (see the section on dead-end microRl-
tration) until the time the steady-state Sux is reached Figure 11 Feed-and-bleed operational configuration for cross-
and then use the steady-state Sux predicted by flow microfiltration.

Figure 12 Microfiltration flux when filtering a particle suspen-
sion, with and without backflush.
(about 0.5 bar) for several seconds every few minutes
(Figure 12).
Fouling is reduced by high cross-Sow velocities and
low transmembrane pressures. High cross-Sow vel-
ocities cause high-pressure drops along the membrane,
which cause the P to be undesirably high at the
entrance of the membrane module. Therefore microRl-
tration processes have been developed which facilitate
a cross-Sow both on the feed side and on the permeate
side. The pressure drops on both sides are similar in
magnitude, guaranteeing a uniform transmembrane
pressure. This method of operation has been shown to
be effective in many dairy applications.
Other process techniques to reduce fouling are the
use of pulsed Sow, gas sparging, and electric or acous-
tic Relds, and the use of Sow geometries that create
secondary Sows or vortices resulting in high shear
rates (e.g. the use of ‘turbulence promoters’ or curved
channels).
Conclusions
Over the last 70 years, microRltration has developed
from a small specialized technology used only in
Pervaporation
H. E. A. BruK schke and N. P. Wynn,
Sulzer Chemtech GmbH, Neunkirchen, Germany
Copyright ^ 2000 Academic Press
Development
In 1917 PA Kober published a paper in which he
described his observation that ‘a liquid in a collodion
II / MEMBRANE SEPARATIONS / Pervaporation 1777
laboratories to a multibillion dollar industry for
separation and puriRcation of liquid and gas
streams. Especially since the 1980s, exciting new
applications have become possible, due to improved
membranes (for example, ceramics) and improved
technologies (for example, backpulsing, uniform
transmembrane pressure). Still, great challenges
exist, for example in the processing of beverages,
such as fruit juices, milk, and beer, where
membrane fouling seriously impairs the economy of
the process.
To overcome these problems, researchers and en-
gineers are becoming increasingly interested in hybrid
and combined processes. Combining microRltration
with good pre- and post-treatments or with other
separation processes may result in better and more
economic separations.
See also: II/Membrane Separations: Filtration.
Further Reading
Belfort G, Davis RH and Zydney AL (1994) The behavior
of suspensions and macromolecular solutions in cross-
Sow microRltration. Journal of Membrane Science 96:
1}58.
Bowen WR and Jenner F (1995) Theoretical descriptions of
membrane Rltration of colloids and Rne particles: an
assessment and review. Advances in Colloid and Inter-
face Science 56: 141}200.
Ho WSW and Sirkar KK (1992) Membrane Handbook.
New York: Van Nostrand Reinhold.
Howell JA, Sanchez V and Field RW (1993) Membranes
in Bioprocessing } Theory and application, 1st edn.
London: Chapman and Hall.
Mulder M (1992) Basic Principles of Membrane Techno-
logy, 1st edn. Dordrecht: Kluwer Academic Publishers.
Scott K (1995) Handbook of Industrial Membranes.
Oxford: Elsevier Science.
Zeman LJ and Zydney AL (1996) MicroTltration
and UltraTltration. Principles and Applications.
New York: Marcel Dekker.
bag, which was suspended in the air, evaporated,
although the bag was tightly closed’. Kober was not
the Rrst researcher to observe this phenomenon, but
the Rrst to realize its potential for the separation of
liquid mixtures which otherwise are difRcult to separ-
ate, and to separate them under moderate conditions.
He introduced the terms ‘Pervaporation’, and ‘Perstil-
lation’, and the Rrst term is now in use to describe in
1778 II / MEMBRANE SEPARATIONS / Pervaporation
general a process in which one component out of
a Suid mixture selectively permeates through a dense
membrane, driven by a gradient in partial vapour
pressure, leaving the membrane as a vapour, and
being recovered in a condensed form as a liquid.
In the years following Kober’s publication a num-
ber of papers were published describing membranes
and processes for pervaporation. Especially during
the 1950s, the interest focused on pervaporation
membranes and processes for the separation of differ-
ent classes of hydrocarbons and of isomeres and nu-
merous patents were granted. Membrane materials
disclosed were natural and synthetic rubbers, cellu-
lose esters and ethers, and several treated and un-
treated polyoleRnes. None of this early membrane,
however, was used in any industrial process, owing to
insufRcient Sux and selectivity.
Pervaporation, vapour permeation and gas per-
meation are very closely related processes. The
driving force is always a gradient in partial vapour
pressure, and transport through the membrane can
best be described by a so-called ‘Solution-Diffusion-
Mechanism’. In this mechanism it is assumed that
a component of the feed having a high afRnity to the
membrane is easily and preferentially absorbed and
dissolved in the dense membrane. Following a con-
centration gradient it migrates through the membrane
by a diffusion process and is desorbed at the down-
stream side of the membrane. The separation charac-
teristic of the membrane is thus governed primarily
by the solubility of components in the membrane
material and, to a lesser extent, by its diffusivity
which even may counteract against the solubility sep-
aration.
In pervaporation and vapour permeation processes
the partial vapour pressures of the components at the
feed side are Rxed by composition and temperature of
the feed; they can be inSuenced only by increasing the
temperature. Therefore, the driving force for the
transport of matter through the membrane is applied
by reducing the partial vapour pressure at the per-
meate side.
Different means have been proposed in order to
effect this reduction of the permeate side partial va-
pour pressure:
E The permeate side of the membrane is swept with
an inert gas in which the partial vapour pressure of
the critical (preferential permeating) component is
kept sufRciently low. If the gas stream cannot be
wasted it has to be reconditioned and recycled.
E All permeating vapour is removed by means of
a vacuum pump. The vapour may be condensed
after recompression at the downstream side of the
pump.
E The permeated vapour is condensed at sufRciently
low temperatures. As the condenser surface will be
installed at a certain distance to the permeate side
of the membrane all non-condensable gases have to
be removed from the permeate compartment in
order to minimize permeate side pressure losses.
In most industrial installations the last has been
proven to be the most effective and economical pro-
cess.
In a pervaporation process the feed is applied as
a liquid and all partial vapour pressures of the com-
ponents in the feed mixture are at saturation level.
Within the limits of membrane stability and process
requirements, temperature and pressure on the feed
side are free adjustable parameters.
In vapour permeation a vaporous feed mixture is
applied, with at least the partial vapour pressure of
the preferential permeating component at or close to
saturation conditions. Temperature and pressure of
the feed are linked by vapour}liquid equilibrium and
can be chosen within these limits only.
In gas permeation all partial vapour pressures at
the feed side are below saturation and the permeate
can no longer be condensed. By increasing the total
feed side pressure the driving force for the transmem-
brane transport can be adjusted.
Pervaporation treatment liquid feed mixtures is in-
sofar unique compared with other membrane pro-
cesses as the transport of matter across the membrane
is coupled with a phase change from liquid to vapour.
The heat of evaporation is extracted from the liquid
feed and transported through the membrane, too. As
a consequence the temperature of the feed is reduced,
which reduces driving force and transmembrane
Sux. Different means such as heated modules have
been proposed to replace the lost heat of evaporation.
In general, the total membrane area is split into
a number of segments (stages) arranged in series with
intermediate heat exchangers between each two seg-
ments or stages.
Membranes and Modules
Membranes
With the much broader knowledge of membrane
structure and membrane manufacture accumulated in
the development of desalination membranes in the
1970s pervaporation processes gained new interest.
The separation characteristic of a membrane process
is determined by the difference in transport rates of
the components through the membrane only, not by
liquid}vapour equilibria, and azeotropic mixtures
can easily be separated. Since only the heat of

evaporation of the permeate vapour is lost a single
step membrane process saves energy compared with,
e.g., distillation.
Two different types of pervaporation membranes
were developed at about the same time in the begin-
ning of the 1980s:
E Hydrophilic membranes, with a preferential per-
meation for water, used mainly for the removal of
water from organic solvents and solvent mixtures,
with an emphasis on azeotropic mixtures.
E Organophilic membranes for the removal of
volatile organic components from water and gas
streams.
In both applications composite membranes are used,
allowing for very thin separation layers but with
sufRcient chemical, mechanical and thermal stability.
Because the composite structure Sat sheet conRgura-
tions are preferred. The substructure of both types of
pervaporation membranes is very similar: a porous
support membrane with an asymmetric pore struc-
ture is laid onto a carrier layer of a woven or non-
woven textile fabric and a basic ultraRltration mem-
brane is formed. On the free side of this porous
substructure the pores have diameters in the order of
20}50 nm which widen up to the fabric side to the
micrometre range. Polyester, polypropylene and sim-
ilar Rbres are used for the textile carrier layer;
structural polymers such as polyacrylonitrile,
polyetherimide, polysulfone, polyethersulfone and
polyvinyliden Suoride form the porous support.
On this substructure a thin dense layer (in the range
of 0.5}5
m thick) is coated, which effects the separ-
ation. Different coating techniques are in use, most
commonly a solution of the respective polymer in an
appropriate solvent is spread onto the porous sub-
structure. The solvent is then evaporated, followed
by further treatment to effect cross-linking of the
polymer.
In hydrophilic membranes the separating layer
is made from cross-linked polyvinyl alcohol (PVA),
from polyimides, or natural polymers such as
chitosan or cellulose acetate (CA), with PVA domi-
nant. For organophilic membranes, the separation
layer is formed mostly from siloxanes such as
polydimethylsiloxane (PDMS), or polyoctylmethyl
siloxane (POMS).
In recent years new efforts have been made in
academia and industry to develop new membranes
for organic}organic separation. Of speciRc interest
are the separation of oleRns from parafRns, e.g.
propene from propane, aromatics such as benzene or
toluene from aliphatic hydrocarbons or the separ-
ation of the xylene isomers. To date, no industrial-
ization has been achieved. The only industrial
II / MEMBRANE SEPARATIONS / Pervaporation 1779
processes in this area are the separation of the light
alcohols methanol and ethanol from their mixtures
with hydrocarbons, ethers and esters. The mem-
branes in use are, however, more of the hydrophilic
type, in which the more polar alcohols replace the
water.
To date, only polymeric membranes have been
applied in pervaporation and vapour permeation pro-
cesses. Thermal, mechanical and chemical stability of
the porous substructure are limiting the operation
range of this type of membrane, more than the stabil-
ity of the separating layer. Demand for higher opera-
tion temperatures and chemical resistance have
stimulated the development of inorganic substruc-
tures, and porous ceramics in particular. These can be
coated by cross-linked polymeric separating layers
similar to those on polymeric substructures. In more
recent developments organic separation layers are
applied, either by coating the porous substructure
with a layer of zeolites or by reducing the size of the
pore to molecular dimensions. The separation mech-
anism of these membranes is even more complex than
that of polymeric separating layers, as molecular siev-
ing effects, caused by shape and size of mol-
ecules, and molecule}surface interaction decide
whether a component can pass through the mem-
brane or will be retained.
Modules
Design of modules for pervaporation and vapour
permeation processes was based on the experience
gained in water treatment by membranes, such as
ultraRltration and reverse osmosis processes. How-
ever, signiRcant modiRcations had to be made be-
cause of the speciRc requirements of pervaporation
and vapour permeation processes.
The partial vapour pressure at the permeate side
has to be reduced in both processes to fairly
low values, especially when low Rnal concentrations
have to be reached in the retentate. Therefore any
pressure losses, even in the range of a few millibars
have to be avoided at the permeate side. Since any
feed mixture will contain organic components at high
concentration, mostly at elevated temperatures,
the chemical stability of all module components,
such as spacer and potting material and glues is criti-
cal. To date, two types of modules are most widely
applied:
E Plate modules, mainly used for dehydration ap-
plications, with permeate channels as open as ap-
plicable. Stainless steel is used as a construction
material for support plates for the membranes and
for spacers. The permeate channels are preferably
open over the circumference of the modules which
1780 II / MEMBRANE SEPARATIONS / Pervaporation
Figure 1 Vapour}liquid equilibrium curves for common phar-
maceutical solvents which azeotrope with water. All can be dehy-
drated using pervaporation.
are assembled inside a special vacuum vessel that
also house the permeate condenser. Alternative de-
signs are very similar to plate heat exchangers, in
which the supported membrane replace the heat
exchanger plates. These modules are closed to the
outside, with internal ducts feed and retentate, and
for permeate removal.
E Spiral wound modules with stainless steel central
tubes, but otherwise similar to those known from
the conventional membrane processes, are mainly
used for organophilic membranes. One or several
of the spiral wound modules are housed inside
a pressure tube and assembled in conventional
skids. In a special design, the sandwich structures
of membranes and permeate and feed spacer are
welded together and not spirally wrapped around
the central tube but arranged as Sat sheets on the
central tube for the removal of the permeate.
Very rarely, hollow Rbres are used, generally with the
feed Sow inside the bore of the Rbre. For the more
conventional arrangement } feed Sow at the shell side
of the Rbre } permeate pressure losses inside the bore
may become detrimental for the process.
Applications
Organophilic Membranes
Organophilic membranes are mostly applied for the
removal of volatile organic components (VOCs) from
a gas stream such as waste air or nitrogen. The main
applications are the treatment of streams originating
from the evaporation of solvents in coating processes
in Rlm and tape production, purging of products such
as polymers, by which unreacted monomers are re-
moved, or from breathing of storage tanks for sol-
vents, especially from loading and unloading of petrol
tanks in tank farms. In many installations the feed
stream received at atmospheric pressure is com-
pressed in order to increase the feed side partial va-
pour pressure. Partial condensation of the component
to be removed is a wanted side effect, since then
condensation on the permeate side under vacuum and
at low temperature can be avoided. The permeate is
simply slightly compressed by the vacuum pump and
let into the inlet of the feed compressor. In speciRc
cases the installation of a vacuum pump will not be
necessary and the permeate is obtained at atmos-
pheric pressure.
The economy of the process is usually determined
by the value of the components recovered. Emission
regulations in all industrial countries demand for very
low Rnal concentrations if the gas stream is released
to the atmosphere, therefore the retentate from the
gas puriRcation by the membrane is either recycled or
followed by an additional polishing step.
Although considerable efforts in research and de-
velopment have been devoted to the removal of
Table 1 Solvents routinely dehydrated using pervaporation/
vapour permeation
Isopropanol, Standard applications for pervaporation, typ-
ethanol ically dehydrated from their azeotropes to
fractions of a percent of water. Many continu-
ous, batch and vapour permeation units are
operating around the world.
Ethyl acetate, Form azeotropes in the miscibility gap and
butyl acetate were traditionally dehydrated by two
distillation columns and a phase separator,
however with a massive recycle. Esters de-
compose in contact with zeolites. Pervapora-
tion/vapour permeation is easily the best
technique for dehydration.
Acetone Does not azeotrope with water but when dis-
tilled a large reflux is required to get a half dry
product. Pervaporation is ideal for final dehy-
dration or for debottlenecking existing distil-
lation systems.
Tetrahydrofuran Easily dehydrated by pervaporation down to
a few hundred ppm water. Traditional caustic
washing is operationally messy, requiring
a redistillation of the product. Pressure swing
distillation requires high pressures and large
recycles.
Methyl ethyl Pervaporation is again the preferred tech-
ketone nique. Distillation is only possible with an
entrainer because the azeotropic composi-
tion is nearly identical to the miscibility limit.
N-butanol, Form azeotropes with high water content so
n-propanol the distillation/phase separation process in-
volves massive recycle streams. Pervapora-
tion plants are less costly to build and easier
to operate.
 II / MEMBRANE SEPARATIONS / Pervaporation
Figure 2 Typical flow diagram for recovery of solvent from mother liquors.
VOCs from aqueous stream, this technique has not
yet been introduced into the industry. Potential mix-
tures which could be treated are more complex, the
economical value of the recovered substances are low,
and competing processes such as biological treatment
of wastewater are cheaper. Applications may be
found in the future in biotechnological processes
where high-value products can be separated from
a fermentation broth and be concentrated and puri-
Red in the same step.
Figure 3 (See Colour Plate 51). Vapour permeation unit for
recovering ink solvent.
1781
Hydrophilic Membranes
The largest industrial installations of pervaporation
and vapour permeation processes are equipped with
hydrophilic membranes which are used for the re-
moval of water from organic solvents and solvent
mixtures.
Solvent dehydration Organic solvents are used for
a variety of purposes in the chemical industry, e.g. for
synthesis of pharmaceuticals, to precipitate materials
from aqueous solutions, for cleaning purposes and for
drying Rnal products. Spent solvents nearly always
contain some water. Dehydration is therefore an
essential step in their recovery but difRcult since most
solvents from azeotropes with water. Final water
removal by distillation is then impossible or complic-
ated. Entrainer use is not an option for pharmaceut-
ical or Rne chemical production, where stringent pro-
cess certiRcation rules out adding potential sources of
contamination.
Pervaporation enables solvents to be dehydrated
without using any third substance or entrainer, sim-
ply, cheaply and without problems and irrespective of
vapour/liquid equilibria (Figure 1). On-site solvent
recovery using pervaporation and vapour permeation
is thus becoming standard practice in the pharma-
ceutical and chemical industries (Table 1).
Often, pervaporation and vapour permeation
plants are designed for operation with a number of
solvents and with mixtures of solvents. Pervaporation
and vapour permeation offers the following bene-
Rts when dehydrating solvents:
E No introduction of additional chemicals, complete
solvent dehydration by pervaporation membranes
irrespective of azeotrope formation and no possi-
bility of contamination.
1782 II / MEMBRANE SEPARATIONS / Pervaporation
Figure 4 Carbon bed adsorption using steam regeneration and vapour permeation for final solvent dehydration.
E A choice of batch or continuous pervaporation
systems, or continuous vapour permeation, de-
pending on the duty.
E Able to dehydrate esters without any decomposi-
tion.
E Low energy consumption.
Solvent recovery from mother liquors Spent sol-
vents (mother liquors) typically contain some water
and are often saturated with dissolved material. They
cannot be re-used without puriRcation. Evaporation
combined with vapour permeation is a powerful tech-
nique for purifying and dehydrating mother liquors
(Figure 2).
The feed of spent solvent is evaporated and the
resulting vapour is fed directly to a vapour per-
meation unit. Water vapour selectivity permeates the
membrane and is condensed under vacuum. The
water-free solvent vapour leaving the vapour per-
meation unit is condensed and is stored for re-use
(product). A blowdown is taken from the evaporator
to prevent buildup of dissolved solids. This purge can
Figure 5 Use of vapour permeation to recover solvent from
nitrogen desorption circuit.
be treated to recover valuable components. Combin-
ing evaporation with vapour permeation gives the
following beneRts: only vapour is fed to the mem-
branes } no possibility of fouling and no possibility of
solids carryover into the recovered solvent; both the
evaporation and the vapour permeation process steps
are carried out in a single unit.
Solvent recovery from carbon bed adsorbers Biodeg-
radable solvents such as alcohols and esters are used
in many speciRc applications in coating and printing.
Typically, a solution of the coating material is applied
to the surface and the solvent is evaporated into an air
stream, leaving a uniform Rlm of coat material. The
use of volatile solvents speeds the drying process.
The solvent-laden air stream cannot normally be
discharged. It must be cleaned up and because the
solvent loading is usually quite low, carbon bed ad-
sorbers are most commonly used. These adsorbers are
periodically regenerated, using either steam or nitro-
gen. Solvent is then recovered from the condensate
Figure 6 Vapour}liquid equilibrium diagram for acetone}water.

Figure 7 Debottlenecking a single pinched column using pervaporation (acetone}water example).
or nitrogen stream for re-use in the coating/printing
process (Figure 3).
To evaporate quickly, the recycled solvent has to be
substantially dry and because most of the organics
used form azeotropes with water, distillation is not
sufRcient for Rnal dehydration. Pervaporation or va-
pour permeation provides a dry solvent at minimal
cost.
Vapour permeation for solvent dehydration in print-
ing and coating If carbon bed adsorbers are
regenerated with steam, the condensate is typically
steam distilled up to the azeotrope (Figure 4).
Continuous coating operations use continuous distil-
lation to concentrate condensate from the bed regen-
eration. A vapour permeation system is normally con-
nected directly to dehydrate net overhead vapour
Figure 8 (See Colour Plate 52). Standard unit for batch dehy-
dration of rinse alcohol.
II / MEMBRANE SEPARATIONS / Pervaporation 1783
from the distillation. This situation is shown sche-
matically below. In this case, no additional energy is
required for the Rnal dehydration by vapour per-
meation.
Vapour permeation for solvent removal from circula-
ting nitrogen If nitrogen is used to regenerate the
carbon bed, solvent vapour can be continuously re-
moved by vapour permeation through an or-
ganophilic membrane. This is much more economical
than cooling the vapour to condense the solvent;
because the solvent loading is low (Figure 5).
The use of vapour permeation for solvent recovery
in printing and coating operations provides the fol-
lowing beneRts: ethanol, isopropanol, ethyl acetate
and other biodegradable solvents are recovered and
dehydrated without entrainers and economically,
Figure 9 Batch pervaporation process for dehydration of rinse
alcohol.
1784 II / MEMBRANE SEPARATIONS / Pervaporation
Figure 10 Water removal by distillation and vapour per-
meation.
even at a modest scale; solvent can be economically
recovered from nitrogen streams with minimal cool-
ing requirement.
Debottlenecking Distillations
Debottlenecking pinched distillations Distillation
processes are driven by volatility differences. If these
volatility differences are small, or become small un-
der certain conditions, then columns need to operate
with high reSux to achieve the desired separation.
Because pervaporation/vapour permeation processes
separate irrespective of volatility differences, they can
Figure 11 Progression of a batch esterification (with equilibrium constant K"4) with continuous water removal by pervaporation.
be used very effectively to debottleneck pinched dis-
tillations.
Consider for, example, the system acetone}water
(Figures 6 and 7). Acetone is concentrated in the
vapour phase at low concentrations so stripping of
acetone from water is easy. At high concentrations
this is not the case. Complete dehydration of acetone
is difRcult.
Debottlenecking entrainer distillation systems Exist-
ing entrainer distillation systems can also be effec-
tively debottelenecked using pervaporation/vapour
permeation. Normally, the rectiRcation column will
be operating to give a product as close to the azeo-
trope as possible, running with a high reSux. To
debottleneck the system, reSux in the rectiRcation
column is reduced, giving more overhead product,
but with a higher water content. The pervaporation
unit is sized to remove enough water that the
subsequent entrainer column is also unloaded. Both
columns can then realize a signiRcant capacity in-
crease.
The pervaporation unit required for debottleneck-
ing is relatively small since the driving force for water
permeation is high. Adding a pervaporation/vapour
permeation system to a pinched distillation can give
providing higher product capacity and reduced re-
 II / MEMBRANE SEPARATIONS / Pervaporation 1785

Figure 12 Some methanol azeotropes which can be separated using pervaporation vapour permeation.

Sux, signiRcantly higher products purity and reduced
energy costs.
Dehydration and Puri\cation of Rinse Alcohol
Many metal components used in the electronics in-
dustry undergo a rinse process using ethanol or iso-
propanol. Typically, the surface is Rrst treated in
another way to remove contaminants and then
washed with water. The Rnal rinse with alcohol dis-
places the water and any remaining contaminants and
also wets the surface completely. The volatile alcohol
then dries uniformly leaving a clean unmarked sur-
face. Because the rinse alcohol displaces water it grad-
ually becomes diluted and loses its drying qualities.
Either fresh alcohol must be purchased or the alcohol
must be dehydrated and puriRed to restore its perfor-
mance.
The purity of the rinse alcohol is critical for com-
ponent performance; wafer fabrication, for example,
sets p.p.b. limits on certain metal ions (Figure 8).
Various standardized pervaporation/vapour per-
meation systems are now used economically and at
a site scale for dehydration and puriRcation of
rinse alcohol. A batch of used alcohol is continuously
circulated from a buffer tank via a recuperator,
heater and pervaporation module and water vapour
is continuously removed via the vacuum pump
(Figure 9). The batch is processed until the
required degree of dryness is reached. Standardized
units economically treat batches as small as
1 m3 day\1.
Use of pervaporation and vapour permeation
units to recover rinse alcohol gives the following
beneRts: minimal alcohol losses, very high product
purities can be reached, and economical recovery at
site scale.
Continuous Water Removal from Condensation
Reactions
Condensation reactions such as esteriRcations,
acetalizations and ketalizations produce water as
coproduct; removal of this water from the reaction
will shift the equilibrium in favour of the desired
product.

Figure 13 Methanol recovery by azeotrope breaking (methanol}ethyl acetate example).

1786 II / MEMBRANE SEPARATIONS / Pervaporation
Figure 14 Methanol removal by pervaporation of column side-
draw.
Distillation is often used to remove water from
condensation reactions. However complete water
removal is difRcult because alcohols, esters and acids
typically azeotrope with water. Boiling the reaction
mix also removes alcohol, which is normally the most
volatile component including a vapour permeation
unit after the distillation step avoids the problems of
azeotrope formation. Water can be completely re-
moved (Figure 10).
Water removal by pervaporation only } membrane
reactors Removing water directly from the reaction
mix is more effective } the reaction can even be run
under stoichiometric conditions. Reactor conRgura-
tion is simpler and energy consumption is much lower
(Figure 11).
Using pervaporation/vapour permeation units to
continuously remove water from condensation reac-
tions gives the following beneRts: complete conver-
sion, maximum yield, minimum reagent consumption
and costs; maximizes reaction kinetics, reactor efR-
ciency and productivity; minimizes product puriRca-
tion costs; works irrespective of azeotrope formation.
Methanol Recovery
Methanol is commonly used as both solvent and
reactant in the chemical industry. However, it forms
azeotropes with many substances, particularly esters
(Figure 12). Methanol often cannot be removed from
spent solvents or from reaction mixtures with simple
distillation. Some quite complicated processes have
been developed to get around this problem.
Industrial pervaporation/vapour permeation units
are now used for separation of methanol, either stand
alone or in combination with distillation.
Methanol recovery by azeotrope breaking By way
of example, a separation scheme for a methanol-rich
methanol}ethyl acetate mixture is shown below. The
mixture is distilled to the azeotrope, taking pure
methanol out as bottom product. The overhead
stream is passed directly to a vapour permeation unit
which permeates a methanol-rich stream. This stream
is condensed and passed back to the methanol column
via the feed buffer. Retentate from the vapour per-
meation unit, strongly depleted in methanol, can be
fed directly to the ethyl acetate column. Pure ethyl
acetate leaves this column as bottom product while
overhead azeotrope is sent to the vapour permeation
unit (Figure 13).
Many solvent or ester/methanol mixtures can be
separated using a similar scheme. If the feed is close to
the azeotrope then the methanol column can be dis-
pensed with. If the capacity is small the puriRcation
column for the second component may not be re-
quired, depending on the desired purity.
Institut Franc7 ais du PeH trole (IFP) has developed
a process where pervaporation of methanol is used to
debottleneck MTBE production. In the debutanizer
columns used in MTBE processing, the MTBE}meth-
anol azeotrope results in a concentration of methanol
at a point midway between the feed tray and the
reboiler (Figure 14). Pervaporating methanol out of
the process from a side-draw taken at this point
results in methanol free MTBE as debutanizer bottom
product.
Separation systems based on pervaporation/vapour
permeation of methanol offer the following beneRts:
problem-free separation of methanol/organic mixtures
irrespective of azeotrope formation; avoids water wash
for methanol removal; minimum energy costs.
See Colour Plates 51, 52.
Further Reading
Bakish Material Crop (1985}1995) Proceedings of the
(1st to 7th) International Conference on Pervaporation
Processes in the Chemical Industry. Englewood:
Bakish Material Corp.
BoK ddeker KW (ed.) (1995) The early history of membrane
science, Journal of Membrane Science 100.
Huang RY (ed.) (1991) Pervaporation Membrane Separ-
ation Processes, I Membrane Science and Technology,
Series 1. Amsterdam: Elsevier.
Mulder M (1991) Basic Principles of Membrane Tech-
nology. Dordrecht: Kluwer Academic Publishers.
 II / MEMBRANE SEPARATIONS / Reverse Osmosis
Polymer Membranes
See II / MEMBRANE SEPARATIONS / Gas Separations with Polymer Membranes
Reverse Osmosis
U. Spohn, Institute of Biotechnology,
University of Halle, Germany
This article is reproduced from Encyclopedia of
Analytical Sciences, Copyright Academic Press 1995
Dialysis and reverse osmosis
Dialysis is a separation process with increasing areas
of application in clinical, biochemical and environ-
mental analysis. A donor and an acceptor solution are
separated by a semipermeable membrane (Figure 1).
Figure 1 Dialysis between quiescent and stirred solutions. C,
concentration; D, donor solution; A, acceptor solution; M, mem-
brane; dA, dD and dM; thicknesses of the corresponding phase
layers,
A and
D; thicknesses of the diffusional boundary layers.
1787
When the solutions have different solute activities
a diffusional membrane transport from the more to
the less concentrated solution takes place to establish
the thermodynamic equilibrium. In most applications
mass transfer to and away from the separation mem-
brane is accelerated convectively by stirring or by the
use of Sow-through separation cells.
When the solutions have different solvent activities
an osmotic pressure is built up, which causes a solvent
Sow to the solution of the lower solvent activity.
This process is termed osmosis. Reverse osmosis
is deRned as a process during which an outer pressure
is applied to force the solvent through a membrane,
which is permeable to the solvent and rejects the
solute. Reverse osmosis is applied to purify water
for laboratory use and is very promising as a
preconcentration technique in trace and environ-
mental analysis.
Fundamentals
Thermodynamic Aspects
During dialysis the activities of the solute i in the
donor and the acceptor solutions differ. The differ-
ence between the chemical potentials i,D and i,A is
the free enthalpy per mole, which propels the dialysis.
The dialysis Rnishes when the thermodynamic equi-
librium shown in eqn [1] is reached:
i,D"i,A [1]
The equilibrium constant KH i depends on the activity
coefRcients fi,D and fi,A and the absolute temperature
T according to eqn [2]:
 
ai,A fi,A;ci,A 0i,D!0i,A
KHi " " "exp [2]
ai,D fi,D;ci,D RT
where 0i,D and 0i,A are the chemical potentials of the
solute under standard conditions. c is concentration
and a is activity. R is the molar gas constant.
An effective separation ratio Ki, which is deRned
1788 II / MEMBRANE SEPARATIONS / Reverse Osmosis

according to eqn [3], is of analytical interest: where VL,D and VL,A are the partial molar volumes of
the solvent which are equal to VL for dilute solutions,
ci,A and  refers to the osmotic pressure. It therefore
Ki" [3]
ci,D follows that:

 
The activity coefRcients fi depend on the solvent, the VL
ionic strength I"zjcj of the solution and the con- KHL "exp (D!A) [12]
RT
centrations of all nondissociated solutes. The general
index, j, labels the ions in the system with the electric with the osmotic pressure difference "
charge zj. Because in most cases: D!A. A volume change is effected to equilibrate
the donor and the acceptor solutions.  can be cal-
0i,D!0i,A"0 [4]
culated for every solution according to eqn [13]:
it follows that:
"!(RT/VL) ln ax,L [13]
KHi "1 [5]
where ax,L is the activity (xL  fL) of the solvent with
and: respect to the mole fraction xL of the solvent. There-
fore the osmotic pressure difference  has to be taken
fi,D into consideration to calculate precisely almost all
Ki" [6]
fi,A dialysis equilibria. The osmotic pressure difference
causes a solvent Sow through the separation mem-
Ki differs from unity when the dialysis equilibrium is branes up to the state at which the hydrostatic back-
coupled to other equilibria and/or the activity coefR- pressure compensates for the osmotic pressure. e.g. in
cients fi,D and fi,A are different. Without coupled push- the batch-type arrangement shown in Figure 2. If the
ing and trapping reactions the enrichment E and the pressure ph is greater than the osmotic pressure the
puriRcation factor P can be calculated according to osmosis will be reversed. This reverse osmosis reaches
eqns [7] and [8]: an equilibrium, which can be described by eqn [14]:

ci,A (ci,D,0!ci,D)VD/VA
E" " [7]
ci,D,0 ci,D,0
ci,D,0!ci.D (ci,A!ci,A,0)VA/VD
P" " [8]
ci,D,0 ci,D,0
where ci,D,0 and ci,A,0 are the initial solute concentra-
tions in the donor and in the acceptor solution, re-
spectively, and VA and VD are the volumes of the
corresponding solutions. The maximum enrichment
factor is around 0.5 for VA4VD. The puriRcation
factor increases up to unity with the increasing ratio
of VA to VD.
The chemical activities of the solvent are equal at
dialysis equilibrium. According to eqn [9]:
 
VL
KHL "exp (!ph) [14]
RT
Obviously the equilibrium constant decreases with
increasing pressure ph. When the solute is rejected
by the membrane it can be enriched in the donor
solution.
To take into consideration the electric charges zi
of the solutes and of the separation membrane the
Donnan effect has to be considered. The chemical
potential i is extended by a term for the electric
potential gradient, . When a membrane separates

L,D"L,A [9]

and because:

0L,D"0L,A [10]

it follows for the solvent L that:

Figure 2 The principles of osmosis and reverse osmosis. M,

 
1 membrane; ph, outer pressure; , osmotic pressure difference.
KHL "exp VL,DD!VL,AA [11]
RT , Aqueous salt solutions; , water.
 II / MEMBRANE SEPARATIONS / Reverse Osmosis
two solutions of a dissociating salt BA and one solu-
tion contains a rejected ion X\, an electric potential
difference " D! A is built up between the solu-
tions. This Donnan potential inSuences the effective
separation ratio K. The equilibrium constant is
shown in eqn [15] with the Faraday constant F:
i "exp[(
KH !
D )ziF/RT]
A [15]
where F is the membrane area that is in contact with
the solutions.
An outer electrical voltage U generates an elec-
tromigration of the anions to the positively charged
anode and of the cations to the negatively charged
cathode (Figure 3). The process is termed elec-
trodialysis. If the potential difference between the
electrodes U is smaller than the voltage UD of the
water decomposition a new electrochemical equilib-
rium is built up with the constant shown in eqn [16]:
KHi "exp[( D! #U)ziF/RT]
A [16]
In many practical applications the dialysis equilib-
rium is coupled with chemical equilibria, e.g.
acid}base, redox, complexation and precipitation
equilibria. It should be noted that distribution equi-
libria between two different solvents and phases can
also be exploited to shift the overall distribution ratio
Ki between the donor and acceptor solutions. The
enrichment factor E can be increased by several or-
ders of magnitude. The thermodynamics of the separ-
ation processes enable the attainable maxima of the
enrichment and the puriRcation factors to be esti-
mated.
Figure 3 Domain dialysis across a microporous membrane. ,
Donnan potential; E, electrodes; U, outer voltage.
1789
Mass Transport
Dialysis For quiescent donor and acceptor solutions
the equilibration time ranges from minutes to several
hours and is dependent on the geometrical size of the
donor and acceptor chambers and the membrane
permeability. In this case the equilibration is domin-
ated by the slow diffusional analyte transport. The
equilibration can be accelerated by convective mass
transport according to Figure 1. Intensive stirring of
both the acceptor and the donor solutions establishes
diffusional boundary layers of thickness
on the
membrane. The diffusional boundary layer can also
be established in Sow-through dialysis cells or on
rotating dialysis membranes.
The overall Sux of the substance i with its molar
amount ni from the donor to the acceptor solution
can be described according to eqn [17]:


dni KAM
Ji" "kF c !
i,D i,D c
i,A i,A [17]
dt KDM
where i,D and i,A are the fractions of the substance
i in the donor and in the acceptor solutions respective-
ly, that can permeate through the separation mem-
brane. KDM and KAM are the distribution constants
between the donor solution D and the membrane
M and between the acceptor solution A and the
membrane, respectively. The overall mass transfer
coefRcient k can be derived from eqn [18]:


1 1 kD D 1
" #
k kD kD D#kD(1! D) kMKDM


KAM kA A
# [18]
kAKDM kA A#kA(1! A)
where 1! D is the rejected fraction of substance i,
1! A is the corresponding fraction that is trapped in
the acceptor solution, and kD, kM and kA are the mass
transfer coefRcients of the permeating fractions of
substance i for the donor, the membrane and the
acceptor phases, respectively. kA and kD are the mass
transfer coefRcients for the so-called inactive form of
substance i, which cannot permeate the membrane. In
many cases the mass transfer across the phase bound-
ary is to be considered additionally, e.g. for homo-
geneous membranes, supported liquid membranes
(SLM) and gas-Rlled microporous membranes. Then
the expression shown in eqn [19] should be added to
the right term of eqn [18]:
1 1 KAM
a" # [19]
kpDM kpAM KDM
1790 II / MEMBRANE SEPARATIONS / Reverse Osmosis

kpDM and kpAM are the phase transfer coefRcients from The distribution ratios KDM and KAM are correlated to
the donor solution into the membrane phase and the concentrations i,D;ci,1,D and i,A;ci,1,A of the
from there into the acceptor phase, respectively. volatile forms in the donor and in the acceptor solu-
From the general equation some cases of analytical tions, respectively:
interest can be deduced.
ci,g,D pi,D
Example 1: Dialysis Through Hydrophilic and KDM" + [26]
i,Dci,1,D i,DRTci,1,D
Microporous Membranes. Because D" A"1,
KDM"KAM"1 and no phase transfer takes place,
ci,g,A pi,A
eqn [18] can be simpliRed to eqn [20]: KAM" + [27]
i,Aci,1,A i,ARTci,1,A

1 1 1 1
" # # [20] The right-hand terms of eqns [26] and [27] are ap-
k kD kM kA
proximations for low partial pressures pi,D and pi,A.
For large concentration gradients (cD!cA)/dM and ci,g,D and ci,g,A are the concentrations of the analyte in
very intensive stirring in the set-up shown in Figure 1 the gas phase at the interface between the donor or
or in Sow-through dialysis cells with high Sow-rates the acceptor solution, respectively, and the mem-
the membrane diffusional transport becomes rate- brane gas phase. For very small partial pressures
determining particularly for relatively thick mem- pi,D and pi,A, KAMKDM (pi,Dpi,A) and without
branes with small pores. Eqn [21] then follows: hydrodynamic transport limitations, eqn [18] can be
simpliRed to eqn [28]:
k"kM [21]
1 1 1
" # #a [28]
Example 2: Dialysis of Volatile Substances Through k kD kMKDM
Hydrophobic and Microporous membranes. To sep-
arate a nonvolatile base B\, its corresponding volatile For fast-Sowing donor solutions with 1/kDP0 the
acid BH is produced according to the equilibrium: mass transport is determined by the gas diffusion
through the membrane, the partial pressure of the
B\#H#8BH analyte in the donor solution and the phase-transfer
resistances. The partial pressure of the volatile
The donor pH value is chosen according to eqn [22] analyte can be increased by decreasing the partial
so that there is a 99.9% degree of conversion into the pressure of the water using high ionic strengths in the
permeable form of the analyte: donor solution. For highly volatile analytes, thin and
highly porous membranes, and fast-Sowing solutions,
10\pH the overall mass transport is controlled by the phase
i,D " '0.999 [22]
10\pH#Ka transfer resistance (k"1/a).

Ka is the acid-dissociation constant of BH. Reverse osmosis The driving force of reverse osmo-
The acceptor pH value should be adjusted accord- sis is the difference between the outer pressure ph and
ing to eqn [23] to trap the analyte in its nonvolatile the osmotic pressure difference . The mass transfer
form: can be described according to eqn [29]:

Ka JL"PF (ph!fR) [29]

1! i,A " '0.999 [23]
Ka#10\pH
where JL is the mass Sux of the solvent through
Since only the volatile part of the analyte amount can the separation membrane and P is the water
traverse the membrane, it follows that: permeability of the membrane. The osmotic pressure
difference is multiplied by the reSection coefRcient
Ji"kFci,D [24] fR, which is a measure of the solute rejection by
the membrane. During the enrichment process in
with: the donor solution the osmotic pressure difference
 increases. The driving force decreases. When the
1 1 1 KAM rejection is sufRciently high, the reSection coefRcient
" # # #a [25]
k kD kMKDM kAKDM fR approximates to unity. The rejection ratio is
 II / MEMBRANE SEPARATIONS / Reverse Osmosis
Figure 4 Reverse osmosis with concentration polarization on
the asymmetric separation membrane. Ji , mass flux of the solute;
p D, outer pressure from the donor solution; pA, outer pressure from
the acceptor solution; D and A, osmotic pressures of the donor
and the acceptor solution respectively;
c, thickness of the polariza-
tion layer; ci,D , solute concentration in the donor solution; ci,DM,
solute concentration on the membrane; ci,M, solute concentration in
the separation membrane. See text for further explanation.
deRned by eqn [30]:
ci,A
R"1! [30]
ci,D,0
where ci,A is the solute concentration in the Rltrate,
and ci,D,0 is the initial concentration in the donor
solution. The rejected solutes accumulates on the
membrane surface (Figure 4). This is the so-called
concentration polarization phenomenon, which can
be described approximately according to eqn [31]:
JL"kL  ln[(ci,DM!ci,A)/(ci,D!ci,A) [31]
where ci,DM is the solute concentration on the mem-
brane surface and kL is the mass transfer coefRcient.
The concentration up to the saturation level will
cause the precipitation of the solute. The precipitated
solute forms a secondary layer on the membrane,
which reduces the solvent mass transfer JL. Therefore
the concentration polarization must be reduced by
a forced convective Sow.
Table 1 Dialysis procedures
Objectives in a microanalytical scale Donor solution
Purification Quiescent or slowly flowing, small
sample volume
Enrichment Flowing or stirred, large sample volume
Reagent addition Stirred or flowing, large reagent volume
Separation Quiescent or slowly flowing, small
sample volume
1791
The analytical usefulness is based on the high en-
richment factor E, which can be achieved following
by eqn [32]:
ci,D VD,0
E" " [32]
ci,D,0 VD,0!VA
where VD,0 is the initial volume of the donor solution.
Geometric Aspects
The geometric shape and extent both of the donor
and the acceptor chambers is decisive for the effec-
tiveness and time of the entire separation process. The
geometry has to be adapted to the particular analyti-
cal task (Table 1). To minimize the separation time
the thickness of the donor solution layer should be as
thin as possible. The ratio of the membrane exchange
area to the donor solution volume should be maxi-
mized. To maximize the enrichment factor for dialy-
sis with enhanced selectivity the volume ratio be-
tween the donor solution and the acceptor solution
has to be maximized.
In this respect, thin hollow-Rbre membranes are
especially useful both for enrichment and puriRcation
procedures. Thin-layer chambers with Sat mem-
branes are also useful and enable a greater variety of
different membrane materials to be used. The minia-
turization of the membrane exchange area up to the
micro or the ultramicro scale enables reproducible
sampling from quiescent or slowly Sowing solutions
to be performed. This is of great importance for
in vivo sampling with microdialytic probes.
Figure 5 shows frequently used hollow-Rbre and
Sat-membrane set-ups. Table 1 summarizes the most
useful procedures for dialysis.
The Separation Membrane
The dialytic transport across thin membranes can be
described in eqn [33]:
dni
Ji" "kMF(ci,MD!ci,MA) [33]
dt
ci,MD and ci,MA are the solute concentrations in the
membrane at the interfaces with the donor and the
Acceptor solution
Flowing or stirred, large volume
Slowly flowing or gently stirred, small volume
Quiescent or slowly flowing, small volume
Quiescent or slowly flowing, small volume
1792 II / MEMBRANE SEPARATIONS / Reverse Osmosis

across gas-Rlled membranes or SLMs enables an

Figure 5 Frequently used dialysis set-up: (A) meander cell with
a flat membrane, (B) dialysis probe with a flat membrane M, (C)
hollow fibre membrane cell, (D) hollow fibre dialysis probe. ID, IA,
inlets to the donor and the acceptor chamber; OA, OD, outlets from
the acceptor and the donor chamber.
acceptor solutions, respectively. Linear concentration
gradients can be assumed in thin membranes.
The separation membrane should be considered
particularly with regard to selectively but also with
regard to the overall mass transfer kinetics. The mem-
brane material determines the transport mechanism,
which inSuences the selectivity of the separation pro-
cess in particular. Table 2 gives an overview about
the most important membrane materials and the
dominant transport mechanisms. Classic dialysis
through microporous membranes causes a loss of
sensitivity with respect to the following detection or
determination procedure. So-called selective dialysis
analyte enrichment to be performed. The selectivity
of the SLM technique can be enhanced by the addi-
tion of selectively reacting ligands to the liquid mem-
brane phase. When charged ions are complexed and
transported through these membrane systems elec-
troneutrality must be maintained. In many cases ion
pairs with selected counter ions are transported
through the membrane. When the ligand is dissolved
in the liquid membrane phase and the counter ion
cannot transverse the membrane the analyte ion
transport is coupled with a back-diffusion of an ion
with the same electric charge. A similar situation can
be found in ion-exchange membranes, which are used
to enrich ions by Donnan dialysis.
Gas dialysis through hydrophobic and micropor-
ous membranes is a fast transport process compared
with the other transport mechanisms. The diffusion
constants in the gas phase are several orders of magni-
tude greater than in liquid and solid phases. The
selectivity of the membrane transport is determined
by the ratio of the partial pressure pi of the analyte to
the total pressure p in the membrane pores. In small
pores the condensation and adsorption kinetics of the
gases also have to be taken into account.
Gas dialysis across homogeneous membranes is gen-
erally more selective. The different solubilities of the
gases in the membrane material are additional selec-
tion factors. The mass transport rate is considerably
smaller than those in microporous membranes.
Applications
Dialysis
Dialysis is mainly used in Sow analytical methods to
purify, dilute and condition sample solutions. It can
also be used to add reagents.

Table 2 Membrane transport and selectivity

Membrane material Transport mechanisms Factors which determine the selectivity

Hydrophillic and porous Diffusion through micropores Sieve effect

Hydrophobic and porous membranes,
filled with an organic solvent
As before, but with selective ligands
in the solvent
Hydrophobic and gas-filled porous
membranes
Ion exchange membranes
Homogeneous and hydrophobic
membranes
(kM"Dm / d M)
Diffusion through micropores
Diffusion through micropores
Gas diffusion and flow
Retarded diffusion through micropores
Solvation and diffusion of gases
and hydrophobic solutes with small
molecular masses
Solubility and sieve effect
Solubility, complexing and sieve effect,
co-ion transport
Volatility
Ion exchange and Donnan exclusion
Solubility in the membrane material

, Membrane porosity; , membrane tortuosity; dM, membrane thickness; Dm, diffusion coefficient.
 II / MEMBRANE SEPARATIONS / Reverse Osmosis
The dialysis module can be placed prior to or in the
sample insertion unit, between the sample insertion
unit and the reaction/separation zone and also into
the detector zone. Figure 5A shows a typical Sow-
through dialysis cell, which is inserted in many online
conRgurations with liquid chromatography (LC) and
Sow injection analysis (FIA), e.g. as shown in
Figure 6. In Figure 6A the dialysis cell is working
continuously. The acceptor stream is connected to the
injection valve, which adds the prepuriRed sample
into a nonsegmented carrier Sow stream. This conRg-
uration is frequently used in LC systems. The sample
substances are continuously separated from higher
molecular weight substances, precipitations and
microorganisms. Therefore dialysis is useful to pre-
vent blocking and prolong the lifetime of the separ-
ation column. The continuous dialytic sample pret-
reatment opens up an effective way to use FIA sys-
tems for the online process monitoring of animal cell
cultures and other industrial bioprocesses. In many
cases food, clinical and environmental samples can be
analysed without cumbersome ofSine sample pret-
reatment. To circumvent the increased detection
limit, which is caused by the inherent dilution, the
dialysis cell can be used as a Sow-through reactor, in
which the analyte is trapped as a more sensitively
detectable derivate. The acceptor stream can be stop-
ped for different times to control the reaction time.
The separated analytes, e.g. phenol derivatives or
aSatoxins, can also be reconcentrated on a solid-
phase extraction column, which is inserted into the
sampling loop of the injection valve. Also highly
speciRc preconcentration columns, e.g. with immobi-
lized antibodies, can be used.
In Figure 6B the dialysis membrane is contacted
only during short concentration impulses with the
sample solution. The probability of membrane foul-
Figure 6 Continuous (A) and pulsed dialysis (B) in flow analyti-
cal set-ups. IV, injection valve; DC, dialysis cell; B, carrier solu-
tion. See text for further explanations.
1793
ing and clogging is decreased considerably. The
sample solution is precisely diluted. This pulsed FIA
dialysis is frequently used to adapt ion-selective elec-
trodes, biosensors and miniaturized enzyme reactors
to biological sample matrices.
A very promising and expanding Reld of applica-
tion was opened up by the so-called microdialysis
technique. Miniaturized dialysis probes with tip dia-
meters smaller than 1 mm are implanted into differ-
ent tissues of living animals for sampling low molecu-
lar weight substances from the tissue Suid. The
low molecular weight analytes are separated across
small membrane (cellulose acetate, cut-off 1000-
60 000 Da) windows into a Sowing acceptor stream,
as shown in Figure 5B. The enrichment factor is con-
trolled by the Sow rate (0.5}25
L min\1). The
microdialysis probes are coupled online to micro-
column LC systems, to capillary FIA set-ups and
recently also to capillary electrophoresis devices. In
many cases the separated substances, e.g. lactate and
glucose, are converted by enzymes which are dis-
solved together with the cofactors and cosubstrates in
the acceptor solution.
Donnan Dialysis
Donnan dialysis across ion exchange membranes is
widely used as an efRcient suppression technique
in ion chromatography. When a cation exchange
membrane separates the eluent, which is an aqueous
Na2CO3 solution, from a reservoir of strong
acid, protons are transported into the eluent. Since
the membrane is impermeable to anions, the cations
from the eluent must diffuse simultaneously into the
suppressor reservoir. The chromatographically separ-
ated anions are combined with completely disso-
ciated acids, which have a high equivalent conductiv-
ity. The CO23\ ions are converted to the weakly
dissociated carbonic acid. In cation chromatography
the eluent, e.g. aqueous hydrochloric acid, has to
be suppressed. Therefore an anion exchange mem-
brane is used to separate the eluent from a sodium
hydroxide suppressor solution. The anions of the
eluent are replaced with hydroxide ions to increase
the conductivity in the separation peaks and to sup-
press the eluent conductivity by neutralization to
water. Donnan dialysis across ion exchange mem-
branes can also be used to neutralize strong alkaline
and acid samples in an online sample pretreatment
for ion chromatography.
Donnan dialysis is also used to enrich low molecu-
lar weight ions. When a cation exchange membrane
separates a high ionic strength solution from a low
ionic strength sample solution, cations are trans-
ported from the high ionic strength solution to the
1794 II / MEMBRANE SEPARATIONS / Reverse Osmosis
Table 3 Enrichment and sample pretreatment by Donnan dialysis (Cox (1992) and cited references)
Analyte
Cations
Pb(II), Cd(II), Tl(I)
Cd(II)
Cu(II)
Cu(II), Zn(II)
Ca(II)
Fe(II), Ni(II), Cr(III)
Co(II), Cd(II), Mn(II), Cu(II), Ni(II)
La(III), Nd(III)
Anions
AsO34\, PO34\
3 , SO4\ PO4\
2 3
Cl\, NO\
Pyruvate, chloroacetate
IC, ion chromatography; FAAS, flame atomic absorption spectrometry; ICP-AES, inducutively coupled plasma-atomic emission
spectrometry; ASV, anodic stripping voltammetry; DPASV, differential pulse ASV; CSV, cathodic stripping voltammetry; V, voltam-
metry.
low ionic strength solution. Since the membrane is
almost impermeable to anions, cations from the di-
lute solution must diffuse back to the more concen-
trated solution to maintain the electroneutrality. An-
ions can be analogously separated by means of an
anion exchange membrane.
Ions from dilute solutions, e.g. ground and
drinking water samples, can be enriched into high
ionic strength solutions with volumes which are
smaller than the sample volume. The acceptor solu-
tions can be adapted to whatever determination pro-
cedure is to be used. Both Sat and tubular ion ex-
change membranes are used, as shown in Figure 5A,B
and D.
Tubular ion exchange membranes with a small
inner diameter are particularly useful because of their
high surface area to internal volume ratios. Such
concentrators can readily be combined with Sow ana-
lytical set-ups, e.g. Sow injection analysis-atomic
absorption spectrometry (FIA}AAS) and ion
chromatography (IC) systems.
Because of the high osmotic pressure difference
between the donor and the acceptor solution and the
porosity of the ion exchange membranes a slow water
Sow takes place, which can be neglected in many
analytical applications (Table 3).
It should be noted that the interaction between the
diffusing ions and the Rxed counterions in the mem-
brane retards the diffusion, which can be accelerated
by an alternating electric Reld with a frequency
around 1 MHz. Multiply electrically charged ions,
e.g. Mg2# and Al3#, decrease the interaction be-
tween the sample cations and the Rxed counterions,
whereby the diffusional transport of the sample ca-
tions is accelerated.
Enrichment factor Detection
'100 FAAS
DPASV
'100 FAAS
ICP-AES
22 FAAS
40}80 ICP-AES
80}1000 IC
50 ICP-AES
5}10 CSV
15 IC
5}10 V
Selective Dialysis Across Solid and
Liquid Membranes
Selective dialysis is deRned here as a separation of
substances from an aqueous donor phase into an
aqueous acceptor phase, the phases being separated
by solid and liquid membranes. The analyte transport
through such membranes is based on a solva-
tion/diffusion mechanism in a lipophilic phase. The
liquid membranes have to be supported by a micro-
porous and hydrophobic layer to enable practical
applications to be performed. Again the basic set-up
of the Sow separation cells shown in Figure 5 can be
used to apply these membranes for separation pro-
cedures that can be coupled online to LC, gas
chromatography (GC) and FIA set-ups.
Thin silicon rubber membranes can be used to separ-
ate e.g. phenols and chlorphenols in a ‘push}pull’
procedure. The sample solution is acidiRed to shift the
chemical equilibrium to the nondissociated phenol/
chlorphenol species, which are dissolved in the silicon
membrane. The phenol molecules are trapped as phen-
olate ions in an alkaline acceptor solution. The max-
imum enrichment factor is determined by the pH
values in the donor and the acceptor solution.
Nonpolar organic substances can also be separated
from a sample donor solution into an acceptor solu-
tion, but the enrichment factor can only be increased
to values greater than 0.5 by addition of an organic
solvent to the acceptor solution.
The SLMs have the advantages of faster membrane
transport and easier modiRcation of the liquid phase,
which determines the transport mechanism and
the separation selectivity. Dialysis across SLMs
has a wide and growing Reld of application in
 II / MEMBRANE SEPARATIONS / Reverse Osmosis 1795

Table 4 Applications of supported liquid membranes

Analytes Conversion with Membrane transport Trapping by Detection

Mode A
Alcohols n-Heptane, supported by GC, LC, SnO2-sensor
microporous PTFE

Mode B
Aminesa OH\ n-Decane, supported H# GC, LC, ph
by microporous PTFE
Phenols H# n-Dodecane, OH\ LC, GC, ph
supported by
microporous PTFE
Carboxylates H# n-Nonane, OH\ LC, ph
supported by
microporous PVDF
Thiolates H# n-Dodecane, OH\ a, GC
supported by PVDF

Mode C
Cu(II)b PAR n-Pentane, which H# ph
contains di-2-ethyl-
hexylphosphoric
acid and is supported
by microporous PVDF
Pb(II)c Anions Phenylhexane, EDTA AAS
which contains bis(1-hydroxyheptyl-
cyclohexano)-18-crown-6,
supported by microporous PTFE
a
JoK nsson JA, LoK vkvist P, Audunsson G and Nilve G (1993) Mass transfer kinetics for analytical enrichment and sample preparation
using supported liquid membranes in a flow system with stagnant acceptor liquid. Analytica Chimica Acta 277: 9}24.
b
Barnes DE and Van Staden JF (1992) Flow injection analysis in the evaluation of supported liquid membranes. Analytica Chimica Acta
261: 441}451.
c
Izatt RM, Bruening RL, Bruening ML et al. (1989) Modelling diffusion } limited, neutral } macrocycle } mediated cation transport in
supported liquid membranes. Analytical Chemistry 61: 1140}1148.
a, amperometric: AAS, atomic absorption spectrometry; EDTA, ethylenediaminetetraacetic acid; GC, gas chromatography; LC, liquid
chromatography; PAR, 4-(2 -pyridylazo)resorcinol; ph, photometric; PTFE, polytetrafluoroethylene; PVDF, polyvinylidene fluoride.

environmental, food and clinical analysis. Table 4
summarizes some examples of application. Three
modes of selective separation are used:
1. The extraction of a hydrophobic substances into the
supported organic liquid phase and the following
back-extraction into the aqueous acceptor stream.
2. ‘Push}pull’ separation. The analyte is converted
into a membrane-soluble substance, which dif-
fuses through the membrane and is trapped as
a substance that is insoluble in the membrane.
3. Co-ion mediated transport on the basis of a carrier
substance which is dissolved in the liquid mem-
brane phase. The carrier molecules take up the
analyte molecules or ions, whereby a hydrophobic
complex or an ion-pair is formed.
Gas Dialysis
Gas dialysis can also be used for FIA procedures and
other Sow analytical methods to enhance their selec-
tivity. The conRgurations shown in Figure 5 can be
used to separate and enrich volatile or nonvolatile
analytes, which can be converted into a volatile sub-
stance. Table 5 gives an overview of the applications
of the gas dialysis technique to determine inorganic
substances.
The application range can be extended to volatile
and water-soluble organic compounds, e.g. lower al-
cohols (methanol, ethanol, propanol), formaldehyde,
acetaldehyde, acetone, ethylene oxide, propylene
oxide, ethyl acetate. Several nonvolatile species. e.g.
acetate, propionate, can be separated after acidiRca-
tion. Gas dialysis membranes can separate aqueous
solutions with very different pH values and ionic
strengths, which enables also extreme sample ma-
trices to be adapted to originally unsuitable detection
procedures. Microporous PTFE or polypropylene
membranes are used in most cases. However, it should
be noted that, for example, surfactants and many
water-soluble organic compounds can be adsorbed on
the membrane surface, which then becomes increas-
1796 II / MEMBRANE SEPARATIONS / Reverse Osmosis

Table 5 Applications of gas dialysis

Analyte Conversion to Trapping as or by Detection

Without conversion
ClO2, Cl2, Br2, I2 A colour or chemiluminescence a, ph, c, pot
reaction or reduction
NO, NO2 Oxidation to NO\3 pot (ISE)
N2H4 Oxidation a, ph, c

Conversion by acid}base reaction

CN\ HCN CN\, Ag(CN)\ 2 c, pot (ISE)
SCN\ HSCN SCN\, colour forming reaction pot, ph
3 , CO3\
2
CO2, HCO\ CO2 HCO\3 c, pot (ISE), ph
NH3, NH\ 4 NH3 NH#
4 , colour forming reaction or c, pot (ISE), pH
oxidation ph, ch, a

NO\2 NO NO\2 , NO\

3 or colour forming reaction ph, pot (ISE)
H2S, HS\, S2\ H2S S2\ pot (ISE), a, c
ph
3 , SO3\ SO2, SO24\ or colour forming reaction
2
HSO\ SO2 c, ph
F\ HF F\ pot (ISE)

Conversion by redox reactions

Cl\, OCl\, ClO\2 , ClO\
3 , Br\, BrO\
3 , l\ Cl2, Br2, I2 I\, Br\ and I\ or Cl2, Br2, I2 pot (ISE), a

a, amperometric: c, conductometric: ch. chemiluminometric: ISE, ion selective electrode: ph, photometric; pot, potentiometric.

ingly hydrophilized. Water penetrates into the mem-
brane pores and changes the mass transfer coefRcients
considerably with time. In some of these situations
homogeneous membranes, e.g. different silicone rub-
ber membranes, can be used to circumvent such inter-
ferences. Silicone rubber membranes are permeable to
hydrogen sulRde, hydrogen cyanide, carbon dioxide
and many volatile organic compounds.
Osmosis and Reverse Osmosis
Diluted sample solutions can be concentrated by both
osmosis and reverse osmosis. The concentration pro-
cess is based on a pressure gradient over a membrane,
which rejects the analyte molecules. High molecular
weight substances are already rejected by ultraRltra-
tion membranes. But the typical application of re-
verse osmosis is the separation of low molecular
weight substances from aqueous solutions to purify
the water or to concentrate the substances which are
to be determined.
To concentrate transition and heavy metal ions
from dilute aqueous solutions by osmosis the sample
solution is separated from a high ionic strength solu-
tion by a membrane which is permeable only to the
water. An osmotic pressure is built up, which then
propels the water into the acceptor (Rltrate) solution
and concentrates the donor solution. In a thin-layer
Sow-cell, which is similar to the cell shown in
Figure 5A and has a mechanically supported separ-
ation membrane (reverse osmosis membrane of cellu-
lose triacetate) preconcentration factors of up to 8}10
for copper(II), cadmium(II), manganese(II), nickel(II)
and zinc(II) can be achieved in a countercurrent Sow
regime.
To implement reverse osmosis, an outer pressure
(Figure 2) is applied to propel the water through the
membrane. An interesting practical aspect is the pos-
sibility of reducing the necessary outer pressure by an
osmotic pressure difference which has the same direc-
tion. This is implemented by high ionic strength ac-
ceptor solutions.
The concentration factor E (eqn [32]) increases
considerably with increasing reSection factors
(eqn [29])). Highly diluted sample solutions can be
concentrated to values that can be determined with
the available determination methods. The sample solu-
tion can be concentrated up to the precipitation of the
solute. Then an additional Rlter layer is used, which
can be exchanged and directly analysed, e.g. by X-ray
Suorescence spectrometry. Transition metals could be
analysed in drinking water up to the microgram per
litre level. Organic contaminants, e.g. chlorobenzene
and phenols in alkalized sample solutions, could also
be concentrated by reverse osmosis. After this precon-
centration traces of the contaminants could be ana-
lysed by LC and GC after their redissolution.
Further Reading
Cox JA (1992) Membrane methods for preconcentration of
trace components of solutions. In: Zeev B and Wai CM

(eds) Preconcentration Techniques for Trace Elements,
pp. 301}331. Boca Raton: CRC.
Cox JA and Twardowski Z (1980) Electric Reld enhance-
ment of Donnan dialysis. Analytical Letters 13(A14):
1283}1291.
Dasgupta P (1988) Approaches to ion chromatography. In:
Tarter JG (ed.) Ion Chromatography, pp. 191}338.
New York: Marcel Dekker.
JoK nson AJ, LoK vkvist P, Audunsson G and Nilve G (1993)
Mass transfer kinetics for analytical enrichment and
sample preparation using supported liquid membranes
in a Sow system with stagnant acceptor liquid. Analytica
Chimica Acta 277: 9}24.
Ultra\ltration
M. Cheryan, University of Illinois, Urbana, IL, USA
Copyright ^ 2000 Academic Press
Introduction
UltraRltration (UF) is a Rltration process that employs
a membrane to fractionate liquid mixtures containing
molecules that range in size from about 1000 daltons
in molecular weight to 500 000 daltons. The mem-
brane, made of either polymeric or inorganic mat-
erials, is a semipermeable barrier containing pores of
a certain size distribution that are used to retain or
‘reject’ components of the feed mixture that are larger
than the rated pore size while allowing molecules that
are smaller than the pores to pass through the mem-
brane. This separation process is very simple
(Figure 1) involving only the pumping of Suids. The
membrane is assembled in a particular conRguration
and placed in a module, and the feed stream is
pumped through the module over the membrane sur-
face in a cross-Sow mode. The pressure forces solvent
Figure 1 Cross-flow ultrafiltration. Particles in the feed that are
larger than the rated pore size of the membrane are retained in
the retentate stream while smaller particles pass through into the
permeate. (Adapted from Cheryan (1998) with permission from
Technomic.)
II / MEMBRANE SEPARATIONS / Ultra\ltration 1797
Robinson T and Justice JB (1991) Microdialysis in the
Neurosciences. New York: Elsevier Science.
Spohn U, Eberhardt R, Joksch B, et al. (1991) Enzymatic
multichannel-FIA methods for on-line fermentation
monitoring and control. In: GBF Monograph, vol. 14,
pp. 51}62. Weinheim: Verlag Chemie.
Stec RJ, Koirtyohann SR and Taylor HE (1986)
Preconcentration of trace elements from aqueous
solutions by osmosis. Analytical Chemistry 58: 3240}
3242.
Valcarcel M and Luque de Castro MD (1991) Non-
chromatographic Continuous Separation Techniques.
Cambridge: Royal Society of Chemistry.
(e.g. water) and solute molecules smaller than the
pores on the membrane surface through the mem-
brane into the ‘permeate’ stream while larger solutes
are rejected and retained in the ‘retentate’ stream. The
retentate is recycled through the module until the
required degree of puriRcation, separation or concen-
tration is achieved.
UltraRltration is similar in concept to other pres-
sure-driven membrane processes such as microRltra-
tion, nanoRltration and reverse osmosis. However, as
shown in Figure 2, the size range of the solutes that
are retained by each membrane is different. Reverse
osmosis (RO) membranes are designed to retain all
components except for the solvent (e.g. water). It is
essentially a concentration process. Owing to the
osmotic pressure of the solutes retained by RO mem-
branes, pressures needed to operate RO systems are
typically 30}60 bar (450}900 lb in\2). NanoRltra-
tion (NF) membranes have slightly larger pores and
are designed to allow monovalent salts such as
sodium chloride to pass through, but retains divalent
salts, disaccharides and dissociated organic acids.
Pressures are usually lower, about 15}25 bar. Micro-
Rltration (MF) membranes retain components that
are in suspension or in colloidal form, and is essential-
ly a clariRcation process. Pressures are usually
1}4 bar.
UltraRltration, on the other hand, is designed to
retain macromolecules and other solutes in the size
range of 1}50 nm, or with equivalent molecular
weights of 1000 to 500 000 daltons. It also operates
at low pressures (2}6 bar) and can simultaneously act
as a concentration, puriRcation and fractionation
process, depending on the components in the feed and
the membrane properties. It has several advantages
over other separation or concentration techniques.
Unlike freeze concentration or evaporation, there is
1798 II / MEMBRANE SEPARATIONS / Ultra\ltration
Figure 2 Examples of compounds of various sizes separated
by different membranes. (Adapted from Cheryan (1998) with
permission from Technomic.)
no change in phase of the solvent and thus energy
consumption is much lower. Being a nonthermal pro-
cess, there are no extremes of temperature and feed
solutions can be concentrated by UF with little or no
thermal damage to heat-sensitive components. Since
pores are large enough to allow passage of soluble
salts, acids and alkalis, the microenvironment of
the solution remains largely unchanged during the
process.
There are several factors that affect ultra-
Rltration applications: the membrane material,
properties of the membrane, process engineering
parameters, design of the membrane module,
fouling and cleaning and process design. The most
important performance parameters in UF are Sux and
rejection. Flux is the volume of permeate per unit
time per unit membrane area. Higher Sux means
lower capital and operating costs. Rejection is
a measure of a membrane’s separating capabilities. It
is deRned as:
Concentration of solute in the permeate
1!
Concentration of solute in the retentate
Membrane Material
Membranes have been made from over 150 different
polymers or inorganic materials, but only about
a dozen have achieved widespread commercial use for
UF. The most common are polymers such as
polysulfone, polyethersulfone, polyvinylidene Suor-
ide, polyacrylonitrile, cellulose acetate and regen-
erated cellulose as well as inorganic materials such
as alumina, zirconia and titania. Most polymeric
UF membranes are asymmetric in structure, i.e.
they have a thin ‘skin’ 0.1}0.2
m thick on the sur-
face of the membrane. This skin contains the pores
of the required size and determines the separation
characteristics of the UF membrane. Polymer layers
under the skin usually consist of voids which support
the skin layer. The skin and void layer are one struc-
ture and one polymer when made by a phase-inver-
sion process, but they could be two or more different
polymers in composite membranes. The membrane is
then laid on a backing such as polyester or polypropy-
lene and then formed into the module. In some cases,
such as hollow Rbres, a concentrated solution of the
polymer is spun or extruded to form self-supporting
single polymer hollow tubes with the pores on the
inside surface of the tube.
Inorganic membranes have considerably widened
the range of membrane applications, particularly in
food processing, waste treatment, recovery of chem-
icals and biotechnology applications, where high
temperature, acid and/or alkali stability, steam steril-
izability and cleanability are important. A macropor-
ous substrate of a Rne dispersion of the powder is Rrst
formed, e.g. by thermal sintering of an extruded paste
of the powder. If a tubular geometry is used, pastes
from two powders of different grain sizes may be
co-extruded, with the Rner grain being closer to the
axis. After baking at high temperatures ('10003C),
the inside may be coated by slip casting with the Rnal
Rne grain powder. A series of such layers may be
necessary to obtain the asymmetric-type ultrastruc-
ture. The membrane is Rnally set by a series of press-
urizing, drying and baking steps. The most common
ceramic materials are
-alumina, zirconia and titania.
Composites of zirconia or titania membranes
on alumina, carbon or stainless-steel supports are
available.
Most inorganic membranes are available in tubular
form, either as a single channel tube or multi-channel
element, the latter containing up to 60 individual
circular channels, depending on the relative diameters
of the channel and the element. The inner diameter of
individual channels vary from 2 to 6 mm and lengths
from 0.8 to 1.2 m. As many as 99 of these elements
may be put together in a single housing, resulting in

8}12 m2 per module. Normal process ratings are
15 bar and 1503C.
Inorganic membranes have several desirable prop-
erties. They are inert to common chemicals and sol-
vents and have wide temperature limits. Depending
on the seals and type of housing, some inorganic
membranes can be operated as high as 3503C and
within wide limits of pH from 0.5 to 13.5. The
biggest advantage is their extended operating life-
times. Operating life of membranes is most affected
by the frequency and nature of the cleaning regime. In
contrast to polymeric membranes which typically
have 9}18-month lifetimes with normal daily clean-
ing cycles, inorganic membranes are able to tolerate
frequent aggressive cleaning regimes. Many are still
operating 10}14 years after installation with the Rrst
set of membranes. One major limitation is that they
are 10}30 times more expensive than polymeric
membranes.
Membrane Properties
Pore size is the most important property of a UF
membrane. Pores can be visualized using electron
microscopy. Surface porosity (the proportion of the
membrane surface occupied by pores) is less than
10% for many UF membranes. In an ideal membrane,
all pores should be of the same size. In reality, there is
a distribution of pore sizes, as shown in Figure 3. This
makes it difRcult to get a sharp separation of similarly
sized molecules by UF. A common method to charac-
terize UF membranes is to challenge the membrane
with several macromolecules of known molecular
sizes. Since proteins of different molecular weights
are usually used as molecular markers, UF mem-
branes are characterized in terms of their ability to
retain proteins of a particular molecular weight. Fig-
ure 3 is a graphical representation of solute rejection
data for ideal and real membranes. No membrane
will display the sharp pore-size distribution shown
for the ideal membrane. MF membranes are given
‘absolute’ ratings which is the largest particle that will
be retained by the membrane, based on actual tests
under standard conditions. In contrast, UF mem-
branes are given ‘nominal’ ratings which refer to the
molecular weight of a test solute (ideally it should be
a globular protein) which is 90% rejected by the
membrane under standard conditions. This rating is
termed the molecular weight cut-off (MWCO) of the
membrane.
Proteins are not ideal compounds to use for this
purpose, since their molecular size can be affected by
pH, ionic strength and interactions with buffer com-
ponents. Proteins can have different isoelectric
points, solubility and hydrophobicity, thus causing
II / MEMBRANE SEPARATIONS / Ultra\ltration 1799
them to interact with and foul the membrane to
different extents, which affects measured rejections.
In addition, proteins which differ by 10 times in MW
may only differ by three times in size in their globular
form. Owing to the difRculty of Rnding proteins that
are sufRciently pure (and inexpensive) to conduct
MWCO evaluations, other compounds such as poly-
ethylene glycols (PEG) and dextrans have been used
because they are water soluble and can be readily
obtained with well-deRned and narrow-size distribu-
tions. Since the shapes of these various compounds
are different, the MWCO proRle of a membrane will
also differ depending on the solute test marker used.
Environmental conditions such as pH and ionic
strength also affect shape and conformation of mol-
ecules which can affect rejection.
Other components in the feed solution could affect
the separation of the target compound. For example,
with UF membranes, low-molecular-weight solutes
(such as sugars and salts) have molecular sizes much
smaller than the smallest pore on the membrane.
These compounds will be freely permeable, i.e. they
will have zero rejection, unless they interact with or
bind to impermeable compounds in the feed.
Changes in operating conditions will not affect their
permeability. On the other hand, large solutes that
are much bigger than the pores will be completely
rejected (i.e. 100% rejection). Its rejection prop-
erties will also be relatively unaffected by operating
conditions or if other compounds are present. How-
ever, if the solute has a size that is of the same order of
magnitude as the pore, its rejection may be
affected in the presence of the large molecule. This is
because the large molecule forms a secondary
Figure 3 Typical molecular-weight profile of ideal and real
membranes. Relationship shown is between molecular size of
a solute in the feed stream and rejection of the solute by the
membrane. (Adapted from Cheryan (1998) with permission from
Technomic.)
1800 II / MEMBRANE SEPARATIONS / Ultra\ltration

dynamic membrane on the original membrane that

inhibits passage of the smaller molecule. Operating
conditions that change the shape or conformation of
the solute will also affect its rejection.
As a general rule, fractionation of polymers can be
accomplished if there is at least a 10-fold difference in
molecular weight. Separation of similarly sized mac-
romolecules can be enhanced by diluting the feed
to minimize solute}solute interactions and solute-
membrane interactions.
Other factors affecting separation are operating
parameters such as pressure and cross-Sow rate.
These control the degree of turbulence and the thick-
ness of the boundary layer and extent of concentra-
tion polarization (deRned below), which in turn affect
permeability of smaller solutes.

Operating Parameters
Separation of solutes by UF membranes occurs by
a sieving mechanism. The transport of Suids through
the pores is modelled as laminar Sow through chan-
nels, with Sux directly proportional to applied trans-
membrane pressure. However, it has frequently been
observed that under certain operating conditions, Sux
becomes independent of pressure as shown in
Figure 4. This is owing to ‘concentration polariza-
tion’ which is shown in Figure 5. Molecules or par-
ticles that are partially or completely retained by the
membrane accumulate on the surface of the mem-
brane during ultraRltration. This build-up of solids
will cause a concentration gradient within the bound-
ary layer, resulting in back-transport of solute into
the bulk stream owing to diffusion. Eventually
a steady state is reached where the two phenomena
balance each other. Solute concentration reaches
a maximum at the ‘gel concentration’. This con-
Figure 4 Effect of operating conditions on flux of an ultrafiltra-
tion system. (Adapted from Cheryan (1998) with permission from
Technomic.)
Figure 5 Concentration polarization in ultrafiltration.
solidated gel layer is the reason that pressure indepen-
dence in Figure 4 is observed. Flux is no longer con-
trolled by pressure but by the mass-transfer charac-
teristics of the system which in turn depends on
the diffusion coefRcient of the rejected molecules in
the boundary layer, turbulence in the Sow channel,
viscosity and density of the Suid stream. Higher tem-
peratures lead to higher Sux because of its favorable
effect on diffusivity and viscosity. In the pressure-
independent region, Sux decreases in a semi-logarith-
mic manner with bulk feed concentration and in-
creases with higher turbulence (usually achieved by
higher Sow rates through the module).
Module Design
There are basically six different designs of membrane
modules: tubular (with channel diameters greater
than 3 mm), hollow Rbre or capillaries (self-support-
ing tubes, usually 2 mm or less internal diameters),
plates, spiral-wound, pleated sheets and moving mod-
ules (e.g. rotating discs or cylinders). Figure 6 shows
the more common types of modules. The selection of
a particular design depends on (a) the physical prop-
erties of the feed stream and retentate, especially
viscosity and osmotic pressure, (b) particle size of
suspended matter in the feed, (c) fouling potential of
the feed stream, and (d) sanitation requirements,
such as cleanability and sterilizability. The viscosity
of feed streams containing macromolecules such as
polymers or proteins will increase nonlinearly with
concentration above a certain value. This will require

Figure 6 Schematic of ultrafiltration membrane modules:
tubular, plate and spiral-wound.
high pressure drops for pumping and require the use
of modules that can withstand high pressures, elimin-
ating most hollow Rbre/capillary modules. On the
other hand, these modules have extremely high pack-
ing densities (surface area : volume ratios) and com-
paratively low energy consumption, making them
useful in applications where the feed is of relatively
low viscosity and low in suspended matter.
Spiral-wound modules and some plate modules in-
corporate a spacer in the feed channel to keep the
membrane sandwich apart. This spacer can add con-
siderable turbulence to the Suid Sow and thus in-
crease the Sux. However, this spacer causes a para-
sitic drag and creates dead spots in the feed channel,
which can cause suspended particles to block the Sow
channel, resulting in high pressure and cleaning prob-
lems.
Fouling and Cleaning
Fouling manifests itself as a decline in Sux with time
under constant operating conditions. The sieving
properties of the membrane may also change. This is
owing to irreversible interactions between feed com-
ponents and the membrane, causing a layer of foulant
on the membrane, blinding of the pores and an in-
creased resistance to Suid Sow through the mem-
brane. Many membrane materials listed earlier
are relatively hydrophobic (e.g. polysulfone, poly-
vinylidene Suoride) and tend to foul more than
hydrophilic membrane materials (e.g. cellulosics, cer-
II / MEMBRANE SEPARATIONS / Ultra\ltration 1801
amics). Many feed components interact strongly with
membranes, e.g. oils through hydrophobic interac-
tions with hydrophobic membranes, proteins by hy-
drogen bonding, charge interactions or hydrophobic
interactions, and salts by precipitation or charge in-
teraction.
A fouled membrane has to be cleaned according to
the nature of the foulant. Proteins can be effectively
cleaned with alkaline solutions, salts are removed
with acid cleaners. The quality of the water is very
important in ensuring a membrane can be effectively
cleaned in the shortest time possible.
Flux can be enhanced by periodic backwashing,
pulsating Sows, uniform transmembrane pressure or
co-current permeate Sow techniques. These have
been found to be effective in maintaining high Suxes
with feed streams containing colloidal or suspended
matter and less effective with foulants that are in
solution.
Applications of Ultra\ltration
Table 1 is a listing of ultraRltration applications. The
food industry has been one of the most successful
users of UF, starting from the early 1970s when it was
used to treat cheese whey to recover the protein.
Another successful application has been electrocoat
painting, where the UF system is used to maintain the
ionic balance of the painting system and to recover
paint that has been washed off. Biotechnology has
beneRted tremendously by UF, where it Rnds its
greatest use in the production of pyrogen-free water
and for fractionation, puriRcation and concentration
Table 1 Applications of ultrafiltration
Food industry
Concentration of protein and fat for cheese manufacture;
fractionation of cheese whey for whey protein concentrates; clari-
fication of fruit juices (apple, apricot, citrus, cranberry, grape,
peach, pear, pineapple; gelatin concentration and de-ashing;
eggs concentration and reduction of glucose; animal blood
concentration; soybean protein concentrates and isolates; clarifi-
cation of protein hydrolysates; vegetable oils (degumming,
deacidification, bleaching, removal of metals, dewaxing; clarifying
frying oils); sugar refining; dextrose clarification; alcoholic
beverages
Chemicals and wastewater
Electrocoat paint; oily wastewater; stillage from bioethanol plants;
caustic and acid recovery; brine recovery; printing ink; laundry
wastewater; textile industry; latex emulsions; pulp and paper
industry; tanning and leather industries; fish processing; poultry
industry
Biotechnology
Separation and harvesting of microbial cells; enzyme recovery;
affinity ultrafiltration; membrane bioreactors
1802 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation
of proteins and other macromolecules. Continued
advances in membrane science and manufacture and
engineering improvements to modules and systems
will allow a greater penetration of this technology in
a variety of industries in the future.
See also: II/Membrane Separations: Filtration; Micro-
filtration.
Further Reading
Bhave RR (ed.) (1991) Inorganic Membranes. Synthesis,
Characteristics and Applications. New York: Van Nos-
trand Reinhold.
Cheryan M (1998) UltraTltration and MicroTltration
Handbook. Lancaster, PA: Technomic.
Cheryan M and Alvarez J (1995) Membranes in food pro-
cessing. In: Noble RD and Stern SA (eds) Membrane
PARTICLE SIZE SEPARATION:
Electric Fields in Field Flow Fractionation
See II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Electric Fields
PARTICLE SIZE SEPARATION
Electrostatic Precipitation
J. J. Harwood, Tennessee Technological University,
Cookeville, TN, USA
Copyright ^ 2000 Academic Press
Introduction
Electrostatic precipitators (ESPs) are the most com-
monly used devices for the removal of Rne particles in
exhaust from industrial and utility processes. Wire-
plate ESPs consist of three or more sections of arrays
of large (e.g., 15 m;5 m), grounded metal collector
plates between which are situated wire or other nar-
row, high voltage electrodes (Figure 1). Less com-
monly, a wire-cylinder electrode conRguration is
used. Particles entering the Rrst section are quickly
charged by ions generated by the plasma coronas
around the wires. (Current does pass between the
Separations. Technology, Principles and Applications,
p. 415. Amsterdam: Elsevier.
Cheryan M and Rajagopalan N (1998) Membrane
treatment of oily streams. Wastewater treatment and
waste reduction. Journal of Membrane Science 151:
15}38.
Ho WSW and Sirkar KK (eds) (1992) Membrane Hand-
book. New York: Chapman and Hall.
Hsieh HP (1996) Inorganic Membranes for Separation and
Reaction. Amsterdam: Elsevier.
Lloyd DR (ed.) (1985) Materials Science of Synthetic Mem-
branes. Washington, DC: American Chemical Society.
Matsuura T (1994) Synthetic Membranes and Membrane
Separation Processes. Boca Raton, FL: CRC Press.
Mulder M (1991) Basic Principles of Membrane Technol-
ogy. Norwell, MA: Kluwer Academic Publishers.
Singh N and Cheryan M (1998) Membrane technology in
corn reRning and bio-products processing. Starch/
Sta( rke 50: 16}23.
electrodes, hence the term ‘electrostatic’ is not really
accurate, but indicates the small current-to-electrode
area.) The charged particles are drawn toward and
deposit upon the collector plates, which are period-
ically cleaned by mechanical ‘rapping’. This method
is very efRcient in removing particles in the
1}'10
m range. The most common use of ESPs is
in control of exhaust from coal combustion utilities.
Precipitators are also used in the cement, ore smelt-
ing, steel production, pulp and paper manufacturing,
and chemical processing industries, and in waste
combustion utilities. Small units are used in cleaning
domestic and workplace air, and have been con-
sidered for use in animal production facilities.
M. HolReld Rrst demonstrated the removal of par-
ticles by electrostatic charging in 1820. HolReld
showed that tobacco smoke can be cleared in a bottle
by applying a spark-producing voltage to a pointed
electrode inserted in the bottle. In 1850, C. F. Guitard
observed that a steady corona discharge is effective in
dissipating smoke. Sir Oliver Lodge Rrst attempted
 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation
Figure 1 Rigid frame electrostatic precipitator. (Reproduced with permission. Copyright Wheelabrator Air Pollution Control Inc.)
the commercial application of this phenomenon by
constructing an electrostatic precipitator to control
fumes at a lead smelter. Unfortunately, lead oxide is
too resistive to allow sufRcient particle charging with
a constant DC discharge, and the application failed.
The Rrst successful commercial application of elec-
trostatic precipitation was by F. G. Cottrell, who
designed an ESP to remove acid mist from sulfuric
acid plants. Cottrell facilitated the establishment of
the technique by developing the Rrst suitable high-
voltage power supplies. ESP particle control became
widely used between 1910 and 1930.
Removal of small particles from gas streams is
important for process, health, and environmental rea-
sons. Centrifugal separators are used to remove par-
ticles larger than 10
m; fabric Rlter ‘baghouses’ are
used to remove Rner particles and constitute a viable
alternative to ESPs. Capital costs are similar for bag-
houses and ESPs; however, maintenance, including
energy cost, is lower with ESP units.
The increased control of sulfur and nitrogen
oxides in the 1970s, and recent calls for the control
of very Rne particles, challenge scientists and de-
signers concerned with the ESP technique. Use of low
sulfur coal reduces Sy ash resistivity below that ap-
propriate for conventional ESP operation. Treatment
by the addition of limestone to coal increases the
particle load on precipitators. Conventional precipi-
tators do not efRciently remove submicron particles.
As will be discussed in this chapter, researchers are
developing means of adapting ESP design and opera-
1803
tion to meet the challenges of improved environ-
mental control.
Characteristics of Fly Ash
Hart et al. recently performed a comprehensive in-
vestigation of the composition and mineralogy of Sy
ash from three utility boilers. Using instrumental neu-
tron activation analysis and X-ray Suorescence spec-
troscopy, they found Si, Al, Fe and Ca to account for
more than 90% of major elements from all three
boilers. Scanning electron microscopy with energy
dispersive spectroscopy revealed successive ESP ashes
to be composed mainly of spherical particles which
decrease in average diameter with increasing distance
from the boiler. Concentrations of As, Co, Cr, Ni,
Mo and Sb increased from bottom ash through the
sequence of ESP ashes. These trace elements are vola-
tilized and transported to cooler regions, where they
condense or are adsorbed onto Sy ash particles. The
major Sy ash mineral phase found by these and other
researchers is quartz (SiO2) with magnetite (Fe3O4),
anhydrite (CaSO4), and mullite (Al6Si2O13) among
other minerals commonly present.
Resistivity (
) is the most important property of
material to be collected by an ESP. The optimum
range of resistivity is 104}1011 -cm. On collection,
low resistivity particles (
4103 -cm) release charge
to the collector plate and may be re-entrained. Par-
ticles with
'1011 -cm insulate the collector plate,
ultimately producing a sufRciently large electric Reld
1804 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation

within the dust layer to cause a counterproductive
‘back corona’.
Two types of resistivity may be important in par-
ticle collection in an ESP. Ions collected at the surface
of particles control ‘surface resistivity’, which domin-
ates at temperatures below 2503C. As indicated in
Figure 2, particle resistivity Rrst increases with tem-
perature, then decreases. Removal of the surface Rlm
(adsorbed water) by heating in vacuo at 3603C
eliminates this initial increase in resistivity. Above
2003C, removal of adsorbed material no longer af-
fects resistivity, and at higher temperatures resistivity
is attributed to ions in the bulk of the particles,
‘volume resistivity’. Both types of resistivity are prim-
arily functions of Na# and Li# ion concentrations,
and in some cases K# and I\ ions.
Two aspects of the
}T relationship affect ESP
collection efRciency. First, the maximum resistivity
of Sy ash occurs within the range of temperature at
which ESPs are commonly operated, 130}1803C.
Second, SO3, produced from sulfur in coal,
adsorbs onto the Sy ash particles and has traditionally
been responsible for lowering the resistivity of the
particles to the optimum range for collection. Present
use of low sulfur coals ((1% S) leads to inadequate
collection.

Figure 2 Effect of surface film resistivity on flue-dust resistivity. (Reproduced with permission from Busby HGT and Darby K (1963)
Journal of the Institute of Fuel 36(268): 184. Copyright The Institute of Fuel).
 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation 1805

Operation at non-optimal temperature can be time (s) and:

avoided by lowering the temperature, but this re-
quires energy input to a cooling device, and also can "40/N0eb
lead to difRculties with corrosion due to condensa-
where N0 is the number density of particles (m\3); e is
tion. The temperature of the exhaust is normally
the electronic charge (C); b is the ion mobility
cooled by heat exchange in an air pre-heater prior to
(m2 s\1 V\1). (This and other formulae are in the
injection into the precipitator, hence the name ‘cold-
form presented in Oglesby and Nichols (see Further
side’ precipitator. ‘Hot-side’ precipitators operate at
Reading.) The corresponding expression for diffusion
temperatures as high as 3703C; resistivity is often
charging presented by White is:
reduced to a desirable level of 2;1010 -cm at above

3153C. However, difRculties are encountered with
the greater volume of the hot gas, and these units
require more careful construction.
More commonly, resistivity of high-resistance ash
is lowered by chemical conditioning.
Theoretical and Practical Aspects
of Electrostatic Precipitation
Charging of Particles
The corona surrounding a discharge electrode of an
ESP is a gas plasma of electrons and cations. The
cations are drawn into the electrode, which is biased
negative. The electrons are repelled into the interelec-
trode space where, within one centimetre travel, they
ionize gas particles. These gas particles are drawn
toward the grounded collector electrodes. On the


akT avN0e2t)
q(t)" ln 1#
e kT
where k is the Boltzmann’s constant (J K\1); v is the
rms thermal velocity of ions (m s\1); T is the absolute
temperature (K).
Smith and McDonald derived an expression for
combined diffusion and particle charging. They
divided the surface of a particle being charged into
three regions: region I, where the electric Reld of the
particle and that within the ESP duct intersect, sweep-
ing ions onto the particle surface; region II, where the
radial component of the electric Reld of the particle
dominates; and region III, where the electric Reld of
the particle and duct are in the same direction, sweep-
ing ions away from the particle. Giving the electric
Reld about the particle, Er, as:


way, many of them collide with and attach to the !1 a3 ne
Er"E0 cos  1#2 3 !
particles to be collected, charging them. The charged #2 r 40r2
particles are then drawn towards the collector elec-
trodes. This particle charging is efRcient. Typical where Er is the radial component of electric Reld
charging time is 5 ms; typical residence time in an ESP (V m\1); E0 is the external Reld (V m\1); r is the
is 2}5 s. radial distance to point of interest (m);  is the azi-
Charging processes are of two types: Reld charging, muth angle measured from the z axis (toward the
where ions moving with the electric Reld charge the discharge electrode) (rad); ne is the particle charge
particles, and diffusion charging, where charging is (C); these researchers produced a combined charging
due to collision with ions moving with random ther- rate equation:
mal motion. The types are distinguished to facilitate
mathematical description. Field charging is dominant dq dq dq dq
" # #
in charging particles with radius (a) greater than dt dtI dtII dtIII
0.5
m; diffusion charging is dominant with


(0.2
m particles. Both mechanisms are important n 2

in the intermediate size range. "(N0Anse2/40) 1!

ns
The charge on an ion at time t (q(t)) resulting from
Reld charging, derived by Pauthenier and Moreau- L/2

 

Hanot is: ea2vN0 ne2(r0!a)
# exp!
2 40kTar0
 t F0
q(t)"12 0a2E0


#2 t# [3ar20!r30(#2)#a3(!1)]eE0 cos 
# sin  d
kTr20(#2)
where  is the relative dielectric constant; 0 is the
permittivity of free space (C2 N\1 m\1); a is the par- ea2vN0
# exp(!ne2/40akT)
ticle radius (m); E0 is the electric Reld (V m\1); t is the 2
1806 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation

Figure 3 Modes of particle charging. }䉬}, field charging; }䊏}, diffusion charging; }䉱}, total charge; };}, Boltzman charge AC.
(Data from Kanazawa et al. (1993) Journal of Electrostatics 29: 193. Copyright: Elsevier Science.)

where A is the surface area of the particle on
which ions may impinge (m2); ns is the saturation
charge; 0 is the greatest angle of region from z axis
(rad).
This expression agrees well with measured charg-
ing rates.
Representative saturated particle charge as a func-
tion of particle radius is presented in Figure 3.
Chemical Conditioning
Conditioning can enhance collection efRciency by re-
ducing resistivity to prevent back corona, increasing
cohesivity of the collected layer, enhancing the elec-
tric Reld by increasing the space charge contribution
(see below), and increasing agglomeration of small
particles. Some exhaust gas conditioning methods
which have been used to enhance ESP performance
are cooling by in-leakage of cold air, heat exchange to
a waste heat boiler, evaporative spray, and chemical
conditioning. The latter has proven most popular in
utility applications.
Most commonly, sulfur is combusted and the SO2
produced is catalytically converted to SO3. On com-
bination with the combustion exhaust, SO3 hydro-
lyses to produce H2SO4, sulfuric acid. The acid de-
posits efRciently on Sy ash particles, decreasing the
particle surface resistivity. SO3 is added at levels of
1}10 ppm; up to 30 ppm can be added without signif-
icant increase in sulfur emission from the ESP. SigniR-
cant SO2 is produced on combustion of even low
sulfur coal, but only about 1% is converted to the
useful SO3. Knowledge of the importance of SO3 was
initially gained through experience with conditioning
of smelter dusts by sulRde ores.
Ammonium salts were found to be effective when
H. J. White was consulted on the problem of de-
creased ESP efRciency which resulted from a change
in petroleum cracker catalyst. He found that replac-
ing ammonia, released by the initial catalyst,
 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation
recovered ESP performance. As NH3 increases collec-
tion efRciency with both low and high resisitivity
boiler ash, and may or may not decrease resistivity, it
is thought that this conditioner functions by furnish-
ing salt particles which increase the space charge
contribution to the electric Reld and, by increasing
dust cohesiveness, reduce rapping losses.
Characteristics of the Corona Discharge
As voltage applied to a wire discharge electrode is
increased, visible discharge points, ‘tufts’, begin to
appear dispersed along the wire. Eventually the tuft
pattern stabilizes, appearing as Rngers which elongate
and retract at 1}10 Hz and remain at somewhat Rxed
locations along the wire. This stabilization of the
corona is indicated by a distinct increase in the
slope of the electric Reld vs. applied voltage curve
(Figure 4). Peek determined a semi-empirical expres-
sion for the corona onset voltage (Vc) in air:
Vc"3;106 am(1#0.03(/a)ln(b/a)
where Vc is the corona onset voltage (V); a is the
radius of discharge electrode (m); m is the wire rough-
ness factor (0.5}0.9);  is the relative air density; P is
the air pressure (Pa).
This discharge produces a current density (with
respect to the area of the collector electrode) in
the order of 150 nA cm\2. While the ionic composi-
tion of the interelectrode space has not been well
studied, ionization properties indicate that this cur-
rent is carried by oxygen, sulfur dioxide, and water,
which are negatively charged by the electron current
Figure 4 Electric field on the collector electrode just under the
corona wire as voltage is applied; d, discharge electrode dia-
meter. 2D"20 cm; *, d"0.5 mm; 䉭, d"0.71 mm; 䊐,
d"1.09 mm. (Reproduced with permission from Ohkubo et al.
1985.)
1807
Figure 5 Pilot ESP V}I curves before and after development of
severe back corona. *, 1 hour after startup; 䊐, 8 days after startup.
(Reproduced with permission from DuBard and Nichols 1990.)
near the corona, and by particles ionized by these
primary ions.
The Electric Field
Current vs. voltage curves (Figure 5) are commonly
used in monitoring the operation of ESP. For best
collection efRciency, ESP units are operated at the
maximum voltage which produces a stable current.
A discussion of operation monitoring of ESPs is given
by DuBard and Nichols.
Much effort has been given to the task of determin-
ing the electric Reld within an ESP. Besides the burden
of inability to solve known expressions analytically
for the Reld in a wire-plate electrode system, this task
is made difRcult by uncharacterized changes in the
collector electrode as the dust layer collects on it. The
advent of high-speed computing has allowed satisfy-
ing solution of the Reld equations for a clean system.
Cristina et al. present a modern method of deter-
mining the electric Reld. These researchers use the
following expressions for charge and electric
potential:
) ( )"!Q
) (Q )"0
where  is the permittivity of air (C2 N\1 m\1); Q is
the ionic charge density (C m\2); is the electrical
potential (V).
These equations have a unique solution for DC
ion Sow Relds. An iterative numerical procedure
1808 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation
is used to balance the electric Reld and charge density
of elements in an array of triangular elements super-
imposed on the inter-electrode space. These elements
are each speciRed to have a constant ionic charge
density. The contribution of space charge by ionized
particulate matter is disregarded in this treatment. In
the more dust-laden regions of an ESP, particulate
matter is found to alter electric Reld slightly.
Equipotential Reld maps developed by this treat-
ment show a fairly linear decrease in potential from
the discharge to the collector electrode, with circular
Relds to about one-third distance across the channel
(from 57.5 kV to 20 kV), then equipotential Reld
lines parallel to the collector. Near the collector elec-
trode, at about Rve-sixths the channel width, the
potential has decreased to 5 kV. Electric Relds of
adjacent discharge electrodes interact, producing
bell-shaped electric Reld patterns radiating from each
wire to the adjacent collector plates.
Back Corona
In working ESPs, a portion of the applied potential
is dropped across the resistive layer of the collected
dust. As the dust layer thickens the resistance,
hence voltage drop, across the layer increases. If
the dust resistivity is high (
'1010 -cm), eventually
the Reld becomes great enough to cause electrical
breakdown of gas within the dust layer, generating
a back corona. Electrons are ripped from the gas, and
the resulting positive ions Sow into the inter-electrode
space, where they attach to negatively charged par-
ticles. The diminished net charge of the particulates
leads to greatly diminished collection efRciency; re-
entrainment also occurs. A back corona occurs with
an electrical Reld strength across the dust layer (resist-
ance of layer times current) of 8}12 kV cm\1.
A back corona can be visually observed as a glow
on the collection surface. It is also characterized by an
increase in corona onset voltage followed by an in-
crease in current for a given applied voltage as com-
pared with a correctly functioning system (Figure 5).
The dust layer acts as a resistive-capacitative element
in having a lag time before discharge. Both intermit-
tent and pulse energization take advantage of this lag
time to allow the application of high voltage without
the occurrence of back corona.
Current Density
The electric Reld at ESP collection surfaces is not
uniform due to both the electrode}collector geometry
and to irregular resistance resulting from uneven dust
layer thickness. The parameter determined in study-
ing the situation at the collection surface is current
density, j (nA cm\2). With poor distribution of cur-
rent over the dust layer, collection efRciency is dimin-
ished. Re-entrainment can occur in regions of low
current density, and high local currents can lead to
back corona even with a low average current. Land-
ham et al. measured the current density distribution
in a pilot scale ESP by inserting a segmented copper
electrode sensor board into a collector electrode.
Conventional, intermittent, and pulse energization
with both wire and barbed-strip discharge electrodes
were investigated. (Barbed-strip electrodes are used
by some manufacturers to force a more uniform cur-
rent distribution at the collector.)
Intermittent energization (IE) is achieved by inter-
rupting the rectiRer circuit output for one to twenty
half cycles of the supply power line. A baseline volt-
age is maintained, with relatively broad pulses super-
imposed. In pulse energization (PE), pulses of 1
s to
1 ms width and up to !75 kV are superimposed on
a DC voltage set just below the ‘spark limit’. Both
techniques have been found to save energy and to
reduce the incidence of back corona. PE is more
effective in countering problems with low resistivity
dust. Dubard and Nichols set IE to one full AC cycle
on and four off (duty cycle 0.2), and PE base voltage
!20 kV with 50 Hz, &125
s width pulses of
!25 kV. Conventional full-wave rectiRed energiz-
ation (CE) was set at about !35 kV.
These researchers found that CE and PE give Gaus-
sian current density proRles across the sensor plate
surface, while IE gives a more even, but lower magni-
tude distribution. Under good operating conditions
(clean collector surface), PE had less collector area
above dust layer breakdown current than CE or IE,
with 96% useful area as compared to 83% and 86%,
respectively. Under severe back corona conditions, PE
maintained 84% useful area, CE 12%, and IE 54%.
Similar results were obtained with both barbed-strip
and wire electrodes. The results indicate a linear rela-
tionship between the increase in collector area receiv-
ing useful values of current density and collection
efRciency.
Among alternative discharge electrode designs,
barbed plate discharge electrodes appear more
effective in producing uniform current distribu-
tions (see McKinney and Davidson). This design
may also have some advantage with respect to Sow
turbulence.
Flow
Besides electric Reld and current density distribution,
gas Sow is important in ESP performance. ESP Sow
velocities vary between 0.2 and 2.0 m s\1. Schwab
and Johnson suggest an optimum velocity of between
 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation
1.22 and 1.52 m s\1. Lower velocities diminish the
turbulent mixing which brings small particles into the
low Sow region near the collector electrode; higher
velocities overwhelm electrostatic attraction of
particles to the collector electrode, and can lead to
re-entrainment.
Flow within an ESP is inherently turbulent due to
high gas Sow velocity, which typically results in
a Reynolds number of about 10 000 } Rve times that
at which turbulent Sow replaces laminar Sow. Turbu-
lence in ESPs is complicated by Sow obstructions
} discharge electrodes and collector stiffening bafSes
and connectors, and by a non-uniform Sow proRle at
the ESP inlet. At low velocities (:0.5 m s\1) turbu-
lence resulting from the ion current between dis-
charge and collector electrodes (ionic wind) contrib-
utes signiRcantly. The Sow regime in an ESP is thus
seen to be quite complicated. BafSes within the ESP
duct and porous plates at the inlet and outlet are
employed to create uniform Sow.
Schwab and Johnson have produced a computer
model based on Navier}Stokes Suid Sow equations as
an alternative to traditional reduced scale physical
models for Sow design. The model can be used to
determine inlet plate perforation patterns which pro-
duce uniform Sow without high Sow near the duct
walls.
Collection Ef\ciency
Most important in ESP operation is the particulate
collection efRciency, , the fraction of particulates
removed. Around 1920, Deutsch and Anderson inde-
pendently derived an expression for , now com-
monly referred to as the ‘Deutsch equation’:


A
"1!exp ! e 40kaE0EpCDf
Q "
where A is the collector electrode area (m2); Q is the
volumetric gas Sow rate (m3 s\1); e is the
effective particle migration velocity (m s\1).
EfRciency is also expressed as fractional particle
penetration, P"1!. A/Q is the speciRc collection
area (SCA), the parameter of importance in sizing an
ESP. Assumptions of the Deutsch equation are that
particles are instantaneously fully charged and accel-
erated, turbulent and diffusive forces cause particles
to be distributed uniformly in any cross section of the
precipitator, the velocity of the gas stream does not
affect the migration velocity of the particles, particle
motion is governed by viscous drag where Stokes’
Law applies, the effect of collision between ions and
neutral gas molecules can be neglected, and there are
1809
no disturbing effects such as erosion, re-entrainment,
uneven gas Sow distribution, sneakage, or back co-
rona. A number of researchers have adjusted the
Deutsch equation to account for neglected factors.
Especially questionable is the assumption of complete
turbulent mixing (for instance, see Zhibin and
Guoquan). The Deutsch model represents one mixing
extreme, with the other extreme being laminar Sow (no
turbulence). The Deutsch model tends to underestimate
for particles with diameter less than about 1
m, and
overestimates the collection of larger particles.
The Matt}OG hnfeldt modiRcation attempts to ac-
count for variation in particle size distribution:


m
A
"1!exp ! k
Q
where k is the particle migration velocity (m s\1);
m is an exponent which depends on particle size
distribution.
An m of 0.5 is commonly used with Sy ash or
cement dust; increasing values can be used with Rner
size distributions.
In comparing ESP performance under varying
conditions, the migration velocity is more useful than
, being independent of ESP collection area and gas
Sow rate. An important consideration in using
values is that migration velocity of large particles
decreases at a much lower rate with increasing gas
Sow (or decreased plate size) than that of small par-
ticles, hence though small particles will be less efR-
ciently collected, measured can increase with gas
Sow rate.
Migration of charged particles can be treated
theoretically. Hall presents the expression:
6
where k is 3/(#2);  is the dielectric constant of ash
(C N\1 m\1); E0 is the particle charging electric Reld
strength (V m\1); Ep is the particle collecting electric
Reld strength (V m\1); a is the particle radius (m); C is
the Cunningham slip correction; D is the factor to
account for diffusion charging of Rnes; f is the factor
accounting for gas turbulence and free electron charg-
ing; is the gas viscosity (Pa s); is the fractional
particle saturation charge.
C is used to adjust diffusion of particles with size
comparable to mean-free path of gas molecules. The
reader can understand that empirical estimation of
values, from measured efRciency, is often more
productive than theoretical estimation.
1810 II / PARTICLE SIZE SEPARATION / Electrostatic Precipitation
The Deutsch model is often used in determining
a minimum speciRc collection area in sizing precipita-
tors. Extrapolation from performance of pilot or
other ESPs should be made only with similar designs
and similar dusts.
Computer Modelling
Several comprehensive models of particle collection
in wire-plate ESPs are available, including the
Southern Research Institute (SRI), Research Triangle
Institute Sectional (RTIS), Italian Electricity Board
(ENEL), and Electric Power Research Institute
(ESPM) models. As an example, the ESPM model will
be brieSy described (see Lawless and Altman).
In ESPM, the applied voltage for a given current
density is Rrst determined for a position directly be-
neath the wire:


z#1
V"Vc#Ecrw z!1!ln
2
where Ec is the electric Reld at wire surface at corona
onset (V m\1); rw is the radius of wire (m) and:


2
jb b
z" 1#
0 Vw
where b is the wire-plate separation (m); is the ion
mobility (m2 s\1 V\1).
Current densities at other plate locations are cal-
culated as:
j()"j0cosn()
where  is the angle from between wire and plate
(rad); n is a function of j, 2 (near corona onset) to 4.
Ion mobility is based on that of SO2 and SO\ 2 .
Space charge resulting from the charging of particles
is treated as an equivalent ion current and used to
adjust the operating voltage.
Non-dimensional parameters are used in models to
add generality to results and to simplify mathematical
treatment. For instance, the initial particle charging
rate in an ESP model (ESPM), combining both diffu-
sion and Reld charging, is:


d 3w  2
" 1! #1
d 4 3w
where  is the non-dimensional charge;  is the non-
dimensional exposure time; w is the non-dimensional
electric Reld.
Intermediate and saturation charging rates are
modelled by different expressions. Back corona
charging can be modelled with ESPM, though better
understanding of this process is needed to assure
accurate results.
Particle collection is modelled with a modiRed
Deutsch expression:


Nw
A/Q
P" 1!
Nw
Nw is a parameter to adjust for turbulence: a high
value approaches Deutschian total mixing, a low
value approaches laminar collection (laminar when
Nw"1). To correct for uneven gas Sow, penetration
is calculated at several gas velocities and the results
combined by weighted averaging. Rapping loss can
be speciRed or estimated by applying conventional
dynamics to the falling of agglomerated ‘cake’ during
collector cleaning. The model treats each ESP section
separately, ‘remixing’ particulates exiting from each
section.
ESPM can be used to predict V}I curves, ‘grade
efRciency’ (penetration as a function of particle dia-
meter), and other aspects of ESP performance
under varying conditions.
Innovative ESP Designs
Several new means have been developed to facilitate
collection of submicron and highly resistive particles
by ESPs.
Pulse energization is discussed above (see Current
density). By superimposing a brief, high voltage pulse
on a DC voltage set just below the back corona level,
efRcient Rne particle charging and collection are ob-
tained with highly resistive dusts.
Enhanced particle agglomeration can also facilitate
Rne particle collection. For instance, Kanazawa et al.
have accentuated particle agglomeration with
a precharging section, in which the gas stream is
divided and particles in the two sub-streams given
either a positive or a negative charge. On recombin-
ing the sub-streams, particles agglomerate through
electrostatic attraction. Watanabe et al. have devised
an alternative method utilizing a three sector design.
Large particles are removed in the Rrst ESP sector.
A ‘modiRed quadrupole’ electrode system in the sec-
ond sector applies an AC Reld to enhance particle
collision, hence agglomeration. The third section is
again a conventional ESP, which removes the agglom-
erated particles.
 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Electric Fields
Fabric Rltration is the surest means of removing
Rne particles. EPRI has devised the ‘Compact Hybrid
Particulate Collector’ (COHPAC). This design simply
places a baghouse after an ESP. The ESP removes
much of the particulates, easing the load on the
baghouse, hence reducing maintenance. The bag-
house reduces pollution due to rapping loss, and is
insensitive to changes in fuel.
EPRI has also developed an ‘EPRICON’ process
which can replace conventional chemical condition-
ing of Sy ash from low sulfur coals. In this process,
a portion of the gas stream is diverted to a vanadium
oxide-based catalytic unit, which efRciently converts
SO2 to SO3. Recombination of the treated stream
with the bulk results in the necessary conditioning of
the Sy ash.
See also: Particle Size Separation: Hydrocyclones for
Particle Size Separation; Sieving/Screening.
Further Reading
Busby HGT and Darby K (1963) EfRciency of electrostatic
precipitators as affected by the properties and combus-
tion of coal. Journal of the Institute of Fuel 36(268):
184.
Field Flow Fractionation: Electric Fields
S. N. Semenov, Institute of Biochemical Physics,
Russian Academy of Science, Moscow, Russia
Copyright ^ 2000 Academic Press
Introduction
Field Sow fractionation (FFF) represents a class of
separation techniques, which use a force Reld perpen-
dicular to the direction of separation to control the
longitudinal velocity of particles injected into the
system. It is achieved by particle redistribution in the
Sow with a parabolic velocity proRle due to the ac-
tion of a transverse force. This transverse force may
be due to an electric Reld, a centrifugal or gravity
Reld, etc. In electric FFF (ElFFF), the transverse move-
ment of the particles is caused by an electric Reld.
The transverse particle velocity, U, is deRned by
1811
Dubard JL and Nichols GB (1990) Diagnosis of electrical
operation of electrostatic precipitators. Journal of Elec-
trostatics 25: 75.
Hart BR, Powell MA, Fyfe WS and Ratanasthien B (1995)
Geochemistry and mineralogy of Sy-ash from the Mae
Moh lignite deposit, Thailand. Energy Sources 17: 23.
Kanazawa S, Ohkubo T, Nomoto Y and Adachi T (1993)
Submicron particle agglomeration and precipitation by
using a bipolar charging method. Journal of Electro-
statics 29: 193.
Landham EC Jr., Dubard JL and Piulle W (1990) The effect
of high-voltage waveforms on ESP current density distri-
butions. IEEE Transactions on Industry Applications
26(3): 515.
McKinney PJ, Davidson JH and Leone DM (1992) Current
distributions for barbed plate-to-plane coronas. IEEE
Transactions on Industry Applications 28(6): 1424.
McKean KJ (1988) Electrostatic precipitators. IEE Pro-
ceedings Pt. A. 135(6): 347.
Oglesby S Jr. and Nichols GB (1978) Electrostatic Precipi-
tation. New York: Marcel Dekker Inc.
Tachibana N (1989) Back discharge and intermittent ener-
gization in electrostatic precipitation of Sy ash. Journal
of Electrostatics 22: 257.
Watanabe T, Tochikubo F and Koizumi Y et al. (1995)
Submicron particle agglomeration by an electrostatic
agglomerator. Journal of Electrostatics 34: 367.
Zhibin Z and Guoquan Z (1994) Investigations of the collec-
tion efRciency of an electrostatic precipitator with turbu-
lent effects. Aerosol Science and Technology 20: 169.
the expression:
U"b ) E [1]
where b is the particle electrophoretic velocity, and
E is the electric Reld strength in the channel interior,
which is available both for the particles and the Sow
of the carrier liquid. The particle electrophoretic mo-
bility is related to the particle electrokinetic potential
(zeta-potential):


 R
b" f [2]
4 
where  is the dielectric constant of the carrier liquid,
is the carrier liquid viscosity,  is the particle elec-
trokinetic potential, R is the particle diameter,
and  is the Debye length characterizing the screening
1812 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Electric Fields
of the electrostatic interaction in an electrolyte.
f (R/) is a function changing monotonously from
1 for particles of R to 1.5 for small particles, when
the zeta-potentials are small. For higher zeta-
potentials, this function approaches a minimum of
less than one. Thus, the particle electrokinetic poten-
tial and electrophoretic mobility represent the para-
meters slowly changing with the particle size and
depending mainly on the surface properties of the
particle. For small objects like macromolecules and
low-molecular-weight ions, the theory of the elec-
trophoretic mobility is absent.
For the characteristic relaxation time, the Boltz-
mann transverse particle distribution is established
in the system by forcing injected particles toward
the wall of the channel and their thermal (diffusion)
motion. In ElFFF (reported as the method for protein
separation), particles of the same size with higher
electrophoretic mobility or zeta-potential will accu-
mulate more closely to the wall, while particles of
lower zeta-potential will form a more diffuse layer
that extends further into the Sow of the carrier liquid.
Proteins still represent most of the ElFFF sample
materials. For particles with about the same zeta-
potential, the thickness of this layer may also be
different, if the particles have different diffusion
coefRcients, D. Particles with higher diffusion
coefRcient (i.e., with smaller size) will accumulate
in a more extensive layer due to more intensive ther-
mal movement. Zeta-potential is an important para-
meter interrelated to the particle surface charge den-
sity, and characterizing the particle surface properties
and the possible exchange of substances between the
particle and the surrounding liquid, e.g. in cellular
processes, including transport through cell mem-
branes, antigen}antibody interactions, and hormonal
control.
ElFFF is carried out in a thin channel of rectangular
cross-section with the width to thickness ratio (aspect
ratio) about 100 (thickness about 10}100 microns). It
allows the separation to approximate to a laminar
liquid Sow between inRnite parallel plates, which is
characterized by a parabolic velocity proRle, where
the Suid velocity at the channel walls is zero and
reaches a maximum in the centre of the channel.
Thus, if a group of particles maintain an average
distance from the wall different from another group
of particles, their velocities along the channel will be
different and they will leave the channel at different
times, related to the particle zeta-potential and size,
which deRnes the particle diffusion coefRcient. In FFF
systems, the same types of Relds are used as in the
so-called ‘direct Reld methods’ (centrifugation, elec-
trophoresis, etc.), but there is no requirement of com-
plete fraction resolution in the Reld direction, and
Reld strengths may be lower. In principle, all mixtures
separated by direct electrophoresis may be effectively
analysed by ElFFF, if they have a size large enough to
form a layer of thickness smaller than the channel
thickness, even when its electrophoretic mobility is
too small for electrophoretic analysis. FFF systems
are elution methods and allow the collection of frac-
tions during a separation. Since the theory of FFF
dynamics is well developed, the separation times for
a given sample can be directly related to the physical
parameter of the particles. This parameter represents
the effective particle charge qH, which deRnes the
thickness of the Boltzmann particle distribution
+exp (qHE ) x/kT) (x is the transverse coordinate in
the channel) in the transverse electric Reld applied to
the ElFFF channel. Using the known Einstein rela-
tionship, this effective particle charge may be deRned
as the ratio of the particle electrophoretic mobility
multiplied by the thermal energy kT, to its diffusion
coefRcient:
b
qH"kT [3]
D
In principle, this effective charge itself represents
a new separation parameter, which may be used for
particle and macromolecule characterization, if the
theory is developed. This theory should relate the
effective charge and the particle and macromolecule
physicochemical parameter important in speciRc ap-
plications, for example, the surface density of charged
groups raised in dissociation or ion adsorption. In
turn, this effective charge could be used for the elec-
trophoretic mobility or zeta-potential determination,
if the particle diffusion coefRcient is determined inde-
pendently, and the system temperature is known.
Another possibility is to separate particles with the
same surface properties (i.e. zeta-potentials) but dif-
ferent sizes, where the sample selectivity is only due to
the differences in diffusion coefRcient. Of course, the
real applications of ElFFF are deRned by speciRc
experimental conditions, opportunities and advant-
ages rather than by method theory, but, without
a clear physicochemical understanding of macro-
molecule and particle behaviour in ElFFF, the method
applications will be very limited.
A focusing (or hyperlayer) mode of operation using
isoelectric focusing in a pH gradient across the chan-
nel has been reported by a number of authors (see
Further Reading) with a channel of trapezoidal cross
section. However, the latter separation mode loses
the high resolution characteristic for the FFF family
due to high hydrodynamic dispersion interrelated to
the shape of the cross section.
 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Electric Fields
Limitations of Channel Con\guration
and Construction
Planar electrical FFF systems have been constructed,
where the walls are deRned by membranes permeable
to the carrier liquid but not to the particles to be
separated. In later versions these membranes are sup-
ported by porous or perforated plates. The mem-
branes prevent the loss of the sample while allowing
the passage of an electric current. Platinum wire
electrodes lying outside the channel provide the
transverse electric Reld. The carrier liquid represents a
buffer, and a solution of identical composition is
circulated through the chambers housing the elec-
trodes in order to remove electrolysis products and
reduce electrode polarization. Other conRgurations
such as a hollow Rbre systems and an annular porous
glass channel have been reported. Presently, graphite
or gold-plated glass channel walls are used, which
minimize these effects. In a micro-machined channel
for ElFFF, the electrode walls were of titanium and
gold.
The resolution increases with the effective voltage
drop across the channel, and increasing the applied
voltage will have a positive effect on the resolution.
Unfortunately, ElFFF is mainly carried out in aqueous
solutions, and voltages above 1.7 V applied across the
Suid-wall interface will cause signiRcant electrolysis
and bubble formation. Since the ElFFF system re-
quires a stable Sow velocity proRle, bubbles cause
serious Sow disturbances, and electrolysis must be
avoided. High Sow velocities can limit the formation
of bubbles and allow voltages above 1.7 V, but the
available voltage is still small. Thus bubble formation
in the electrochemical reactions at the wall electrodes
is the limiting factor for the applied voltage.
Another voltage-related difRculty is the potential
drop in the electric double layers near the channel
walls. An electric double layer will cause most of the
voltage drop to occur very close to the channel walls.
As a consequence, the effective Reld which may be
used for particle redistribution and separation, in the
ElFFF system, is greatly reduced. Experimental data
indicate that the effective Reld strength in the Sow
inside the channel is in the range of 0.25%}1% of the
applied Reld depending on the composition of the
buffer. Though this effect greatly reduces the perfor-
mance of the ElFFF system, it is still able to perform
separations. Due to the low electric Reld strength,
Joule heating is not expected to be a signiRcant prob-
lem in ElFFF, in contrast to electrophoresis systems.
One problem that arises in the separation in FFF
systems is the symmetric parabolic velocity proRle of
the Sow that performs the separation function. In-
deed, particles with equal and opposite zeta-potential
1813
will elute at the same time from the channel. This can
be a problem for samples containing both positively
and negatively charged particles. Though such par-
ticles are usually prone to coagulation or aggregation,
these processes may be slowed by a steric stabiliz-
ation, for example, as in polyelectrolyte solutions or
particles with adsorbed polymers. The coagulation or
aggregation kinetics may then be studied by ElFFF.
Most samples, and especially biological samples, are
of a uniform charge type. Biological samples contain
mostly negatively charged particles (at least the par-
ticles of interest). For samples containing both types
of particle charge, the asymmetrical Sow velocity
proRle and an additional particle velocity asymmetry
may be arranged by the application of a longitudinal
electric Reld in a channel having walls with different
zeta-potentials, which can cause electroosmotic Sow
with a non-uniform velocity proRle.
The establishment of Boltzmann equilibrium distri-
bution in the channel requires a relaxation time for
a particle to migrate from one channel wall to the
other in the applied electric Reld. If the drift velocity
U is constant, the relaxation time will be found using
eqn [1]:
w
r" [4]
U
In conventional ElFFF systems, the relaxation time
typically is over 5 min, but in micro-machined ElFFF
systems it may be less than 3 s.
ElFFF and Related Electrophoretic
Methods
The nearest relatives of ElFFF are different elec-
trophoresis techniques in liquids. However, elec-
trophoresis systems often require very high Reld
strengths for resolution, and the high voltages are
limited by the Joule heating. Particle electrophoretic
mobilities should also be high enough to have an
acceptable resolution in common capillary lengths.
Another type of electrophoresis separation system,
free-Sow electrophoresis, utilizes an electric Reld
across a curtain of buffer between two vertical plates
(a principle very close to FFF) and allows for continu-
ous sample injection, but limits the detection and
collection systems to the certain number of fractions.
Resolution in free-Sow electrophoresis is limited
by fraction spreading in the Suid stream caused by
the parabolic Sow proRle. Other methods of separat-
ing molecules and cells are needed for applications,
where these limitations preclude the use of existing
systems. Field Sow fractionation is the solution
for some applications, where particle electro-
phoretic mobilities are too low for the conventional
1814 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Electric Fields
electrophoresis systems and high electric Reld
strengths are not desirable. ElFFF has all the advant-
ages of FFF systems, i.e., the ability to separate cells,
large molecules, colloids, emulsions, and structure,
which are too delicate for electrophoretic separation
such as liposomes, both in the ‘original’ condition
and after surface modiRcation. Unlike the free-Sow
electrophoresis systems, elution in FFF systems is
zonal and proceeds through one outlet port; for this
reason, it is capable of signiRcantly higher resolution.
Therefore, anticipated applications of ElFFF systems
include cell separations, characterization of emul-
sions, liposomes, and other particulate vehicles for
intravenous drug administration with respect to size,
charge, and stability, diagnostic tests for speciRc mol-
ecules in colloidal suspensions, quick and accurate
separations of molecules, environmental water
monitoring, tests for sample contamination, and fur-
ther research involving zeta-potentials. ElFFF systems
also Rnd application as sample pretreatment systems
by performing an initial separation on a sample,
which is later collected for further testing by another
analysis system.
The resolution of ElFFF is determined in the stan-
dard way for chromatography, i.e. by the comparison
of the sum of the peak dispersions for two neighbour-
ing peaks to the difference of their maxima. The
resolution of ElFFF is inversely proportional to the
separation distance of the electrodes; thus, the smaller
the distance between the channel walls, the higher the
resolution between two distinct particles, making
ElFFF an ideal application for using micro-machining
techniques. The resolution increases with the square
root of the channel length, so the longer the channel
the better the resolution, but the time required for the
improved resolution increases, which is not generally
desirable.
Combined Electric-Thermal FFF
(El-ThFFF)
An interesting combined technique represents the
application of an electric Reld across the channel
for Thermal FFF, where a temperature gradient is
used to separate the analysed particles. After a
potential drop of about 2 V is applied, the ThFFF
retention is apparently changed. This combination
allows a more exact examination of the particle sur-
face properties, since ThFFF, like electrophoresis,
represents a surface kinetic phenomenon deRned by
the surface force potential. Also, this combined FFF
method gives information on particle electrokinetic
properties in non-aqueous solution. El-ThFFF may
allow distinction between electrostatic and non-
electrostatic interactions in surface layer by
programming of the electric Reld strength during the
separation.
Conclusion
ElFFF is a method for the separation of charged par-
ticles and macromolecules according to their effective
charge. This parameter is not measured by elec-
trophoretic methods and may be obtained directly
from ElFFF experiments. The effective charge may be
used for the characterization of surface particle prop-
erties and macromolecule charged groups, when the
required theory has been developed. ElFFF has electric
Reld strength limitations due to electrochemical reac-
tions at the channel walls and potential interface drop.
See also: II/Particle Size Separation: Field Flow Frac-
tionation: Thermal; Theory and Instrumentation of Field
Flow Fractionation.
Further Reading
Andreev VP and Ste panov YV (1997) Field Sow fractiona-
tion with asymmetrical electroosmotic Sow. II. Charged
particles. Journal of Liquid Chromatography & Relative
Techniques 20: 2873}2886.
Andreev VP, Stepanov YV and Giddings JC (1997) Field
Sow fractionation with asymmetrical electroosmotic
Sow. I. Charged particles. Journal of Microcolumn Sep-
arations 9(3): 163}168.
Caldwell KD and Gao Y-S (1993) Electrical Reld Sow
fractionation in particle separation. 1. Monodisperse
standards. Analytical Chemistry 65: 1764}1772.
Caldwell KD, Kesner LF, Mayers MN and Giddings JC
(1972) Electrical Reld Sow fractionation of proteins.
Science 176: 296}298.
Gale BK, Caldwell KD and Frazier AB (1998) A micro-
machined electrical Reld Sow fractionation (
-EFFF)
system. IEEE Transactions in Biomedical Engineering
45(12): 1459}1468.
Giddings JC, Shiundu PM and Semenov SN (1995) Ther-
mophoresis of metal particles in a liquid. Journal of
Colloidal Interface Science 176: 454}458.
Liu G and Giddings JC (1991) Separation of particles in
nonaqueous suspensions by thermal}electrical Reld Sow
fractionation. Analytical Chemistry 63(3): 296}299.
Martin M (1998) Theory of Reld Sow fractionation.
Advances in Chromatography 39: 1}138.
Martin M and Williams PS (1992) Theoretical basis of Reld
Sow fractionation. In: Dondi F and Guiochon G (eds)
Theoretical Advancement in Chromatography and
Related Separation Techniques, NATO ASI Series C:
Mathematical and Physical Sciences, vol. 383,
pp. 513}580. The Netherlands: Kluwer.
Schimpf ME, Russel DD and Lewis JK (1994) Separation of
charged latex particles by electrical Reld Sow frac-
tionation. Journal of Liquid Chromatography 17:
3221}3238.
 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Thermal
Field Flow Fractionation: Thermal
S. N. Semenov, Institute of Biochemical Physics,
Russian Academy of Science, Moscow, Russia
Copyright ^ 2000 Academic Press
Introduction
Field Sow fractionation (FFF) represents a class of
separation techniques, which are one-phase methods.
They are preferable for the separation and character-
ization of mixtures such as high molecular weight
polymers which might be modiRed or damaged in
two-phase separation methods like chromatography.
FFF uses a force Reld perpendicular to the direction of
separation to control the longitudinal velocity of par-
ticles injected into the system. It is achieved by the
particle redistribution in a Sow with a parabolic velo-
city proRle due to the transverse force action. In
Thermal FFF (ThFFF), the transverse movement of
the particles is caused by the particle ‘thermophoresis’
in the temperature gradient. The transverse particle
velocity, U, is deRned by the expression:
U"bT )
T [1]
where bT is the particle thermophoretic velocity, and

T is the transversal gradient of the temperature in
the channel. The particle thermophoresis is com-
monly related to the osmotic pressure gradient pro-
duced in the surface layer due to the temperature
gradient. This excess osmotic pressure is established
in the particle surface layer due to accumulation of
the solvent or dissolved molecules or ions in the
particle surface layer. This accumulation is related to
the particle surface potential . This surface potential
may have the electrostatic nature, when the particle
surface carries electric charge, or represent some kind
of dipole}dipole interaction, when ion adsorption or
surface group dissociation is impossible. The latter
situation should be characteristic for organic sol-
vents, where dispersion interaction between the par-
ticle surface and solvent molecules should play the
main role. The theory of thermophoresis is developed
mainly for particles larger in size than the character-
istic thickness of the surface layer and having moder-
ate surface potential of several kT (k is the Boltzmann
constant), which interacts with dissolved ions or mol-
ecules. These ions or molecules should be present at
a concentration low enough to avoid the excluded
volume effects in their accumulation in the particle
surface layer. In this situation, the particle ther-
1815
mophoretic mobility may be written as


 
3kc0  y 
bT" y dy 1#n e\(/kT) #1 !1
(2#n) 0 R kT
[2]
where c0 is the solute concentration in the carrier
liquid,  is the carrier liquid viscosity, R is the particle
diameter, n is the particle-to-liquid thermal conduct-
ivity ratio, and y is the transverse coordinate in the
surface layer. The immediate physical factor for the
particle thermophoresis is the ‘slip’ liquid Sow in the
particle surface layer due to the osmotic pressure
gradient, which is related to the temperature gradient
in the particle surface layer established along the
macroscopic temperature gradient in the liquid. Thus,
the main physical events in thermophoresis take place
near the particle surface, though the temperature
gradient near the particle surface playing the role of
the driving force for the particle is deRned by the
particle and liquid thermal conductivity, which are
bulk properties. However, one can expect that these
bulk properties will be the same for small particles
and larger samples of the material, and the particle
thermal conductivity can be obtained from literature
data on thermal conductivity of the material or inde-
pendent experimental determination. It means that
particle thermophoresis is mainly related to particle
surface properties. It becomes absolutely true for
metal particles with very high thermal conductivity,
when the parameter n in eqn [1] is very large. For
metal particles, the particle thermophoretic mobility
is a function of the particle surface properties only.
For smaller particles with higher surface potentials,
eqn [1] is not true due to intensive solute transport in
the particle surface layer.
This surface transport should be compensated by
the solute diffusion outside the surface layer, and
which, in turn, leads to the solute concentration
gradient and related electric Reld establishment (in
electrolytes) around the particle (so-called concentra-
tion polarization). However, for a particle with mod-
erate size and thermal conductivity having a surface
potential about two to four kT, we can state that
the particle thermophoretic mobility is deRned by the
particle surface properties and does not depend on its
size. For emulsion droplets, the thermophoretic
mobility in the absence of concentration polarization
1816 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Thermal

is determined as: mophoretic mobility will be accumulated more close-

ly to the wall, while particles of lower thermophoretic



3kc0R )
T  y mobility will form a more diffuse layer that extends
bT"! 1#n
(2#3i)(2#n) R further into the Sow of the carrier liquid. For particles
\ with about the same thermophoretic mobility, the


; e\(C/kT)

kT  
#1 !1 y dy
where i is the viscosity of the liquid inside the drop-
let. For homopolymer chains, it is shown by ThFFF
experiments, that chain thermophoretic mobility
does not depend on chain length and branching, and
one can expect that eqn [2] will deRne it accurately,
where R will be the monomer radius. The theory of
particle thermophoresis may be true, if some solutes
present at low concentration, for example, salt ions,
are accumulated around the monomers. However, in
true polymer solutions, where no dissolved extrinsic
solutes are present, excluded volume effects cannot be
neglected, and eqns [2] and [3] cannot be used for the
description of thermophoretic behaviour.
For calculation, the Boltzmann exponent indexes in
eqns [2] and [3] may be simpliRed using the approxi-
mation:
[3] thickness of this layer may also be different, if par-
ticles have different diffusion coefRcients, D. Particles
with higher diffusion coefRcient (i.e., with smaller
size) will be accumulated in a more extensive layer
due to more intensive thermal movement. The ther-
mophoretic mobility related to the surface potential is
an important parameter interrelated to the particle
surface charge density (where it represents the elec-
trostatic potential) and characterizing the particle
surface properties and the possible exchange of sub-
stances between the particle and the surrounding
liquid. Also, the separation of particles of the same
material but with different sizes may be important in
the characterization (molecular mass distribution) of
commercial latex and polymer particles.
ThFFF is carried out in a thin channel of rectangu-
lar cross-section with a width to thickness ratio
(aspect ratio) of about 100 (thickness about 10}100
microns). It allows the separation system to
approximate to the laminar Sow between inRnite


 x
+! 1!
kT h velocity proRle, where the Suid velocity at the channel
where  is the depth of the surface potential well in kT
units, and h is the characteristic width of this well.
Typical orders of values for different kinds of surface
potentials are present in Table 1, where A is the
Hamaker constant, d is the solute radius, q is the
solute electric charge,  is the particle zeta-potential,
and  is the Debye length.
For the characteristic relaxation time, the Boltz-
mann transverse particle distribution is established in
the system by forcing injected particles toward the
wall of the channel and by their diffusion motion. In
ThFFF, particles of the same size with higher ther-
parallel plates, which is characterized by a parabolic
[4]
walls is zero, and a maximum in the centre of the
channel. Thus, if a group of particles are accumu-
lated, maintaining an average distance from the wall
different from another group of particles, their vel-
ocities along the channel will be different. As a conse-
quence, they will leave the channel at distinct times,
related to the particle thermophoretic mobility and
size, which deRnes the particle diffusion coefRcient.
There are no other direct methods, where temper-
ature gradients are used for particle, droplet or mac-
romolecule separation. FFF systems are elution
methods and allow the collection of fractions during
a separation.

Table 1

Surface potential Analytical expressions for the  (y),  and h The ranges of values for parameters  and h

(y)  h  h
Van der Waals forces !A(d /r )6
A/kT d/3 5}50* 10\8 cm
(low-molecular surfactant)
Coulomb electrostatic forces !q e\(y/) q/kT  0}10 10\7!10\4 cm**
(aqueous electrolytes)
Adsorption forces None None None 0}10 +10\7 cm
Structure forces None None None 0}10 +10\5 cm

*The maximum values of Hamaker constant are reached for metals

**The maximum value of the Debye length is calculated for the deionized water.
 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Thermal
Channel Con\guration and
Construction Limitations
In thermal FFF the temperature gradient across the
channel thickness is maintained by the electrical heat-
ing of a polished metal block (usually a chrome or
nickel-plated copper block) forming one wall and
a cooled similar block forming the opposite wall. The
plating improves resistance to corrosion, the factor
limiting the range of permitted solvents and the separ-
ated particles and macromolecules. Cooling is usually
accomplished by passing cold water through longitu-
dinal holes bored in the block. To prevent thermo-
gravitational convection, it is the upper block that is
heated and the bottom one cooled. Thermocouples
are mounted in the blocks to control their temper-
atures. In the bottom block, capillaries for the intro-
duction and elution of the solvent and the sample are
placed. The copper blocks are separated by a spacer
of a polymer material with a low thermal conductiv-
ity (Mylar, Kevlar, etc.) to provide high temperature
gradients. The channel constructions are described
more exactly in the literature (see Further Reading).
One of the main advantages of the FFF family stems
from the uniform open channel geometry and the
well-deRned Sow proRle. As a consequence, retention
can be related directly to the physicochemical para-
meters of the analyte material and carrier liquid.
Possible deviations of the Sow proRle and the poly-
mer parameters due to the non-uniform temperature
distribution have been described.
Polymer Characterization
Progress in ThFFF instrumentation and methodology
has allowed a systematic study on the thermal diffu-
sion of polymer solutions. The success of these studies
is provided by the ability of ThFFF to produce accu-
rate values of thermal diffusion parameters using
small quantities of polymer (a few hundred micro-
grams). The values of the thermal diffusion coefR-
cients (parameters equal to the thermo-
phoretic mobility) have been obtained for 17 polymer}
solvent systems and are about 10\8}10\7 cm2/s ) K,
and their molecular masses are from about several
tens to about hundred Daltons. The results show the
correlation of the polymer thermophoretic mobility
with several polymer and solvent parameters, the
thermal conductivity of the polymer and solvent, the
polymer density, and the viscosity and viscous activa-
tion energy of the carrier liquid. Studies also demon-
strated a correlation of the polymer thermophoresis
parameters with the solvatation properties of the sol-
vent. Though conventional diffusion in polymer solu-
tions is well deRned, the thermal diffusion of poly-
1817
mers in liquids is not exactly understood and not well
characterized. Although there are equations relating
retention to experimental parameters and transport
coefRcients of polymers, values of polymer ther-
mophoretic mobility are not commonly available,
and a model for predicting them from physicochemi-
cal parameters is in progress only. Therefore, a calib-
ration is necessary for characterizing the molecular
weight distribution (MWD) of polymers by ThFFF
(although a single calibration point can be used, when
the dependence of the diffusion coefRcient on mo-
lecular weight is known). Calibration is simple in the
analysis of homopolymers because well-characterized
molecular weight standards are available for a variety
of polymers.
The characterization of copolymers presents
more problems because of the overlapping effects of
composition and molecular weight distribution
(MWD). Often it is necessary to characterize both the
MWD and the compositional distribution. In this
case the commonly used method of size exclusion
chromatography (SEC) is not adequate because the
separation is governed by size alone. Thus, molecular
weight fractions with different compositions may be
eluted in SEC simultaneously. In contrast, ThFFF
may separate polymers by both chemical composition
and size, and is therefore capable of yielding both size
and compositional information on copolymers. Sep-
aration by size in ThFFF is governed by differences in
the diffusion coefRcient of the polymers, while separ-
ation by chemical composition may result from dif-
ferences in the thermophoretic mobility.
The results on ThFFF of random and block
copolymers of polystyrene (PS) and polyisoprene (PI)
in tetrahydrofuran and cyclohexane show that for
random copolymers and block copolymers with
a random conRguration in solution, the ther-
mophoretic mobility is a linear function of monomer
composition. It may be a basis for obtaining composi-
tional information on such copolymers by ThFFF. For
copolymers with a radial segregation of monomers,
thermophoresis is determined mainly by monomers
located in the outer region of the polymer coil. The
dependence of retention on the radial distribution of
monomers provides a basis for evaluating bonding
arrangements in copolymers. The further progress in
copolymer characterization by ThFFF is related to
progress in the theory of the polymer thermophoresis.
Particle Characterization
Though most ThFFF samples are polymers, the abil-
ity of ThFFF to retain and separate both submicron
and micron size particles (latex and silica) suspended
in various organic carrier liquids has also been
1818 II / PARTICLE SIZE SEPARATION / Field Flow Fractionation: Thermal
demonstrated by Shiundu, Lee and Giddings. In
their article, the dependence of particle retention
on various factors (solvent properties, amount of
added electrolyte, particle size and composition,
and cold-wall temperature) is evaluated and
discussed.
Thermophoretic mobilities of several latex-
solvent combinations have been obtained from the
ThFFF retention data. The studies were carried out in
polar organic solvents, cyclohexane and aqueous car-
riers. As a rule, the thermophoretic mobilities of par-
ticles range from 10\8 to 10\7 cm2/s ) K, while the
particle size ranges from about 0.04 to about 1 mi-
cron.
The retention of colloidal particles in ThFFF dem-
onstrates a strong dependence on the chemical com-
position of the particles or their surfaces. These
results are observed among both similar particles
(such as latex particles) and different particles (in-
cluding latex particles, and inorganic and metallic
colloids). These compositional effects are observed
for particles suspended in both aqueous and non-
aqueous carrier liquids. Also, metal particles (e.g.
palladium) are less retained than silica particles, with
latex particles most retained. The resolution of par-
ticles of equal size in ThFFF experiments is also dem-
onstrated. Surface compositional effects were also
demonstrated in this study. These effects conRrm the
possibility of colloid particle surface analysis by ther-
mal FFF.
Combined Electric-Thermal
FFF (El-ThFFF)
An interesting combined technique represents the
application of an electric Reld across the channel
for Thermal FFF, where a temperature gradient is
also established. For details see Particle Size Separ-
ation / Field Flow Fractionation: Electric Fields.
Conclusion
ThFFF is a method for the particle and macro-
molecule separation in electrolyte and non-electrolyte
solvents according to their interaction with the sol-
vent molecules or/and ions of the added salt. For
particles, these are surface interactions, though the
particle/solvent thermal conductivity ratio is also
important. Thermophoretic mobility may be cal-
culated immediately from ThFFF experimental data,
when the particle or macromolecule diffusion coefRc-
ient is known. This parameter may depend on both
electrostatic and non-electrostatic (dispersion) inter-
actions and can be used for their characterization.
Also, the thermophoretic mobility may be used for
the characterization of surface particle and macro-
molecule properties, when the respective theory is
developed.
See also: II/Particle Size Separation: Field Flow Frac-
tionation: Electric Fields; Theory and Instrumentation of
Field Flow Fractionation. III/Colloids: Field Flow Frac-
tionation. Polymers: Field Flow Fractionation.
Further Reading
Giddings JC, Shiundu PM and Semenov SN (1995) Ther-
mophoresis of metal particles in a liquid. Journal of
Colloid Interface Science 176: 454}458.
Janca J (1992) Field-Flow Fractionation. Analysis of Mac-
romolecules and Particles. New York: Dekker.
Liu G and Giddings JC (1991) Separation of particles
in nonaqueous suspensions by thermal electric Reld-
Sow fractionation. Analytical Chemistry 63(3):
296}299.
Martin M (1998) Theory of Reld-Sow fractionation. Ad-
vances in Chromatography 39: 1}138.
Martin M and Williams PS (1992) Theoretical basis of
Reld-Sow fractionation. In: Dondi F and Guiochon
G (eds) Theoretical Advancement in Chromatography
and Related Separation Techniques, NATO ASI Series
C: Mathematical and Physical Sciences, vol. 383, pp.
513}580. The Netherlands: Kluwer.
Schimpf ME and Giddings JC (1987) Characterization of
thermal diffusion in polymer solutions by thermal Reld-
Sow fractionation: effects of molecular weight and
branching. Macromolecules 20: 1561}1563.
Schimpf ME, Wheeler LM and Romero PF (1993)
Copolymer retention in thermal Reld-Sow fractionation.
Dependence on composition and conformation. In:
Provder T (ed.) Chromatography of Polymers. Charac-
terization by SEC and FFF (ACS Symposium Series) vol.
521, pp. 63}76. Washington DC: American Chemical
Society.
Semenov SN (1997) Thermophoresis and thermal FFF in
electrolytes. Journal of Microcolumn Separations 9(4):
287}294.
Shiundu PM, Lee G and Giddings JC (1995) Separation of
particles in nonaqueous suspensions by thermal Reld-
Sow fractionation. Analytical Chemistry 57(15):
2705}2713.
 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation 1819

Hydrocyclones for Particle Size Separation

J. J. Cilliers, UMIST, Manchester, UK The operating principle is simple: the Suid, carry-
Copyright ^ 2000 Academic Press
ing the suspended particles, enters the cyclone tangen-
tially, spirals downward and produces a centrifugal
Reld in free vortex Sow. Larger particles move
Introduction through the Suid to the outside of the cyclone in
a spiral motion, and exit through the spigot with
The hydrocyclone is a static, continuous particle size
a fraction of the liquid. Due to the limiting area of the
separation device that can also be used for phase
spigot, an inner vortex, rotating in the same direction
separations, including solid}liquid, liquid}liquid and
as the outer vortex but Sowing upward, is established
liquid}gas separations and has been used for various
and leaves the cyclone through the vortex Rnder,
classiRcation duties since the 19th century.
carrying most of the liquid and Rner particles with it.
Hydrocyclones are attractive for industrial use as
If the spigot capacity is exceeded, the air core is closed
they have no moving parts, a small footprint, relative-
off and the spigot discharge changes from an um-
ly low capital and operating costs, and are simple to
brella-shaped spray to a ‘rope’ and a loss of coarse
operate. On the other hand, their operation is rather
material to the overSow.
inSexible once installed and single-stage efRciencies
The diameter of the cylindrical section is the major
may be low, especially for particles Rner than 10
m.
variable affecting the size of particle that can be
This article describes the mode of operation of
separated, although the outlet diameters can be
hydrocyclones, and the motion of Suid and solid
changed independently to alter the separation
particles in the classiRer. Quantifying the separation
achieved. While early workers experimented with cyc-
is followed by the effect of the major design and
lones as small as 5 mm diameter, commercial hydro-
operating variables on the efRciency.
cyclone diameters currently range from 10 mm to
Two modelling approaches are introduced: a fun-
2.5 m, with separating sizes for particles of density
damentally based model, including computational
2700 kg m\3 of 1.5}300
m, decreasing with in-
Suid dynamics (CFD), and empirical models, which
creased particle density. Operating pressure drop
are still in general use.
ranges from 10 bar for small diameters to 0.5 bar for
The article concludes with aspects of further
development.

Description
Hydrocyclones are cono-cylindrical in shape,
with a tangential feed inlet into the cylindrical section
and an outlet at each axis. The outlet at the cylin-
drical section is called the vortex Rnder and extends
into the cyclone to reduce short-circuit Sow directly
from the inlet. At the conical end is the second outlet,
the spigot. For size separation, both outlets are gener-
ally open to the atmosphere. Hydrocyclones are gen-
erally operated vertically with the spigot at the lower
end, hence the coarse product is called the underSow
and the Rne product, leaving the vortex Rnder, the
overSow. Figure 1 schematically shows the principal
Sow and design features of a typical hydrocyclone:
the two vortices, the tangential feed inlet and the
axial outlets. Except for the immediate region of the
tangential inlet, the Suid motion within the cyclone
has radial symmetry. If one or both of the outlets are
open to the atmosphere, a low pressure zone causes
a gas core along the vertical axis, inside the inner
vortex. Figure 1 Principal features of the hydrocyclone.
1820 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation
large units. To increase capacity, multiple small hydro-
cyclones may be manifolded from a single feed line.
Although the principle of operation is simple, many
aspects of their operation are still poorly understood,
and hydrocyclone selection and prediction for indus-
trial operation are largely empirical.
Liquid Velocity Distributions
Kelsall, in 1952, performed a classic series of experi-
ments measuring axial and tangential Suid velocity
proRles in a hydrocyclone using an ingenious experi-
mental system with rotating objectives. The radial velo-
city was calculated by continuity. The velocity proRles
are shown in Figure 2. More recently, velocity proRles
measured using laser Doppler velocimetry (LDV) were
found to correspond closely to those of Kelsall.
The Suid velocity in the cyclone has tangential,
axial and radial components. The axial velocity is
negative (downward) close to the walls in the cone
and positive (upward) near the air core, increasing
towards the spigot. This results in a locus of zero
vertical velocity between the two vortices, which
roughly follows the proRle of the cyclone. Toroidal
rotation in the inlet Sow and interaction between the
vortices result in multiple Sow reversals.
The tangential velocity increases toward the axis,
reaching a maximum near the air core, thereafter
decreasing in a forced vortex region. It is the tangen-
tial velocity component that generates the centrifugal
force, which separates coarser particles from Rner
ones. The radial velocity, which is two orders of
magnitude smaller than the axial or tangential vel-
ocities, is directed toward the centre of the cyclone
and increases toward the apex.
Particle Motion
Particles entering the cyclone move radially, depend-
ing on their mass, either outward due to tangential
liquid motion, or inward due to radial Suid motion.
In the radial and axial directions, the particle motion
is assumed to equal the Suid motion.
Direct measurement of particle motion and solids
concentrations at positions in the hydrocyclone can
be performed using phase Doppler anemometry.
Electrical impedence tomography has been used to
measure the position of the air core and the solids
concentration proRle in a plane through industrial
hydrocyclones.
Classi\cation Performance and the Partition
Curve
The partition curve (also called a performance curve,
efRciency curve or Tromp curve) is used to quantify
hydrocyclone particle size separation performance. It
quantiRes the weight fraction (or percentage) of each
particle size fraction in the feed reporting to the
underSow product. For any particle size d, the parti-
tion number, p(d), is calculated from:
U.u(d)
p(d)" [1]
F.f (d)
U and F are the mass Sow rates of solids (in the same
units) and u(d) and f(d) are the weight fractions of
particle size d in the feed and underSow streams,
respectively. The size at which the partition number
equals 0.5 (or 50%) is called the cut size (d50).
A fraction of Rne particles always report to
the underSow, hence experimentally observed parti-
tion curves do not asymptote to zero but to a min-
imum, called the bypass. This can be interpreted as
a fraction of all particles in the feed bypassing classi-
Rcation and reporting directly to the underSow
stream. Short-circuiting of feed material to the over-
Sow stream may cause the partition curve not to
reach a partition value of 1 (100%): this is not com-
mon. The effect of bypass on classiRcation perfor-
mance is taken into account by correcting the parti-
tion value:
p(d)!r(d)
c(d)" [2]
1!r(d)
where c(d) is the corrected partition value and r(d) the
fraction of material of size d bypassing classiRcation.
The particle size at which the corrected partition
number is 0.5 (50%) is called the corrected cut size
(d50c). It is often found that the bypass equals the
water recovery from the feed to the underSow (RF),
although there is no fundamental reason why this
should be so.
Figure 3 schematically shows an observed and cor-
rected partition curve.
A so-called Rshhook may occur in the observed
partition curve when, for particle sizes Rner than that
at the minimum partition value, progressively higher
partition numbers are observed. This is more com-
monly observed for smaller diameter hydrocyclones
and is thought to be the result of turbulent dispersion.
In such cases the water recovery may be signiRcantly
lower than the lowest partition value observed. Ap-
plying the correcting concept to such partition curves
is meaningless.
Mathematically Describing the Partition Curve
Corrected partition curves have a sigmoidal shape
that can be represented using two-parameter
 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation 1821

Figure 2 (A) Axial, (B) tangential and (C) radial velocity profiles in a hydrocyclone. (Reproduced with permission from Kelsall
(1953).)

functions such as the exponential sum, the Rosin}
Rammler and the log-logistic expressions. The two
parameters determine the cut size and the sharpness
of separation, respectively. The Rshhook partition
curve can be modelled using the sum of a corrected
partition curve and an inverted partition curve multi-
plied by a bypass fraction.
The observed partition curve gives a complete de-
scription of the selective separation of all sizes of
solids entering a cyclone and can be used to predict
1822 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation
Figure 3 The observed (continuous line) and corrected (dashed
line) partition curves of a hydrocyclone with a bypass of 20%.
the product size distribution and solids recovery for
changes in feed size distribution. If the bypass is
assumed to equal the water recovery, the liquid and
volumetric balances can also be estimated.
Hydrocyclone Geometry
The hydrocyclone diameter is the main design vari-
able, and affects both capacity and cut size. The
broad operating range available for any hydrocyclone
diameter is narrowed down by Rxing the inlet and
outlet dimensions. It is not generally possible to select
independently all the design variables; however, there
are reasonable ranges in relation to the hydrocyclone
diameter, Dc. Figure 4 shows the approximate cut
size and throughput range that can be achieved using
cyclones of different diameters.
The cone angle for classiRcation of hydrocyclones
is 15}303, with smaller angles for Rner cut sizes, and
larger angles for coarser cut sizes, respectively. The
free vortex height is the distance between the bottom
of the vortex Rnder and the spigot. Increasing hydro-
cyclone height improves both capacity and separation
efRciency, and generally varies between 3 and 8 Dc.
Figure 4 Cut size and throughput for different cyclone diameters.
The inlet opening is usually rectangular with
a height to width ratio of 2 and an equivalent circular
diameter of 0.14}0.33 Dc. The inner wall, outer wall
or centre of the hydrocyclone inlet may be designed to
be tangential to the cyclone body, and may also scroll
downwards.
The outlet dimensions are the most important
physical parameters used to alter the operation. Vor-
tex Rnder diameters of 0.13}0.43 Dc are commonly
used. Spigot diameters in the range 0.1}0.2 Dc are
used, but the ratio to the vortex Rnder is more impor-
tant. In general, the vortex Rnder diameter is greater
than that of the spigot. Equal diameters should be
avoided.
The Effect of Operating and
Design Variables
Table 1 summarizes the effect that changes to the
major design and operating variables have on the
capacity, cut size and sharpness of classiRcation.
The effect of pressure drop on the sharpness of
separation depends on the operating range, as an
increase in pressure drop increases the throughput
and hence the separation efRciency, but decreases the
volumetric Sow to the underSow. Of particular inter-
est is the effect of feed solids concentration, which has
a signiRcant effect on the classiRcation. Figure 5
shows clearly that an increase in solids concentration
increases the cut size and reduces the sharpness of
separation.
Hydrocyclone Models
The modelling of hydrocyclones is performed by
either describing the Suid Sow and particle motion
within the cyclone, or by developing empirical
 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation 1823

Table 1 Cyclone design and operating variable effectsa Suid and particle motion is estimated from solution
of the Navier}Stokes equations have been developed
Increasing Throughput Cut size Sharpness of more recently.
(Q) (d50) classification

Cyclone diameter, Dc ! ! ! Equilibrium orbit theory It can be postulated that

Vortex finder diameter,! ! ! particles will Rnd an equilibrium orbit in the hydro-
Do cyclone where their terminal settling velocity radially
Spigot diameter, Du !
outward is equal to the radial velocity of the liquid
Feed inlet, Di !
Cone angle Not comparable! ! inward. A particle will report to the spigot if its
Free vortex height, h ! ! ! equilibrium orbit is in the downward axial liquid Sow
Pressure drop, P ! ! or and to the vortex Rnder if in the upward axial Sow.
Volumetric feed solids ! ! The cut size is deRned by particles that have an
concentration,
equilibrium orbit that coincides with the locus of zero
! increase;
a
decrease. vertical velocity and therefore have an equal prob-
ability of reporting to either product streams. An
equilibrium orbit may not be achieved due to the
(or semi-empirical) relationships between operating short residence times and high solids concentrations
variables and measured responses. Fundamental in the hydrocyclone.
models are appealing from a rigorous standpoint but
have difRculty in describing satisfactorily the com- Residence time theory This theory determines
plex particle}particle and particle}Suid interactions whether the residence time in the hydrocyclone
for hydrocyclones operating at higher solids concen- allows a particle entering the cyclone at the centre of
trations. the inlet to settle to the cyclone wall and enter the
Empirical or semi-empirical models, which relate boundary layer Sow to the underSow.
the parameters of the partition curve to cyclone de-
sign and operating variables, are generally used for Crowding theory At higher feed concentrations, it is
industrial hydrocyclone modelling and simulation. found that the separation size is primarily determined
A number of general models, particularly for larger by the discharge capacity of the spigot and the feed
diameter hydrocyclones, have been developed (see size distribution. By controlling the outlet dimen-
Further Reading). sions, it is thought that any cut size within the feed
size distribution can be obtained.
Fundamentally Based Hydrocyclone Models
Early attempts at understanding the physical prin- Computational Wuid dynamics (CFD) solutions
ciples that govern size separation in hydrocyclones This is the preferred approach for fundamentally
yielded theories based on equilibrium, residence time based modelling of hydrocyclone performance.
and crowding. More complete simulations in which Complete Sow modelling of the hydrocyclone

Figure 5 Effect of feed solids concentration on hydrocyclone separation. Circles, 2.68 vol%; squares, 11.11 vol%; triangles, 17.54
vol%; diamonds, 23.75 vol%. (Reproduced with permission from Braun and Bohnet (1989). Copyright: Wiley-VCH.)
1824 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation
involves predicting the liquid-phase velocities, the
slurry concentration proRle, the turbulent viscosities
and the slip velocities of particles with respect to the
liquid phase for a range of particle sizes before predic-
ting the partition curve. The solution is complex,
because the governing Suid Sow equations are
nonlinear, simultaneous partial differential equa-
tions.
Chakraborti and Miller (1992) have published an
extensive review of Suid Sow modelling in hydrocy-
clones. They describe the Sow models in detail, giving
particular attention to models based on the
Navier}Stokes equation and the treatment of Suid
turbulence. They further discuss techniques for Sow
measurement and visualization and give a brief sum-
mary of pressure drop correlations and measure-
ments. This paper is an essential reference for the
Suid Sow modelling approach.
The general approach to develop a complete
CFD-based model of a hydrocyclone must include
a wide range of components. If it is assumed that
variations of local density and viscosity are small
for dilute slurries and that particle}particle
interactions are negligible, the Suid and particle
modelling can be decoupled. Liquid velocities are
predicted by combining the Suid transport equations
for vorticity, stream function and angular spin
velocity with a modiRed Prandtl mixing length
model, which varies both radially and axially, for
the turbulent viscosity. The set of simultaneous,
nonlinear partial differential equations are solved by
overlaying the hydrocyclone dimensions with a rec-
tangular grid and using appropriate boundary condi-
tions at the solid walls and liquid}air interface, to
solve for conditions within each cell of the grid. By
balancing all the forces on the particle, the particle
motion with respect to the Suid can be computed.
The particle trajectories are found by calculating
axial and radial slip velocities with respect to the
Suid. Size classiRcation performance is determined by
following a particle of a given size from the inlet until
it exits. This computation is repeated for each particle
size across the inlet diameter yielding the partition
curve.
For concentrated slurries, liquid-phase velocities
are affected by local density and viscosity, which in
turn are affected by local solid concentration and
particle size distribution. Since particle motion deter-
mines the concentration and size distribution at each
location, this being determined from liquid velocities,
an iterative solution is required so that local slurry
property changes can be estimated and liquid-phase
velocities recalculated.
Advances in CFD methods such as computation
grid generation, numerical methods and computing
resources are increasing the applicability of this mod-
elling technique to improve designs.
Empirical Models
At present, empirical models are the most commonly
used technique for hydrocyclone selection and perfor-
mance prediction. Empirical hydrocyclone models
use the partition curve as a basis for describing size
separation. Suitable equations are developed from
experimental results to relate the parameters of the
corrected partition curve to physical variables. In
general, empirical hydrocyclone models consist of
four relationships that describe the cut size, the sharp-
ness of separation, the water balance around the
hydrocyclone and the throughput}pressure drop
relationship.
An empirical hydrocyclone model was described in
1976 that is still commonly used to predict separation
performance. This model was the Rrst to document
an empirical form for the sharpness of separation and
therefore allow direct simulation of expected perfor-
mance without any testwork. This model form is
often used as a basis for the development of models
that include further variables, such as, for example,
angle of inclination, or for an operating range in
which the model has not been tested.
The Rosin}Rammler function describes the re-
duced partition curve:
ci"1!exp (!0.693xm
i ) [3]
where m indicates the sharpness of separation and
xi is:
di
xi" [4]
d50c
In SI units, and using the symbols in Table 1, the Plitt
equation for the cut size is:
50.5 D0.46
c D0.6
i D1.21
o exp[6.3 ]
d50c" [5]
Du h Q (
s!
l)0.5
0.71 0.38 0.45
where
s,
l and
p are the densities of the solid, liquid
and pulp, respectively.
To describe the water balance, Plitt develops a rela-
tionship for the volumetric Sow split between the
overSow and underSow streams, S, rather than the
bypass:
3.28(Du/Do)3.31h0.54(D2u#D2o)0.36
0.24
p exp[0.54 ]
S"
P0.24D1.11
c
[6]
 II / PARTICLE SIZE SEPARATION / Hydrocyclones for Particle Size Separation
The relationship for the sharpness of separation is
given by:
 
D2ch 0.15
m"1.94 exp[!1.58 Rv] [7]
Q the separation and CFD models will play a greater
where Rv, the recovery of slurry to the underSow, is
related to the Sow split by:
S lone inefRciencies. With increased resolution and
Rv" [8]
1#S
The relationship between the pressure drop across the
cyclone and the throughput is given by:
1.88 Q1.78exp[0.55 ]
P" 0.37 0.94 0.28 2 [9]
Dc Di h (Du#D2o)0.87
Roping is affected by the spigot diameter and the
volumetric solids concentration in the underSow;
however, there is no satisfactory method for predic-
ting operating limit.
It must be emphasized that empirical models, al-
though developed from an extensive database, should
be used with caution.
Future Developments
The extremely wide range of hydrocyclones available
and separation applications for which they can be
used assures their future role in particle classiRcation.
However, signiRcant obstacles remain before they
can be used to replace more efRcient methods for Rne
classiRcation purposes, such as centrifuges. ClassiRca-
tion inefRciencies, in particular the large bypass, limit
their application. The potential of very small dia-
meter hydrocyclones for sub-micron particle separ-
ation, especially in multistage conRguration, is enor-
mous, if these inefRciencies can be reduced.
Instrumentation of Field Flow Fractionation
See II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation
Sedimentation
See II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation
1825
Hydrocyclone modelling has advanced signiR-
cantly with the use of CFD. Empirical hydrocyclone
models are convenient ways of describing experi-
mental data but do not enhance the understanding of
role in hydrocyclone simulation. Nonintrusive
measurement techniques such as laser Doppler anem-
ometry (LDA), laser Doppler velocimetry (LDV) and
tomography have indicated the source of hydrocyc-
combined with CFD models, this will improve hydro-
cyclone unit design.
Hydrocyclone operations will beneRt from novel
methods for monitoring which are currently being
developed. Industrial tomography is becoming af-
fordable, and the potential of visual and sonic tech-
niques has been illustrated.
See also: II/Particle Size Separation: Electrostatic
Precipitation.
Further Reading
Bradley D (1965) The Hydrocyclone. Oxford: Pergamon
Press.
Braun T and Bohnet M (1990) InSuence of feed solids
concentration on the performance of hydrocyclones.
Chem. Eng. Technol. 13: 15}20.
Chakraborti N and Miller JD (1992) Fluid Sow in hydro-
cyclones: a critical review. Mineral Processing and
Extractive Metallurgy Review 11:
Heiskanen K (1993) Particle ClassiTcation. London:
Chapman & Hall.
Kelsall DF (1953) A study of the motion of solid particles in
a hydraulic cyclone. Transactions of the Institution of
Chemical Engineers 30: 87}104.
Plitt LR (1976) A mathematical model of the hydrocyclone
classiRer. CIM Bulletin December 114}122.
Svarovsky L (1984) Hydrocyclones. London: Holt,
Rinehart and Winston.
1826 II / PARTICLE SIZE SEPARATION / Sieving / Screening
Sieving/Screening
J. Skopp, School of Natural Resource Sciences,
University of Nabraska, Lincoln, NE, USA
Copyright ^ 2000 Academic Press
Introduction
Sieving is one of the oldest and most commonly used
method of sorting materials. Yet, when improperly
carried out, sieving can provide misleading informa-
tion or biased separation. Sieving has widespread
application to industries as diverse as mining, phar-
maceutical production and agriculture. The goal is
typically to control or measure the particle size distri-
bution. Sieving may be a direct part of a production
process, a quality control procedure or a sample char-
acterization. Regardless of the purpose, an under-
standing of sieving is necessary to optimize and
accurately use this technique.
Sieving has the clear advantage of being a simple,
readily understood and relatively inexpensive
method. This method also has the ability to provide
reproducible results. This has tended to boost conR-
dence in the method, even where it is not warranted.
Some of this conRdence is a result of ignorance as to
the actual errors involved in a given sieving operation
or with a particular set of sieves. While these errors
can be described and measured, the difRculty of doing
so largely detracts from the attractiveness of the
method. Thus, most people rely on standard proced-
ures and the reproducibility of the method to reassure
themselves as to the quality of the data obtained.
Overview of Sieving Process
Sieving was the earliest means of particle size frac-
tionation. Basically, the process of sieving is that of
placing the particles to be fractionated on a pattern of
openings or holes. The individual openings are refer-
red to as the aperture. Small particles may fall
through or the sieve retains the larger particles. Separ-
ation requires agitation and time. A variety of mecha-
nisms exist to provide agitation either to the sieve or
to the particles to be fractionated. Typically, com-
mercial equipment varies in the manner in which
agitation is created or the Suid used to support the
particles. Either air or water may be used to support
the particles as they sort on the sieve. Dry sieving has
a lower practical limit of 50
m, while wet sieving
can separate smaller particle sizes when using special
sieves or small volumes of particles.
We brieSy examine three applications of sieving
processes before continuing in more detail with a de-
scription of the sieving process, the time dependence
of sieving and lastly sources of error in sieving.
Applications of Sieving
A single sieve may provide a straightforward classi-
Rcation (or gradation) of particles where a clear thre-
shold is desired. Crushing operations typically seek to
ensure that all particles are below such a threshold
and sieves provide a convenient means of doing so.
However, in some operations it is equally important
to remove particles less than a given size. The Rrst
case is referred to as scalping (removing oversize
particles) while the second case is referred to as re-
moval of Rnes. Sometimes it is desirable to do both.
Thus, it can be useful to employ a set (or nest) of
sieves varying in aperture size. Unsorted particles are
applied to the topmost sieve and agitation begun. As
the particles sort, smaller ones pass through from
upper sieves to lower ones. This cascade of particle
sorting continues as the smallest particles only
gradually make their way past the smallest sieve.
A nest of sieves with a sufRcient variation in aper-
ture sizes may be used to construct a particle size
distribution. Such distribution may in turn be used to
describe the behaviour of the material under study.
A particle size distribution obtained in this way
should recognize that the errors associated with the
top sieve (the coarsest fraction) may be different from
the errors associated with the bottom sieve (the Rnest
fraction). This occurs in part due to differing oppor-
tunity times for the particles to pass through the sieve
that they are ultimately retained upon.
A third application of sieving is the use of sieving
curves to develop a morphological characterization of
the material being screened. Sieving curves represent
the time dependence of the sieving process and are
discussed in more detail in a later section. The goal is
to infer particle shape or other morphological fea-
tures (e.g. agglomeration) from the speed with which
sieving proceeds. Unfortunately, these techniques are
still in the developmental stages.
Description of the Sieving Process
Sieve Construction
Sieves are typically constructed in one of two ways:
Rrst, through the use of a wire mesh or cloth. This
 II / PARTICLE SIZE SEPARATION / Sieving / Screening
Table 1 Sieve mesh equivalent sizes: US standard testing
sieves designated for wire cloth at some principal ISO sizes
Mesh number: Sieve opening
US standard testing sieves (mm)
5 4.00
7 2.80
10 2.00
14 1.40
18 1.00
25 0.710
36 0.500
45 0.355
60 0.250
80 0.180
120 0.125
190 0.090
325 0.045
results in square apertures. The diameter of the wire
controls the size of the aperture and per cent of the
total area that is open. Second, openings or circular
holes are created by perforating a plate or Sat disc.
The number of perforations controls the amount of
open area in this sieve. These two methods result in
openings that differ in shape and in their ability to
pass particles. Thus, it is important to specify which
kind has been used to obtain a particular set
of data.
Sieves are available in standard sizes (Table 1)
from a number of companies. Sieve openings are
given in a mesh number or nominal diameter of the
opening. Surprisingly, no standard sieve is readily
available for the 50
m cut-off between the sand and
silt separates (as deRned in some systems of particle
nomenclature). Consequently, sieving cannot distin-
guish this class boundary directly using standard
sieves. However, such boundaries (or other speciRc
sizes) can be estimated by determining the particle
size distribution with a nest of sieves that bracket the
desired size.
Industry standards have focused on maintaining
consistency in the manufacture of sieves. This allows
the comparison of sieving operations provided the
materials and conditions of the sieving are also com-
parable. The focus of many practitioners of sieving
has been on standard procedures to maintain repro-
ducibility and consistency.
Sieve Agitation
Particles placed on a sieve may not sort or pass
through the openings unless some form of agitation is
used. Typically, a large mass of soil placed on a sieve
1827
will allow particles to bridge, thus restricting their
passage through the openings. Agitation breaks the
bridges, shufSes the particles and provides an oppor-
tunity for particles to present themselves to an open-
ing.
An undersize particle may still fail to pass through
a sieve. Figure 1 illustrates how identically shaped
particles may approach the aperture at different
angles. Some angles may allow passage of the particle
while others may restrict passage. Agitation increases
the number of times that a particle approaches the
aperture and the velocity of approach and alters the
orientation of the particle as it approaches the aper-
ture. The Rrst two factors depend on shaking intensity
and competition among particles. The third factor
depends on the shape of the particle and the shape of
the sieve opening. These factors inSuence not only the
speed with which fractionation occurs, but the reten-
tion of speciRc particles.
Alternative commercial instruments attempt to
agitate the particles in different ways. Vibration is
commonly used, either as a jarring action or an oscil-
latory motion. This motion may have a large or small
amplitude and may be restricted to horizontal or
vertical motion. The efRciency of separation usually
increases as vibration amplitude increases, but it may
reach a maximum. EfRciency declines as the increased
agitation merely serves to suspend particles without
giving them an opportunity to pass through an aper-
ture.
One means of agitating particles is by displacing
the Suid (air or water) surrounding the particles. In
the case of air, this may involve ultrasonic oscillation
or mechanical Sow of air counter to the direction of
sieving. In an extreme case, the sieve may be placed in
Figure 1 How particle shape influences the efficiency of sieving
by delaying passage through the sieve.
1828 II / PARTICLE SIZE SEPARATION / Sieving / Screening
a vertical orientation as air Sow lifts the particles and
attempts to pass through the sieve. Particles retained
on the sieve would need to be mechanically removed.
Such a manipulation could be maintained in opera-
tion continuously.
Sieved Materials
The particle inSuences the time and efRciency of siev-
ing. Issues relate to the sample’s mass, particle size
distribution, particle density, particle shape, friabil-
ity, the tendency to aggregate and electrostatic prop-
erties. SpeciRc procedures have been developed for
particular materials.
Obtaining a representative sample or subsample
for seive analysis is a key step; this is a particular issue
where nonuniformity is extreme or segregation of the
sample may occur. It may sometimes be necessary to
settle for reproducible sampling with the goal of
making comparisons on a relative basis.
Sample mass inSuences the efRciency of sieving and
the time needed for a sieving operation. At high
loadings, blinding of the sieve may be more important
as well as increased breakdown of friable samples.
Particle size distribution has several effects on the
efRciency of sieving. These effects, if they exist, may
depend on the choice of sieves. Particles which are
slightly oversize may lodge in the sieve, blocking the
aperture from participating in the sieving process.
This ‘blinding’ of the sieve reduces the overall efR-
ciency of sieving or increases the time needed to
achieve separation.
Figure 2 Example of three sieving curves modelled with the equation P"100% [1!exp(!kt n)], n"0.8. Dashed line, k"0.025;
dotted line, k"0.050; continuous line, k"0.10.
Sample friability can also inSuence separation.
Friability is the tendency of particles to break apart.
The process of agitation and sieving may change the
particle size distribution for friable samples. Longer
sieving times or larger amplitude agitation increase
these effects. Ultimately, the goal of the sieving opera-
tion must be compared to the effect of sieving on the
underlying size distribution.
Time Dependence of Sieving
Sieving Curves
The mass passing through the sieve can be observed
by collecting this material on a balance or strain
gauge. This can be done manually or the sensor can
be connected to a computer and the progress of siev-
ing continuously monitored as a function of time. The
graph of mass remaining on the sieve or the mass that
has passed the sieve is referred to as a sieving curve
(Figure 2). The graph shown in Figure 2 shows the
mass passing through the sieve as a function of time.
Note that complete separation may never be
achieved.
Sieving Equations
Sieving curves like that shown in Figure 2 are the
result, in part, of particle shape. Extracting this in-
formation requires a quantitative description of the
sieving curve. Models describing the sieving curves
can be classiRed as empirical or mechanistic.
The most common empirical model uses a power
 II / PARTICLE SIZE SEPARATION / Sieving / Screening
function. The rate at which material passes through
the sieve for short times is nearly constant and can be
expressed by a power function model given as:
M"atb
where M"cumulative mass of the material passed
at sieving time t (grams), t"sieving time (seconds),
b"a constant nearly equal to 1 and a"sieving rate
constant, with units of g s\1b. This model has had
some practical utility but its validity in describing
data, even for short times, is questionable and de-
pends on how the data are truncated.
One mechanistic model starts by assuming that the
probability of particle passage at time t is directly
proportional to the mass of material on the sieve.
Thus, the mass of material remaining (Mt) at sieving
time t is described in differential form by:
dMt
"!k(Mt!Mr)
dt
where Mt"total mass of the material remaining on
the sieve at time t (grams), Mr"material mass which
does not pass the sieve (i.e. the sieve residue after
inRnite sieving time) (gram) and k"passage prob-
ability of the material with a given size at time t, or
passage rate factor (s\1).
This model assumes that, after a time, k becomes
independent of time and particle size. This occurs
when the smallest particles have passed through the
sieve and those left on the sieve are of a size close to
the sieve opening. However, the sieving constant (k)
is not guaranteed to be constant. For example, it
may change with the depth of material on the sieve
and hence change as M changes. Thus, one
difRculty in using the above equation lies in describ-
ing how k changes with experimental conditions or
time.
Integrating the above equation (setting Mr"0)
with the application of initial conditions yields:
M"M0(1!e\kt)
where M"cumulative mass of material passed at
sieving time t (grams) and M0"cumulative mass of
the material that would pass at inRnite sieving
time (grams). Note that M#Mt"M0. The rate
constant can be determined empirically or deduced
from physical principles. In the latter case k is deRned
as:
k"A[1!2R(1!fs)/D]2
1829
where A is a constant depending on the sieve and
sieving conditions, R is a radius of the particle, D is
the sieve opening and fs is a grouping of factors that
describe particle shape. Tests of this model have not
been extensive. It should work best for longer times.
An alternative or modiRed proportional model has
been used with some success over a wide range of
sieving times. This model starts by assuming that the
constant k in the differential equation above is a func-
tion of time, k"kH/tm. Then the differential equation
can be written as:
dMt !kH(Mt!Mr)
"!
dt tm
where kH"passage probability of the material with
a given size at time t, or passage rate factor, (s\n) and
n"1!m where m is a constant.
Upon solving this equation for M and substituting
1!n for m, the resulting model is:
n
M"M0(1!e\kHt )
This model appears to describe both the initial and
long time sieving behaviour. It is empirical in nature,
so its coefRcients need to be related to particle mor-
phology or other sample characteristics.
Sources of Error
The Sample
The major concern with sieving operations is the
efRciency and time of each sieving step. This becomes
particularly important when we use a nest of sieves to
perform several separations simultaneously. The use
of words such as effective or nominal diameters with
sieves is in recognition of the imperfect separation
that may occur. What is less readily recognized is that
the sample may contribute to imperfect separation.
Placement of a soil sample on a sieve does not result
in instantaneous separation. Several factors inSuence
the time to achieve a Rxed level of segregation. These
factors include sample size, shaking intensity, particle
shape, particle size and hole geometry. No one set of
sieving times applies for all conditions, but for many
soils with small samples (&100 g) a waiting time of
about 3}5 min for coarse fractions and 10 min for
Rne fractions gives acceptable results. A typical nest
of sieves (with 3 in (7.5 cm) diameter) operating for at
least 15 min is desirable for separation. Since samples
vary in their sieving characteristics, it is best to run
a trial sample at several times. This is particularly true
with a nest of sieves to ensure adequate separation
past the smallest sieve opening.
1830 II / PARTICLE SIZE SEPARATION / Sieving / Screening
The single most important factor changing the efR-
ciency of sieving is the initial sample mass. It is faster
and more efRcient to split a large sample into several
smaller ones. A useful rule of thumb is to keep the
depth of material on the sieve to less than 1 cm.
A better rule is to run a test sieving curve. In general,
this will show that as the sieve opening decreases
smaller masses are needed. For 8 in (20 cm) diameter
sieves these typically range from 200 to 30 g for sieve
openings varying from 2 mm to 45
m.
Particle shape also inSuences the efRciency of siev-
ing. Rougher surfaces with elongated shapes are ex-
pected to require longer sieving times than smooth
surfaces with a more spherical shape. It may be pos-
sible to use equations like the one presented earlier to
predict the sieving rate constant. However, the in-
verse problem of determining the particle shape from
sieving curves appears to be ambiguous.
Sieve Construction and Cleaning
The manufacture of sieves is subject to error. This
error takes the form of a variation in aperture open-
ing within a sieve as well as from sieve to sieve. This
error also varies with the size of the opening, with the
coarser sieves generally being more consistent. It is
expected that for the Rner sieves (less than 100
m),
this error could approach 10%. Thus, all sieve open-
ings represent nominal diameters. The determination
of the diameter of material that passes a particular
sieve must be determined by calibration, as discussed
in the next section.
Sieve cleaning represents one means of ensuring
the reproducibility of a sieve. This typically
involves use of a brush with the coarser sieves.
Finer sieves may require reverse Sushing with water
or use of an ultrasonic cleaning bath. Ultimately,
sieves must be inspected at periodic intervals to make
sure that the mesh has not been distorted or damaged
in use.
Calibration
A sieve calibration is used to establish the size of
separation achieved by a particular sieve. This in-
volves the use of standard materials and microscopic
examination of the material passing through the
sieve. The material is placed on the sieve and sieving
proceeds until the mass passing appears unchanged.
The material which has passed is examined under
a microscope to determine the size of particles. This
information is used to Rnd an average equivalent
volume diameter which represents the effective size of
the sieve.
Conclusion
Sieving is a valuable and widely used tool for both
sorting and particle size determination. It is a relative-
ly inexpensive procedure that, with the use of stan-
dard methods, can provide reproducible results. The
use of sieves for novel materials requires the deter-
mination of sieving curves under a variety of condi-
tions such as loading. These curves are used to opti-
mize the conditions for efRcient sieving. Even where
only slight variations in the material to be analysed
exist, it may be desirable to determine the appropriate
sieving conditions rather than relying on standard
methods.
See also: II/Particle Size Separation: Hydrocyclones for
Particle Size Separation.
Further Reading
Allen T (1990) Particle Size Measurement, 4th edn. Powder
Technology Series. London: Chapman and Hall.
ASTM Committee E-29 on Particle Size Measurement
(1985) Manual on Test Sieving Methods. ASTM
Special Technical Publication 447B. Philadelphia, PA:
ASTM.
Beddow JK (1981) Particulate Science and Technology.
New York, NY: Chemical Publishing.
Beddow JK (1984) Particle Characterization in Techno-
logy. I. Applications and Microanalysis. Boca Raton,
FL: CRC Press.
Beddow JK (1984) Particle Characterization in Technology.
II. Morphological Analysis. Boca Raton, FL: CRC Press.
Fayer ME and Otten L (eds) (1984) Handbook of Powder
Science and Technology. New York: Van Nostrand
Reinhold.
Herdan G and Smith ML (1953) Small Particle Statistics; an
Account of Statistical Methods for the Investigation of
Finely Divided Materials. Amsterdam: Elsevier.
Irani RR and Clayton FC (1963) Particle Size Measure-
ment, Interpretation and Application. New York, NY:
Wiley.
Lauer O (1966) Grain Size Measurements on Commercial
Powders; A Guide for Experts, Worked out in Alpine’s
Experimental Station. Augsberg: Alpine.
Leschonski K (1979) Sieve analysis: the Cinderella of par-
ticle size analysis methods? Powder Technology 24:
115}124.
 II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation
Split Flow Thin Cell (SPLITT) Separation
C. Contado, University of Ferrara, Ferrara, Italy
Copyright ^ 2000 Academic Press
Introduction
SPLITT fractionation (SF) is a relatively new family
of separation techniques primarily } but not exclus-
ively } applicable to macromolecules and particles.
The SF techniques utilize a thin ribbon-shaped Sow
cell and achieve fractionation by differential trans-
port across the thin (transverse) axis of the cell. Since
the cell is only a few hundred micrometres thick, the
separation path } which may be less than the nominal
channel thickness } is extremely short and the separ-
ative transport is correspondingly rapid. Separation is
typically accomplished in only a few minutes. This is
a particularly valuable feature for example for fragile
species that must be fractionated rapidly to avoid
degradation (e.g. biological samples). The Suid that
carries dissolved or suspended components through
the SPLITT cell is divided at both ends by thin Sow
splitter elements (see Figure 1). The inlet splitter ele-
ment allows for the smooth merging of two incoming
laminae, one carrying the suspended feed material
and the other generally containing only the pure car-
Figure 1 Side view of a generic SPLITT cell.
1831
rier liquid. Differential transport of feed components
between the two laminae (after they are brought
into contact) then occurs as a result of a transverse
driving force or gradient. At the outlet end, the
Sowing liquid volume is divided at a predetermined
position by a second splitter element, thus producing
two sub-streams that are enriched or depleted in the
desired components as a result of the differential
transport.
Both preparative and analytical fractionation pro-
cess can occur in the SPLITT cells, depending on the
injection procedure. A continuous (CSF) process of
feeding the cell is advantageous for preparative frac-
tionation (gram, kilogram), offering rapid through-
put, minimum holdup volumes, and a sharp separ-
ative cut off; examples of continuous fractionation
can be found in the separation of mineralogical, in-
dustrial and food samples.
The analytical version of SPLITT fractionation
(ASF) is often more practical to operate. The injection
of discrete pulses can be made, if desired, to follow
one another in close sequence. In this use, separation
is performed for the measurement of quantitative
properties of sample components and the fractiona-
tion is termed ‘analytical SPLITT fractionation’;
1832 II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation

examples of quantitative determinations include Theory

diffusion coefRcients of proteins, settling velocity
The theory of separation by SPLITT cell was for-
and the relative content of oversized particles
mulated by J. C. Giddings in terms of experimentally
above a cutoff diameter in a particulate material.
controllable Sow rates in the inlet and outlet sub-
Moreover because of its ease of theoretical interpreta-
streams and it has been developed and implemented
tion, the SPLITT cell can be used for the rapid
through the years; SPLITT fractionation theory can
measurement of transport-related properties such
now be found in numerous publications. The separ-
as particle size and particle size distribution.
ation is performed inside a thin channel, where the
The throughput of SF is proportional to many vari-
behaviour of a sample particle depends on the bal-
ables such as the sample concentration in the feed
ance between the external force Reld and frictional
stream, the volumetric Sow rate of the sample stream,
forces (as for Reld Sow fractionation techniques
the applied Reld strength and the SPLITT cell area.
(FFF)), combined with the action of the Suxes opera-
For preparative applications there is obviously
tive within the cell.
a trade-off between the resolution and throughput in
In Figure 1 the sample, suspended in a suitable
the operation of SF: maximizing the throughput and
carrier Suid, is introduced through the top inlet a
at
maintaining an acceptable resolution is the common
a predetermined volumetric Sow rate VQ (a
). At the
choice.
same time, pure carrier Suid enters through the bot-
The effectiveness of the SPLITT process can be
tom inlet b
at a Sow rate VQ (b
); where the two inlet
modulated by simply varying the Sow rates of the
streams join to form a single stream we have the ISP.
inlet and outlet substreams, which determine the
When the Suid stream reaches the end of the channel,
position of the inlet splitting plane (ISP) and the
it is mechanically divided into two fractions by the
outlet splitting plane (OSP) and controlling the thick-
outlet splitter.
ness of the transport region; sometimes, unwanted
The differential displacement of the particles oc-
displacements of a few tens of micrometres may be
curs towards wall B, based on the driving force
difRcult to discern but they are sufRcient to interfere
exerted on each type of particle by the applied Reld
with effective separation. In some cases, the efRciency
and the frictional resistance offered by the carrier
of the SPLITT process can be controlled by altering
Suid to particle motion. Thus different types of par-
the strength of the Reld or gradient driving the separ-
ticle occupy different laminae while the Sow through
ative transport.
the channel displaces them axially towards the outlet
The efRcacy of SF separation depends instead, on
end.
the hydrodynamic integrity of the SPLITT cell. Effec-
The total volumetric Sow rate VQ in the channel can
tive separation is based in fact, on two central re-
be written both in terms of inlet Sow rates or outlet
quirements:
Sow rates

E there must be no hydrodynamic mixing across
stream planes; and
E the splitters must be absolutely perfectly aligned so
that they are capable of splitting the Rlm of Sowing
liquid evenly along the stream plane.
Success in fulRlling these requirements is not easy to
judge because of the thinness of the cell and the
shortness of the transport path.
VQ "VQ (a
)#VQ (b
)"VQ (a)#VQ (b) [1]
and since the walls A and B are parallel and their
dimensions much larger than w (thickness) and
b (width), the velocity proRle is essentially two-
dimensional (see Figure 2). The mean Suid velocity
vN can be computed as VQ /bw, and the velocity para-
bolic proRle across the cell thickness is described by
the equation:




The selectivity of SPLITT fractionation comes from x x x x x
the applied force. The principal transverse driving v "6vN 1! "4vmax 1! [2]
w w w w w
forces used include gravity, centrifugation, diffusion,
electrical potential gradients, magnetic gradients and where vmax is the maximum Suid velocity found at the
hydrodynamic lift forces. The geometry of all the midplane (x"w/2) of the cell.
different cells is similar to that depicted in Figure 1, By looking at the Sow stream components (see
except for the curvature characteristic of the centrifu- Figure 1) it is possible to Rnd a relation which takes
gal SPLITT cell. into account the Suid Sow proceeding in the transport
The simplicity of the SPLITT cell leads to rather layer VQ (t):
rigorous theoretical guidelines on the conditions ne-
cessary to achieve a given level of separation. VQ "VQ (a
)#VQ (t)#VQ (b) [3]
 II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation
Figure 2 Upper view of the ‘Splitter’ used in the gravitational SPLITT cell.
VQ (t) can be obtained by combining eqns [1] and [3]
as:
VQ (t)"VQ (a)!VQ (a
)"VQ (b
)!VQ (b) [4]
The separation of sample components is achieved,
as previously described, by their different rates of
transport toward the opposite wall B under the inSu-
ence of the transverse Reld. It is important to be able
to calculate the distance that a component has to
travel. The ratio of sample Sow rate VQ (a
) to the total
Sow rate VQ deRnes the position of the ISP as distance
wa, from wall A:
VQ (a
) w2a w3a
"3 2Y!2 3Y [5]
VQ w w
The feed stream is then conRned to the lamina thick-
ness wa , between wall A and the ISP.
Y
By assuming that, during their residence in the
SPLITT cell, the particles are driven from wall A to
wall B at constant velocity U, the volumetric Sow rate
VQ of the lamina traversed by a particle is simply
given by
VQ "bLU [6]
which contains the physical dimensions of the chan-
nel b, and L (length); U is the velocity of the induced
transverse transport. In fact, a number of different
driving forces have been utilized to implement CSF:
gravitational, centrifugal, magnetic, electrical. U as-
sumes different expressions depending on the Reld
inducing the transverse transport, i.e.
gd2
Ugravitational" [7]
18o
2rd2
Ucentrifugal" [8]
18o
1833
H2d
Umagnetic" [9]
48o
Uelectric"E [10]
where d is the particle diameter (assumed spherical),
 is the difference between particle density s (com-
pact particles) and carrier density l, o is the viscosity
of the carrier, g is the acceleration of gravity (earth
Reld eqn [7]);  is the angular velocity (rad s\1) and
r is the radius of the rotation (centrifugal acceleration
eqn [8]);  is the difference between the magnetic
susceptibilities of the particles, p, and the carrier, l ,
and H is the drop in magnetic Reld strength
(eqn [9]);  is the electrophoretic mobility and E is
the electric Reld strength (eqn [10]). In some cases,
experimental studies have been based on diffusive
transport or by using hydrodynamic lift forces (which
can be used in unique ways because of their
nonuniformity).
In order to settle whether particles exit the channel
through outlet a or b, the relative values of VQ and
VQ (t) are of critical importance. For a sample intro-
duced close to the ISP, particles exit from outlet a if:
VQ )VQ (t) [11]
and from outlet b if:
VQ 'VQ (t) [12]
In the case of a SPLITT cell operating under
a gravitational Reld, the relation which allows one to
set the diameter cutoff of a spherical solid particle can
be obtained by combining eqns [4], [6] and [7]. In
fact, if VQ (t)"VQ and:
bLgd2
VQ " [13]
18o
1834 II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation
then the diameter at which 50% of the particles exit
outlet b is called the cutoff diameter dc, expressed as:

18o(VQ (a)!0.5VQ (a
))
dc" [14]
bLg
in which only half of VQ (a
) is considered (cf. eqn [4]).
Once dc has been chosen for a given channel, the
difference between VQ (a) and 0.5VQ (a
) is set according
to eqn [14]. However, the two constituents Sow rates
VQ (a) and VQ (a
) are not uniquely deRned by this equa-
tion, but some criteria useful for setting the Sow rates
are available in the optimization of SPLITT opera-
tions literature. In general, in order to obtain a separ-
ation with a good resolution, the practical rule which
states the following necessary but not sufRcient
condition:
VQ (b
)VQ (a
) or VQ (a)VQ (b) [15]
can be followed. This condition ensures that regions
I and II will be narrow, which automatically increases
the transport region, i.e. the cell space in which the
separation process occurs. For maximum resolution,
typical experimental conditions, are chosen to obtain
a VQ (a
)/VQ (t) ratio within 0.1}0.3. From a theoretical
point of view, the high resolution in the operative
transport mode is, as already stated, contingent on
compression of the feed sub-stream, a
, into a thin
lamina near wall A, and the sharpness of the separ-
ation, as in chromatography, can be judged by the
number N of the theoretical plates generated during
transport. Generally, for Reld (gravitational) driven
migrations, the effective N is given by the ratio of two
energies:
Fwt d3gwt
N" " [16]
2kT 12kT
where F is the force on the particle inducing its trans-
port, wt the length of the transport path, and kT the
thermal energy. Generally values of N*102 are re-
quired to assure achievable resolution, and each type
of macromolecule or particle should be checked
against this criterion.
Resolution in SF can be related to channel Sow
rates, deRning an index which measures the relative
breadth of the unresolved region. It has been demon-
strated that for sedimentation particles the equation
of the resolving power has the following form:
d1 d1 VQ (a)
" c(x, 0)"0 for wa )x)w
2 [17]
d d1!d0 VQ (a
)
where d0 and d1 are respectively the diameters of the
particles exiting from a and b. The particles falling
between these two sizes (d1 and d0), which are not
fully resolved, exit from both outlets in different
proportions, so the difference should be small. Ac-
cording to eqn [17] the ratio of VQ (a) to VQ (a
) allows
control of the range of unresolved particles which exit
both outlets a and b.
Speci\c Applications
Determination of the Diffusion Coef\cient D
CSF and ASF can be successfully applied to the separ-
ation of macromolecules such as proteins and lipo-
somes. The transport region in this case is seen as
a diffusion barrier, which acts as a dialysis mem-
brane. In order to be able to determine the diffusion
coefRcient D, an important parameter useful for char-
acterization of the sample components, the theory has
to be looked at in a deeper way, with equations which
explicitly contain D.
If a component enters inlet a
as a steady stream
and its transport through the SPLITT system is gov-
erned by the simultaneous displacements of diffusion
and parabolic Sow, mass conservation requires that:


2 2
c c c c
"!v(x) #D 2# [18]
t z x z2
c/ t is the rate of change of concentration at an
arbitrary point within the channel, x is the transverse
distance into the cell measured from wall A, z is the
distance into the channel along the length measured
from the inlet splitter and v is the local stream velo-
city. In order to simplify eqn [18], it can be observed
that all transverse transport in the channel occurs by
diffusion and that all the components are transported
along the cell (z-direction) by convention, so that the
diffusion term (D 2c/ z2) can be neglected.
Moreover, if the system operates in CSF mode, the
component concentrations throughout the cell will
attain a steady state, i.e. c/ t"0, so that eqn [18]
reduces to:


2
c D c
" [19]
z v(x) x2
where c(x, z) is subject to the boundary conditions:
c(x, 0)"co for 0)x)wa
Y
Y
c/ x"0 for x"0 and x"w at all z
 II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation 1835

In order to calculate D, a dimensionless diffusion By recalling the basic properties of the sedimentation
time D"Dto/w2 has been developed. The para- process different expressions can be obtained, which
meters to and w2 are known because w is Rxed by the contain the density parameter. During the SPLITT
geometry of the channel and to is the elution time of fractionation, under a gravitational Reld, the sample
a species dispersed in the total volume of the channel components are subject to two forces: the gravi-
and is related to the total Sow rate VQ by: tational force Fg"meff g and the frictional force
Ff"fU (meff is the effective mass, and f the friction
Vo bLw coefRcient). Usually, the stationary state is estab-
to" " [20] lished very rapidly and the two forces balance each
VQ VQ
other out and thus:
where Vo is the cell void volume expressed as a prod-
uct of channel dimensions b, L and w. The dimen- g
U"meff [24]
sionless time parameter to is thus related to D, VQ , and f
the channel dimensions by:
Spherical particles In the case of compact spherical
Dto DbL particles, by assuming that the particles do not under-
D" " [21] go any shrinking or swelling meff"m!mb or
w2 wVQ
meff"Vss!Vsl, where m is the real mass and
mb the buoyant mass while Vs and s are, respectively,
Consequently D is given by:
the volume and density of the particle and l the
density of the liquid. More explicitly
wVQ D
D" [22]
bL
meff"16
d3 [25]

This equation is used to obtain experimental

The   accounts for positive or negative mass values in
D values, once D is found. The total procedure re-
eqn [25] corresponding, respectively, to a falling or
quires the following steps: (i) to compute theoretically
a Soating particle. The friction coefRcient f can be
the concentration proRle c(x, L) over the lateral coor-
expressed by Stokes law f"3
d, where  is the
dinate x at outlet (z"L) by using the Crank}Nicol-
viscosity of the suspension Suid, which can be ap-
son numerical method. In this way, the retrieval of
proximated by the carrier viscosity, 0 . By combining
a component at each outlet substream Fa (in i) and
eqns [23], [24] and [25] one obtains the classical
Fb (in ii) can be calculated from c(x, L); (ii) to
expression (see eqn [13]):
construct a graph of Fa versus D for different VQ (a)/VQ
ratios; (iii) to determine experimentally the retrieval
factor at outlet sub-stream a (Fa) from the relative bLgd2
VQ (t)" [26]
strength of the detector signal (this value depends on 18o
the Sow ratio VQ (a)/VQ used in the experiment); (iv) to
compute the correspondent D value from the graph In the case of porous particles, porosity is deRned
Fa versus D by Rnding the correspondence between as:
the experimental and theoretical Fa values; (v) to use
eqn [22] to calculate D using the D value, and the Vp Vp
geometrical dimensions b, L, w and VQ . " " [27]
Vs#Vp Vptot
Effect of Particle Shape and Density in the
Gravitational SPLITT where Vp is the volume of the pore, Vs the volume of
the solid and Vptot the total volume of the particle.
The SPLITT cell has been largely applied for the Then eqn [25] changes into:
separation of environmental samples. Since the natu-
ral matter particles have different properties (poros-
ity, density, shape), the basic SPLITT equations have meff"16
d3(1! ) [28]
to be revised to Rt the relevant particle properties.
The basic relationship for the SPLITT cell, pre- and the correspondent eqn [26] into:
viously derived (see eqn [6]) can be written as:
bLgd2(1! )
VQ (t)" [30]
VQ "VQ (t)"bLUgravitational [23] 18o
1836 II / PARTICLE SIZE SEPARATION / Split Flow Thin Cell (SPLITT) Separation

The porosity can be expressed also in terms of ‘appar- Nonspherical particles There are two main effects
ent density’: to be accounted for with nonspherical particles: the
Rrst is related to the particle volume expression and
app"(1! )s# l [31] the second to Stokes Law. The parameter d contained
in eqn [7] has to be changed depending on the kind of
from which the differential apparent density is de- data available.
Rned as: The Rrst effect is accounted for by using the volume
equivalent diameter, i.e. the diameter of a sphere
app"app!l [32] having the same particle volume VSp, i.e.
dv"3((6VSp/
) instead of the sphere diameter, d.
By combining "s!l with eqns [31] and [32] This quantity sometimes can be related to true geo-
one has: metrical dimension of the particle if its geometry is
known.
app"(1! ) [33] An irregular shape affects the behaviour of the
particle while it is moving within the Suid. Stokes
which can be substituted in eqn [30].
Law takes account of this by substituting the dia-
When the ‘mass porosity’ is available:
meter d with the ‘drag diameter’, i.e. the diameter of
a sphere having the same resistance to the motion
Vp
p" [34] within the Suid
m
f"3 odd [39]
where m is the real mass of the particle, i.e. m"Vss
and by using eqn [27] one can show that:
In order to conclude this section, an appropriate
combination of all the above cases is necessary for


1
"(1! ) [35] irregular porous particles.
ps#1

which gives: List of Symbols

a "aspect ratio


bLgd 2 1
VQ (t)" [36] b "width of the SPLITT channel
18o ps#1 d "diameter of the sphere
dc "cutoff diameter
Alternatively, one can employ the ‘bulk density’, dd "drag diameter
which is the ratio between the amount of the porous dv "diameter of an equivalent sphere
material, mtot, and the total volume occupied by the D "diffusion coefRcient
packed particles Vtot, including both the inter-particle E "electrical Reld
volume, Vex, the particle volumes, V tot
p (total volume f "frictional coefRcient
of the pores) and V tot
s (total solid volume): Fa "retrieval of a component from outlet a
Fb "retrieval of a component from outlet b
mtot mtot Fg "gravitational force
bulk" tot " [37]
Vp #Vs #Vex Vtot
tot
Ff "frictional force
g "gravity acceleration
In this instance, both the bulk and ex depend on the h "thickness of the transport layer
degree of packing. The relative apparent density, hc "high of a cylindrical particle
app can be obtained by combining eqns [33] and L "length of the SPLITT channel
[37]: m "real mass of a particle
mb "mass corrected for buoyancy
bulk meff "effective mass
app"  [38]
s(1! ex) mtot "total mass of all particles present in the
container
From eqn [33] it is apparent that, in this case, one p "mass porosity
must know the value of ex under the same experi- r "radius of rotation
mental conditions which bulk was determined. Lack- tr "crossing time, i.e. the time a particle
ing this information, it is only possible to make takes to pass through the cross sectional
a rough estimate of app. area bh in the h-direction.
 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation
U "particle migration velocity
vN "mean Suid velocity
vmax "maximum Suid velocity
VQ "total volumetric Sow rate through cell
VQ (a
) "volumetric Sow rate at outlet a

VQ (b
) "volumetric Sow rate at outlet b

VQ (a) "volumetric Sow rate at inlet a
VQ (b) "volumetric Sow rate at inlet b
VQ (t) "volumetric Sow rate of the transport
region
Vex "external volume between particles
Vp "pore volume of a particle
Vtot
p "total pore volume for all the particles
Vs "volume of the solid part of a particle
Vtot
s "total solid volume for all the particles
VSp "volume of a sphere
Vtot "total volume occupied by all particles
present in the container
Vptot "total volume of a particle
w "thickness of the SPLITT channel
wa "thickness of the Suid lamina between wall
Y tionation: new technique for separation of magnetically
A and ISP
wt "thickness of the transport region
l "magnetic susceptibility of the carrier
p "magnetic susceptibility of a particle
 "internal porosity
 "suspension viscosity
o "carrier viscosity
 "electrophoretic mobility
app "apparent density
bulk "bulk density
l "density of the liquid
s "density of the spherical particle
 "angular velocity
D "dimensionless diffusion time
See also: II/Particle Size Separation: Field Flow Frac-
tionation: Electric Fields; Theory and Instrumentation
of Field Flow Fractionation. III/Polymers: Field Flow
Fractionation.
Theory and Instrumentation of Field Flow Fractionation
J. Janc\ a, Universite& de la Rochelle, La Rochelle,
France
Copyright ^ 2000 Academic Press
Principle
Field-Sow fractionation (FFF) is one of the important
analytical methodologies, suitable for the separation
and characterization of particles in the submicron
1837
Further Reading
Allen T (1981) Particle Size Measurement, 3rd edn. Lon-
don: Chapman and Hall.
Contado C, Dondi F, Beckett R and Giddings JC (1997)
Separation of particulate environmental samples by
SPLITT fractionation using different operating modes.
Analytica Chimica Acta 345: 99}110.
Contado C, Riello F, Blo G and Dondi F (1999) Continuous
split-Sow thin cell fractionation of starch particles. Jour-
nal of Chromatography A 845: 303}316.
Dondi F, Contado C, Blo G and Martin SG (1988) SPLITT
cell separation of polydisperse suspended particles of
environmental interest. Chromatographia 48: 643}654.
Fuh CB and Giddings JC (1995) Isolation of human blood
cells, platelets, and plasma proteins by centrifugal
SPLITT fractionation. Biotechnology Progress 11:
14}20.
Fuh CB and Giddings JC (1997) Separation of submicron
pharmaceutic emulsion with centrifugal split-Sow thin
(SPLITT) fractionation. Journal of Microseparation 9:
205}211.
Fuh CB and Chen SY (1998) Magnetic split-Sow thin frac-
susceptible particles. Journal of Chromatography A 813:
313}324.
Fuh CB, Levin S and Giddings JC (1993) Rapid diffusion
coefRcient measurements using analytical SPLITT frac-
tionation: application to proteins. Analytical Biochemis-
try 208: 80}87.
Levin S, Myers MN and Giddings JC (1989) Continuous
separation of proteins in electrical split-Sow thin
(SPLITT) cell with equilibrium operation. Separation
Science and Technology 24(14): 1245}1259.
Provder T (ed.) (1991) Particle Size Distribution. II. Assess-
ment and Characterization. ACS Symposium Series 472.
Washington DC: American Chemical Society.
Yong J, Kummerow A and Hansen M (1997) Preparative
particle separation by continuous SPLITT fractionation.
Journal of Microseparation 9: 261}273.
Zhang J, Williams PS, Myers MN and Giddings JC (1994)
Separation of cells and cell-sized particles by continuous
SPLITT fractionation using hydrodynamic lift forces.
Separation Science and Technology 29(18): 2493}2522.
and micron ranges. The effective Reld generates the
Sux of the separated particles and forms a concentra-
tion gradient of each particular species across the
ribbon-shaped separation channel. The concentration
gradients are counter-balanced by a diffusion Sux. At
equilibrium, a stable concentration distribution of
each particular species is established in the direction
across the channel. Simultaneously, a Sow velocity
proRle is formed across the channel due to the viscous
1838 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation
Figure 1 Schematic representation of the general principle and
of the experimental arrangement of FFF: (1) carrier liquid reser-
voir; (2) pump; (3) injector; (4) separation channel; (5) detector;
(6) computer; (7) external field; (8) hydrodynamic flow. Detail
shows the schematic representation of two fundamental separ-
ation mechanisms: polarization FFF and focusing FFF.
drag in the longitudinal Sow of the carrier liquid. As
a result, each particle is carried along the channel
with a velocity corresponding to an instantaneous
position of the particle within the Sow velocity pro-
Rle. The carrier liquid thus elutes each species with
a mean velocity which corresponds roughly to the
position of the centre of gravity of the Reld-induced
concentration distribution across the channel of that
species. This principle is schematically demonstrated
in Figure 1.
The separation is usually governed by the differ-
ences in size of the separated components of a poly-
disperse sample. If the appropriate relationship be-
tween the retention parameters and the size of the
particles is known or found empirically by using
a suitable calibration procedure, the fractograms can
be used to calculate the particle size distribution
(PSD) and the average values of the particle size of the
fractionated species. However, the intensive proper-
ties (such as the electrical charge, density, etc.) can
inSuence the separation based on size differences.
Theory of Separation
Two distinguished separation mechanisms, either
polarization or focusing, govern the separation. The
separated particles can be differently compressed to
the accumulation wall of the channel according to
their sizes or focused at different levels across the
channel according rather to an intensive property
(see Figure 1).
The polarizing Reld force, F, and the velocity of the
Reld-induced migration of the fractionated particles,
U, are usually constant and independent of the posi-
tion in the direction of the Reld action:
FO0 and UO0 within 0(x(w
where w is the thickness of the FFF channel in the
direction of the Reld action (x-axis); x"0 is situated
at the accumulation wall of the channel. The steady-
state concentration distributions of the sample com-
ponents across the channel are exponential:


x
ci(x)"ci(0) exp !
li
where li"Di/Ui is the mean layer thickness, Di is the
diffusion coefRcient and ci is the concentration of the
ith species. Larger particles are usually concentrated
more closely to the accumulation wall. As a result, the
order of the elution is from the small species to larger
ones.
The focusing Reld force and the corresponding
velocity U are position dependent:
F"f (x), U"f (x) within 0(x(w
F(x)"0, U(x)"0 for x"xmax , 0(xmax(w
The coordinate xmax corresponds to the position at
which the concentration distribution of a focused
sample is maximal. Each sample component is
focused around its proper xmax position. The steady-
state concentration distribution is close to the Gaus-
sian distribution:

 
1 dF(x)
c(x)"cmax exp ! (x!xmax)2
2kT dx x"xmax
where k is the Boltzmann constant and T is the
temperature. In some cases the polarization and the
focusing mechanisms can act simultaneously.
As mentioned above, a real separation channel is
usually ribbon-shaped. However, two parallel inRnite
planes represent a good approximation of this form.
The Sow velocity proRle established in such a hypo-
thetical channel is parabolic under isoviscous
conditions:
Px(x!w)
(x)"
2L
 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation 1839

where (x) is the linear velocity of a Sow streamline
at the position x, P is the pressure drop along the
channel of length L, and  is the viscosity of the
carrier liquid.
To describe conveniently the retention of the separ-
ated particulate species, the dimensionless retention
ratio R is deRned:
spreading should be applied. It is based on the decon-
volution of an experimental fractogram h(V) of
a polydisperse particulate sample which is a super-
position of the true PSD g(Y) and the spreading func-
tion G(V, Y) representing the zone of uniform par-
ticles having the elution volume Y:



w w
h(V)" g(Y)G(V, Y) dY
0c(x) (x) dx dx 0 0
R" w
0c(x) dx w0 (x) dx
where V and Y are then the elution volumes. This
equation, called the Tung integral equation, is the
R is the ratio of the average velocity of a retained
basis for all well-known correction methods and can
sample component divided by the average velocity of
be solved analytically under the condition that the
the carrier liquid. The integration gives the relation-
spreading function is uniform. In this case, the convo-
ship for the polarization FFF
lution integral to be solved is:

  
1 
R"6 coth !2 h(V)" g(Y)G(V!Y) dY
2 0

where "l/w. The analogous approximate relation- In a number of practical cases, the spreading function
ship for the focusing FFF is: can be approximated by the normal Gaussian func-
tion. The application of the correction of an experi-
R"6 ( max ! 2
max ) mental fractogram is demonstrated in Figure 2.
The true PSD can be expressed as a number of the
where max"xmax/w, is the dimensionless coordinate particles of a given diameter n relative to the number
G
of the maximal concentration of the focused zone. of all the particles in the sample:
R can be experimentally determined as the ratio of
the retention volume (or the retention time) of an ni
Ni"
unretained sample component (equal to the volume 
ni
of the channel) divided by the retention volume (re- 0
tention time) of the retained sample component. The
simple and known relationship between the and the or as the mass of the particles m of a given diameter
G
particle size make it possible to calculate the PSD d relative to the total mass of the sample:
G
from the experimental retention data.
Each fractionation is based on transport processes mi
Mi"
which lead to the formation of the concentration 
mi
gradients. From the thermodynamic point of view, the 0
general entropic tendency of a closed system is to erase
such gradients by molecular motion. As a result, the The PSD can be used further to calculate various
spreading of the zones due to dispersion processes average particle sizes such as the mass average par-
occurs. The zone spreading can be quantitatively de- ticle diameter:
scribed by the height equivalent to a theoretical plate H:
 
midi hidi


 2
H"L dM m" 0 "0
VR  
mi hi
where VR is the retention volume and  is the stan- 0 0

dard deviation of the zone of uniform size particles.

or the number average particle diameter:
The elution curve (fractogram) of a polydisperse
sample thus reSects the contribution of the spreading
 
processes superposed over the fractionation accord- nidi hi
ing to the PSD. Mdn" 0 " 0

In order to calculate a true PSD from the experi-  

ni hi/di
mental raw fractogram, a correction for the zone 0 0
1840 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation
Figure 2 Schematic representation of a procedure for the treat-
ment of a raw experimental fractogram to correct for zone
broadening.
where hi is the normalized detector response to the ith
particle diameter. The polydispersity of the frac-
tionated sample can be characterized, for example, by
the index of polydispersity:
dM m
I"
dM n
The above basic theory and data treatment can be
applied independently of a particular FFF method or
technique.
Instrumentation
Polarization FFF
In particle size separations by FFF, the nature of the
applied Reld (physical or chemical forces) determines
each particular method or technique of polarization
FFF and, consequently, the appropriate instrumenta-
tion. The most important polarization FFF methods
at the present time are:
E sedimentation FFF
E Sow FFF
E electric FFF
E thermal FFF
The basic experimental devices as well as speciRc
instrumentation are described here for each particular
FFF method or technique.
Independent of the method or technique, all FFF
apparatuses are composed of a system of solvent
delivery (reservoir, pump), injector sample (injection
valve, syringe-septum, etc.), separation channel (dif-
ferent construction for each method), detector (re-
fractive index detector, spectrophotometer, molar
mass detector, etc.) and a data acquisition and treat-
ment system (computer). With the exception of the
FFF separation channel, all other components, and
the system as a whole, are practically the same as
a conventional liquid chromatography system.
Schematic representation of the separation channel
for sedimentation FFF is shown in Figure 3(A). The
separation channel is coiled inside a centrifuge rotor.
A delicate part of this separation unit is the rotating
seal which must permit the Sow-through of a carrier
liquid and the connection to the injector at the entry
to the channel proper and of a detector at the exit.
However, this technical problem is solved and the
rotors for sedimentation FFF are commercially avail-
able. On the other hand, a home-built solution is
also possible providing that some technical skill is
available.
If the particles to be separated are relatively large
or dense and, consequently, the gravitational force is
enough to generate the formation of sufRciently
strong concentration gradients, the construction of
the separation channel is much simpler, as shown in
Figure 3(B). In this case, the channel is composed of
two sandwiched glass plates, one of them is provided
with holes and capillaries for carrier liquid entry and
exit and a thin foil in which the channel proper is cut.
The whole channel must be positioned horizontally to
avoid casual parasite convections which could cause
the separation to deteriorate.
The channel for Uow FFF is schematically demon-
strated in Figure 4(A). It is formed between two par-
allel, semipermeable membranes Rxed on porous sup-
ports. The cross-Sow of the carrier liquid is super-
posed perpendicularly to the Sow of the carrier liquid
in a longitudinal direction inside the channel. The
cross-Sow acts as an external Reld of hydrodynamic
forces which generate a uniform Sux of all particles.
 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation 1841

The channel for electric FFF is usually formed by

semipermeable membranes as in Sow FFF (see
Figure 5). The reason for such a solution is to de-
couple the separation channel proper from the elec-
trode chambers and thus to avoid the contamination
of the channel by products of electrolysis (gas bub-
bles). However, channels of simpler construction in
which the metal or graphite electrodes form the
channel walls and thus are not decoupled from the
separation space have been constructed and work
quite well under carefully chosen experimental condi-
tions. The channel for thermal FFF is constructed
in such a manner to allow a temperature difference
between two metallic bar walls with highly polished
surfaces. The walls are separated by a spacer in which
the channel proper is cut. The upper bar is heated by
using appropriate electrical cartridges and the lower
bar is cooled by circulating water. Both bars should
be equipped with several holes to accommodate the
thermocouples for temperature control. Schematic
representation of a channel for thermal FFF is shown
in Figure 6. In some cases, when the temperature of

Figure 3 Simplified schemes of the construction of the sedi-

mentation FFF channels used in a centrifuge and in natural
gravitational field. (A) Sedimentation FFF channel: (1) channel;
(2) direction of the flow; (3) rotation; (4) flow inlet; (5) flow outlet.
(B) Gravitational FFF channel: (1) channel walls; (2) foil spacer;
(3) inlet and outlet.

The carrier liquid passes through the membranes but

the separated particles should not, due to the conve-
niently chosen porosity of the membranes. The uni-
formity of the cross-Sow is, however, not necessary to
achieve high performance separation. If only one of
the main channel walls is semi-permeable, a non-
uniform hydrodynamic Reld is generated in such an
asymmetrical Sow FFF channel. The dependence of
the separation resolution on particle size in such
a channel is different compared with a channel
equipped with two semi-permeable walls, but high
performance particle size separation is also
achieved.
A classical type of rectangular cross-section chan-
nel has sometimes been substituted with a circular
cross-section capillary with an overpressure applied
inside or by applying an external cross-Sow in a more
standard manner, as shown in Figure 4(B). The sim-
plicity of the construction of such a ‘channel’ is the Figure 4 (A) Construction of a rectangular cross-section chan-
nel for flow FFF: (1) porous supports; (2) cross-flow inlet and
main advantage of this conRguration. The theoretical
outlet; (3) membranes; (4) foil spacer; (5) longitudinal flow inlet;
description of the separation is complex, however, (6) longitudinal flow outlet. (B) Circular capillary for flow FFF with:
and, moreover, the probability of the formation of (1) overpressure applied from the inside; (2) cross-flow applied
parasite Sows degenerating the separation is higher. externally.
1842 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation
Figure 5 Construction of a channel for electric FFF:
(1) electrodes and electrolyte inlet and outlet; (2) membranes;
(3) foil spacer; (4) longitudinal flow inlet; (5) longitudinal flow
outlet.
the heated wall is above the boiling point of the
carrier liquid used, the channel must be sealed so as to
operate under high-pressure conditions. The thick-
ness of the channel can be as low as few micrometers
which permits performing high-speed and high-res-
olution fractionations. The separation can be ac-
complished in just a few seconds.
Focusing FFF
Focusing FFF methods have been classiRed according
to various combinations of the driving Reld forces and
gradients:
E effective property gradient of the carrier liquid,
E cross-Sow velocity gradient,
E lift forces,
E shear stress, and
E gradient of the non-homogeneous Reld action.
Figure 6 Construction of a channel for thermal FFF: (1) electric
heating cartridge; (2) cooling liquid inlet and outlet; (3) foil spacer;
(4) holes for thermocouples; (5) longitudinal flow inlet; (6) longitu-
dinal flow outlet.
While this classiRcation scheme is perfectly consistent
with fundamental separation mechanisms and related
driving forces, particular focusing FFF methods and
techniques are more often called according to experi-
mental procedure. The instrumentation will be de-
scribed for each implemented focusing FFF method or
technique.
The channels for sedimentation}Sotation focusing
Reld-Sow fractionation (SFFFFF) or isoelectric
focusing Reld-Sow fractionation (IEFFFF) are
either of standard rectangular cross-section or of
modulated cross-sectional permeability (for example,
of trapezoidal or triangular cross-section), as
shown in Figures 7(A) and (B). While the Sow
velocity proRle in channels of rectangular cross-
section are symmetrical (e.g. parabolic), the
modulated cross-sectional permeability channels
allow formation of Sow velocity proRles which
are not symmetrical. The advantage of these
channels is that almost all zones focused symmetric-
ally regarding the central longitudinal axis of the
separation channel can be separated. If the Sow velo-
city proRle is symmetrical, the zones focused at the
opposite sides regarding the central axis of the chan-
nel can be confused.
Both above-mentioned methods belong to the
Rrst category in which an effective property
gradient of the carrier liquid represents the
major driving force. The focusing in these cases
can appear to be due to the effective property gradi-
ent of the carrier liquid in the direction across the
channel combined with the primary or secondary
transverse Reld.
It has been shown that the gradient of the
effective property of the carrier liquid can be
performed at the beginning of the channel. For
example, the step density gradient can easily be
formed by pumping the carrier liquids of various
densities through several inlet capillaries into the
channel. Such an arrangement can effectively be used
for continuous preparative fractionation providing
that the separation channel is also equipped with
several outlet capillaries to continuously collect the
fractions which are focused at different levels. Sche-
matic representation of such a channel is shown in
Figure 8.
The elutriation focusing Reld-Sow fractionation
(EFFFF) method belongs to the category in which the
focusing is due to the gradient of transversal Sow
velocity of the carrier liquid which opposes the action
of the external Reld. The longitudinal Sow of the
carrier liquid is acting simultaneously. A trapezoidal
cross-section as well as a rectangular cross-section
channels can be used in this case. Schematic repres-
entation of such a channel for elutriation FFF is
 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation 1843

Figure 8 Continuous preparative channel for focusing FFF in

preformed step density gradient: (1) gravitational field; (2) flow
inlets; (3) flow outlets.

Very few experiments have been published on FFF

exploiting the hydrodynamic lift forces at high carrier
Sow rates which, with the high shear gradient, result
in the deformation of soft particles and their sub-
sequent displacement and focusing. Similarly little
has been published on FFF using a non-homogeneous
high gradient external Reld. Although these methods
can, in principle, use one of the types of channel
described above for other focusing FFF methods, no
experimental proof for this currently exists.

Conclusion
Figure 7 (A) Schematic representation of a channel for sedi-
mentation flotation focusing FFF in coupled electric and gravi- A large number and variety of homemade
tational fields: (1) flow in; (2) flow out; (3) electrodes forming the channels exist which conRrms that in most cases, the
channel walls; (4) spacer. (B) Schematic representation of a trap- construction of a channel is not extremely difRcult.
ezoidal cross-section channel for isoelectric focusing FFF: (1) Pt
anode; (2) Pt cathode; (3) anolyte; (4) catholyte; (5) ampholyte;
(6) sample; (7) to detector; (8) trapezoidal cross-section channel;
(9) membranes.

shown in Figure 9. The channel shown has a trap-

ezoidal cross-section which causes formation not only
of a convenient, axially asymmetrical Sow velocity
proRle but, providing the volumetric transversal Sow-
in and Sow-out are equal, a linear velocity gradient is
established across the channel. In combination with
different constant velocities of different size-separ-
ated particles the conditions for the focusing phenom- Figure 9 Schematic representation of a channel for elutriation
enon to appear are established. focusing FFF: (1) field force; (2) cross-flow; (3) longitudinal flow.
1844 II / PARTICLE SIZE SEPARATION / Theory and Instrumentation of Field Flow Fractionation
However, commercial FFF apparatus is increasingly
available which could further stimulate interest in
applying this high performance separation methodo-
logy in routine laboratory practice.
See also: II/Particle Size Separation: Field Flow Frac-
tionation: Electric Fields. III/Cells and Cell Organelles:
Field Flow Fractionation.
Further Reading
Barth HG (ed.) (1984) Modern Methods of Particle Size
Analysis. New York: John Wiley.
Janc\ a J (1987) Field-Uow fractionation: analysis of
macromolecules and particles. New York: Marcel
Dekker.
Janc\ a J (1995) Isoperichoric focusing Reld-Sow fractiona-
tion based on coupling of primary and secondary Reld
action In: Provder T, Barth HG and Urban MW (eds)
Chromatographic Characterization of Polymers, Hy-
phenated and Multidimensional Techniques. Advances
in Chemistry Series 247. Washington DC: American
Chemical Society.
Janc\ a J (1999) Field-Sow fractionation. In: Pethrick RA and
Dawkins JV (eds) Modern Techniques for Polymer
Characterisation. New York: John Wiley.

ACIDS
Gas Chromatography
G. Gutnikov and N. Scott, California State Polytechnic
University, Pomona, CA, USA
Copyright ^ 2000 Academic Press
Introduction
The Rrst separation of acids by gas chromatography
(GC) coincides with the inception of GC itself. In
1952 James and Martin pioneered GC by demon-
strating the separation of the C1 to C12 aliphatic acids
on a stationary phase of silicone oil DC 550 contain-
ing stearic acid or H3PO4 and quantifying using
a special titrimetric detector. Since then, the GC anal-
ysis of acids has been extended to a very wide variety
of species and samples. To enable ready application
of GC, the acids are usually converted to suitable
volatile derivatives for resolution on efRcient col-
umns. As they are eluted they must be identiRed by an
appropriate technique, the most deRnitive being mass
spectrometry (MS). Various applications are present-
ed in this article.
Derivatization
It was noted early on that separation of free acids is
frequently hampered by their relatively low volatility,
molecular association and, particularly, their adsorp-
tion on the stationary phase support with the result-
ant tailing, peak distortion and ghosting. Although
special columns (FFAP, OV-351, SP-1000) have been
developed since then for the separation of short and
medium chain free (underivatized) aliphatic acids, the
majority of carboxylic acids (especially those contain-
ing additional polar substituents) are insufRcient-
ly volatile for analysis by GC. Therefore, the carboxyl
and other polar groups are usually converted to less
polar derivatives to improve their chromatographic
properties.
Carboxylic Acids
Both free and bound carboxyl groups are almost
exclusively derivatized to volatile esters } predomi-
nantly silyl and methyl } by a variety of methods.
These employ a number of silylation reagents, acid-
and base-catalysed reactions, on-column pyrolysis,
III / ACIDS / Gas Chromatography 1847
diazomethane and other reagents. Each has its ad-
vantages, limitations and special applications.
Silyl esters Silylation is now one of the most exten-
sively used techniques for esterifying free acids prim-
arily because of its speed, convenience and the simul-
taneous derivatization of other polar functional
groups containing an active hydrogen (}OH, }SH,
}NH2). The trimethylsilyl (TMS) group is the most
commonly introduced substituent by the many
silylating agents available, of which N,O-bis-
(trimethylsilyl)triSuoroacetamide (BSTFA) is the
most widely used. It reacts with all the common polar
functionalities and yields volatile by-products that
are usually eluted with the solvent. Even more
volatile by-products are produced by substituted re-
agents, e.g. N-methyl-N-trimethylsilyltriSuoroacetam-
ide (MSTFA), which are also more reactive toward
the polar functional groups. Although all silylating
reagents and their products are sensitive to moisture,
considerably greater hydrolytic stability is exhibited
by t-butyldimethylsilyl (TBDMS) derivatives that are
best prepared with N-t-butyldimethylsilyl-N-methyl-
triSuoroacetamide (MTBSTFA), which can also serve
as its own solvent. It yields excellent results with both
volatile and nonvolatile carboxylic acids (Figure 1).
A limitation of silylation is that bound acids such
as lipids (triacylglycerols) are not converted and
their derivatization to methyl (or other alkyl) esters is
necessary.
Alkyl esters Methyl esters are most frequently pre-
pared by acid-catalysed reactions with methanol. The
principal advantage of this method is the concurrent
esteriRcation of free acids and the transesteriRcation
of bound ones. The most extensively used catalysts
are BF3, HCl and H2SO4, usually as 14%, 5% and
2% solutions, respectively. The reaction is fastest
with BF3, requiring the mixture to be boiled for 2 min
for free acids and 30}60 min for lipids. With HCl and
H2SO4 about twice the time is required. The higher
concentration of BF3 used compared to the other
catalysts may be responsible not only for the faster
reaction, but also for partial degradation of un-
saturated acids and reported artefact formation.
These problems can be reduced by prior saponiR-
cation with methanolic KOH, followed by re-
esteriRcation of the free acids formed under mild
conditions. Several ofRcial methods are based on
this procedure.
1848 III / ACIDS / Gas Chromatography

Figure 1 Chromatogram of a mixture of carboxylic acids as the t-butyldimethylsilyl derivatives. GC conditions: 30 m;0.32 mm i.d.,
DB-1 fused-silica capillary column initially at 603C for 2 min, then programmed to 2803C at 43C min\1; 0.8 L sample, injected with split
ratio of 15 : 1; both injector and detector temperatures at 3003C; nitrogen as the carrier gas at 0.9 mL min\1. Peaks: 1, Formic; 2, acetic;
3, propionic; 4, isobutyric; 5, butyric; 6, isovaleric; 7, valeric; 8, caproic; 9, enanthic; 10, benzoic; 11, caprylic; 12, lactic; 13,
phenylacetic; 14, glycol; 15, oxalic; 16, pelargonic; 17, malonic; 18, capric; 19, succinic; 20, methylsuccinic; 21, undecanoic; 22,
fumaric; 23, 5-phenylvaleric; 24, p-aminobenzoic; 25, lauric; 26, mandelic; 27, adipic; 28, 3-methyladipic; 29, tridecanoic; 30,
phenyllactic; 31, hippuric; 32, myristic; 33, p-hydroxybenzoic; 34, malic; 35, suberic; 36, pentadecanoic; 37, vanillic; 38, palmitic; 39,
syringic; 40, tartaric; 41, margaric; 42,
-resorcylic; 43, p-hydroxymandelic; 44, -resorcylic; 45, stearic; 46, homogentisic; 47,
protocatechuic, 48, nonadecanoic; 49, citric; 50, arachidic acid. (Reproduced with permission from Kim KR, Hahn MK, Zlatkis A et al.
(1989) Simultaneous gas chromatography of volatile and nonvolatile carboxylic acids as tert-butyldimethylsilyl derivatives. Journal of
Chromatography 468: 289.

Substituting microwave irradiation for conven- pyrolytic conversions include (m-triSuoromethyl-

tional heating may substantially reduce reaction
times and lipid degradation. Thus, using the BF3-
methanol reagent, a reaction time of 30 s sufRced
for the transesteriRcation of most lipids to their fatty
acid methyl esters (FAMEs) with less oxidation of the
unsaturated species.
Base-catalysed reactions are used extensively for
the transesteriRcation of lipids because they proceed
faster than those in acid media without degradation
of the unsaturated fatty acids. However, they do not
esterify free fatty acids. The most commonly used
reagents are methanolic solutions of NaOCH3 or
KOH. Transmethylation of lipids is usually complete
in 5 min at room temperature.
Strong organic bases can be used similarly and
possess the great advantage of forming salts which,
unlike their inorganic analogues, can be pyrolysed
to methyl esters at the high temperatures of a GC
injection port. This permits simple one-step deter-
mination of both free and bound acids. The or-
ganic bases that have been recommended for such
phenyl)-trimethylammonium, trimethylphenylam-
monium and trimethylsulfonium hydroxides. The
latter reagent requires the lowest pyrolysis temper-
ature and yields innocuous by-products. It is simply
added to the sample solution, mixed and injected.
EsteriRcation of free acids with diazomethane pro-
ceeds rapidly in high yield under mild conditions,
with minimal side reactions. Special microequipment,
reagents and procedures have been developed that
allow its relatively safe handling despite its toxic and
explosive nature. Other reagents of interest include
alkyl chloroformates that can esterify free acids
even in the presence of a considerable amount of
water (40%). Another reagent, dimethylformamide
dimethylacetal, can be simply mixed with the sample
of acid and injected into the GC; the reaction occurs
in the hot injection port. Silver or potassium salts of
acids can be converted to esters with methyl iodide or
sulfate. Many other reactions have been reported.
Short chain acids are frequently derivatized to
higher esters with butanol or isopropanol and acid
 III / ACIDS / Gas Chromatography 1849

catalysts in order to mitigate losses due to volatility

and substantial water solubility. Higher diazoalkanes
may also be used if the methyl esters are too
volatile.
Enantiomers of optically active carboxylic acids
have been separated following acid-catalysed esteriR-
cation with a chiral alcohol such as S(#)-2-butanol,
R(!)-2-octanol, or (!)-methanol or transesteriRca-
tion with sodium menthylate. Diastereometric esters
have also been prepared from optically active acids by
reaction with O-(!)-menthyl-N,N-diisopropylisourea.
The above silyl and alkyl esters are most commonly
detected by a Same ionization detector (FID). Greater
sensitivity, however, can be achieved by forming
halogenated silyl esters, e.g. chloromethyldimethyl-
silyl, and monitoring with an electron-capture de-
tector (ECD). Similarly, very small amounts of
volatile acids may be detected via their penta-
Suorobenzyl (PFB) esters with an ECD. Special deriv-
atives for this detector include the 2-chloroethyl and
trichloroethyl esters.

Figure 2 Separation of N,O -heptafluorobutyryl amino acid iso-

Other derivatives The silyl and alkyl esters de-
scribed are generally also suitable for detection by
MS. However, special derivatives are necessary for
unsaturated fatty acids to prevent double-bond mi-
gration during fragmentation. The most widely used
derivatives are those of 3-hydroxymethylpyridine
(picolinyl) and 4,4-dimethyloxazoline (DMOX).
Picolinyl esters must be prepared from the acid but
DMOX derivatives can be prepared even from their
esters.
Amino Acids
For amino acids, derivatization is indispensable for
analysis by GC since they all exist in the zwitterion
form. Some also contain other polar functionalities,
including hydroxyl, thiol and imino groups. The dif-
ferent reactivities of these groups greatly complicate
their concurrent derivatization. Silylation offers
the best approach for a single-step attachment of the
same tag to all these functional groups.
The most successful attempt to generate a single
product is by silylation with MTBSTFA to form
TBDMS derivatives. Reaction conditions (heating at
1503C for 2.5 h) were developed for the reproducible
derivatization of amino acids in high yield. TMS
derivatives of the common amino acids, except ar-
ginine, can also be prepared with BSTFA under sim-
ilar conditions.
An alternative method of derivatization of amino
acids entails Rrst esteriRcation and then acylation
to produce various N-acyl alkyl esters (Figure 2). The
most widely used of these combinations is the
butyl ester derivatives obtained from silkworm t -RNA after
deacylation and analysed with FID. GC conditions:
25 m;0.4 mm i.d. capillary column coated with 5% Chromosorb
R and 15% OV-101 SCOT column; carrier gas, hydrogen at a flow
rate of 3 mL min\1; make-up gas, nitrogen at a flow rate of
30 mL min\1; hydrogen flow rate, 27 mL min\1; air flow rate,
350 mL min\1; temperatures: detector, 3203C; no inlet heater
block; column, 803C programmed at 43C min\1. Pulse interval,
15
s; attenuation, 2;102; sample size, 20
L. Peaks: 1, Alanine;
2, glycine; 3, valine; 4, threonine; 5, serine; 6, leucine; 7,
isoleucine; 8, norleucine (I.S.); 9, proline; 10, methionine; 11,
aspartic; 12, glutamic acid; 13, lysine; 14, tyrosine; 15, arginine.
(Reproduced with permission from Chauhan J and Darbre A
(1982) Determination of amino acids by means of glass capillary
gas-liquid chromatography with temperature-programmed elec-
tron-capture detection. Journal of Chromatography 236: 151.
N-triSuoroacetyl-n-butyl ester (TAB) derivative. Es-
teriRcation is performed by one of the methods de-
scribed above and acylation by heating the dried
product with triSuoroacetic anhydride. The selectiv-
ity of the NP detector can be exploited to monitor
amino acids in the presence of interfering matrices,
particularly lipids.
Enantiomeric resolution has been achieved with
a chiral aliphatic alcohol and an achiral acylating
agent such as N-triSuoroacetyl chloride. Alterna-
tively, the amino group has been converted to dia-
stereomeric amides, ureas, thioureas and isoindoles.
Resolution
Since many real samples are complex mixtures of
acids (and other components), high efRciency
1850 III / ACIDS / Gas Chromatography
columns are essential for satisfactory resolution. This
requirement has made packed columns effec-
tively obsolete for such samples and use of capillary
(or open tubular, OT) columns is becoming routine.
The high efRciency of OT columns requires corre-
spondingly less selectivity to gain the necessary
separation. Therefore, relatively few different sta-
tionary phases in OT columns will adequately separ-
ate the majority of mixtures encountered.
Nonpolar stationary phases have the advantages of
greater inertness, thermal stability and operation at
lower temperatures. Since retention times increase
with increasing polarities of the stationary phase and
analyte, the least polar column affording the
necessary resolution should be selected. Silyl deriva-
tives are usually adequately separated on nonpolar
polydimethylsiloxanes (e.g. DB-1, SE-30, OV-101);
for greater selectivity somewhat more polar phases
such as DB-5, SE-54, OV-17 or even OV-1701 may
be used. On the other hand, stationary phases con-
taining hydroxyl groups (such as the polyethylene
glycols, PEGs) should be avoided because they react
with silylation reagents.
Saturated and unsaturated FAMEs are generally
separated on more polar columns because they tend
to cluster together on nonpolar phases, with the un-
saturated ones preceding the saturated. On polar
phases such as PEGs, the unsaturated are eluted after
the saturated with minimal overlap of different
chain lengths. This shift in retention behaviour is
further enhanced on very polar stationary phases such
as the cyanosilicones (CP-Sil-88, OV-275, DB-23)
which are used for resolving cis, trans isomers and
very complex mixtures.
Relatively nonpolar columns are used for the separ-
ation of diastereomeric esters formed from optically
active carboxylic and amino acids. As an alternative
approach, amino acid enantiomers have been separ-
ated as their alkyl N-perSuoroacyl derivatives on
a chiral column, e.g. Chirasil-Val.
Identi\cation
With conventional GC detectors, such as the FID and
ECD, identiRcation of the most commonly encoun-
tered acids is based on comparison of the retention
times obtained with authentic standards. For uniden-
tiRed acid peaks in general, retention index values or,
for FAMEs, equivalent chain lengths (ECL) from the
literature may be helpful. The preferred solution is,
however, MS detection in view of the more deRnitive
structural information it provides. Especially for car-
boxylic acids, the usual data (e.g. molecular weights,
fragmentation patterns, isotopic peak patterns) af-
forded by MS are supplemented by additional struc-
tural information, the most useful being the degree of
unsaturation.
The presence of a double bond can be deduced
from the molecular weight of an ester but its location
cannot be ascertained due to migration during frag-
mentation. Hence, for reliable identiRcation of posi-
tional isomers by GC-MS, two methods are em-
ployed: the on-site method of Rxing the location of
the double bond through its chemical modiRcation,
or the remote group method in which the carboxylic
group is derivatized to a nitrogen-containing product
which restricts double-bond migration. The remote
group method is more convenient and versatile.
Chemical modiRcation involves the addition of
a reagent across the double bond of the acid ester to
generate a product which gives diagnostic fragment
ions. Dimethyl disulRde is a widely used reagent since
it adds to a double bond in a single step at room
temperature and enables identiRcation of positional
and geometrical isomers after separation on an ap-
propriate column. But the picture is less clear with
polyenoic acids, especially when the double bonds are
in close proximity, and with acids containing other
structural features such as cyclopropane rings. Diels-
Alder reactions with cyclopentadiene derivatives can
be applied similarly. The double-bond site may also
be established by treating the unsaturated acid with
OsO4 and converting the resulting diol to the bis-
TMS ethers for GC-MS analysis. Although this
method is suitable for locating the double-bond sites
of polyunsaturated acids, their fragmentation pat-
terns are more complex and careful interpretation is
necessary.
In derivatizing the carboxylic group, the picolinyl
and DMOX compounds are the most commonly gen-
erated nitrogen-containing products. In the mass
spectra of these derivatives, the saturated segments of
the molecules are indicated by the regular separation
of successive peak clusters by 14 amu (corresponding
to the cleavage of a CH2 group), whereas at double-
bond sites the gap is only 12 amu. Furthermore, frag-
mentation on either side of the double bond gives two
ions which are separated by 26 amu. In a branched
acid derivative, the site of branching is shown by
a similar gap of 28 amu.
Geometrical isomers and ring structures are more
reliably identiRed by infrared (IR) spectrometry,
which underscores the utility of GC-Fourier trans-
form IR (FTIR)-MS in the structure elucidation of
acids. However, the inherently lower sensitivity of IR
requires larger sample sizes and columns with a
higher load capacity.
Quantitative analysis of acids by GC-MS is carried
out most sensitively by selected ion monitoring (SIM)
employing an isotopically labelled analogue or a
 III / ACIDS / Gas Chromatography 1851

derivative of a structurally similar acid as internal Applications

standard. The desired sensitivity of detection is a criti-
cal factor in the choice of the derivative. For increased
sensitivity ion currents must be intensiRed by reduc-
ing fragmentation. Hence, TBDMS derivatives are
preferred to those of TMS. Moreover, TBDMS de-
rivatizes the amino acids arginine and glutamine,
whereas TMS fails to do this. (However, the preferred
method for quantiRcation of amino acids involves the
butyl perSuoroacyl derivatives.) Fragmentation may
also be reduced by increasing molecular stability via
cyclic derivatives, as illustrated by quinoxalinol com-
pounds utilized in the GC-MS analysis of 2-oxo-
acids. An excellent method of augmenting sensitivity
is performing negative ion mass spectrometry via
derivatives (e.g. p-nitrobenzyl, pentaSuorobenzyl)
with high electron afRnity.
These methods have allowed the determination of
a variety of acids by GC-MS at pg levels. Even mix-
tures of acids can be analysed quantitatively by
monitoring several characteristic ions. Programmable
SIM, which optimizes the selectivity at various points
in a chromatogram and the desired sensitivity of
analysis, has been invaluable in this regard.
Examples of GC analysis of acids in real-world sam-
ples are so numerous and diverse as to permit only
representative cases from more signiRcant Relds to be
cited.
Carboxylic acids present at abnormal levels in
plasma and urine may indicate various metabolic
disorders. Hence, their monitoring is vital for diag-
nostic purposes. GC has simpliRed such analysis by
expediting the separation and determination of very
low concentrations of acids present in these complex
matrices (Figure 3). For example, C27 and C29 bile
acid levels provide the basis for a screening test for
a genetic condition characterized by peroxisomal dys-
function syndrome and are measured by GC-MS as
methyl-silyl derivatives. Elevated levels of certain
acylcarnitines may signify a potentially lethal condi-
tion caused by the deRciency of an enzyme which is
essential for -oxidation of fatty acids. Their quantiR-
cation by GC-MS has been achieved by the ready
conversion to volatile acyloxylactones. Metabolic
products of amino acids whose presence in urine
at unusually high levels may be symptomatic of

Figure 3 Chromatogram of a 3-hydroxy-dicarboxylic aciduria. GC conditions: 30 m;0.32 mm i.d. column coated with OV-1701;
temperature-programmed from 70 to 2703C at a rate of 53C min\1. Detector: FID. Some important peaks are indicated: 1, lactic di
TMS;2, oxalic di TMS; 3, 3-hydroxy-propionic di TMS; 4, 3-hydroxybutyric di TMS; 5, 3-hydroxy-isobutyric di TMS; 6, 2-methyl-3-
hydroxybutyric di TMS; 7, 3-hydroxy-isovaleric di TMS; 8, internal standard; 9, 3-hydroxy-adipic lactone mono TMS; 10, adipic di TMS;
11, hexenedioic di TMS; 12, triglycine mon TMS; 13, 4-hydroxy-phenylacetic di TMS; 14, octenedioic di TMS; 15, 3-hydroxy-adipic tri
TMS; 16, suberic di TMS; 17, 3-keto-adipic enol tri TMS; 18, aconitic tri TMS; 19, citric tetra TMS; 20, hippuric mono TMS; 21,
decenedioic di TMS; 22, 3-hydroxy-octenedioic tri TMS; 23, 3-hydroxy suberic tri TMS; 24, sebacic di TMS; 25, 4-hydroxy-phenyllactic
tri TMS; 27, 3-hydroxy-decendioic tri TMS; 28, 4-hydroxy-phenolpyruvic enol tri TMS; 29, 3-hydroxy-sebacic tri TMS; 31, 3-hydroxy-
dodecadienedioic tri TMS; 32, 3-hydroxy-dodecenedioic tri TMS; 33, 3-hydroxydodecenedioic tri TMS; 34, 3-hydroxy-dodecanedioic tri
TMS; 37, 3-hydroxy-tetradecadienedioic tri TMS; 38, 39, 3-hydroxy-tetradecentedioic tri TMS; 40, 3-hydroxy-tetradecanedioic tri TMS;
Ph"phosphoric tri TMS. (Reproduced with permission from Lefevere MF, Verhaeghe BJ, Declerk DH et al. (1989) Metabolic profiling
of urinary organic acids by single and multicolumn capillary gas chromatography. Journal of Chromatographic Science 27: 23.
1852 III / ACIDS / Gas Chromatography

metabolic disorders, e.g. hydroxyproline in collagen
metabolism, and -carboxyglutamate in blood coagu-
lation and bone metabolism. These compounds are
converted to N-isobutyloxycarbonyl methyl deriva-
tives prior to measurement.
Prostaglandins, which are indicators of several dis-
eases, are a class of acidic biomolecules whose
measurement in biomatrices still presents a formid-
able analytical challenge. They are present in urine at
concentrations as low as few pg mL\1 and require the
quantiRcation of several structurally closely related
compounds. The difRculties are further com-
pounded by their extreme sensitivity to acids, bases
and oxygen. The determination of prostaglandin
E2 has been achieved by negative ion chemical ioniz-
ation (NICI)-GC-MS following methylation and de-
rivatization of other functionalities. There are several
methods reported for the determination of other pros-
taglandins by isotope dilution GC-MS. GC-MS has
been of immense utility in elucidating the role of
-aminobutyric acid as a neurotransmitter via its 15N-
labelled derivative. Catecholamines and their acidic
metabolites such as homovanillic, vandillomandelic,
5-hydroxyindole-3-acetic and phenylacetic acids, are
implicated as etiological factors in affective dis-
orders. They have been determined by NICI-GC-MS
via acetyl-PFB derivatives and by isotope dilution
GC-MS. In clinical research, GC-MS has proved in-
valuable for pharmacokinetic studies of therapeutic
drugs with acidic functionalities. Such studies have
been performed on methylphenidate, which is used in
the treatment of children suffering from hyper-
kinesia, and on the butyl ester-triSuoroacetyl deriva-
tive of isotopically labelled histidine in investigations
of the hereditary metabolic disorder histidinaemia.
Another application is the analysis of the anti-inSam-
matory drug biphenylacetic acid in urine and synovial
Suid by NICI-GC-MS-MS via its PFB ester. Some
therapeutic drugs can lead to a build-up of toxic
metabolites that must be monitored. This is exempli-
Red by GC-MS analysis of patients’ urine and plasma
for 2-n-propyl-4-pentenoic acid, which is a product
of the antiepileptic drug valproic acid.
In analytical microbiology, GC of fatty acids pro-
vides a basis for microbial chemotaxonomy and
a means of identifying genus, species and even strains
of microorganisms (Figure 4). The compounds pro-
Rled may be the nonvolatile C10}C20 fatty acids pres-
ent in cell membranes or the volatile acids which
accumulate in the headspace. The extraction of the
nonvolatile fatty acids and their derivatization to
alkyl esters have been simpliRed by commercially
available automated systems. Fatty acid proRles have
permitted identiRcation of pathogenic bacteria and
even strains of yeast.
The realization that the enantiomers of a chiral
compound may exhibit different bioactivities has
prompted pharmaceutical and other industries to

Figure 4 Reconstructed chromatogram of fatty acid methyl esters from the unicellular alga Tetraselmis suecica obtained by GC-MS.
Chromatographic conditions: 50 m;0.20 mm i.d. methylsiloxane fused capillary column; column temperature, initially at 403C for
1 min, increased to 1203C at 303C min\1 and then to 3103C at 43C min\1; helium carrier gas. (Reproduced with permission from
Volkman JK, Jeffrey SW, Nichols PD et al. (1989) Fatty acid and lipid composition of ten species of microalgae used in mariculture.
Journal of Experimental Marine Biology and Ecology 128: 219.
 III / ACIDS / Gas Chromatography 1853

ascertain the optical purity of products and the meta-
bolic fate of each enantiomer. As a result, industries
and regulatory bodies have evinced interest in reliable
methods for resolving optically active compounds. In
the particular case of chiral acids, GC has proved
invaluable. This is clearly illustrated by the separation
of the optical isomers of the common drug ibuprofen
via diastereoisomeric esters, and by a group of anti-
inSammatory drugs, arylpropionic acids, which are
routinely monitored in biological Suids as their
R(!)/S(#)-amphetamine derivatives.
The differentiation between biogenic and non-
biological urinary carboxylic acids is vital in the for-
ensic sciences to establish the use of illicit drugs.
Cannabis is the most widely used illicit drug in
the world. 11-Nor--9-tetrahydrocannabinol acid
(THCA) is found in urine specimens of cannabis users
at few ng mL\1 levels as a major metabolite of tet-
rahydrocannabinol. THCA may be detected in urine
4}6 days after use of marijuana and even up to
a month in chronic users: its determination by GC,
principally as the TMS derivative, has been the focus
of much research. Benzoylecgonine, which is a car-
boxylic acid produced by de-esteriRcation of cocaine
at physiological pH and temperature, and ecgonine
methyl ester are the major metabolites that appear in
the urine of cocaine users. Both are analysed by either
GC-ECD or GC-FID, after converting the acid to the
TMS derivative.
Toxic haloacids are environmentally signiRcant
and may be present in drinking water and other
beverages. They are monitored by GC-MS or GC-
ECD as the methyl esters. Low concentrations of
pesticide and herbicide residues contaminating fruits
and vegetables present another health hazard, e.g.
residues of the fungicidal metal salts of alkylene-bis-
dithiocarbamic acids. These fungicides are Rrst con-
verted to CS2 for analysis by headspace GC. Traces of
some widely used acidic herbicides, such as chlorin-
ated phenoxycarboxylic acids, are quantiRed in food
samples by GC-MS as their methyl esters.
Carboxylic acids and derivatives are important Sa-
vour and aroma constituents of foods (Figure 5) and
beverages. Volatile fatty acids that are present at low
concentrations also contribute to organoleptic char-
acteristics and can be determined by headspace GC in

Figure 5 Gas chromatogram of free fatty acids (FFAs) from cheese spiked with an FFA reference mixture and short chain FFA (2:0,
3:0, 2-CH3-3:0, 5:0, 3-CH3-4:0 and 7:0). Chromatographic conditions: 25 m;0.32 mm i.d. fused silica capillary column coated with
FFAP-CB; oven temperature-programmed to increase from 65 to 2403C at a rate of 103C min\1; FID detector; helium carrier gas at
a flow rate of 2 mL min\1. Peaks: 1, C2; 2, C3; 3, 2-CH3-C3; 4, C4; 5, 3-CH3-C4; 6, C5;7, C6; 8, C7; 9, C8; 10, C9; 11, C10; 12, C10:1;
13, C11; 14, C12:0; 15, C12:1; 16, C13-iso; 17, C13:0; 18, C14-iso; 19, C14:0; 20, C14:1#C15-iso; 21, C15-anteiso; 22, C15:0; 23,
C15:1; 24, C16-iso; 25, C16:0; 26, C16:1; 27, C17-iso; 28, C17-anteiso; 29, C17:0; 30, C17:1; 31, C18-iso; 32, C18:0; 33, C18:1; 34,
C18:2; 35, C18:2; 36, C19:0; 37, C18:3; 38, C18:2 conjugated; 39, C20:0; 40, C20:1. (Reproduced with permission from de Jong C and
Badings, HT (1990) Determination of free fatty acids in milk and cheese. Procedures for extraction, clean up and capillary gas
chromatography. Journal of High Resolution Chromatography 13: 94.
1854 III / ACIDS / Liquid Chromatography
underivatized form. Fatty acids containing unusual
structural features, such as cyclopropane rings or
epoxy groups, are constituents of some edible veg-
etable oils and are suspected of being health hazards.
Hence they have been analysed in foods by capillary
GC as FAMEs. Such studies have provided a basis for
identifying components in blends of vegetable oils
with potential application to detecting adulteration.
Similar studies have been carried out to determine
brominated acid constituents in vegetable oils that are
added to disperse Savouring constituents in citrus-
based beverages. Clinical and epidemiological Rnd-
ings of the beneRcial effects of Rsh oils have led
to GC methods, effected on polar capillary col-
umns, for determining -fatty acids such as
eicosapentaenoic and docosahexaenoic acids in
foods. Trans isomers of fatty acids have a possible
link with cardiovascular diseases. Hence the occur-
rence of trans isomers in relatively large concentra-
tions in margarines, shortenings and similar food
products has stimulated development of methods for
resolving geometrical isomers. The solution of this
problem is very difRcult by GC alone and has
required the use of very long capillary columns and
preliminary separation steps. It may be cited as an
existing challenge to GC in the analysis of acids.
Conclusion
GC continues to be the method of choice for the
analysis of acids because of its speed, efRciency
and sensitivity. However, very complex mixtures still
pose serious challenges. Future developments may
entail use of shorter, narrower capillary columns for
greater speed and, in conjunction with routine MS
Liquid Chromatography
K. L. Ng and P. R. Haddad, University of Tasmania,
Hobart, Tasmania, Australia
Copyright ^ 2000 Academic Press
Introduction
The determination of carboxylic acids is important in
many areas of application, including environmental
samples, foods and beverages, and pharmaceutical
and biological materials. The modes of high perfor-
mance liquid chromatography (HPLC) used most fre-
quently in the separation of carboxylic acids are ion
suppression chromatography, reversed-phase ion in-
teraction chromatography, ion exclusion chromato-
graphy and ion exchange chromatography.
detection, for more deRnitive identiRcation. Automa-
tion of sample preparation, perhaps in conjunction
with microwave irradiation in lieu of conventional
heating, will shorten derivatization times, relieve the
tedium of manual manipulations and reduce total
analysis times.
See also: II/Chromatography: Gas: Derivatization; De-
tectors: Mass Spectrometry; Detectors: Selective. III/Oils,
Fats and Waxes: Supercritical Fluid Chromatography.
Triglycerides: Liquid Chromatography; Thin Layer
(Planar) Chromatography. Volatile Organic Compounds
in Water: Gas Chromatography.
Further Reading
Blau K and Halket JM (eds) (1993) Handbook of Deriva-
tives for Chromatography, 2nd edn. Chichester: John
Wiley.
Christie WW (1989) Gas Chromatography and Lipids.
Ayr, Scotland: Oily Press.
Christie WW (ed.) (1992}97) Advances in Lipid Methodo-
logy, vols 1}4. Dundee, Scotland: Oily Press.
Clement RE (ed.) (1990) Gas Chromatography } Biochemi-
cal, Biomedical, and Clinical Applications. New York:
John Wiley.
Gutnikov G (1995) Fatty acid proRles of lipid samples.
Journal of Chromatography B 671: 71.
Poole CF and Schuette SA (1985) Contemporary Practice of
Chromatography. Amsterdam: Elsevier.
Shantha NC and Napolitano GE (1992) Gas chromatogra-
phy of fatty acids. Journal of Chromatography 624: 37.
Wittkoski R and Matissek R (eds) (1992) Capillary Gas
Chromatography in Food Control and Research. Ham-
burg, Germany: B. Behr’s Verlag.
Zumwalt RW, Kuo KCT and Gehrke CW (1987) Amino
Acid Analysis by Gas Chromatography, vols 1}3. Boca
Raton, FL: CRC Press.
In ion suppression chromatography, a buffer
of appropriate pH is added to the mobile phase in
order to suppress the ionization of the carboxylic
acids so that they can be retained on nonpolar
stationary phases and eluted in order of increasing
hydrophobicity. Ion interaction (or ion pair)
chromatography has been used for the separation of
carboxylic acids under isocratic or gradient condi-
tions and involves the complete ionization of the
solute and the addition to the mobile phase of an ion
interaction reagent (IIR), consisting of lipophilic ions
of opposite charge to the solute. Ion exclusion
chromatography (i.e. the separation of partially
ionized carboxylic acids on a cation exchange station-
ary phase using amperometry, coulometry, ultra-
1854 III / ACIDS / Liquid Chromatography
underivatized form. Fatty acids containing unusual
structural features, such as cyclopropane rings or
epoxy groups, are constituents of some edible veg-
etable oils and are suspected of being health hazards.
Hence they have been analysed in foods by capillary
GC as FAMEs. Such studies have provided a basis for
identifying components in blends of vegetable oils
with potential application to detecting adulteration.
Similar studies have been carried out to determine
brominated acid constituents in vegetable oils that are
added to disperse Savouring constituents in citrus-
based beverages. Clinical and epidemiological Rnd-
ings of the beneRcial effects of Rsh oils have led
to GC methods, effected on polar capillary col-
umns, for determining
-fatty acids such as
eicosapentaenoic and docosahexaenoic acids in
foods. Trans isomers of fatty acids have a possible
link with cardiovascular diseases. Hence the occur-
rence of trans isomers in relatively large concentra-
tions in margarines, shortenings and similar food
products has stimulated development of methods for
resolving geometrical isomers. The solution of this
problem is very difRcult by GC alone and has
required the use of very long capillary columns and
preliminary separation steps. It may be cited as an
existing challenge to GC in the analysis of acids.
Conclusion
GC continues to be the method of choice for the
analysis of acids because of its speed, efRciency
and sensitivity. However, very complex mixtures still
pose serious challenges. Future developments may
entail use of shorter, narrower capillary columns for
greater speed and, in conjunction with routine MS
Liquid Chromatography
K. L. Ng and P. R. Haddad, University of Tasmania,
Hobart, Tasmania, Australia
Copyright ^ 2000 Academic Press
Introduction
The determination of carboxylic acids is important in
many areas of application, including environmental
samples, foods and beverages, and pharmaceutical
and biological materials. The modes of high perfor-
mance liquid chromatography (HPLC) used most fre-
quently in the separation of carboxylic acids are ion
suppression chromatography, reversed-phase ion in-
teraction chromatography, ion exclusion chromato-
graphy and ion exchange chromatography.
detection, for more deRnitive identiRcation. Automa-
tion of sample preparation, perhaps in conjunction
with microwave irradiation in lieu of conventional
heating, will shorten derivatization times, relieve the
tedium of manual manipulations and reduce total
analysis times.
See also: II/Chromatography: Gas: Derivatization; De-
tectors: Mass Spectrometry; Detectors: Selective. III/Oils,
Fats and Waxes: Supercritical Fluid Chromatography.
Triglycerides: Liquid Chromatography; Thin Layer
(Planar) Chromatography. Volatile Organic Compounds
in Water: Gas Chromatography.
Further Reading
Blau K and Halket JM (eds) (1993) Handbook of Deriva-
tives for Chromatography, 2nd edn. Chichester: John
Wiley.
Christie WW (1989) Gas Chromatography and Lipids.
Ayr, Scotland: Oily Press.
Christie WW (ed.) (1992}97) Advances in Lipid Methodo-
logy, vols 1}4. Dundee, Scotland: Oily Press.
Clement RE (ed.) (1990) Gas Chromatography } Biochemi-
cal, Biomedical, and Clinical Applications. New York:
John Wiley.
Gutnikov G (1995) Fatty acid proRles of lipid samples.
Journal of Chromatography B 671: 71.
Poole CF and Schuette SA (1985) Contemporary Practice of
Chromatography. Amsterdam: Elsevier.
Shantha NC and Napolitano GE (1992) Gas chromatogra-
phy of fatty acids. Journal of Chromatography 624: 37.
Wittkoski R and Matissek R (eds) (1992) Capillary Gas
Chromatography in Food Control and Research. Ham-
burg, Germany: B. Behr’s Verlag.
Zumwalt RW, Kuo KCT and Gehrke CW (1987) Amino
Acid Analysis by Gas Chromatography, vols 1}3. Boca
Raton, FL: CRC Press.
In ion suppression chromatography, a buffer
of appropriate pH is added to the mobile phase in
order to suppress the ionization of the carboxylic
acids so that they can be retained on nonpolar
stationary phases and eluted in order of increasing
hydrophobicity. Ion interaction (or ion pair)
chromatography has been used for the separation of
carboxylic acids under isocratic or gradient condi-
tions and involves the complete ionization of the
solute and the addition to the mobile phase of an ion
interaction reagent (IIR), consisting of lipophilic ions
of opposite charge to the solute. Ion exclusion
chromatography (i.e. the separation of partially
ionized carboxylic acids on a cation exchange station-
ary phase using amperometry, coulometry, ultra-
 III / ACIDS / Liquid Chromatography 1855

violet, refractive index and both suppressed and

nonsuppressed conductivity detection) is the most
commonly used mode of liquid chromatography for
the separation of carboxylic acids. Finally, anion
exchange chromatography can be used for the separ-
ation of carboxylic acids, after conversion of these
species to anions. Detection is usually achieved by
suppressed or nonsuppressed conductivity or by in-
direct photometry.

Ion Suppression Chromatography

Background
Ion suppression chromatography is a technique for the
separation of ionizable solutes which functions by sup-
pressing the ionization of these solutes, thus increasing
their retention on nonpolar stationary phases. In the
separation of carboxylic acids, an acidic buffer is added
to the mobile phase to suppress the ionization of the
solutes, which are then separated on nonpolar poly-
meric or silica-based (usually C18) stationary phases.
This method is only applicable to those acids for
which the ionization can be suppressed using buf-
fers having pH values in the range 3}8, since the
C18 stationary phases are unstable outside this pH
range. However, these restrictions do not apply to the
use of polymeric stationary phases, which can be used
for the separation of a wider variety of solutes. The
mobile phase is usually an acidic buffer of the appro-
priate pH. Commonly used buffers include phos-
phoric acid, sodium or potassium phosphate, sodium
hydrogen sulfate, acetic acid and citric acid. Organic
modiRers such as methanol or acetonitrile can also be
added to the mobile phase to improve the separation.
Manipulation of Retention of Acids
in Ion Suppression Chromatography
Solute retention results from solvophobic effects oc-
curring between the mobile phase, the stationary
phase and the solutes. For the separation of monocar-
boxylic acids, the pH of the mobile phase inSuences
the retention behaviour according to the following
equation:
Ka
k0!k 1 #
\ [H ]
k" [1]
Ka
1# #
[H ]
where k0 is the retention factor of the undissociated
acid, k 1 is the retention factor of the conjugate base,
\
and Ka is the acid dissociation constant in the mobile
phase. This retention behaviour is illustrated in
Figure 1, which shows the retention factor of a weak
acid versus (pH}pKa). The curve is sigmoidal in shape
Figure 1 Plot of the retention factor of a weak monoprotic acid
vs. (pH!pKa).
and the inSection point is located at the point where
the pH of the mobile phase is equal to the pKa of the
solute in the mobile phase. At pH values substantially
less than its pKa value, the acid is present in its neutral
form and has a large retention factor. Further de-
creases in mobile-phase pH show no effect on
retention factor, since there will be no further
change in the ionization of the acid. Conversely,
mobile-phase pH values substantially greater than the
pKa value will result in complete ionization of the
solute, leading to a small retention factor. At inter-
mediate pHs, the solute charge, and hence its reten-
tion, will be dependent on the particular pH used and
its proximity to the pKa value.
In the case of dicarboxylic acids, the shape of the
curve is largely determined by the difference be-
tween the two pKa values. When pKa1 and pKa2 are
very close, sigmoidal curves are obtained and the
behaviour of dicarboxylic acids is almost the same as
that of monocarboxylic acids. When the two pKa
values are well separated, the curve is a composite of
two sigmoidal curves.
Both the ionic strength and organic modiRer con-
tent of the mobile phase may be varied in order to
manipulate retention in ion suppression chromato-
graphy. Increasing the ionic strength of the mobile
phase causes an apparent increase in the dissocia-
tions, leading to a decrease in the retention factor.
1856 III / ACIDS / Liquid Chromatography

This effect is more pronounced in nonaqueous media.
In the approximate range of ionic strengths from 0 to
0.5 mol L\1, the higher the ionic strength of the mo-
bile phase, the greater the increase in pKa. The
addition of organic modiRers inSuences retention be-
haviour in two ways. Firstly, increasing the organic
modiRer content of the mobile phase decreases the
retention factor, as is generally the case in reversed-
phase liquid chromatography. However, the apparent
pKa of the solute increases as organic modiRer is
added to the mobile phase, leading to an increase in
the degree of ionization of the solute and therefore
reduced retention.
Applications
The utility of ion suppression on polymeric stationary
phases may be appreciated by considering the separ-
ation of the homologous series of aliphatic carboxylic
acids. Neither ion exchange nor ion exclusion
chromatography yields a complete separation of these
species. However, ion suppression coupled with
gradient elution and conductivity detection enables
the separation of butyric through to stearic acid, as
illustrated in Figure 2. The gradient used involved an
increase in the percentage of organic modiRer in the
mobile phase and a decrease in mobile-phase pH.
Carboxylic acids more hydrophilic than butyric acid
are eluted in a single peak at the column void volume.
Ion Interaction Chromatography
Background
Ion interaction chromatography involves the addition
of an ion interaction reagent (IIR) to the mobile
phase. The IIR is usually a lipophilic ion of opposite
charge to the analyte ions. In the case of the separ-
ation of carboxylic acids, cationic IIRs such as tet-
raalkylammonium salts are used.
The mechanism of ion interaction chromatography
is considered to begin with the establishment of a
dynamic equilibrium between IIR in the mobile phase
and IIR adsorbed onto the stationary phase:
IIR# #
(M) B IIR(S) [2]
where the subscripts M and S refer to the mobile and
stationary phases. This results in the formation of an
electrical double layer at the stationary phase surface.
The adsorbed IIR ions constitute a primary layer of
charge, to which is attracted a diffuse, secondary
layer of oppositely charged ions. This secondary layer
of charge consists chieSy of the counter-ions of the
IIR. The double layer is shown schematically in Fig-
ure 3A. A solute anion can compete for a position in
the secondary charged layer, from which it will tend
to move into the primary layer as a result of electro-
static attraction and, if applicable, reversed-phase
solvophobic effects. The presence of such a solute
anion in the primary layer causes a decrease in the
total charge of this layer, so to maintain charge bal-
ance a further IIR ion must enter the primary layer.
The result is that solute retention involves the adsorp-
tion of a solute anion accompanied by the adsorption
of an IIR ion, shown schematically in Figure 3B.
Typical stationary phases used in ion interaction
chromatography include neutral poly(styrene-
divinylbenzene) (PS-DVB) polymers and bonded sil-
ica materials with C18, C8, phenyl and cyanopropyl
groups as the chemically bound functionality. The

Figure 2 Gradient elution ion suppression chromatography of carboxylic acids, obtained on a polymeric reversed-phase column.
A Dionex MPIC-NS1 column was used with a gradient of 100% mobile phase A (t"0) to 100% mobile phase B (t"20 min), with
maintenance of mobile phase B after this time. Mobile phase A was 24% acetonitrile and 6% methanol in 0.03 mmol L\1 HCl; mobile
phase B was 60% acetonitrile and 24% methanol in 0.05 mmol L\1 HCl with detection by suppressed conductivity. The baseline
conductance for a blank gradient has been subtracted in the chromatogram shown. (Reprinted with permission from Slingsby RW
(1986) Gradient elution of aliphatic carboxylic acids by ion chromatography in the ion-suppression mode. Journal of Chromatography
371: 373}382.)

Figure 3 Schematic illustration of the ion interaction model for
the retention of anionic solutes in the presence of a lipophilic
cationic IIR. The solute and the IIR are labelled on the diagram.
The long hatched boxes represent the lipophilic stationary phase,
the black circles with negative charges represent the counter-
anion of the IIR, whilst the white circles with positive charges
represent the counter-cation of the solute. (Reprinted with per-
mission from Haddad and Jackson, 1990.)
choice between stationary phases is usually based on
considerations such as chromatographic efRcien-
cy, pH stability and particle size. However, the elu-
tion position of certain ions can differ between
different stationary phases. Further factors to be
considered in the selection of a stationary phase for
ion interaction chromatography are speciRc interac-
tions existing between the stationary phase and either
the IIR or the solutes, and the role of residual silanol
groups on silica-based stationary phases.
The most important component of the mobile
phase in ion interaction chromatography is the IIR
itself. The requirements of the IIR are that its charge
is unaffected by mobile-phase pH, it has suitable
lipophilicity to permit adsorption onto nonpolar sta-
tionary phases, and it is compatible with other mobile-
phase components and the desired detection system.
In the separation of carboxylic acids by dynamic
coating ion interaction chromatography, moderately
hydrophobic strong base cations, such as tetra-
butylammonium ions, are used as the IIR. The IIR is
present at a constant, speciRed concentration in the
mobile phase in order to maintain a desired concen-
tration of IIR on the stationary phase. The lipophilic-
ity of the IIR governs the degree to which it is adsor-
bed onto the stationary phase, which in turn governs
the effective ion exchange capacity of the column
and hence the retention times of solute ions.
An alternative to the above method is permanent
coating ion interaction chromatography, where
a very lipophilic IIR is used initially to equilibrate the
stationary phase and is then removed from the mobile
phase in the separation step. The coating persists for
long periods of subsequent use. Permanent coating of
the column is achieved by passing a solution of the
III / ACIDS / Liquid Chromatography 1857
IIR (approximately 10\3 mol L\1) in dilute (5%)
methanol or acetonitrile through the column for
about 20 min. The purpose of the organic solvent is
to wet the surface of the lipophilic stationary phase in
order to improve binding of the IIR.
The counter-ion of the IIR is important in dynamic
coating ion interaction chromatography of anionic
solutes. This counter-ion usually acts as an ion ex-
change competing anion and is responsible for the
elution (and in many cases also the detection) of the
solute anions. The nature of the counter-ion deter-
mines the type of separation which is required and the
mode of detection applicable.
Manipulation of Retention of Acids
in Ion Interaction Chromatography
The parameters which affect the adsorption of
the IIR onto the stationary phase and hence the reten-
tion of solutes include the nature of the stationary
phase, the lipophilicity of the IIR, the concentration
of the IIR in the mobile phase, the ionic strength of
the mobile phase, the nature and concentration of any
competing ion added to the mobile phase, and the
mobile-phase pH.
The Rrst four of these factors will determine the
surface concentration of the IIR on the stationary
phase, and hence the surface charge density and
the effective ion exchange capacity. The higher
the surface concentration of IIR, the greater the solute
retention. Thus, retention times will increase as the
lipophilicity of the IIR is increased and as the percent-
age of modiRer in the mobile phase is decreased.
Solute retention generally increases with the concen-
tration of IIR in the mobile phase, but there is a
threshold concentration above which solute retention
decreases with further increase in the concentration
of IIR. The stationary phase becomes saturated with
IIR and any further addition to the mobile phase
results in decreased retention because of the increased
concentration of the IIR counter-ion.
The nature and concentration of any competing ion
added to the mobile phase will determine the reten-
tion times and elution order for solute ions. Increases
in the concentration of the mobile phase competing
ion will result in decreased solute retention, in the
same manner as observed for ion exchange separ-
ations. Finally, the mobile phase pH may inSuence
the charges on the competing ion and the solutes. An
example of this effect is the inSuence of pH in an
ion interaction chromatographic system using tet-
rabutylammonium as the IIR and phthalate as the
competing anion. Increases in mobile-phase pH over
the range 4.0}6.0 cause a decrease in the solute reten-
tion as a result of increased ionization of phthalate,
1858 III / ACIDS / Liquid Chromatography

leading to the formation of a strong, divalent compet-

ing anion.
Applications
Carboxylic acids are usually separated by ion interac-
tion chromatography using a reversed-phase column
with quaternary ammonium salts as the IIR and
water}methanol or water}acetonitrile as the mobile
phase. The more lipophilic the quaternary am-
monium ion, the more the acid is retained on non-
polar stationary phases. Such separation systems have
been used for the determination of ascorbic acid in
fruits and vegetables, as well as carboxylic acids in
beverages such as wine, beer and fruit juices.
Gradient elution ion interaction chromatography
is also possible. The concentration of the organic
modiRer or the pH of the mobile phase may be varied
to optimize the separation. Figure 4 shows an
example of the separation of carboxylic acids on
a reversed-phase column by gradient elution using
tetrabutylammonium hydroxide as the ion interac-
tion reagent.

Ion Exclusion Chromatography

Background
Ion exclusion chromatography was Rrst introduced
by Wheaton and Bauman in 1953. In this mode of
chromatography, the negatively charged, partially
dissociated carboxylic acids are separated on cation
exchange resins comprising silica or a polymer with
chemically bound anionic sulfonate or carboxylate
functional groups. This is the opposite situation
to that occurring in normal ion exchange chromatog-
raphy.
The chromatographic system consists of three
phases: the mobile phase, the resin phase and an
occluded liquid phase. The mobile phase passes
through the interstitial volume existing between the
beads of the ion-exchange resin. An occluded liquid
phase is formed by mobile phase that becomes trap-
ped within the pores of the resin phase. This trapped
liquid acts as the stationary phase of the system. The
resin phase is the solid resin network and function-
lized groups, which can be considered to be
a semipermeable ion-exchange membrane separating
the Sowing mobile phase from the stationary oc-
cluded liquid inside the resin. The three phases are
illustrated schematically in Figure 5.
Fully ionized species (A\) are completely excluded
from the interior of the resin due to electrostatic
repulsion by the Rxed anionic functional groups, in
accordance with the Donnan exclusion effect. There-
fore, these species are not retained and are eluted at
Figure 4 Ion interaction chromatography of carboxylic acids on
a LiChrosorb RP-8 column with a mobile phase of aqueous
tetrabutylammonium hydroxide (1 g L\1) and methanol using
gradient elution. Detection was at 254 nm. Carboxylic acids are:
1, ascorbic; 2, oxalic; 3, pyruvic; 4, fumaric; 5; maleic. Chromato-
gram courtesy of Alltech Chromatography Catalog (1997) 610.
the column void volume. Partially ionized species like
weak carboxylic acids (pKa"2.5}6.5) permeate se-
lectively into the stationary phase (the occluded liquid
trapped within the pores of the resin), resulting in
some retention of these species, which are then eluted
some time later than the fully ionized solutes.
Ion exclusion chromatography was Rrst performed
on large particle size, high capacity, fully functional-
ized PS-DVB polymers. However, separations have
also been performed on ploymethacrylate copolymer
resins, as well as on silica. Separations of carboxylic
acids by modern ion exclusion chromatography are
usually carried out on a cation exchange column
containing sulfonated functional groups (}SO\ 3 ) or
mixed sulfonate and carboxylate functional groups,
with the resin most commonly being used in the
hydrogen form.
Ion exclusion columns are usually quite large be-
cause most sample species are eluted with retention
volumes intermediate between the interstitial volume
(V0) and V0#Vi, where Vi is the occluded (or intra-
particle) liquid volume. Large columns contain more

Figure 5 Schematic representation of the ion exclusion mechanism, showing the retention of a weak acid (HA) in the occluded liquid
phase and the exclusion of the acid anion (A\).
resin, thus increasing the amount of occluded liquid
and hence also the capacity of the stationary phase.
A typical column would be 30 cm in length, with an
internal diameter of 7 mm or more.
The mobile phases used in ion exclusion
chromatography are often very simple in composi-
tion. Most of the early work was performed using
water as the mobile phase. However, water has lim-
itations as stronger acids or bases show too great
a degree of ionization to be retained and for weaker
acids such as carboxylic acids, the separation is slow
and the peaks are unsymmetrical.
In modern ion exclusion chromatography it is com-
mon to use dilute solutions of strong mineral acids for
the elution of carboxylic acids. The dilute mineral
acid solution suppresses the ionization of the acids so
that they can partition into the occluded liquid phase,
resulting in longer retention times and better separ-
ation between the stronger carboxylic acids. The
choice of acid used in the mobile phase is usually
determined by the form of detection being used. Sul-
fonic acids are used for conductivity detection with-
out suppression since mineral acids have a high back-
ground conductance. Sulfuric acid is often used with
ultraviolet detection and hydrochloric acid is used
with conductivity detection after the mobile phase
has been passed through a suitable suppressor. Weak
acids such as benzoic acid, phosphoric acid, salicylic
acid and carbonic acids have also been used as mobile
phases in ion exclusion chromatography when con-
ductivity detection is utilized.
Manipulation of Retention of Acids
in Ion Exclusion Chromatography
The dominating factor which determines retention is
the degree to which the acid is ionized. Separation is
based on the electrostatic repulsion between the
III / ACIDS / Liquid Chromatography 1859
solute ions and Rxed functional groups of the resin.
Therefore, ionic species are excluded from the sta-
tionary phase while partially ionized or uncharged
species partition between the mobile phase and the
occluded liquid within the resin pores. Assuming this
is the only mechanism, the solute retention time, tR, is
given by:
tR"t0#DAti [3]
where t0 is the time taken for the interstitial volume of
mobile phase (i.e. the volume of mobile phase Sowing
between the resin beads) to be eluted, ti is the time
taken for the volume of occluded liquid inside the
pores of the resin to be eluted, and DA is the distribu-
tion constant for the solute between the interstitial
mobile phase and the occluded liquid. The value of
DA is dependent on the degree of ionization of the
solute.
If a solute cannot enter the stationary phase be-
cause it is fully ionized (ion exclusion), DA"0.
Therefore, the retention time of fully ionized solutes
is equal to t0, whilst for an uncharged solute which is
free to enter the stationary phase, DA"1, and its
retention time is equal to ti. Thus, in the separation of
carboxylic acids, the retention times of the acids de-
pend on their Rrst dissociation constants (pKa). Since
the fraction of the ionized solute molecules increases
with increasing pH, an increase in the mobile phase
pH will reduce the retention time.
The retention times of monocarboxylic acids larger
than acetic acid, and dicarboxylic acids larger than
succinic acid, show an increase with increasing car-
bon number, even for solutes with similar pKa values.
This increased retention can be attributed to hydro-
phobic adsorption of the solutes on to the neutral,
unfunctionalized regions of the polymeric resin, in
a manner similar to that observed in reversed-phase
1860 III / ACIDS / Liquid Chromatography
HPLC. Hydrophobic adsorption effects can be
expected to increase in magnitude as the alkyl chain
length of the solute is increased, leading to larger
retention times. In the case of aromatic acids, the
interaction of -electrons of the benzene ring of the
acid with those of the ion exchanger (such as styrene-
divinylbenzene packing materials) leads to much
higher retention times than expected from their pKa
values. The existence of hydrophobic adsorption ef-
fects creates the possibility for manipulation of solute
retention by adding typical reversed-phase organic
modiRers, such as methanol or acetonitrile, to the
mobile phase.
Ion exclusion chromatography is usually carried
out on a high-capacity sulfonated PS-DVB resin in the
H# form. However, recently work has also been
carried out using a polymethacrylate resin in the
H# form with carboxylate functional groups, bare
silica (where the silanol group on the surface of
the silica acts as the anionic functional group) and
also on silica-based cation exchangers functionalized
with alkylsulfonic acid or phenylsulfonic acid groups.
Since silica gel is chemically stable and inert to or-
ganic solvents, silica-based cation exchangers of-
fer the advantage that high concentrations of organic
modiRers can be used. Also, aromatic acids which
adsorb strongly on to PS-DVB resin due to -electron
interactions between the aromatic ring and the solid
resin network are eluted earlier when using a silica gel
column.
Other factors which play a part in the retention
process of carboxylic acids in ion exclusion
chromatography include the addition of other mobile
phase modiRers such as polyols, sugars and inclusion
compounds (e.g. -cyclodextrin), as well as resin
characteristics such as the pore size, the degree of
cross-linking, the ion exchange capacity and the ionic
form of the resin.
Applications
The separation of carboxylic acids is the most com-
mon application of ion exclusion chromatography.
When coupled with spectrophotometric detection at
low wavelength (e.g. 210 nm), ion exclusion
chromatography yields excellent separations and
relatively clean chromatograms for a wide variety of
complex sample matrices, such as urine, plasma,
foods and beverages and pharmaceuticals. Figure 6
shows a chromatogram for a urine sample, with-
out sample pretreatment. Ion exclusion chromatogra-
phy has also found increasing use for the determina-
tion of anions of weak inorganic acids. It is especially
attractive as an adjunct to ion exchange chromatog-
raphy since the selectivities obtained by these two
techniques are quite different. Solutes such as
Suoride, carbonate, cyanide, borate and sulRte have
been determined using this approach. Interference
from strongly ionized species is minimal because
these solutes are unretained and appear at the
column void volume. Ion exclusion chromatogra-
phy can therefore readily separate weakly ionized
solutes in samples containing high concentrations
of ionic species, e.g. sea water and oil reservoir
brines.
Ion Exchange Chromatography
Background
Ion exchange chromatography of carboxylic acids
can be performed using an anion exchange stationary
phase. The capacity of this anion exchanger is impor-
tant since the capacity needs to be sufRciently
high to separate carboxylic acids of similar charge,
but low enough for the ionic strength of the mobile
phase to permit the use of conductivity detection. The
development of new high-efRciency, low- and
medium-capacity columns combined with a new gen-
eration of micromembrane suppressors capable of
handling concentrated mobile phases has made the
determination of carboxylic acids by anion exchange
chromatography a practical proposition.
One disadvantage of anion exchange chromatogra-
phy is that groups of mono-, di- and tricarboxylic
acids must be analysed separately. However, the use
of gradient elution (water, sodium hydroxide and
methanol) has made it possible to separate these com-
pounds in a single run, as well as simultaneously
separating inorganic anions (Figure 7).
Ion exchange chromatography of carboxylic acids
has been performed on anion exchange resins in the
hydroxyl, carbonate, sulfate, chloride, nitrate, for-
mate, acetate or borate form. The mobile phase usu-
ally consists of an alkaline solution such as sodium
hydroxide or sodium carbonate and sodium hydrogen
carbonate, with detection achieved using a suppressor
column and conductivity. Solutes are usually low-
molecular-weight, saturated or unsaturated acids and
hydroxy acids. Factors which affect retention
include molecular dimensions, pKa and speciRc ad-
sorption of organic acid molecules on the organic
matrix of the ion exchanger.
Manipulation of Retention of Acids
in Ion Exchange Chromatography
Apart from the usual electrostatic effects which
govern retention in ion exchange chromatography,
one of the main factors affecting retention of
carboxylic acids is the molecular adsorption of the
acid on the anion exchange resin. The presence of
 III / ACIDS / Liquid Chromatography 1861

Figure 6 Analysis of human urine using ion exclusion chromatography. An Interaction ORH-801 column was used with a mobile
phase comprising 10 mmol L\1 H2SO4 containing 10% methanol. Detection was at 254 nm. Solute identities: 1, oxalic acid; 2,
oxaloacetic acid; 3, -ketoisovaleric acid; 4, ascorbic acid and -keto--methyl-n-valeric acid; 5, -phenylpyruvic acid; 6, uric acid; 7,
-ketobutyric acid; 8, homoprotocatechuic acid; 9, unknown; 10, unknown; 11, hydroxyphenylacetic acid; 12, p-hydroxyphenyllactic
acid; 13, homovanillic acid. (Reprinted with permission from Woo DJ and Benson JR (1984) American Clinical Products Review Jan:
20.)

double bonds in carboxylic acids leads to higher
retention factors, probably due to stronger hydropho-
bic interactions of the double bond with the poly-
meric matrix of the resin and also stronger electro-
static interactions between ionic groups. The pres-
ence of hydroxy groups in carboxylic acids increases
the polarity of the acid and results in stronger interac-
tions both with the aqueous mobile phase (leading to
lower retention factors for the acids) and any alkanol
substituent of the quaternary ammonium functional
group of the anion exchange resin (leading to higher
retention factors). Since the adsorption of carboxylic
acids plays such an important role in the retention of
these acids in ion exchange chromatography, the pKa
values of the acids are also important, as are any
parameters which inSuence the dissociation of the
acid, such as the pH of the mobile phase and the
concentration of any organic solvent. Additionally,
the pH of the mobile phase may also affect its
elution strength and hence affect retention as well.
Applications
Compared to other separation methods such as ion
exclusion chromatography, anion exchange provides
improved selectivity within the three groups of acids:
mono-, di- and tricarboxylic acids. This is particu-
larly true among the stronger acids such as most of
1862 III / ACIDS / Liquid Chromatography

Figure 7 Example of a gradient separation of inorganic and organic acid anions by anion exchange chromatography. A Dionex
IonPac AS11 column was used with a mobile phase comprising water and NaOH as a gradient. Detection was by conductivity in the
suppressed mode. Solute identities: 1, isopropylethylphosphonic acid; 2, quinate; 3, fluoride; 4, acetate; 5, propionate; 6, formate; 7,
methylsulfonic acid; 8, pyruvate; 9, chlorite; 10, valerate; 11, monochloroacetate; 12, bromate; 13, chloride; 14, nitrite; 15, tri-
fluoroacetate; 16, bromide; 17, nitrate; 18, chlorate; 19, selenite; 20, carbonate; 21, malonate; 22, maleate; 23, sulfate; 24, oxalate; 25,
ketomalonate; 26, tungstate; 27, phthalate; 28; phosphate; 29, chromate; 30, citrate; 31, tricarballylate; 32, isocitrate; 33, cis-aconitate;
34, trans-aconitate. Chromatogram courtesy of Dionex Corporation Product Selection Guide (1997}98) 48.

the di- and tricarboxylic acids. Of the di- and tricar-
boxylic acids which are in the Krebs cycle or are
commonly found in foods, there are only two groups
of co-eluting acids: malic and malonic, and isocitric
and cis-aconitic. Another advantage of anion ex-
change separation is the possibility of simultaneous
determination of some inorganic ions, such as
Suoride, chloride, and sulfate, with the carboxylic
acids.
Applications of anion exchange chromatography
of carboxylic acids include the quantiRcation of short
chain organic acids and inorganic anions for the
biotechnology, chemical or power industries, the sep-
aration of the Krebs cycle acids in foods and bever-
ages, and also the separation of aromatic carboxylic
acids in chemical process solutions and as impurities
in precursors in the polymer industry.
Conclusion
Four modes of HPLC used in the separation of car-
boxylic acids have been discussed. Ion suppression
chromatography, using a buffer to suppress the ioniz-
ation of the acids, is the simplest separation system
for carboxylic acids. Ion interaction chromatography
offers the greatest variety of parameters to alter
the selectivity of the separation system by changing
the properties of the ion interaction reagent. Ion ex-
clusion chromatography is the most commonly used
method in the separation of carboxylic acids due to its
compatibility with a wide range of sample matrices.
Ion exchange chromatography provides improved
selectivity within groups of acids but the technique
requires the use of gradient elution.
See also: I/Ion Exchange. II/Chromatography: Liquid:
Mechanisms: Ion Chromatography. Ion Exchange:
Theory.
Further Reading
Bruzzoniti MC, Mentasti E, Sarzanini C and Hajos P
(1997) Ion chromatographic separation of carboxylic
acids, Prediction of retention data. Journal of
Chromatography 770: 13}22.
Coenen AJJM, Kerkhoff MJG, Heringa RM and van
der Wal Sj (1992) Comparison of several methods for the
determination of trace amounts of polar aliphatic mono-
carboxylic acids by high-performance liquid chromato-
graphy. Journal of Chromatography 593: 243}252.
Ding MY, Koizumi H and Suzuki Y (1995) Comparison of
three chromatographic systems for determination of or-
ganic acids in wine. Analytical Science 11: 239}243.
Haddad PR and Jackson PE (1990) Ion Chromatography
} Principles and Applications. Amsterdam: Elsevier.
Robards K, Haddad PR and Jackson PE (1994) Principles
and Practice of Modern Chromatographic Methods.
London: Academic Press.
Rocklin RD, Slingsby RW and Pohl CA (1986) Separation
and detection of carboxylic acids by ion chromatogra-
phy. Journal of Liquid Chromatography 9: 757}775.
Schmuckler G (1987) High-performance liquid ion-ex-
change chromatography. Journal of Liquid Chromato-
graphy 10: 1887}1901.
Schwarzenbach R (1982) High-performance liquid
chromatography of carboxylic acids. Journal of
Chromatography 251: 339}358.
 III / ACIDS / Thin^Layer (Planar) Chromatography 1863

Thin-Layer (Planar) Chromatography

J. H. P. Tyman, Brunel University, Uxbridge,
Middlesex, UK
Copyright ^ 2000 Academic Press
Introduction
The thin-layer chromatography (TLC) of aliphatic
and aromatic acids having a wide range of structures
has proved to be of great practical value in the chem-
istry and biochemistry of this large group of organic
compounds. This review of the TLC properties of
acids is Rrstly conveniently divided into a discussion
of qualitative aspects of the relative hRF values of
various classes of aliphatic and aromatic carboxylic
acids having a wide range of structures on different
layers and in different solvent systems. Some mention
is made of compounds with other acidic functions.
Secondly, there is a selective account of applications
on the quantitative determination of acids in typical
current synthetic and some natural sources.
Acyclic and Cycloaliphatic
Compounds
Alkanoic, Alkanedioic, Hydroxy, Keto, Unsaturated,
Arylalkanoic Acids and Other Related Acids of
Biological Signi\cance
n-Alkanoic acid The separation of these acids by the
technique of TLC with respect to the lower homolog-
ous fatty acids has a historic precedent in that their
separation in the vapour phase on a column coated
with a stationary phase was the Rrst published
example of gas chromatography.
Although it might be generally considered that gas
chromatography is more suitable than TLC for the
separation of alkanoic acids, Table 1 shows some
simple conditions that have been used in this series
typical of a partition separation. Many of the values
quoted in the ensuing tables have been adapted from
extensive published information by Hanai (see Fur-
ther Reading). For comparison, the hRF values of the
dibasic acids malonic, succinic, glutaric and adipic in
solvent a are 9, 14, 18 and 22 respectively, and that
of glycolic acid, 38. The Rrst four acids in the table
have also been examined on crystalline cellulose
impregnated with sodium bicarbonate in ethanol}
water (100 : 20) and detection by dicyclohexyl
carbodiimide to separate formic acid, acetic,
propionic and butanoic acids having the hRF values
31, 37, 45 and 52 respectively.
n-Alkanedioic acids The saturated dibasic acids
have been more widely studied on a variety of layers
and solvents, as illustrated in Table 2 which again, as
with the monobasic series, shows partition separ-
ations. In cases where a considerable number of sol-
vents have been listed, the optimum conditions for
the series of compounds have been given. For com-
parison, the hRF value of glycolic acid under the
conditions of g was 38. In another separation on silica
gel (sil G25, Macherey Nagel) with the solvent n-
pentyl formate}chloroform}formic acid (70 : 15 : 15)
and detection by bromocresol green, nonlinearity was
found in that malonic, succinic, glutaric and adipic
acids had hRF values of 40, 43, 54 and 48 respective-
ly. Folic acid, which may be regarded as a 2-
acylamino derivative of glutaric acid, had an hRF
value of 0 compared with 78 for nicotinic acid on
silica gel G in water as developing solvent.
Hydroxy acids It is convenient to classify this group
of saturated acids as monohydroxy, monohydroxy-

Table 1 hRF values of homologous alkanoic acids on starch and on cellulose layers

Alkanoic acid Conditions Detection

a b a b

Formic 52 8 Fluorescein UV 254 nm Methyl red

Acetic 56 19 Fluorescein UV 254 nm Methyl red
Propionic 66 28 Fluorescein UV 254 nm Methyl red
Butanoic 71 37 Fluorescein UV 254 nm Methyl red
Pentanoic (valeric) 78 48 Fluorescein UV 254 nm Methyl red
Hexanoic (caproic) 85 59 Fluorescein UV 254 nm Methyl red

a, Ethanol}water}concentrated ammonia (78 : 20 : 13), rice starch; b, light petroleum

(40}603C)}acetone (2 : 1) 95% saturated with ethane}1.2-diol, cellulose and Dowex威
50 W. (With acknowledgement to Hanai, 1982.)
1864 III / ACIDS / Thin^Layer (Planar) Chromatography

Table 2 hRF values of n-alkane-
,-dioic acids (dibasic acids) and the polarity of the developing solvent has to be
on various layers increased by the use of ethanol. The meso and DL
Dibasic acid Conditions forms of tartaric acid show a small difference of
hRF which can be enhanced by the use of silica gel
a b c d e f g impregnated with boric acid. It is also possible to
Oxalic (C2) 16 0 6 separate the enantiomers of racemic hydroxy acids by
Malonic (C3) 21 52 20 7.5 9 the incorporation of a chiral additive in the adsorbent
Succinic (C4) 37 63 38 27 25 59 14 layer. The role of impregnated layers has been re-
Gultaric (C5) 46 71 47 32 31 74 18 viewed by Hauck et al. (see Further Reading).
Adipic (C6) 55 82 55 37 38 84 22
Pimelic (C7) 50 94 Keto acids The hRF values of a number of mono
Suberic (C8) 58 100
Azelaic (C9) 67 keto derivatives of monobasic and dibasic acids are
Sebacic (C10) 72 given in Table 4. The compounds shown from top to
Undecyl (C11) 82 bottom in the table are glyoxylic, pyruvic, 2-
oxobutanoic, 2-oxovaleric, 2-oxoisocaproic, oxalo-
a, Ethanol}concentrated ammonia}water (150 : 8 : 40), cellulose
(Merck 5552); b, 2-ethyl-1-butanol}formic acid}water (40 : 12 : acetic and 2-oxoglutaric acid. The need of formic acid
48); c, diethyl ether}light petroleum}CCl4}water}formic acid in high concentration to effect a separation is
(50 : 20 : 20 : 8 : 1); polyamide 6; d, ethanol}concentrated ammo- illustrated in d compared with f. For comparison, the
nia}water (100 : 16 : 12), cellulose MN300; e, di-n-butyl ether} hRF values under conditions d of citric and malic
formic acid}water (65 : 25 : 2.2), cellulose (Merck 5716); f, acids were 44 and 56 respectively. Intramolecular
toluene}propionic acid}water (47 : 47 : 4.9), silica gel (Merck
5721); g, ethanol}concentrated ammonia}water (78 : 13 : 20), rice hydrogen bonding may account for the higher hRF
starch. (With acknowledgement to Hanai, 1982.) The use of values of the monobasic compounds. The cis and
formic acid diminishes streaking sometimes found in the TLC of trans 2,4-dinitrophenylhydrazones of a range of keto
acids in neutral solvents. It is thought that in acidic solvents the acids have been examined.
formation of a dimeric intermolecularly hydrogen-bonded species
is then favoured in the equilibrium with the monomeric form, while Unsaturated monobasic dibasic and polybasic acids
in basic solvents the monomeric anion is largely present. Acidic
adsorbents may likewise simulate acidic solvents.
The unsaturated acids are a large group which have
technical and medicinal uses. The majority are either
di- or tribasic. Table 5 summarizes the hRF values of
a selection of compounds. Extensive details of separ-
ations have been described by Hanai and also in early
work a limited range of monobasic keto-, hydroxy
acids and of dibasic acids was studied. The separation
of cis and trans isomers, for example maleic and
(Modified with permission from Hanai, 1982.)
fumaric acids, appears to be generally straightfor-
ward and free of the requirement for argentation
dibasic, monohydroxytribasic, dihydroxydibasic and TLC, as in the case of unsaturated fatty acids. The
polyhydroxy types. Table 3 lists the hRF values of stereochemistry of the glutaconic acid described in
a number of acids with this functionality. For com- Table 5 was not stated. The formulae of (1) trans-
parison, the hRF value of malonic acid under condi- aconitic acid, (2) itaconic acid, (3) trans-glutaconic
tion f was 40 and in the aromatic series that of acid, (4) mesaconic acid (trans) and (5) citraconic
mandelic acid (
-hydroxyphenylacetic acid) was 57. acid (cis) are depicted.
In general, cellulose has been used as adsorbent in
examples a to e and silica gel in f. In early work, silica
gel G-kieselguhr (1 : 1), kieselguhr impregnated with
polyethylene glycol and polyamide layers were also
employed. It is possible that in acidic developing
solvents certain of these acids are present as intra-
molecularly hydrogen-bonded structures and that
Rve-membered are likely to be more stable than
six-membered rings. Thus glycolic and lactic acids
would be expected to have high hRF values whereas
acids having hydrogen-bonded rings and additional
acidic groups would have lower values. Under basic
conditions with ammonia the solutes are more polar
 III / ACIDS / Thin^Layer (Planar) Chromatography 1865

Table 3 hRF values of hydroxyacids on various layers

Acid Conditions

a b c d e f

Glycolic, HOCH2CO2H 67 46 50 31
Lactic, HOCH(CH3)CO2H (DL) 76 72 73 89 36
Malic, HO2CCH2CH(OH)CO2H (DL) 29 30 32 35 50 26
Citramalic, HO2CCH2C(Me)(OH)CO2H 65
Citric, HO2CCH2C(CO2H)(OH)CH2CO2H 16 11 18 23 42 22
IsoCitric, HO2CCH(OH)CH(CO2H)CH2CO2H 40
Glyceric, HOCH2CH(OH)CO2H 60 32 24 36
Tartaric, HO2CCH(OH)CH(OH)CO2H (DL) 24 19 18 31 19
Quinic, 1R,3R,4R,5R-Tetrahydroxycyclohexane carboxylic 18
Ascorbic 15

a, Diisopropyl ether}formic acid, (3 : 1), cellulose MN 300HR, detection by UV;

b, ethanol}concentrated ammonia}water, (150 : 8 : 40), cellulose, (Merck 5552), detection
by bromocresol green or starch-iodine reagent; c, 2-ethyl-1-butanol}formic acid}water,
(40 : 12 : 48), cellulose, (Merck 5552), detection as in b; d, diisopropyl ether}formic
acid}water, (65 : 25 : 10), cellulose (Merck 5716), detection by aniline-xylose, furfural;
e, propanol}methyl benzoate}90% formic acid}water, (7 : 3 : 2:1), cellulose, detection by
PaH skovaH and Munk reagent; f, n-pentyl formate}chloroform}formic acid, (70 : 15 : 15),
silG25, detection by bromocresol green. (With acknowledgement to Hanai, 1982.)

Arylalkanoic acids Prior to an account of the TLC

properties of aromatic acids it is of interest to note
those of the semi-aromatic group typiRed by
phenylacetic acid and its homologues and analogues.
The hRF values of a range of these compounds are Table 4 hRF values of keto acids on various layers
shown in Table 6. The need to use the least polar
Keto acid Conditions
combination of solvents is illustrated by the condi-
tions with c and d where the latter is ineffective a b c d e
while the former affords a separation of homo-
logous compounds. In the case of the unsaturated OHCCO2H 50 37 55 43
compound, the separation in conditions d would CH3COCO2H 68 25 60
CH3CH2COCO2H 78
almost certainly be improved with argentated silica CH3CH2CH2COCO2H 45
gel. (CH3)2CHCH2COCO2H 86
HO2CCOCH2CO2H 18 86 53
Acidic Compounds of Biosynthetic and HO2CCH2CH2COCO2H 36 74 50 45
Biological Importance
a, Ethyl formate}light petroleum (60}803C)}acetic acid (50 : 50 :
A number of polyfunctional cyclohexanyl derivatives 7), silica gel; b, ethanol}concentrated ammonia}water (78 : 13 :
classiRable in several of the above groups are (6) 20), rice starch, detection by fluorescein and UV; c, water-
saturated diethyl ether}88% formic acid (7 : 1), silica gel G, aniline
shikimic acid, (7) mevalonic acid and (8) abscisic
ribose reagent; d, chloroform}methanol}formic acid (80 : 20 : 1),
acid, all of which have biological signiRcance. Their silica gel G, aniline ribose; e, n-pentyl formate}chloroform}formic
TLC properties in a number of solvents have been acid (70 : 15 : 15), sil G25, bromocresol green. (With acknow-
described. ledgement to Hanai, 1982.)
1866 III / ACIDS / Thin^Layer (Planar) Chromatography

Table 5 hRF values of unsaturated di- and tribasic acids on Table 7 The hRF values of cyclohexane- and dienecarboxylic
various layers acids (dihydro- and tetrahydro-derivatives of benzoic acid)

Acid Conditions Compound Conditions

a b c d e f g h i a b

Maleic 3 18 30 27 22 Cyclohexanecarboxylic acid 83 92

Fumaric 32 87 31 83 47 37 82 49 72 Cyclohexa-1-enecarboxylic acid 91 95
Itaconic 49 45 79 53 Cyclohexa-3-enecarboxylic acid 91 95
Mesaconic (trans) 82 88 62 Cyclohexa-1,4-dienecarboxylic acid 77 83
Citraconic (cis) 36 39 Cyclohexa-2,5-dienecarboxylic acid 77 82
Glutaconic 44 56 Benzoic acid (cyclohexa-1,3,5-trienecarboxylic) 77 88
Hex-3-ene dicarboxylic 54 2-Hydroxycyclohexanecarboxylic acid 54 21
cis-Aconitic 1 4 65
trans-Aconitic 9 35 9 78 57 a, Benzene}dioxane}acetic acid (90 : 25: 4), kieselgel G, detec-
tion by autoradiography; b, light petroleum}diethyl ether}acetic
a, Toulene}propionic acid}water, (47 : 47 : 4.9), cellulose (Merck acid (50 : 50 : 1), as before in a. (With acknowledgement to
5716), detection by aniline-xylose, furfural; b, diisopropyl Hanai, 1982.)
ether}formic acid (3 : 1), cellulose MN300HR, detection by di-
chlorofluorescein; c, diethyl ether}formic acid}water, (10 : 2 : 1),
cellulose (DC Fertigplatten) detection by fluorescence; d, 95% 32. The hRf value of the keto hydroxyacid, mevalonic
ethanol}25% ammonia}water (8 : 2 : 1), same layer and detection acid, in diethyl ether}formic acid (7 : 1) on silica gel
as c; e, diisopropyl ether}light petroleum}carbon tetrachloride} (Eastman) was 29 and that of abscisic acid in n-
water}formic acid (50 : 20 : 20 : 8 : 1), polyamide 6, detection by K propanol}25% ammonia}water (80 : 10 : 10) on
ferricyanide, ferric ammonium sulfate; f, n-pentyl formate}
chloroform}formic acid (20 : 70 : 10), sil G25, detection by bromo- kieselgel (HF254) was 57. The rooting hormone,
cresol green; g, propanol}methyl benzoate}90% formic acid}water indole-3-acetic acid, under the same conditions
(7 : 3 : 2 : 1), layer not stated but probably cellulose, detection by was 45.
PaH skovaH and Munk reagent; h, butyl formate}ethyl acetate}formic The hRF values of other cyclic compounds which
acid (82 : 9 : 9), polyamide, bromocresol green; i, diisopropyl are metabolites of benzoic acid and also structurally
ether}formic acid}water, (90 : 7 : 3) silica gel, bromocresol
green. (With acknowledgement to Hanai, 1982 and to Copius- related to shikimic acid are given in Table 7 alongside
Peereboom, 1969.) the reference compound benzoic acid. Isomeric
compounds were not separable, although by the use
of argentation TLC this may be possible.

Aromatic Acidic Compounds

Substituted Benzoic Acids
In this category the compounds under consideration
Shikimic acid (6), a hydroxy unsaturated cyclic are those in which the carboxyl group is directly
compound, in the solvent g (Table 5) had an hRF of attached to the aryl ring.
The isomeric hydroxybenzoic acids have been
Table 6 hRF values of derivatives and homologues of phenyl- listed in the section on phenols. In Table 8 the TLC
acetic acid properties of the aminobenzoic acids are given
alongside the reference compounds benzoic acid,
Acid Conditions
2-hydroxybenzoic acid, 2,4-dihydroxybenzoic acid,
a b c d 3,4,5-trihydroxybenzoic acid (gallic acid) and
phthalic acid.
Phenylacetic 68 74 54 95 The hRF values of a wide range of other phenolic
4-Phenylbutanoic 75 95
acids have been recorded, as have those of more
4-Phenylbut-3-enoic 71 95
Phenoxyacetic 64 63 complex compounds, the polycyclic series of lichen
trans-Cinnamic 67 95 acids. In the case of cis and trans isomers of aro-
matic acids having an unsaturated side chain, separ-
a, n-pentyl formate}chloroform}formic acid, (70 : 15 : 15), sil G25, ations do not seem to be difRcult. Thus, on silica
bromocresol green; b, n-pentyl formate}chloroform}formic acid
gel 60 (F254) with diethyl ether}hexane}chloro-
(20 : 70 : 10), sil G25; c, light petroleum}acetic acid, (49 : 1), silica
(Eastman), bromocresol green; d, light petroleum}diethyl ether} form}acetic acid (12 : 38 : 50 : 0.5) the hRF values
formic acid, (45 : 5 : 1), silica (Eastman). (With acknowledgement of cis- and trans-3,4-dihydroxycinnamic acid were
to Hanai, 1982.) 18 and 27 and of the corresponding isomers of
 III / ACIDS / Thin^Layer (Planar) Chromatography
Table 8 hRF values of aminobenzoic acids and some hydroxy-
benzoic acids
Acidic compound Conditions
a b c d e f
Benzoic 53 15 72 69 79
2-Aminobenzoic 38
3-Aminobenzoic 28
4-Aminobenzoic 24
2-Hydroxybenzoic 55 72 78 54
2,4-Dihydroxybenzoic 19 30
3,4,5-Trihydroxybenzoic 4 11
Phthalic 17 55 41
a, Ethanol}butanol}water}cocentrated ammonia (40 : 30 : 15 : 15),
rice starch, detection by UV; b, n -hexane}acetic acid (96 : 4), SIF
silica gel sheet, UV; c, same as b, cellulose-TLC alumina, UV;
d, n -pentyl formate}chloroform}formic acid (70 : 15 : 15), sil G25,
detection by bromocresol green; e, n -pentyl formate}chloro-
form}formic acid (20 : 70 : 10), same as d, UV; f, 2-butanone}
methyl phenyl ketone}50% acetic acid (5 : 5 : 4), poly-N-vinylpyr-
rolidone-gypsum, detection by molybdate, diazotized sulfanilic
acid, phloroglucinol. (With acknowledgement to Hanai, 1982.)
4-hydroxy-3,5-dimethoxycinnamic acid, 43 and 55
respectively.
The effect of the aryl nucleus, Ar, on the hRF
value of the acid ArCO2H is seen with benzoic, naph-
thalene-2-carboxylic and diphenyl-2-carboxylic
acids, which are 53, 70 and 75 respectively with
the solvent ethanol}butanol}water}conc ammonia
(40 : 30 : 15 : 15) and the adsorbent, rice starch. The
water-soluble vitamin, nicotinic acid (pyridine-3-
carboxylic acid) on sil G25 had an hRF value of
5 in n-pentyl formate}chloroform}formic acid
(70 : 15 : 15), while that of benzoic acid was 69.
The hRF values of the isometric benzenedicar-
boxylic and certain polybasic reference acids were the
subject of early studies under a variety of conditions
and are shown in Table 9.
More recent experiments on the separation of the
dicarboxylic acids have been carried out with chloro-
form}tetrahydrofuran (2 : 1) on silufol in the presence
of an ion pair reagent but the separations described
earlier were just as effective.
The inSuence of the substituent position on the hRF
values of substituted benzoic acids has been studied
with reference to amino, nitro, chloro, hydroxy and
carboxy compounds, although nothing appears to
have been described on the separation of the isomeric
toluic acids. However, there has been extensive work
by Guinchard et al. (see Further Reading) on (1)
benzoic acid compared with (2) 2-chloro, (3) 3-
chloro, (4) 4-chloro, (5) 2-hydroxy, (6) 3-hydroxy,
(7) 4-hydroxy, (8) 2-nitro, (9) 3-nitro, (10) 4-nitro,
(11) 2-amino and (12) 4-aminobenzoic acid. Figure 1
1867
depicts the effect on the RF of small changes in
the aqueous formic acid concentration in benzene
when this range of acids was chromatographed on
silica gel G.
Figure 2 depicts the effect on the RF value of
the same series when run in benzene containing di-
ethyleneglycol monoethyl ether with various concen-
trations of formic acid. Separation of all 12 acids can
be achieved in either system with the appropriate
formic acid concentration. The more polar com-
pounds have lower RF values than the less polar
ones.
By contrast, under reversed-phase conditions with
benzene as the developing solvent, Figure 3 shows the
separations of a number of 2-substituted benzoic
acids on silanized silica gel (RP-8) with various aque-
ous organic solvents (organic solvent}water, 40 : 60,
v/v) containing 0.1 mol L\1 tetramethylammonium
bromide. With this system the more polar solutes
have higher RF values.
An extensive range of adsorbents and solvents for
a variety of aromatic carboxylic acids have been sum-
marized by Tyman (see Further Reading).
While carboxylic acids have been the main group
of acidic compounds studied, by contrast sulfonic
acids RSO3H, of both aliphatic and aromatic origin,
have received little attention. 1-N-Acylamino-8-hy-
droxynaphthalene-3,6-disulfonic acid derivatives of
interest for anti-human immunodeRciency virus
activity have been studied on S III Chromarods
with methanol or methanol}chloroform}ammonia
(35 : 55 : 10) as solvents.
Sulfuric esters, ROSO3H of substituted phenols,
have been examined on silica gel G with benzene}
butanone}ethanol}water (30 : 30 : 30 : 10). The less
acidic group, for example the sulfonamides, NH2
C6H4SO2NHR (where R comprises a wide variety of
Table 9 The hRF values of carboxy derivatives of benzoic acid
Acid Conditions
a b c d
Phthalic (1,2) 51 30 39 36
Isophthalic (1,3) 75 64 71 59
Terephthalic (1,4) 0 69 81 0
Trimellitic (1,2,4) 41 13 14 13
Pyromellitic (1,2,4,5) 0 2 2 0
Hexahydrophthalic 60 65 79 66
a, Diisopropyl ether}formic acid}water (90 : 7 : 3), silica gel;
b, same as b but saturated with polyethylene glycol M 1000
kieselguhr impregnated with polyethylene glycol; c, diisopropyl
ether}light petroleum}carbon tetrachloride}formic acid}water
(50 : 20 : 20 : 8 : 1), polyamide; d, butyl formate}ethyl acetate}for-
mic acid (82 : 9 : 9), same polyamide as c. (With acknowledge-
ment to Copius-Peereboom, 1969.)
1868 III / ACIDS / Thin^Layer (Planar) Chromatography
Figure 1 RF values of aromatic carboxylic acids in benzene containing formic acid. 1, Benzoic; 2, 2-chloro; 3, 3-chloro; 4, 4-chloro;
5, 2-hydroxy; 6, 3-hydroxy; 7, 4-hydroxy; 8, 2-nitro; 9, 3-nitro; 10, 4-nitro; 11, 2-amino; 12, 4-aminobenzoic acids. (Reproduced with
premission from Guinchard et al., 1976.)
groups), has been examined in detail. Monoalkyl
phosphate esters, ROP(O)(OH)2, dialkyl esters, (RO)2
P(O)OH and monoalkylphosphonic acids RP(O)
(OH)2 do not seem to have been examined by TLC.
Visualizing Agents for Aromatic Carboxylic
Acids
In this article, reference has frequently been made to
the detection of acids with bromocresol green and
other systems. Some other reagents for aromatic car-
boxylic acids are hydrogen peroxide or alkaline
potassium permanganate. Several new visualizing
agents and sodium hydroxide (10% aqueous solu-
tion) were compared with respect to the minimum
quantity of acid detectable (ug per spot) and the type
of layer. Generally, of the three layers, silica gel 60
GF254, silica gel}kieselguhr mixtures and polyamide,
the Rrst was preferred. Although the minimum detect-
able amount of solute varied with the 13 differ-
ent solutes and the 12 different visualizing agents
examined, thymol blue detected all the solutes
while bromothymol blue and bromocresol green de-
tected all but 4-hydroxybenzoic acid and 3-hy-
droxycinnamic acid respectively with silica gel as
adsorbent.
Quantitative TLC Determination of
Organic Acids in Synthetic and
Natural Mixtures
Examples of the application of TLC for the quantitat-
ive determination of a variety of acids in edible,
potable and polymeric products are discussed in this
section. Many simple aliphatic acid aromatic acids,
notably benzoic acid, citric and sorbic acids, are em-
ployed in edible materials such as preservatives while
salicylic acid and its acetyl derivative appear in
numerous pharmaceutical preparations. Accordingly,
their quantitative determination is important and for
such analyses planar methods have been widely used.
Some typical quantitative applications are described
in detail.
HPTLC Determination of Organic Acid
Preservatives in Beverages
In a high performance TLC (HPTLC) method sorbic
acid (2,4-hexadienoic acid) and benzoic acid were
determined without preliminary extraction or clean-
up by the chromatography of aliquots of samples and
of standards on preadsorbent silica gel or C18-bonded
silica gel plates containing Suorescent indicator.
 III / ACIDS / Thin^Layer (Planar) Chromatography 1869

Figure 2 RF values of aromatic carboxylic acids. 1, Benzoic; 2, 2-chloro; 3, 3-chloro; 4, 4-chloro; 5, 2-hydroxy; 6, 3-hydroxy;
7, 4-hydroxy; 8, 2-nitro; 9, 3-nitro; 10, 4-nitro; 11, 2-amino; 12, 4-aminobenzoic acid in benzene containing formic acid and diethylene
glycol monoethyl ether. (Reproduced with premission from Guinchard et al., 1976.)

The zones which quenched Suorescence upon UV
irradiation at 254 nm were compared by scanning
densitometry. This procedure was preferred to measure-
ment of densitometry based on UV absorption.
Preadsorbent high-performance LHPKDF silica gel
(Whatman) plates (20;10 mm) with 19 lanes were
used for normal-phase experiments with the solvent
n-pentyl formate}chloroform}formic acid (2 : 7 : 1) in
which the hRF values for sorbic acid and benzoic acid
were 61 and 58. For reversed-phase TLC on (What-
man) C18 LKC18F plates (20;20 mm) with meth-
anol}0.5 mol L\1 sodium chloride (1 : 1), the respect-
ive hRF values for these two acids were 44 and 59. It
was found necessary to apply a stream of warm air
during spotting of samples with a 10
L Drummond
digital microdispenser and, after this stage, to dry the
plates. Development was then effected in
a Camag twin-trough chamber to 7 cm beyond the
sorbent}preadsorbent interface with normal-phase
plates and to 10 cm for C18 plates. The plates were
then dried and the areas of the dark quenched zones
against a Suorescent background were scanned at the
predetermined maximum absorption (between 200
and 370 nm) with a Shimadzu Model 930 den-
sitometer operated in the reSectance mode. From the
chromatography of 0.50, 1.00, 2.00, 4.00, 6.00 and
8.00
L of standards for sorbic and benzoic acids
containing 125}2000 ng and 1.00}16.0
g respec-
tively, linear calibration plots of scan area/weight
were obtained. For quantiRcation, the sample scan
area was compared with that of a closely matching
standard within the linear calibration range and the
corresponding weight found. Recovery analyses were
carried out with beverage samples spiked with sorbic
and benzoic acids, which were compared with the
corresponding unfortiRed samples. They averaged at
98.0% for all analyses.
By the HPTLC method, sorbic and benzoic acids
present separately in a variety of beverages have been
directly quantiRed. The analysis of standards on the
same TLC plate eliminates the requirement for an
internal standard, as in high performance liquid
chromatography (HPLC). By contrast with the
HPTLC and HPLC methods, spectrophotometric
1870 III / ACIDS / Thin^Layer (Planar) Chromatography

Figure 3 RF values of 2-substituted benzoic acids in different solvents. The solvent composition, organic component}water (40:60
v/v) with addition of 0.1 mol L\1 tetramethylammonium bromide (pK value in parantheses): open circles, benzoic acid (4.19); filled
triangles, 2-hydroxy (2.97); open squares, 2-acetoxy (3.5); filled squares, 2-carboxy (2.91/5.59); filled circles, 2-nitro (2.16); open
squares, 2-methyl (3.91); open triangles, 2-amino (6.97); filled/inverted triangles, 2-chloro (2.92). (Reproduced with permission from
Jost et al., 1984.)

analysis requires a preliminary sample preparation by
steam distillation. However, very low concentrations
of benzoic acid are more amenable to HPLC analysis
and when sorbic and benzoic acids are present to-
gether the method is less satisfactory due to sample
streaking, even on a C18-bonded silica gel layer (parti-
cularly at higher loads).
In view of these limitations, a modiRed method was
adopted, involving solid-phase extraction (SPE) on
a C18 cartridge followed by the preceding quantiRca-
tion method established on preadsorbent C18 plates.
The extraction procedure was validated by spiking
commercial samples with known amounts of the
acids in turn and demonstrating the satisfactory re-
covery of each. With this total method, sample inter-
ferences were eliminated and samples too low for
analysis by direct spotting could be analysed. The
whole TLC methodology is considered to be applic-
able to a wide range of solid and syrupy-type
samples containing either or both of the two preserv-
atives at concentrations as low as those measurable
by HPLC.
 III / ACIDS / Thin^Layer (Planar) Chromatography
Quantitative Fluorescence Densitometry for
the Analysis of Rosmarinic Acid
Rosmarinic acid (9), a useful natural antimicrobial
compound of potential interest to the food industry,
occurs in eel grass (Zostera marina) from which it is
extractable together with a number of other phenolic
acids. It has been directly quantitatively and rapidly
analysed by an HPTLC densitometric method which
utilized the Suorescence of the material upon excita-
tion at 366 nm.
Crushed leaves (200 mg) of the natural product
were extracted with 5% acetic acid}methanol (1:2)
accompanied by ultrasonication during 30 min. The
extract was Rltered and then employed for direct
HPTLC on plates (10;20 cm) pre-coated with cellu-
lose without Suorescent indicator. Samples and stan-
dard solutions (2
L) were applied to plates as 7 mm
wide bands with a Linomat IV applicator under
a pressure of 2.5 bar; this was developed in a twin-
trough chamber with 3% sodium chloride in 0.5%
acetic acid}acetonitrile}tetrahydrofuran (100 : 24 : 1)
until the solvent had migrated 4.5 cm. The dried plate
was irradiated with a mercury vapour lamp and the
resultant Suorescence emission measured through
a cut-off Rlter (400 nm) by scanning with a TLC
scanner II (Camag) equipped with CATS software
(version 3.14).
Plots of either peak area or height/concentration
were linear over concentration range 0.1}0.6 mg
mL\1 (i.e. 0.2}1.2
g) and the weight of rosmarinic
acid in unknown samples was readily found.
Densitometric Analysis of Gallic Acid in
Fermentation Liquors
One of the ways used for obtaining gallic acid (3,4,5-
trihydroxybenzoic acid), an important intermediate
in synthesis for the pharmaceutical and food indus-
tries, is by the acid hydrolysis of natural gallotannins,
for example from gall nuts, tara pods or sumac leaves.
In an enzymatic procedure hydrolysis of these types
of raw material with a fungal tannin acylhydrolase
which cleaves depside bonds, the monitoring of
a large number of samples by a simple and rapid TLC
method was investigated as a potential alternative to
HPLC analysis.
Crude samples from enzymatic solutions were di-
luted between one- and 100-fold with methanol and
1871
Rltered through a Minisart NML 0.45
m Rlter and
gallic acid used at known concentrations as an inter-
nal standard. TLC analysis was performed on glass
plates (5;20 cm), coated with a 0.25 mm layer of
RP-18 F254 (Merck 15683); the glass plates were pre-
cleaned with a single development in methanol. Sam-
ples (6
L) were applied with a Linomat IV spotter
and then developed to a distance of 12 cm, with
M aqueous acetic acid}methanol (1 : 1) for 2 h. Den-
sitometry was effected by spectrophotometry and
a mercury light source (254 nm) in the absorbance
mode, to determine extinction of Suorescence, as an
area measurement, with a TLC scanner II (Camag)
controlled by CATS software. Calibration plots were
found to be near to linearity with between 10 and
75
g gallic acid on the plate when the ratios of the
peak area of the acid to the internal standard were
between 0.3 and 1.5, although in practice ratios of
areas between 0.5 and 1.25 (corresponding to gallic
acid between 25 and 62.5
g) were adopted in the
analytical method. An inherent difRculty was found
to be slight inhomogeneity in the coating of the Suor-
escent indicator: to improve on this, the plate was
scanned before an assay to determine any back-
ground Suorescence, which was then subtracted to
‘zero’ the plate. With this proviso and by the use of
the strict linearity range, the values obtained for gallic
acid were 98$2.1% of those found by HPLC.
Determination of Diacetonegulonic acid (DAG)
in Water Samples
DAG (10) is the penultimate intermediate in the syn-
thesis of ascorbic acid (vitamin C) and for many years
was discharged in waste surface waters. This led to
contamination of groundwaters and, although it is
not toxic to humans, it has an inhibitory effect
on the growth of grasses. Current European drinking
water regulations restrict its concentration to 0.1
g
L\1. A fast and efRcient HPTLC method has been
described.
1872 III / ACIDS / Thin^Layer (Planar) Chromatography
Due to the low concentration of DAG, SPE is used
for sample preparation. Because of the sensitivity of
DAG to silica gel and, more particularly to acidic
solutions, it was found necessary to adjust the water
sample for analysis to no less than pH 4 and to
effect SPE with Polyspher RP-18 (a 35
m poly-
styrene-divinylbenzene polymer with C18 side chains)
which gave a 100% recovery. For the extraction
a cartridge (0.2 g) was Rrst conditioned successively
with ethyl acetate, methanol and water at pH 4
(1 cm3 of each), after which the water sample for
analysis adjusted to pH 4 (20 cm3) was aspirated
through the cartridge. The cartridge was dried in
a stream of nitrogen and then eluted with ethyl acet-
ate (2;1 cm3) and the eluate after treatment with one
drop of ammonia evaporated at less than 403C to
leave 0.5 cm3, an aliquot of which was applied to an
PTLC silica gel 60 F254 pre-coated plate (10;20 cm).
In the case of original concentrations of less than 5
g
L\1, the total eluate was used for TLC.
For analysis of sample volumes up to 20
L, mul-
tiple development one-dimensionally with solvent A,
chloroform}methanol (80 : 20) to 8 cm and then after
drying, solvent B (chloroform}methanol}glacial
acetic acid, 80 : 20 : 2) for 6.5 cm was carried out.
Alternatively, two-dimensional development was car-
ried out with the same two solvents, distances and
drying. Spots or streaks were detected by immersion
of the plate in an ethanolic solution of 4-methoxybenz-
aldehyde containing sulfuric acid, followed by drying
and heating at 1303C for 2}3 min to form red Suor-
escent areas which were visible under UV light
(366 nm) and quantiRed with a TLC scanner. Two-
dimensional development was advocated for samples
with less than 5
g L\1 DAG, while for higher con-
centrations, one-dimensional development was ad-
equate. The calibration of peak area/weight DAG
was linear within the range 0.125}1.5
g. It was
found that for the determination of higher concentra-
tions it was essential to apply DAG as streaks to
preserve linearity over the range of concentrations
and it was then established that from 0.25 to 250
g
could be analysed with consistent accuracy.
The SPE procedure followed by TLC appears to be
superior to derivatization followed by GC-MS and it
was considered that very small concentrations of
DAG could even be estimated visually without any
instrumentation, thus generally giving an inexpensive
procedure. Other application of quantitative TLC to
the analysis of humic acids in natural waters, 6-
aminocaproic acid (12), -caprolactam in polyamide-
6 (11) and to uric acid (13), creatine (14) and cre-
atinine (15) mixtures in biological materials have
been described.
Conclusions
Acids of simple and more complex structures are
components of many edible, technical and medicinal
products and TLC affords an ideal approach for
their analysis because no derivatization is required
and a wide variety of detection methods is applicable
for their qualitative and quantitative determination.
It can be envisaged that the use of HPTLC, of special
layers and the employment of combined techniques
will continue to extend and expand the planar ap-
proach to the analysis of acidic compounds.
See also: II/Chromatography: Thin Layer (Planar):
Densitometry and Image Analysis; Ion Pair Thin-Layer
(Planar) Chromatography; Spray Reagents. III/Acids:
Gas Chromatography; Liquid Chromatography.
Further Reading
Ariga N (1972) Thin-layer chromatography of keto acid
2,4-dinitrophenylhydrazones. Analytical Biochemistry
1972, 49: 436.
Barthomeuf C, Regerat F and Combe-Chevaleyrer S (1993)
Densitometric analysis of gallic acid in fermentation
liquors. Journal of Planar Chromatography 6: 245}247.
Copius-Peereboom JW (1969) Thin layer chromatography.
In: Stahl E (ed.) Foodstuffs and their Additives, p.
653. London: G Allen and Unwin.
Eisenbeiss A, Reuke S and TuK rck M (1992) Determination
of diacetoneketogulonic acid in water samples by
HPTLC. Journal of Chromatography 589: 390}393.
 III / AFLATOXINS AND MYCOTOXINS / Chromatography
GaK nshirt H (1969) Synthetic pharmaceutical products. In:
Stahl E (ed.) Thin-layer Chromatography, p. 541. Lon-
don: G. Allen and Unwin.
Guinchard C, Truong TT, Masson JD and Panouse JD
(1976) Migration d’acides aromatiques en chromatog-
raphie sur couche mince de gel de silice en fonction de la
teneur en eau ou en acide formique de solutions creH ant
l’atmosphere de la cuve à chromatographie. Chromato-
graphia 9: 627}629.
Hanai T (1982) Phenols and organic acids. In: Zweig G and
Sherma J (eds), Handbook of Chromatography, vol. 1,
pp. 159}174. Boca Raton, CRC Press.
Hauck HE, Mack M and Jost W (1996) Sorbents and
precoated layers in thin-layer chromatography. In: Sherma
J and Fried BJ (eds) Handbook of Thin Layer Chromatog-
raphy, 2nd edn, p. 101. New York: Marcel Dekker.
Jost W, Hauck HE and Herbert H (1984) Reversed-phase
thin-layer chromatography of 2-substituted benzoic
acids with ammonium compounds as ion-pair reagents.
Chromatographia 18: 512}516.
Kas\ telan-Macan M, Cerjan-Stefanovics and Jals\ ovec D (1992)
Determination of aquatic humic acids in natural river
waters. Water Science and Technology 26: 2567}2570.
Khan SH, Murawski MP and Sherma J (1994) Quantitative
HPTLC determination of organic acid preservatives.
Journal of Liquid Chromatography 17: 855}865.
Klaus R, Fischer W and Hauck HE (1991) Qualitative and
quantitative analysis of uric acid, creatine and creatine
together with carbohydrates in biological materials by
HPTLC. Chromatographia 32: 307}316.
AFLATOXINS AND MYCOTOXINS
Chromatography
R. D. Coker, Natural Resources Institute,
Medway University, Chatham, UK
Copyright ^ 2000 Academic Press
Introduction
Mycotoxins have been deRned as ‘fungal metabolites
which, when ingested, inhaled or absorbed through
the skin, cause lowered performance, sickness or
death in man or animals, including birds’.
Exposure to mycotoxins can produce both acute
and chronic toxic effects ranging from death to
deleterious effects on the central nervous, cardio-
vascular and pulmonary systems, and on the alimen-
tary tract. Mycotoxins may be carcinogenic,
mutagenic, teratogenic and immunosuppressive. The
ability of some mycotoxins to compromise the im-
1873
Madelaine-Dupich C, Azema J, Escoula B, Rico L and
Lattes A (1993) Analysis of N-acylaminonaphthalene
sulphonic acid derivatives with potential anti-human
immunodeRciency activity by TLC and FID. Journal of
Chromatography 653: 178}180.
Petersen HW, Petersen LM, Piet H and Ravn H (1991)
A new HPTLC Suorescence densitometric method for
the quantitative analysis of rosmarinic acid. Journal of
Planar Chromatography 4: 235}236.
Petrowitz H-J (1969) Synthetic organic products. In: Stahl
E (ed.) Thin-layer Chromatography, p. 678. London:
G Allen and Unwin.
Sarbach Ch, Postaire E and Sauzieres J (1994) Simultaneous
determination of
-caprolactam and
-aminocaproic
acid contaminants in polyamide-6. Journal of Liquid
Chromatography 17: 2737}2749.
Smith MC and Sherma J (1995) Determination of benzoic
acid and sorbic acid preservatives. Journal of Planar
Chromatography 8: 103}106.
Tyman JHP (1996) Phenols, aromatic carboxylic acids and
indoles. In: Sherma J and Fried BJ (eds) Handbook of
Thin-layer Chromatography, 2nd edn, pp. 906}907,
912}913. New York: Marcel Dekker.
Wardas W, Pyka A and Jedrzejczak M (1995) Visualising
agents for aromatic carboxylic acids in TLC. Journal of
Planar Chromatography 8: 148}151.
Williams RJ and Evans WC (1975) The metabolism of
benzoate by Moraxella species through anaerobic ni-
trate respiration. Biochemistry Journal 148: 1.
mune system and, consequently, to reduce resistance
to infectious disease, is now widely considered to be
their most important effect.
The mycotoxins attract worldwide attention be-
cause of the signiRcant economic losses associated
with their impact on human health, animal produc-
tivity and both domestic and international trade. It
has been estimated, for example, that annual losses in
the USA and Canada arising from the impact of
mycotoxins on the feed and livestock industries are in
the order of US$5 billion. In developing countries
where the food staples (e.g. maize and groundnuts)
are susceptible to contamination, signiRcant addi-
tional losses amongst the human population are like-
ly, because of morbidity and premature death asso-
ciated with the consumption of mycotoxins.
It is likely that mycotoxins have plagued mankind
since the beginning of organized crop production.
Ergotism (St Anthony’s Fire), for example, which is
caused by the consumption of rye contaminated
with the ‘ergot alkaloids’, is discussed in the Old
1874 III / AFLATOXINS AND MYCOTOXINS / Chromatography
Testament, and reached epidemic proportions in
many parts of Europe in the tenth century.
Mycotoxins of Worldwide Importance
An ‘important’ mycotoxin will have demonstrated its
capacity to have a signiRcant economic impact on the
exposed human and/or animal population. Those
moulds and mycotoxins that are currently considered
to be of worldwide importance are shown in Table 1,
and the chemical structures of the mycotoxins in
Figure 1.
A]atoxins
The term ‘aSatoxins’ was coined in the early 1960s
when the deaths of thousands of turkeys (‘Turkey X’
disease), ducklings and other domestic animals were
attributed to the presence of Aspergillus Uavus toxins
in groundnut meal imported from South America.
The acute and chronic effects of the aSatoxins on
a wide variety of livestock are now well documented,
and include death, decreased productivity, and in-
creased susceptibility to disease. ASatoxin B1 is a hu-
man carcinogen and one of the most potent hepa-
tocarcinogens known. Human fatalities have resulted
from the consumption of heavily aSatoxin-con-
taminated foods, frequently when wholesome food is
in short supply. ASatoxin M1 occurs in milk, and is
produced by the bovine metabolism of aSatoxin
B1 when contaminated feed is ingested by dairy cows.
Trichothecenes
T-2 toxin, deoxynivalenol (and nivalenol) belong to a
large group of structurally related sesquiterpenes
known as the ‘trichothecenes’, which occur primarily
in cereals. T-2 toxin is the probable cause of ‘alimen-
tary toxic aleukia’ (ATA), a disease that affected
thousands of people in Siberia during the Second
World War, and led to the elimination of entire vill-
ages. The symptoms of ATA include fever, vomiting,
acute inSammation of the alimentary tract and a var-
iety of blood abnormalities. The same toxin is also
Table 1 Moulds and mycotoxins of worldwide importance
Mould species Mycotoxins produced
Aspergillus parasiticus Aflatoxins B1, B2, G1, G2
A. flavus Aflatoxins B1, B2
Fusarium sporotrichioides T-2 toxin
F. graminearum Deoxynivalenol
(or nivalenol in some areas)
zearalenone
F. moniliforme Fumonisin B1
Penicillium verrucosum and
A. ochraceus Ochratoxin A
associated with outbreaks of haemorrhagic disease in
animals and with neurotoxic effects in poultry.
An important effect of T-2 toxin (and other tricho-
thecenes) is the immunosuppressive activity which has
been clearly demonstrated in experimental animals.
Deoxynivalenol (DON) is probably the most wide-
ly occurring Fusarium mycotoxin. (The trivial name
of ‘vomitoxin’ has also been accorded to DON be-
cause of outbreaks of emetic (and feed refusal) syn-
dromes, amongst livestock, caused by this toxin.) The
ingestion of DON has caused acute human mycotoxi-
coses in India, China and rural Japan. The Chinese
outbreak, in 1984}85, was caused by mouldy maize
and wheat. Symptoms occurred within 5 to 30 min
and included nausea, vomiting, abdominal pain, diar-
rhoea, dizziness and headache.
Zearalenone
Zearalenone is an oestrogenic mycotoxin that is co-
produced with DON, and which has been implicated,
with DON, in outbreaks of acute human mycotoxi-
coses. In livestock, exposure to zearalenone-con-
taminated maize has caused hyperoestrogenism, espe-
cially in pigs, characterized by vulvar and mammary
swelling and infertility.
Fumonisins
Fumonisin B1 (FB1) occurs in maize produced in a
variety of agroclimatic zones. Two animal species,
horses and pigs, are particularly targetted by FB1.
Exposure to FB1 causes leukoencephalomalacia
(LEM) in horses and pulmonary oedema in pigs. The
presence of fumonisins in maize has been linked with
human oesophageal cancer in the Transkei (South
Africa) and China.
Ochratoxin A
Ochratoxin A (OA) causes nephropathy and im-
munosuppression in several animal species, and is
carcinogenic in experimental animals. OA has
been linked to the human disease Balkan endemic
Main sources
Edible nuts, oilseeds and cereals
Wheat and Maize
Wheat and Maize
Maize
Wheat, barley, coffee beans, vine fruits
 III / AFLATOXINS AND MYCOTOXINS / Chromatography 1875

Figure 1 Mycotoxins of worldwide importance.

1876 III / AFLATOXINS AND MYCOTOXINS / Chromatography
nephropathy, a fatal, chronic renal disease occurring
in limited areas of Bulgaria, the former Yugoslavia
and Romania. It has been suggested that pork prod-
ucts are signiRcant human dietary sources of OA.
Control of Mycotoxins
The control of mycotoxins is summarized in Figure 2.
The interventions that may be employed for the con-
Figure 2 The mycotoxin control system.
trol of mycotoxins are prevention of contamination,
identiRcation and segregation of contaminated ma-
terial (quality control, monitoring and legislation),
and detoxiRcation.
Preventative measures that militate against the on-
set of biodeterioration and, subsequently, the produc-
tion of moulds and mycotoxins, may be introduced
throughout the commodity system. However, the
preharvest control of biodeterioration is somewhat
 III / AFLATOXINS AND MYCOTOXINS / Chromatography 1877

compromised by our inability to control the climate!
Attempts have been made to prevent preharvest con-
tamination by breeding for resistance to moulds and
by ‘biocontrol’ methods, involving the introduction,
to the Reld, of atoxigenic strains of competing fungi.
After harvest, it is important that the crop is dried to
a ‘safe’ moisture level (which will not support the
growth of moulds and mycotoxins) as quickly as
possible.
The identiRcation and segregation of mycotoxin-
contaminated material may be pursued through qual-
ity control and regulatory procedures. More than 50
countries currently impose legal limits on the occur-
rence of mycotoxins (especially the aSatoxins) in
foods and feeds.
Commercial detoxiRcation plants, for the treat-
ment of aSatoxin-contaminated groundnut meal, are
currently operating in Senegal, France and the UK.
The chemical detoxiRcation reagent that is most
widely used is ammonia, both as an anhydrous va-
pour and an aqueous solution.
If the package of control procedures described above
is to be successfully implemented, it is essential that it
is underpinned by an integrated package of sampling,
sample preparation and analytical procedures.
Analysis of Mycotoxins
Worldwide, 5 parts per billion (
g kg\) is the most
common maximum level of total aSatoxins permitted
in foods. Similarly, aSatoxin M1 is regulated in at
least 14 countries, the permitted levels typically fall-
ing within the range 0.05 to 0.5 parts per billion.
Consequently, it is essential that the analytical
methods used for quality control and monitoring
(regulatory control) purposes are accurate and precise
at these extremely low concentrations.
Analytical Sequence
The analysis of mycotoxins may be considered in
terms of a sequence of four operations: extraction,
clean-up, quantiRcation and conRrmation. Some of
the more commonly used procedures associated with
these operations are illustrated in Table 2.
The mycotoxin(s) under investigation must Rrst be
extracted from the complex and variable chemical
milieu of the food or feed under investigation, using
an appropriate extraction solvent. Commonly used
solvent systems include acetone, acetonitrile, meth-
anol, ethyl acetate, chloroform and water, either sing-
ly or as mixtures of two or more solvents. The extrac-
tion is performed either by shaking the mixture of
sample and solvent for 30}45 min or by blending at
high speed for approximately 3 min. The choice of
solvent can signiRcantly affect the extractability
of the mycotoxin. The extraction of the aSatoxins
from corn, for example, is signiRcantly enhanced if
the aqueous extraction solvent contains acetone as
opposed to methanol. Supercritical Suid extraction
is an emerging alternative to liquid extraction, and
has been successfully applied to the extraction of
aSatoxin B1 from corn.
The crude extract, obtained after Rltration of the
shaken or blended mixture, is cleaned-up in order to
remove as much non-mycotoxin material as possible,
since the presence of extraneous compounds can
seriously diminish the efRciency of the analysis.
Clean-up procedures include liquid}solid extraction
(defatting), liquid}liquid partitioning, chemical ad-
sorption and chromatographic methods.

Table 2 The analysis of mycotoxins

Operation Commonly used procedure

Extraction Sample extracted by shaking or blending with chloroform, or mixtures

of water/methanol, water/acetonitrile or water/acetone
Clean-up Liquid}liquid partitioning or liquid}solid extraction
Chemical adsorption
Solid-phase extraction (SPE)
Multifunctional clean-up column
Chromatography
Immunosorbent columns
Quantification Thin layer chromatography (TLC)
High performance thin layer chromatography (HPTLC)
High performance liquid chromatography (HPLC)
Gas chromatography (GC)
Fluorimetry
Confirmation Cochromatography
Visual observation of colour change after derivatization
Spectroscopy (with or without derivatization)
Mass spectrometry
1878 III / AFLATOXINS AND MYCOTOXINS / Chromatography
Solid-phase extraction (SPE) and immunosorbent
columns are examples of recently introduced clean-up
procedures that are now frequently used. SPE car-
tridges are available with a wide variety of polar,
nonpolar and ion exchange bonded phases.
A ‘multifunctional clean-up column’ (MFC), com-
posed of lipophilic, dipolar and anion exchange sites,
reportedly affords the efRcient clean-up of
acetonitrile/water extracts within 10 s. MFC high
performance liquid chromatography (HPLC) analysis
methods have been applied to at least 10 mycotoxins.
The chromatographic quantiRcation techniques
used for the determination of mycotoxins in cleaned-
up extracts include thin-layer chromatography
(TLC), high performance TLC (HPTLC), high perfor-
mance liquid chromatography (HPLC), and gas
chromatography (GC). Worldwide, TLC is the most
common method employed for the estimation of
mycotoxins.
No assay can be considered as complete until the
presence of the presumptive mycotoxin has been con-
Rrmed. This is especially important when an unusual
commodity is under investigation. The ultimate con-
Rrmation involves the comparison of the physico-
chemical characteristics of the presumptive myco-
toxin with those of a standard compound. Such
a course of action is not normally utilized as a routine
procedure. ConRrmatory techniques used in con-
junction with HPLC include mass spectrometry and
ultraviolet spectroscopy. When TLC or HPTLC are
used for quantiRcation, the formation of derivatives
with characteristic chromatographic and Suorescence
properties is commonly employed to conRrm the
presence of the presumptive mycotoxin(s).
Analytical Accuracy
The overall accuracy of the determination of myco-
toxins will be governed by the combined effects
of the sampling, sample preparation and analytical
components of the analytical process. Undoubted-
ly, the sampling component is currently the greatest
source of analytical error. Until effective samp-
ling (and sample preparation) procedures have been
developed for a variety of mycotoxin/commodity
combinations, the accuracy and precision of methods
for the determination of mycotoxins will be severely
compromised.
The reliability of an analytical procedure may be
expressed in terms of the accuracy, precision and
limit of detection of the method. Interlaboratory pre-
cision is determined by the implementation of check-
sample and collaborative studies. The level of inter-
laboratory precision for the determination of
mycotoxins is still disappointing. A review of the
reliability of mycotoxin assays, conducted in 1993,
indicated that little or no improvement in interlabora-
tory precision had occurred over the previous 20
years. The precision of TLC and HPLC methods were
reportedly similar, whereas the precision of enzyme-
linked immunosorbent assay (ELISA) methods was
somewhat poorer. A series of proRciency testing exer-
cises were carried out during 1993 and 1994 involv-
ing those European laboratories who contribute ana-
lytical data on food contamination to the World
Health Organization (WHO) Global Environmental
Monitoring Scheme (GEMS). The tests were per-
formed according to the International Organization
for Standardization/International Union of Pure and
Applied Chemistry/Association of OfRcial Ana-
lytical Chemists (ISO/IUPAC/AOAC) International
Harmonized Protocol, and laboratories were
awarded ‘z scores’ that signiRed their analytical capa-
bility. Eighty eight per cent of the laboratories ob-
tained results of acceptable accuracy for the deter-
mination of the aSatoxins, whereas only 53% of the
laboratories demonstrated acceptable accuracy for
patulin (a mycotoxin produced by Penicillium expan-
sum and other moulds.)
Simple Methods
Methods of quantiRcation employing HPTLC, HPLC
and GC require expensive equipment and skilled per-
sonnel. However, such procedures are not normally
available in the basic analytical laboratories that exist
in, for example, exporting developing countries and
in some food and feed manufacturing plants.
Basic laboratory environments require simple,
robust, low-cost methods that can afford reliable
results in the hands of semiskilled operators. Methods
that have been developed with such an application in
mind include minicolumn and immunodiagnostic
procedures. The minicolumn approach utilizes small
glass columns packed either with selected chromato-
graphic adsorbents or with other inorganic adsorbing
materials. Minicolumns are used either to clean up
the crude extract before quantiRcation; or the my-
cotoxin under test is adsorbed onto the column, as
a band, which is normally visually determined under
ultraviolet (UV) light. Immunodiagnostic procedures
take the form either of immunoafRnity columns
or of solid-phase ELISA methods. ImmunoafRn-
ity columns are used to effect the sample clean-
up before the mycotoxin is quantiRed, either by
adsorption onto a Florisil ‘tip’ or by elution into
a simple Suorimeter.
Solid-phase ELISA methods have been developed
where the mycotoxin antibody is immobilized, for
example, onto a card (about the size of a credit card),
a plastic cup (the ‘immunodot’ approach) or a plastic
probe. The presence of the mycotoxin, above a
 III / AFLATOXINS AND MYCOTOXINS / Chromatography
predetermined level, is indicated by a visually ob-
served colour change within small indentations with-
in the card, cup or probe.
Chromatography of Selected
Mycotoxins
The methods used for the chromatographic analysis
of mycotoxins will now be further illustrated by de-
scribing the determination of the ‘important’ my-
cotoxins listed in Table 1. In each case, ‘ofRcial’
methods that have been approved by an appropriate
internationally recognized body will be described,
together with a selection of recently developed
procedures.
A]atoxins
The chromatographic methods employed for the de-
termination of the aSatoxins (B1,B2,G1,G2,M1,M2)
include TLC, HPTLC and HPLC, usually in combi-
nation with Suorescence detection. The aSatoxins
exhibit an intense Suorescence when subjected to
UV irradiation.
For TLC and HPTLC the intensity of Suoresence
may be estimated either visually (using, for example,
the ‘comparison of standards’ procedure) or den-
sitometrically.
When HPLC methods are employed, the intensity of
the Suorescence and the position of the excita-
tion/emission maxima vary with the composition of
the mobile phase. For example, the aSatoxins B1 and
G1 are much less intense than aSatoxins B2 and G2 in
aqueous or alcoholic solutions. The Suorescence exci-
tation maximum for B1 occurs at 355 and 363 nm in
acetonitrile and water, respectively, whereas the emis-
sion maximum varies from 415 (in chloroform) to
450 nm (in water). In aqueous solutions, the sensitivity
of the Suorescence detection system may be enhanced
by the pre-column treatment of the aSatoxins B1 and
G1 with triSuoracetic acid (TFA), or by post-column
treatment with either iodine or bromine solutions.
HPTLC, involving semiautomated sample applica-
tion and Suorescence densitometry, is sufRciently
robust to have been successfully exploited in labora-
tories in developing countries.
OfVcial methods Those methods that have been
approved by the AOAC and other international
bodies are described in Table 3. Methods 968.22 and
971.24 have also been adopted by the International
Union of Pure and Applied Chemistry (IUPAC);
methods 975.36 and 972.26 by the American Associ-
ation of Cereal Chemists (AACC); and methods
970.45 and 971.24 by the American Oil Chemists
Society (AOCS). It is evident from Table 3 that many
1879
of the ofRcial methods are based upon analytical
procedures that were developed many years ago, us-
ing a combination of silica gel column chromatogra-
phy clean-up and normal phase silica gel TLC.
Recent developments Reversed-phase HPLC, with
post-column derivatization and Suorescence detec-
tion, is now widely used in the developed world for
the analysis of the aSatoxins. Post-column iodination
is performed within a heated reaction coil, where the
column eluent is mixed with iodine-saturated water.
Post-column bromination can be performed where
bromide ion in the mobile phase is converted to
bromine using a commercially available electro-
chemical cell. Sample clean-up is frequently per-
formed using proprietary immunoafRnity or SPE
columns. The AOAC OfRcial Method 991.31,
for example, utilizes the ASatest immunoafRnity
column in combination with reversed-phase C18
HPLC for the determination of the aSatoxins. A sim-
ilar approach was reported in 1995 for the determina-
tion of aSatoxin M1 in cheeses. BrieSy, the di-
chloromethane extract is evaporated to dryness in
a rotary evaporator, redissolved in a mixture of meth-
anol/water/hexane (1 : 30 : 50 v/v), and subjected to
liquid partitioning. The aqueous phases are then
cleaned up using an immunoafRnity column con-
taining monoclonal antibodies against aSatoxin M1.
Reversed-phase (C18) HPLC quantiRcation, in com-
bination with Suorescence detection, affords an
approximately 75% recovery of aSatoxin M1, and
a limit of quantiRcation of 0.02
g kg\1. The Suorimet-
ric excitation and emission wavelengths are 360 and
435 nm. In the EC method (92/95/EEC), the sample
clean-up is performed using a combination of Florisil2+
and C18 SPE columns. The combination of C18 SPE
column clean-up and HPLC quantiRcation, with Su-
orescence detection, is frequently used for the deter-
mination of the aSatoxins in a variety of substrates.
HPTLC, in combination with phenyl bonded phase
SPE and Suorescence densitometry, has been success-
fully applied to the determination of aSatoxins in
a variety of commodities including corn, cottonseed,
sorghum, peanut butter and palm kernels. Typically,
aluminium-backed silica gel HPTLC plates are sub-
jected to bidirectional chromatography using anhyd-
rous diethyl ether and chloroform/xylene/acetone
(6 : 3 : 1 v/v) in the Rrst and second directions, respec-
tively. Interfering components may be removed by
carefully cutting away the upper part of the plate
after the Rrst development, before rotating the plate
through 1803 prior to the second development. The
estimation of aSatoxin B1, by bidirectional HPTLC,
in a variety of commodities is illustrated in Table 4.
HPTLC has also been recently used for the
1880 III / AFLATOXINS AND MYCOTOXINS / Chromatography

Table 3 Official methods for the analysis of aflatoxins; these are AOAC methods unless stated otherwise

a
Method Date Aflatoxin Commodity Extraction Development Stationary Clean-up Detection limit,
no. method solvent solvent/mobile phase method LOD (
g kg\1)/
developed phase Additional
information

HPLC (with fluorescence detection)

b
980.20 1980 B1, B2, G1, G2 Cottonseed Acetone/H2O H2O saturated Silica gel Chemical LOD not specified
products CHCl3/(cyclohexane) adsorption and
CH3CN (25:7.5:1)# silica gel
1.5% abs. ethanol column
or
2.0% isopropanol
986.16 1986 M1, M2 Liquid milk C18 SPE H2O/isopropanol/ C18 column Small silica LOD not specified
CH3CN (80:12:8) (pre-column gel column
derivatization)
990.33 1990 B1, B2, G1, G2 Corn and CH3OH/ H2O/CH3CN/ C18 column Silica gel
peanut butter 0.1M HCl CH3OH (pre-column column 5.0
(700:170:170) derivatization) 10.0, total
(AOAC/IUPAC
method)
991.31 1991 B1, B2, G1, G2 Corn, raw CH3OH/H2O H2O/CH3CN/ C18 column Aflatest 10.0, total
groundnuts and CH3OH (post-column Immunoaffinity (AOAC/IUPAC
peanut butter (3:1:1) derivatization) column method)

c
92/95/ 1991 B1 Animal feeds CHCl3 H2O/CH3OH/ C18 column Florisil and 1.0
EEC CH3CN (post-column C18 SPE
(130:70:40) derivatization)
TLC
b
968.22 1968 B1, B2, G1, G2 Groundnuts CHCl3 /H2O Acetone/CHCl3 Silica gel Silca gel LOD not specified
and their (5:95 to 15:85) column (IUPAC/AOAC
products method; CBd
method)
970.45 1970 B1, B2, G1, G2 Groundnuts CH3OH/H2O/ Acetone/CHCl3 Silica gel Centrifugation LOD not specified
and their hexane (5:95 to 15:85) and liquid (AOCS/AOAC
products partitioning method; BFd
method)
971.23 1969 B1, B2, G1, G2 Cocoa beans CHCl3 /AgNO3 Acetone/CHCl3 Silica gel Defatting and LOD not specified
solution (5:95 to 15:85) silica gel (IUPAC/AOAC
column method; modified
CB method)
971.24 1971 B1, B2, G1, G2 Coconut, CHCl3/NaCl Acetone/CHCl3 Silica gel Silca gel LOD not specified
copra, copra solution (5:95 to 15:85) column (IUPAC/AOCS/
meal AOAC method)
972.26 1972 B1, B2, G1, G2 Corn CHCl3/H2O Acetone/CHCl3 Silica gel Silca gel LOD not specified
(5:95 to 15:85) column (AACC/AOAC
method; based
upon CB method)
972.27 1972 B1, B2, G1, G2 Soya beans CHCl3/H2O Acetone/CHCl3 Silica gel Silca gel LOD not specified
(5:95 to 15:85) column (based upon CB
method)
974.16 1974 B1, B2, G1, G2 Pistachio nuts CHCl3/H2O Acetone/CHCl3 Silica gel Silca gel LOD not specified
(Method 1) (5:95 to 15:85) column (based upon CB
method)
(Method 2) CH3OH/H2O/ Acetone/CHCl3 Silica gel Centrifugation LOD not specified
hexane (5:95 to 15:85) and liquid (based upon BF
partitioning method)
978.15 1977 B1 Eggs Acetone/H2O/ 2D TLC: Silica gel Chemical LOD not specified
saturated (a) anhydrous diethyl adsorption,
NaCl solution ether/CH3OH/H2O liquid
(96:3:1) partitioning and
(b) Acetone/CHCl3 silica gel
(1:9) column
 III / AFLATOXINS AND MYCOTOXINS / Chromatography 1881

Table 3 Continued

a
Method Date Aflatoxin Commodity Extraction Development Stationary Clean-up Detection limit,
no. method solvent solvent/mobile phase method LOD (
g kg\1)/
developed phase Additional
information

980.20 1980 B1, B2, G1, G2 Cottonseed Acetone/H2O Acetone/CHCl3 Silica gel Chemical LOD not specified
products (5:95 to 15:85) adsorption and
silica gel
column
980.21 1978 M1 Milk, cheese CHCl3/NaCl For milk: Silica gel Silica gel LOD not specified
solution CHCl3/acetone/ column
isopropanol
(87:10:3)
For cheese:
2D TLC:
(a) diethyl
ether/CH3OH/H2O
(95:4:1)
(b) CHCl3/acetone/
isopropanol
(87:10:3)
982.24 1981 B1, M1 Liver CH2Cl2/citric 2D TLC: Silica gel Silica gel LOD not specified
acid solution (a) diethyl column
ether/CH3OH/H2O
(95:4:1)
(b) CHCl3/acetone/
isopropanol
(87:10:3)
993.17 1994 B1, B2, G1, G2 Corn and CH3OH/H2O CHCl3 /acetone Silica gel Silica gel 5.0,
groundnuts (9:1) column densitometrically
10.0, visually
Minicolumn
975.36 1975 B1, B2, G1, G2 Food and Acetone/H2O CHCl3/acetone CaSO4, Chemical 5.0, total;
feeds (9:1) Florisil, adsorption almonds
silica gel, 10.0, total: corn,
neutral alumina groundnuts,
peanut butter,
pistachio nuts,
groundnut and
cottonseed meals
15, total, mixed
feeds
Romer method
(AACC/AOAC
method)
979.18 1979 B1, B2, G1, G2 Corn, CHCl3/H2O CHCl3/acetone CaSO4, Liquid 10.0 (Holaday-
groundnuts Florisil, partitioning Velasco method)
(9:1) silica gel,
neutral alumina

a
The minimum contamination level to which the method is applicable: applies to aflatoxin B1, unless otherwise stated.
AOAC classification. c EC Directive. d Scott (1998).
b

determination of aSatoxin M1 in milk. The samples
were extracted with chloroform contained within
a hydrated dialysis tube, before subjecting the con-
centrated extract to HPTLC on silica gel plates. This
method gave a recovery of 96% and Suorescence
densitometry gave a detection limit of 0.002
g L\1.
The excitation wavelength was 350 nm, with an
emission cut-off of below 400 nm.
A recently reported novel approach to the deter-
mination of aSatoxins in corn utilizes silica or
immunoafRnity column clean-up in combination
with capillary electrophoresis, with laser-induced
1882 III / AFLATOXINS AND MYCOTOXINS / Chromatography

Table 4 Estimation of aflatoxin B1 by bidirectional HPTLC

Commodity Extraction solvent Clean-up method Limit of detection

(B1,
g kg\1)

Peanut butter Acetone/H2O Phenyl SPE 2.0

Corn Acetone/H2O Phenyl SPE 1.7
Cottonseed Acetone/0.1N HCl Phenyl SPE 2.7
Sorghum CHCl3/H2O Florisil column 1.3

Suorescence detection. The reported limit of detec-
tion is 0.5
g kg\1 aSatoxin B1, with an average re-
covery of 85% over the range 1 to 50
g kg\1.
Ochratoxins
OfVcial methods OfRcial AOAC methods exist
for the determination of the ochratoxin A in barley,
corn and green coffee. These procedures are
summarized in Table 5. It is evident from Table 5
that both the TLC methods are rather old, whereas
the HPLC procedure is reasonably modern. Each of
the ofRcial methods utilizes the native Suores-
cence of ochratoxin A for detection purposes. On
a silica gel TLC plate, ochratoxin A Suoresces most
intensely under 365 nm UV light. If the plate is
sprayed with alcoholic NaHCO3 solution the Suores-
cence increases in intensity, and changes from
greenish blue to blue in colour. If the TLC plate is
quantiRed densitometrically, the optimum excitation
and emission wavelengths are 310}340 and
440}475 nm, respectively. When employing HPLC,
the recommended Suorescence detection wavelengths
are 333 (excitation) and 460 nm (emission).
The AOAC Method 991.44 has been subjected to
an interlaboratory study involving 12 European
laboratories, under the auspices of the AOAC/
IUPAC/NMKL (Nordic Committee on Food Analy-
sis). The results of the intercomparison are given in
Table 6 for contamination levels, in wheat bran, rye
and barley, of between 2 and 9
g kg\1 ochratoxin A.
The mean recoveries varied from 64 to 72%. The
method has been accepted as an ofRcial NMKL
procedure.
Recent developments Recently developed HPLC
methods for the determination of ochratoxin A
employ silica gel SPE and immunoafRnity clean-
up followed by reversed-phase C8, C18 and C22 HPLC
columns, in combination with Suorescence detection.
The ionization of the phenolic group in the un-
derivatized toxin is suppressed by the presence of
phosphoric or acetic acids in the mobile phase.
An HPLC method (Method 1, Table 7) for the
determination of ochratoxin A in roast and ground
coffee uses a combination of silica gel SPE and
immunoafRnity clean-up in order to ensure a
good recovery (87%) of toxin. (Very low recoveries
were obtained when immunoafRnity clean-up
alone was used.) Fluoresence detection with excita-
tion and emission wavelengths of 333 and 470 nm

Table 5 Official methods for the analysis of ochratoxin A; these are AOAC methods unless stated otherwise

Method Date Commodity Extraction Development solvent/Mobile Stationary Clean-up method Detection limit, LOD
no. method solvent phase phase (
g kg\1)/Additional
developed information

TLC
973.37 1973 Barley CHCl3/ Acetone/CHCl3 (5:95 to 15:85) Silica gel NaHCO3/ LOD not specified
0.1 mol L\1 diatomaceous (IUPAC/AOAC
H3PO4 soln earth column method)
975.38 1975 Green coffee CHCl3 Toluene/ethyl acetate/ Silica gel NaHCO3/ LOD not specified
formic acid (5:4:1) diatomaceous
or earth column
benzene/CH3OH/acetic acid
(18:1:1, two sequential
developments)
HPLC
991.44 1992 Corn and barley CHCl3/ H2O/CH3CN/acetic acid C18 column C18 SPE 10.0 (IUPAC/
a
0.1 mol L\1 (99:99:2) NMKL method)
H3PO4 soln

a
NMKL, Nordic Committee on Food Analysis.
 III / AFLATOXINS AND MYCOTOXINS / Chromatography 1883

Table 6 Interlaboratory study of the official NMKL HPLC
method for the analysis of ochratoxin A
Commodity Coefficient of variation (%)
Intralaboratory Interlaboratory
Wheat bran 21 23}28
Rye 17 20}28
Barley 12 18}31
was employed. The presence of ochratoxin A was
conRrmed by the HPLC determination of its methyl
ester.
HPLC quantiRcation has also been used to deter-
mine the ochratoxin A content of milk (Method 2,
Table 7). The emulsion produced during the chloro-
form/methanol extraction was broken by refrigerated
centrifugation. After clean-up, the puriRed extract
was dissolved in methanol, by ultrasonic treatment,
before application to the HPLC column. The emis-
sion and excitation wavelengths of the Suorescence
detector were set at 330 and 460 nm. The presence of
ochratoxin A, in the range 0.01 to 0.03
g L\1, was
conRrmed by ELISA.
An HPTLC method (Method 3, Table 7) has
recently been developed for the determination of
ochratoxin A in parboiled rice. Extraction was
performed with chloroform and phosphoric acid;
the clean-up involved a combination of partition-
ing into sodium bicarbonate solution and phenyl
Table 7 Contemporary methods for the analysis of ochratoxin A
bonded-phase SPE. Bidirectional HPTLC using
aluminium-backed silica gel plates was employed,
using diethyl ether/methanol (98 : 2 v/v) and tol-
uene/ethyl acetate/formic acid (5 : 4 : 1 v/v) in the
Rrst and second directions, respectively. After remov-
ing the bottom portion of the plate, a third develop-
ment was performed, in the same direction, with
n-hexane/ethyl acetate/acetic acid (18 : 3 : 1 v/v).
Flurodensitometric detection (excitation at 365 nm)
afforded a mean intralaboratory precision of
5.4% over the concentration range 10 to 200
g kg\1
ochratoxin A. The mean recovery and limit of detec-
tion were 83% and 11.6
g kg\1, respectively.
Two intercomparison studies have recently been
performed, within the European Commission,
Measurements and Testing Programme, on the HPLC
determination of ochratoxin A. The Rrst study, using
kidney naturally contaminated at 10
g kg\1 och-
ratoxin A, involved 20 European laboratories. A var-
iety of extraction and clean-up procedures were used,
and recoveries ranged from 43 to 128%. The second
study, involving 26 European laboratories, used
wheat naturally contaminated with approximately
7
g kg\1 ochratoxin A. Again, a variety of extrac-
tion and clean-up procedures were employed. Some
laboratories compared their normal clean-up method
with the use of immunoafRnity columns supplied
from two different sources. The recoveries and
interlaboratory precision obtained using the normal
and immunoafRnity clean-up methods are compared

Method Commodity Date method Extraction Mobile phase/Developing Stationary Clean-up method Detection
no. developed solvent solvent phase limit, LOD
(
g kg\1)

HPLC
1 (1) Roast and 1997 CHCl3/ H3PO4/CH3CN (1:1) C18 column Silica gel SPE# 0.1
ground 0.1 mol L\1 immunoaffinity
coffee H3PO4 soln
2 (2) Milk 1996 CHCl3/CH3OH H3PO4 (0.008 mol L\1/CH3CNa C18 column Centrifugation 0.01
g L\1
(pH 1.6}2) (a) (60:40), (43C)# b
0.03
g L\1
(b) (90:10), (c) (60:40) silica gel SPE
HPTLC
3 (3) Rice 1996 CHCl3/ Bidirectional HPTLC: Silica gel Liquid 11.6
0.1 mol L\1 (a) diethyl ether/CH3OH (98:2) HPTLC plate partitioning#
H3PO4 soln (b) toluene/ethyl acetate/formic phenyl SPE
acid (5:4:1)
(c) n-hexane/ethyl acetate/acetic
acid (18:3:1)

a
Successive mobile phases.
b
Quantitation limit.
(1) Patel S, Hazel CM, Winterton AGM and Gleadle AE (1997) Survey of ochratoxin A in UK retail coffees. Food Additives and
Contaminants 14: 217I222.
(2) Valenta H and Goll M (1996) Determination of ochratoxin A in regional samples of cows milk in Germany. Food Additives and
Contaminants 13: 669I676.
(3) Dawlatana M, Coker RD, Nagler MJ and Blunden G (1996) A normal phase HPTLC method for the quantitative determination of
ochratoxin A in rice. Chromatrographia 42: 25I28.
1884 III / AFLATOXINS AND MYCOTOXINS / Chromatography

Table 8 Intercomparison of clean-up methods used for the
HPLC determination of ochratoxin A in wheat
Clean-up method Coefficient of variation Recovery (%)
(%) (Interlaboratory )
Normal 34 58}114
Immunoaffinity 34 58}114
(first source)
Immunoaffinity 42 4}86
(second source)
in Table 8. A recovery within the range 70 to 110%
was considered to be acceptable. The interlaboratory
coefRcient of variation obtained using normal
(and one immunoafRnity) clean-up methods were
similar to, but slightly greater than, the values ob-
tained by the intercomparison study of the ofRcial
AOAC/IUPAC/NMKL procedure.
Deoxynivalenol
OfVcial methods The two ofRcial AOAC methods
for the determination of deoxynivalenol in wheat
both date from 1986; these are outlined in Table 9.
The TLC procedure (Method 986.17) involves ex-
traction with acetonitrile/water followed by clean-up
using a small column packed with a mixture of char-
coal, alumina and Celite. The deoxynivalenol is ob-
served as a blue Suorescent spot, under UV light, on
the heated, aluminium chloride-treated plate. When
subjected to a collaborative study the reported aver-
age recoveries were between 78 and 96%, with intra-
and interlaboratory precisions (CV%) of 30}64 and
33}87% respectively.
The GC method includes extraction with water/
chloroform/methanol, a silica gel column clean-up
(under centrifugation) and derivatization with hep-
taSuorobutyric acid anhydride (HFBAA). Chrom-
atography is performed on a 3% OV-101 column
(using/argon methane as the carrier gas) with a 63Ni
electron capture detector. A collaborative study of
this procedure afforded an average recovery of
92% and intra- and interlaboratory precisions
(CV%) of 31 and 47%, respectively, for naturally
contaminated samples.
Recent developments In 1992, an intercomparison
study was reported on the determination of
deoxynivalenol in wheat and corn Sours. Fifteen la-
boratories participated, using one- and two-dimen-
sional TLC (Rve participants), GC (four) and HPLC
(six) procedures. Ten of the laboratories used a char-
coal-based clean-up method. A mixture of acetonit-
rile/water was widely used as an extraction solvent.
HPLC quantiRcation was performed using UV detec-
tion at 225 nm, whereas the GC determinations
employed trimethylsilyl, triSuoroacetyl and hepta-
Suorobutyryl derivatives. For all methods the recove-
ries varied between 68 and 116% for wheat and 53
and 100% for corn. There was no discernible dif-
ference in the efRcacy of the various quantiRca-
tion procedures.
Typically, TLC methods for the anlaysis of
trichothecenes involve extraction with acetonitrile or
methanol followed by clean-up using liquid partition-
ing and column chromatography on silica gel or
Florisil. Deoxynivalenol may be visualized on the
TLC plate by spraying with, for example, aluminium
chloride, 4-(p-nitrobenzyl)-pyridine, p-anisaldehyde
or cerium sulfate.
Recently developed methods for the determination
of deoxynivalenol, T-2 toxin (and zearalenone) are
summarized in Table 10.
The HPLC analysis of trichothecenes is frequently
performed using gradients of methanol/water or
acetonitrile/water in conjunction with C18 (or occa-
sionally C8) columns and detection by UV absorption.
Electrochemical detection has also been employed,
together with a variety of derivatization techniques.
The extraction/clean-up step in the HPLC procedure
(Method 1) includes the precipitation of milk protein,
with acetic acid, pH adjustment to 7}8, Extrelut2+
column chromatography and a charcoal}alumina
clean-up column. The recovery, for the concentration
range 25}200
g L\1 deoxynivalenol, was low (57%)

Table 9 Official methods for the analysis of deoxynivalenol; these are AOAC methods unless stated otherwise

Method Date method Commodity Extraction solvent Development solvent Stationary phase Clean-up method Detection limit,
no. developed or carrier gas LOD (
g kg\1)

TLC
986.17 1986 Wheat CH3CN/H2O CHCl3/acetone/ Silica gel Small column; mixture 300
isopropanol of charcoal, alumina
and Celite
GC
986.18 1986 Wheat H2O/CHCl3/ CH4/Ar (5:95) 3%OV-101 (on Quick-Sep silica 350
ethanol 80}100 mesh gel column
Chromosorb WHP)
 III / AFLATOXINS AND MYCOTOXINS / Chromatography 1885

Table 10 Recently developed methods for the determination of deoxynivalenol, T-2 toxin and zearalenone

Method Date method Commodity Extraction Clean-up Development Stationary phase Derivatization Detector/detection
no. developed solvent method solvent/mobile method limit, LOD (
g kg\1)
phase/carrier
gas

HPLC
1 (4) 1994 Cow’s milk Extrelut Centrifugation H2O/CH3CN Reversed-phase N/A UV absorption
column and (96:4) C18 column (220 nm) 25
g L\1
charcoal/alumina (Deoxynivalenol
column only)

GC
2 (5) 1996 Barley, CH3CN/ Celite mixed Helium (1.5% Cross-linked Trimethyl- NICI/MS
mixed feed, H2O charcoal, argon added methyl silicone silylation LOD not reported
sweet corn alumina, Celite for GC/MI capillary column (T-2 and
column and deoxynivalenol,
C8SPE only)

3 (6) 1996 Barley, maize CH3CN/H2O Defatting Helium Cross-linked Trimethyl- EI/SIM/MS
(hexane)# methyl silicone silylation 5
g kg\1
Florisil column capillary column
HPTLC
4 (7) 1998 Corn H2O/CHCl3 Liquid Toluene/ethyl Silica gel HPTLC N/A Fluorodensitometry
partitioning acetate/formic plate 2.6 (Zearalenone,
acid (6:2:1) only)

(4) Vudathala DK, Prelusky DB and Trenholm HL (1994) Analysis of trace levels of deoxynivalenol in cow’s milk by high pressure liquid
chromatography. Journal of Liquid Chromatography 17: 673I683.
(5) Mossoba MM, Adams S and Roach JAG (1996) Analysis of trichothecene mycotoxins in contaminated grains by gas chromatography/matrix
isolation/Fourier transform infrared spectroscopy and gas chromatography/mass spectrometry. Journal of AOAC International
79: 1116I1123.
(6) Ryu JC, Song YS, Kwon OS, Park J and Chang IM (1996) Survey of natural occurrence of trichothecene mycotoxins and zearalenone in
Korean cereals harvested in 1992 using gas chromatography mass spectrometry. Food Additives and Contaminants 13: 333I341.
(7) Dawlatana M, Coker RD, Nagler MJ, Blunden G and Oliver GWO (1998) An HPTLC method for the quantitative determination of
zearalenone in maize. Chromatographia 47: 215I218.

but consistent; the extensive clean-up probably con-
tributed to the loss of toxin.
GC is widely employed for the determination of
trichothecenes, including deoxynivalenol, notwith-
standing the inconvenience of lengthy clean-up and
derivatization steps prior to quantiRcation. Typically,
either the original trichothecene, or the alcohol pro-
duced by alkaline hydrolysis, is determined. The hy-
droxyl group(s) of trichothecenes are normally de-
rivatized in order to attain the required volatility
and sensitivity. Trimethylsilyl (TMS) derivatives
are frequently utilized for the GC of trichothecenes;
heptaSuorobutyryl and pentaSuoropropionyl deriva-
tives are employed for electron capture detection
(ECD) whereas triSuoroacetates are utilized for Same
ionization (FID), ECD and mass spectrometric (MS)
detection. GCMS methods have the advantage of
high sensitivity together with the opportunity of using
mass spectrometry for conRrmation purposes. The
speciRcity of MS detection affords the reliable
detection of toxins in grains, biological Suids and
other matrices. Generally, capillary GC is preferred
to the use of packed columns since the efRciency
of the latter can be compromised by interferences.
Capillary GC has been used for the analysis of
trichothecenes in a variety of commodities.
Both GC/matrix isolation (MI)/Fourier transform
infrared (FTIR) spectroscopy and GCMS have
been used to analyse mixtures of trichothecenes in
a variety of commodities (Method 2, Table 10).
Matrix isolation was performed by adding argon
to the carrier gas and trapping the efSuent on
the outer ring of a slowly rotating gold disc, at
low temperatures. The IR-transparent argon matrix,
containing the isolated trichothecenes, was then
analysed by IR spectroscopy, and the presence of
individual toxins conRrmed by observing the charac-
teristic MI/FTIR bands. Negative ion chemical
ionization (NICI) mass spectrometry was used to
quantify the high levels (67}445 mg kg\1) of
deoxynivalenol found in naturally contaminated
sweet corn. Seven Fusarium mycotoxins (including
deoxynivalenol, T-2 toxin and zearalenone) in barley
and maize have also been determined by GC/electron
impact-selective ion monitoring MS (Method 3,
Table 10). 5-Cholestane was used as an internal
1886 III / AFLATOXINS AND MYCOTOXINS / Chromatography
standard. The mean recovery for the seven myco-
toxins was 92%.
T-2 Toxin
OfVcial methods There are no ofRcial AOAC
methods for the determination of T-2 toxin.
Recent developments Methods available for the de-
termination of T-2 toxin include TLC, GC and HPLC.
T-2 toxin and other type A trichothecenes (charac-
terized by a hydrogen atom or an hydroxyl group
at the C8 position) are visualized on TLC plates
by treatment with sulfuric acid or chromotropic
acid (disodium 4,5-dihydroxynaphthalene-2,7-disul-
fonate dihydrate). Another approach involves the
formation of the diphenylindenone sulfonyl (Dis) es-
ters of trichothecenes and their visualization, as Suor-
escent spots under UV light, by spraying the TLC
plate with sodium methoxide. Using this procedure
20}25 ng per spot of T-2 toxin can be detected.
The HPLC determination of T-2 toxin is compro-
mised by the lack of the enone chromophore pos-
sessed by deoxynivalenol. The successful HPLC de-
termination of T-2 and other Type A trichothecenes
requires effective clean-up and derivatization pro-
cedures. A variety of post-column derivatization
methods have been developed including those involv-
ing the UV detection of p-nitrobenzoate and
diphenylindenone sulfonyl esters of T-2 toxin; the
reported detection limits are approximately 10 and
30 ng T-2, respectively.
The capillary GC-ECD determination of T-2 toxin,
and other Type A trichothecenes, afford detec-
tion limits of about 200
g kg\1 (with one chromato-
graphic clean-up) and 50}100
g kg\1 (with two
chromatographic clean-ups). A similar result has been
reported using a capillary GC-FID method. T-2 toxin
has also been detected in spiked wheat (in combina-
tion with deoxynivalenol), at levels of 1
g kg\1, by
using a GC-NICI MS-MS method. A highly sensitive
Table 11 Official methods for the analysis of zearalenone; these are AOAC methods unless otherwise stated
Method Date Commodity Extraction Development
no. method solvent solvent/mobile
developed phase/carrier gas
TLC
976.22 1976 Corn H2O/CHCl3 Ethanol/CHCl3
a
(5:95)
HPLC
985.18 1985 Corn H2O/CHCl3 CH3OH/CH3CN/H2O
(1.0:1.6:2.0)
a
Or ethanol/CHCl3 (3.5:96.5), acetic acid/benzene (5:95 or 10:90).
method for T-2 in urine employs capillary GCMS (EI
and NICI) with a detection limit of 2}5
g L\1. Cap-
illary GC-PICI MS was employed after clean-up of an
acetonitrile extract on an XAD-2 column and derivat-
ization with TFA.
Recently developed GC/NICI/MS and GC/EI/MS
methods for the determination of T-2 toxin, and
other trichothecenes, are outlined in Table 10
(Methods 2 and 3).
Zearalenone
OfVcial methods There are two ofRcial AOAC
methods (TLC and HPLC) for the determination of
zearalenone in corn (Table 11). The TLC method
(976.22) dates from 1976 and has also been adopted
by the AACC. The HPLC method (985.18) dates
from 1985 and can also be used for the determination
of -zearalenol. No limits of detection are given for
these procedures.
The ofRcial TLC method for zearalenone in-
volves extraction with chloroform/water, clean-up by
silica gel column chromatography and liquid parti-
tioning followed by TLC using either ethanol/chloro-
form or acetic acid/benzene. Zearalenone Suoresces
greenish-blue under 254 nm UV light; and blue under
365 nm UV light after treatment with aluminium
chloride.
The ofRcial HPLC method for zearalenone and
-zearalenol involves extraction with chloroform/
water (in the presence of diatomaceous earth), clean-
up by liquid partitioning and chromatography on
a C18 column using water/acetonitrile/methanol as the
mobile phase. Fluorescence detection is employed.
Recent developments A variety of HPLC methods
have been developed for the analysis of zearalenone
in corn together with methods for milk, blood, urine
and animal tissue. Clean-up procedures include liquid
partitioning and the use of silica gel cartridges. The
mobile phases used for reversed-phase HPLC include
Stationary phase Clean-up method Detection limit,
LOD (
g kg\1)
Additional
information
Silica gel Liquid partioning AACC/AOAC
method
LOD not specified
Reversed-phase Liquid partitioning LOD not specified
C18 column
 III / AFLATOXINS AND MYCOTOXINS / Chromatography 1887

Table 12 AOAC Official First Action HPLC method for the analysis of the fumonisins

Date method Commodity Extraction Mobile phase Stationary phase Clean-up method Detection limit,
developed solvent LOD (
g kg\1)

1990 Corn H2O/CH3OH Na2HPO4 C18 column SPE SAX cartridge 10

(buffered to pH
3.3)/CH3OH

acetonitrile/water, acetonitrile/water/acetic acid, water (3 : 1 v/v) as the extraction solvent followed by

methanol/acetonitrile/water and methanol/water.
Water-saturated dichloromethane containing 2% 1-
propanol has been used for normal-phase HPLC.
Fluorescence detection is most commonly used; other
methods include electrochemical, voltametric and UV
spectroscopic detection.
An HPTLC method (Method 4, Table 10) for the
determination of zearalenone in maize has recently
been developed, based upon the AOAC HPLC pro-
cedure (985.15). The mean recovery is 75.3%, over
the range 10 to 320
g kg\1 zearalenone.
Most of the numerous GC methods for the deter-
mination of zearalenone (and zearalenol) utilize
trimethylsilyl derivatization. A recently developed
GC method for the determination of zearalenone and
other Fusarium toxins, in barley and corn, is shown in
Table 10 (Method 3).
Fumonisins
OfVcial methods An HPLC method has received
OfRcial First Action status from the AOAC Inter-
national (Table 12). The procedure uses methanol/
strong ion exchange (SAX) clean-up and pre-column
derivatization with o-phthaldialdehyde (OPA). The
mobile phase is sodium dihydrogen phosphate solu-
tion (buffered to pH 3.3)/methanol and Suores-
cence detection is employed.
Recent developments Typically, the fumonisins are
determined by TLC, HPLC or GCMS, using ion ex-
change SPE clean-up and quantiRcation, after derivat-
ization of the primary amino group. HPLC is by far
the most widely used quantiRcation method.
A worldwide survey of methods used for the analysis
of the fumonisins was reported in 1996. Of the 32
laboratories included, 91% used HPLC. TLC and
GC/MS methods were each used by 3% of the labor-
atories. (ELISA was utilized by the remaining 3%.)
HPLC methods that are broadly similar to the
AOAC OfRcial First Action method have also been
developed using other clean-up procedures (e.g.
C18 SPE and immunoafRnity columns) and mobile
phases. The latter include mixtures of acetonitrile/
methanol/acetic acid; acidiRed methanol; and sodium

Table 13 Recently developed methods for the analysis of the fumonisins

Method Date method Commodity Chromatography Clean-up method Detection limit, Additional information
no. developed type LOD

HPLC
1 (8) 1996 Corn and corn Ion-pair SAX and C18 SPE 20 ng Derivatization with OPA and
products chromatography N-acetyl-L-cysteine;
fluorescence detection
2 (9) 1998 Corn-based Reversed-phase On-line immunoaffinity 5 ng Electrospray ionization MS
feed column
HPTLC
3 (10) 1998 Rice Silica-gel HPTLC SAX SPE a
250
g kg\1 Derivatization by dipping plate
into 0.17% p -anisaldehyde
solution; fluorescence
densitometry

a
Limit of quantification.
(8) Miyahara M, Akiyama H, Toyoda M and Saito Y (1996) New procedure for fumonisins B1 and B2 in corn and corn products by ion
pair chromatography with o-phthaldialdehyde post column derivatization and fluorometric detection. Journal of Agricultural and
Food Chemistry 44: 842I847.
(9) Newkirk DK, Benson RW, Howard PC, Churchwell MI, Doerge DR and Roberts DW (1998) On-line immunoaffinity capture,
coupled with HPLC and electrospray mass spectrometry, for automated determination of fumonisins. Journal of Agricultural and
Food Chemistry 46: 1677I1688.
(10) Dawlatana M, Coker RD, Nagler MJ and Blunden G (1995). A normal phase HPTLC method for the quantitative determination of
fumonisin B1 in rice. Chromatographia 41: 187I190.
1888 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography
hydrogen phosphate solution/methanol followed by
acetonitrile/water. Although OPA is used as the derivat-
ization reagent by the majority of laboratories, other
reagents have been employed including naphthalendial-
dehyde, Suoronitrobenzofurazan and Suorescamine.
The last reagent is unsatisfactory as it generates two
peaks in the HPLC chromatogram for fumonisin B1.
Three recently developed methods for the deter-
mination of the fumonisins in corn-based commodi-
ties are outlined in Table 13. Method 1 uses a
combination of SAX and C18 SPE clean-up prior to
ion-pair HPLC and Suorescence detection; on-line
derivatization within a reaction coil is employed. The
recovery of the fumonisins ranged from 54 to 110%
at 40 and 80
g kg\1, respectively. Method 2 is an
automated procedure using on-line immunoafRnity
clean-up, reversed-phase HPLC and electrospray ion-
ization MS detection. The protonated molecule signal
(m/z 722) was used to achieve a limit of quantiRca-
tion of 250 pg.
An HPTLC method (Method 3, Table 13), for the
determination of fumonisin B1 in rice, has recently
been reported. A novel derivatization step involved
the brief immersion of the HPTLC plate in a 0.16%
acidic solution of p-anisaldehyde, followed by quan-
tiRcation by scanning Suorodensitometry. The re-
sponse was linear over the range 0 to 5 mg kg\1 (ppm).
An intercomparison study on a variety of methods
for the determination of the fumonisins in maize has
recently been undertaken under the auspices of the
European Commission, Measurements and Testing
Programme. Twenty-four laboratories participated,
using their normal routine procedure for the deter-
mination of fumonisins B1 and B2 in the range 0.5}3.0
and 0.2}1.5 mg kg\1 (ppm), respectively. All labora-
tories used a similar method involving extraction
with methanol/water, clean-up with an SAX SPE col-
umn and HPLC Suorescence quantiRcation of the
OPA derivative. The intra- and interlaboratory pre-
cisions were high (10 and 11%, respectively, for
fumonisin B1; and 11 and 13%, respectively, for
fumonisin B2). However, the recoveries were low
(70$14% and 69$16% for fumonisins B1 and B2,
respectively). Interestingly, higher recoveries were
Thin-Layer (Planar) Chromatography
M. E. Stack, US Food and Drug Administration,
Washington DC, USA
Copyright ^ 2000 Academic Press
The aSatoxins are toxic and carcinogenic metabolites
of the moulds Aspergillus Uavus and A. parasiticus.
associated with extraction by shaking (85$12% for
fumonisin B1) rather than by blending (62$6%).
Conclusions
The continued use of a variety of chromatographic
procedures for the determination of mycotoxins is
envisaged. Although HPLC is the method of choice in
the developed world for a wide range of applications,
it is important that precise and accurate methods
continue to be developed that are appropriate to the
special needs of developing country laboratories.
See Colour Plate 53.
See also: II/Affinity Separation: Immunoaffinity Chrom-
atography. Chromatography: Gas: Detectors: Mass
Spectrometry. Chromatography: Liquid: Derivatization.
III/Aflatoxins and Mycotoxins: Thin Layer (Planar)
Chromatography. Membrane Preparation: Phase Inver-
sion Membranes.
Further Reading
Anon (1993) Some naturally occurring substances: food
items and constituents, heterocyclic aromatic amines
and mycotoxins. IARC Monographs on the Evaluation
of Carcinogenic Risks to Humans, vol. 56. Lyon,
France: International Agency for Research on Cancer.
Betina V (ed.) (1993) Chromatography of Mycotoxins:
Techniques and Applications, Journal of Chromatogra-
phy Library, vol. 54. London: Elsevier.
Coker RD (1997) Mycotoxins and their Control: Con-
straints and Opportunities, NRI bulletin 73. Chatham,
UK: Natural Resources Institute.
Coker RD and Jones BD (1988) Determination of my-
cotoxins. In: Macrae R (ed.) HPLC in Food Analysis.
London: Academic Press.
Horwitz W, Albert R and Nesheim S (1993) Reliability of
mycotoxin assays } an update. Journal of AOAC Inter-
national 76: 461.
Miller JD and Trenholm HL (1994) Mycotoxins in Grain
Compounds Other Than AUatoxin. St Paul, MN: Eagan
Press.
Scott PM (1998) Natural toxins. In: Cunniff (ed.) Of-
Tcial Methods of Analysis of AOAC International, 16th
edn, 4th revision. Washington: AOAC.
They are often found as contaminants of peanuts, tree
nuts, corn and cottonseed. They were discovered as
a result of investigations into Turkey X disease in
Britain, in which 100 000 turkeys and numerous
other poultry died as a result of feeding on peanut
meal which had been contaminated with mould.
1888 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography
hydrogen phosphate solution/methanol followed by
acetonitrile/water. Although OPA is used as the derivat-
ization reagent by the majority of laboratories, other
reagents have been employed including naphthalendial-
dehyde, Suoronitrobenzofurazan and Suorescamine.
The last reagent is unsatisfactory as it generates two
peaks in the HPLC chromatogram for fumonisin B1.
Three recently developed methods for the deter-
mination of the fumonisins in corn-based commodi-
ties are outlined in Table 13. Method 1 uses a
combination of SAX and C18 SPE clean-up prior to
ion-pair HPLC and Suorescence detection; on-line
derivatization within a reaction coil is employed. The
recovery of the fumonisins ranged from 54 to 110%
at 40 and 80
g kg\1, respectively. Method 2 is an
automated procedure using on-line immunoafRnity
clean-up, reversed-phase HPLC and electrospray ion-
ization MS detection. The protonated molecule signal
(m/z 722) was used to achieve a limit of quantiRca-
tion of 250 pg.
An HPTLC method (Method 3, Table 13), for the
determination of fumonisin B1 in rice, has recently
been reported. A novel derivatization step involved
the brief immersion of the HPTLC plate in a 0.16%
acidic solution of p-anisaldehyde, followed by quan-
tiRcation by scanning Suorodensitometry. The re-
sponse was linear over the range 0 to 5 mg kg\1 (ppm).
An intercomparison study on a variety of methods
for the determination of the fumonisins in maize has
recently been undertaken under the auspices of the
European Commission, Measurements and Testing
Programme. Twenty-four laboratories participated,
using their normal routine procedure for the deter-
mination of fumonisins B1 and B2 in the range 0.5}3.0
and 0.2}1.5 mg kg\1 (ppm), respectively. All labora-
tories used a similar method involving extraction
with methanol/water, clean-up with an SAX SPE col-
umn and HPLC Suorescence quantiRcation of the
OPA derivative. The intra- and interlaboratory pre-
cisions were high (10 and 11%, respectively, for
fumonisin B1; and 11 and 13%, respectively, for
fumonisin B2). However, the recoveries were low
(70$14% and 69$16% for fumonisins B1 and B2,
respectively). Interestingly, higher recoveries were
Thin-Layer (Planar) Chromatography
M. E. Stack, US Food and Drug Administration,
Washington DC, USA
Copyright ^ 2000 Academic Press
The aSatoxins are toxic and carcinogenic metabolites
of the moulds Aspergillus Uavus and A. parasiticus.
associated with extraction by shaking (85$12% for
fumonisin B1) rather than by blending (62$6%).
Conclusions
The continued use of a variety of chromatographic
procedures for the determination of mycotoxins is
envisaged. Although HPLC is the method of choice in
the developed world for a wide range of applications,
it is important that precise and accurate methods
continue to be developed that are appropriate to the
special needs of developing country laboratories.
See Colour Plate 53.
See also: II/Affinity Separation: Immunoaffinity Chrom-
atography. Chromatography: Gas: Detectors: Mass
Spectrometry. Chromatography: Liquid: Derivatization.
III/Aflatoxins and Mycotoxins: Thin Layer (Planar)
Chromatography. Membrane Preparation: Phase Inver-
sion Membranes.
Further Reading
Anon (1993) Some naturally occurring substances: food
items and constituents, heterocyclic aromatic amines
and mycotoxins. IARC Monographs on the Evaluation
of Carcinogenic Risks to Humans, vol. 56. Lyon,
France: International Agency for Research on Cancer.
Betina V (ed.) (1993) Chromatography of Mycotoxins:
Techniques and Applications, Journal of Chromatogra-
phy Library, vol. 54. London: Elsevier.
Coker RD (1997) Mycotoxins and their Control: Con-
straints and Opportunities, NRI bulletin 73. Chatham,
UK: Natural Resources Institute.
Coker RD and Jones BD (1988) Determination of my-
cotoxins. In: Macrae R (ed.) HPLC in Food Analysis.
London: Academic Press.
Horwitz W, Albert R and Nesheim S (1993) Reliability of
mycotoxin assays } an update. Journal of AOAC Inter-
national 76: 461.
Miller JD and Trenholm HL (1994) Mycotoxins in Grain
Compounds Other Than AUatoxin. St Paul, MN: Eagan
Press.
Scott PM (1998) Natural toxins. In: Cunniff (ed.) Of-
Tcial Methods of Analysis of AOAC International, 16th
edn, 4th revision. Washington: AOAC.
They are often found as contaminants of peanuts, tree
nuts, corn and cottonseed. They were discovered as
a result of investigations into Turkey X disease in
Britain, in which 100 000 turkeys and numerous
other poultry died as a result of feeding on peanut
meal which had been contaminated with mould.
 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography
Thin-layer chromatography (TLC) played a crucial
part in the discovery and subsequent research on the
aSatoxins and continues to play an important part in
the analytical methods used for control of aSatoxins
in food and feeds. The four major compounds are
aSatoxins B1, B2, G1 and G2. ASatoxins B1 and
B2 have bright blue Suorescence on TLC and G1 and
G2 are bright green-blue. ASatoxin B1 is found in the
largest amounts in samples and is also the most
toxic and carcinogenic of the four. ASatoxin M1 is
found in the milk of animals which have ingested
aSatoxin B1.
For the structures of the aSatoxins see Figure 1.
After the discovery of the aSatoxins other mycotoxins
were discovered and methods of analysis using TLC
have been devised for them.
Preparation of Samples
ASatoxin contamination of food and feeds is usually
in the range of ng g\1 to
g g\1. Sampling error is a
severe problem in aSatoxin determination because
only a few affected kernels can contaminate a
large amount of Rnished product. Amounts as high as
207 000 ng g\1 have been found in individual corn
kernels. This is sufRcient aSatoxin to produce
Figure 1 Structures of aflatoxins B1, B2, G1, G2 and M1.
1889
a level of contamination of 20 ng g\1 in a batch of
10 000 kernels of grain. Sampling plans have been
developed for various commodities. In general, the
larger the unit size of the commodity, the larger the
sample size should be. The sample should be Rnely
ground and mixed before taking out the analytical
test portion. Often the sampling error is larger than
the analytical error.
Various methods of analysis have been devised.
Many of these have been published in the OfTcial
Methods of Analysis of AOAC International, after
collaborative studies by several laboratories. If the
precision and accuracy of the results are acceptable
the method becomes ofRcial.
The three most widely used extraction and clean-
up methods for preparing aSatoxin extracts for TLC
are the CB method, the BF method and the im-
munoafRnity column method. The CB method,
named after the Contaminants Branch of the US Food
and Drug Administration (FDA), uses chloroform
extraction, Rltering through paper, addition to a silica
gel column, washing with hexane and ether, elution
with chloroform}methanol (97 : 3 v/v), and evapor-
ation to dryness to prepare the extract for TLC. The
BF method, named after the Best Foods Company,
uses methanol}water (55 : 45 v/v) extraction, hexane
1890 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography
defatting in a separatory funnel, partition into chloro-
form and evaporation to dryness to prepare the
extract for TLC. The immunoafRnity column method
uses methanol}water (7 : 3 v/v) for extraction, Rlter-
ing through paper, dilution with water, Rltering
through a glass microRbre Rlter, application to a col-
umn upon which antibodies to aSatoxins have been
bound, washing with water, elution with methanol
and evaporation to dryness to prepare the extract for
TLC.
The advantage of the CB method is that it is precise
and accurate when correctly performed. Disadvan-
tages include the acquisition and disposal costs of the
reagents used. The advantage of the BF method is that
it has the lowest cost of any of the methods. The
disadvantage is that it results in a somewhat dirtier
extract. The advantages of the immunoafRnity
column method are its simplicity of performance and
the purity of the aSatoxins in the extract. Its disad-
vantage is the high cost of the columns.
After evaporation, in all three methods, the extract
is carefully transferred using rinses of chloroform to
a small vial. The solvent is again evaporated to dry-
ness in a water bath under a stream of nitrogen. The
residue is dissolved in a small amount of solvent
(200
L), usually benzene}acetonitrile (98 : 2 v/v),
for spotting on TLC. Since the use of benzene is
sometimes prohibited because of its toxicity, other
solvents such as toluene}acetonitrile (9 : 1 v/v) may
be used as well.
A]atoxin Standards
A large source of error in the analysis is due to
incorrectly prepared aSatoxin standards. In a check
sample series, standards were found that contained
more or less than the stated amount of aSatoxins
when compared with a correctly prepared standard
sent out with the study. It is very important to work
with pure and accurate standards if accurate quantit-
ative and qualitative results are to be obtained.
ASatoxin standards may be purchased from chemical
distributors but need to be checked by means of TLC
or liquid chromatography (LC) to ensure that they are
pure. Small quantities of aSatoxin standards are
sometimes available from organizations such as the
FDA without cost.
Crystalline aSatoxin standards should be handled
in a glove box because of their carcinogenicity and
the electrostatic nature of the crystals. Because of the
difRculty in handling the crystalline material, the
aSatoxins are often received as dry Rlms deposited in
a precise amount in the bottom of a glass vial. The
contents of the vial should be dissolved in the solvent
(benzene}acetonitrile, 98 : 2 v/v) and mixed on a vor-
tex mixer for 1 min since the standards do not dis-
solve rapidly. After mixing, the solution is transferred
to a screw-cap vial and the ultraviolet spectra mea-
sured between 370 and 330 nm. The concentration
can then be calculated using the values listed in the
AOAC International OfRcial Method 971.22. The
standards are applied to a TLC plate to verify the
purity. The solutions must be stored in a closed and
sealed vial in a refrigerator at 4}83C. Mixtures of the
four major aSatoxins can be prepared by diluting the
concentrated stock solutions. The mixture most often
used contains aSatoxins B1 and G1 at 1.0
g mL\1
and B2 and G2 at 0.2
g mL\1. This ratio between the
four aSatoxins approximates to the ratio found in
some sample extracts. The aSatoxins in benzene}
acetonitrile (98 : 2 v/v) are stable when stored in
a closed and sealed vial in a refrigerator at 4}83C.
Evaporation or decomposition of the aSatoxins can
be detected using TLC or LC, shown by additional
spots or additional peaks or an increase or decrease in
the Suorescent intensity, as evidenced by unusual area
integration values from the LC detector or TLC den-
sitometer.
Spotting, Development and
Examination of the TLC Plate
The plates most often used for aSatoxin analyses are
20;20 cm glass plates, pre-coated with a 0.25 mm
layer of silica gel 60 (E. Merck, Darmstadt); plates
from other manufacturers may work equally well.
Spotting should be done in subdued incandescent
light to avoid photodecomposition of the aSatoxins.
Using a 10
L syringe, on an imaginary line 4 cm
from the bottom of the plate and 1 cm apart, 2, 5 and
two 10
L spots of the sample extract are applied
together with 2, 5 and 10
L spots of mixed aSatoxin
standards; 5
L of the standard is applied on top of
one of the 10
L spots of sample extract. It is possible
to spot four samples on to each plate.
The plate is developed for less than 90 min with
acetone}chloroform (1 : 9 v/v) until solvent is within
4 cm of the top of the plate. It may be necessary to
adjust the acetone}chloroform ratio to obtain opti-
mum resolution. The plate is removed from the tank
and air-dried in the hood in the dark.
Plates are examined under long wave ultraviolet
light at 365 nm in a cabinet equipped with a Rlter for
protecting the eyes from the ultraviolet light.
ASatoxins appear in order of decreasing RF: B1, B2,
G1 and G2. G1 and G2 are slightly greener than the
blue B1 and B2. The RF values for the aSatoxins in
the sample spots should be the same as those of the
standard spots. The aSatoxins in the sample spot
upon which the standard is superimposed should
 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography 1891

coincide exactly with the standard spots. The inten-

sity of the Suorescence of each of the sample spots
may be compared with that of the standard spots to
estimate the amount of aSatoxin present in the ex-
tract. Separate estimates need to be made for B1, B2,
G1 and G2. If the spots of the smallest portion of the
sample are more Suorescent than the strongest stan-
dard spot it is necessary to dilute the sample extract
and re-chromatograph. The plate may be run on
a densitometer equipped with an ultraviolet light
source set at 365 nm and an ultraviolet Rlter before
the photomultiplier detector. Connecting the TLC
densitometer to a computer permits the integration,
calculation, printing, and storage of results. If more
accurate quantitative results are necessary, the
extract can be re-diluted to a concentration approx-
imately equal to that of the standard and re-
chromatographed in the same manner as above. The
concentration of each aSatoxin in the extract can be
calculated using the formula:
ng g\1"(S;Y;V)/(X;W)
where S"
L of standard spot equal to sample;
Y"concentration of standard in ng
L\1; V"
L of
Rnal dilution of sample extract; X"
L of sample
spot equal to standard; and W"grams of sample
that the extract represents.
Not all blue Suorescent spots in the extracts are
necessarily aSatoxins. Sample extracts may contain
interferences, especially at the RF values of G1 and G2.
Respotting with an alternative solvent system such as
the top phase benzene}ethyl alcohol}water
(46 : 35 : 19 v/v) or with benzene}methanol}acetic
acid (90 : 5 : 5 v/v) often resolves the aSatoxins from
the interferences. Other solvents which are sometimes
used are: ether}methanol}water (96 : 3 : 1 v/v),
chloroform}acetone}water (88 : 12 : 1.5 v/v), or
chloroform}acetone}isopropanol}water (88 : 12 : 1.5 :
1 v/v).
Two-Dimensional TLC
Another powerful technique for resolving the
aSatoxins from interferences is two-dimensional
TLC. In this technique two spots of aSatoxin stan-
dards and one spot of sample extract are spotted on
the plate, as shown in Figure 2. The plate is Rrst
developed with ethyl ether}methanol}water
(96 : 3 : 1 v/v) in the Rrst direction. After develop-
ment and air drying, the plate is redeveloped in the
second direction with acetone}chloroform (1 : 9 v/v).
After development and air drying the plate is exam-
ined under ultraviolet light at 365 nm for aSatoxin
spots. A blue spot should appear at the intersection of
imaginary lines from the two standard spots. The
Figure 2 Two-dimensional TLC plate for aflatoxin analysis.
two-dimensional technique works well for difR-
cult materials such as eggs and spices.
Con\rmation of Identity of A]atoxins
To conRrm the identities of aSatoxins B1 and
G1 a technique has been devised which uses derivative
formation on the TLC plate. The sample extracts and
standards are spotted on the origin line of the plate
and 1
L amounts of triSuoroacetic acid are then
added to each spot. After reacting for 5 min, the
triSuoroacetic acid is removed by blowing air at
35}403C on the plate for 10 min. The triSuoroacetic
acid catalyses the addition of water across the double
bond in the terminal furan ring of aSatoxins B1 and
G1 to form the derivatives called aSatoxin B2a and
G2a, which give lower RF values than the parent com-
pounds. The plate is developed with chloro-
form}acetone (85 : 15 v/v). Upon examination of the
plate under ultraviolet light at 365 nm, sample and
standard will have low RF blue and green spots of the
derivatized aSatoxins. Since aSatoxin B2 and G2 do
not have the unsaturated double bond, they will be
unaffected by the test and will appear at their
normal RF values. For additional conRrmation the
plates can be sprayed with sulfuric acid}water
(1 : 3 v/v), which causes the aSatoxin spots to change
from blue or blue-green to yellow Suorescence.
Mass Spectrometric Con\rmation
of the A]atoxins
The aSatoxins may be conRrmed by negative ion
chemical ionization}mass spectrometry. The
aSatoxin is Rrst puriRed using preparative TLC. The
entire extract is applied along the origin line of a TLC
1892 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography
plate which is developed using chloroform}acetone
(9 : 1 v/v). After drying, the silica gel is scraped from
a band containing the aSatoxin B1. If the silica gel is
scraped into a sintered glass funnel the aSatoxin can
be eluted with chloroform}methanol (2 : 1 v/v). After
evaporation and re-dissolving in acetone, the
aSatoxin can be introduced into the inlet probe of the
mass spectrometer and spectra of sample and stan-
dard aSatoxin compared.
Methods of Analysis for A]atoxin M1
When cows consume aSatoxin in their feed, a small
percentage of it is metabolized and excreted in the
milk in the form of aSatoxin M1. ASatoxin M1 is also
toxic and carcinogenic, so methods have been de-
veloped to detect it in milk. Since infants and children
are major consumers of milk products, the levels of
concern for M1 in milk are set quite low by various
countries, in the range of 0.05}0.5
g L\1. Analyses
of milk and cheese samples at these low levels are
more difRcult. One method of analysis uses partition
from the milk into chloroform and silica gel column
clean-up before the TLC determination. Another
method used extraction from the milk on to
a C18 solid-phase extraction column and clean-up on
a silica gel column before TLC or LC determination.
An immunoafRnity column clean-up can also be used.
TLC is accomplished on 10;10 cm or 20;20 cm,
0.25 mm layer thickness silica gel 60 plates developed
with chloroform}acetone}isopropanol (87 : 10 :
3 v/v). Other solvent systems which have been used
are ether}methanol}water (95 : 4 : 1 v/v) and
ether}hexane}methanol}water (87 : 10 : 4 : 1 v/v).
A two-dimensional TLC method for aSatoxin
M1 has been developed for liver but also works for
milk and cheese extracts. The plate is spotted in
a similar manner to the two-dimensional plate for
aSatoxin B1 and developed in the Rrst direction with
ether}methanol}water (95 : 4 : 1 v/v) and after devel-
opment and drying is developed in the second direc-
tion with chloroform}acetone}isopropanol (87 : 10 :
3 v/v). The developed plate is examined under ultra-
violet light at 365 nm for a blue spot at the intersec-
tion of imaginary lines from the two standard spots.
The conRrmatory technique using triSuoroacetic
acid works for aSatoxin M1 as well but is performed
using two-dimensional TLC and requires heating the
plate in an oven at 753C for the reaction to occur.
TLC Determination of Other
Mycotoxins
Mycotoxins can be generated by a large number of
mould species. Several books review the incidence
and toxicity of the most common mycotoxins. The
interest of the regulatory authorities has been focused
on relatively few of these metabolites that cause prob-
lems in human and animal health. The mycotoxins of
regulatory interest are currently the aSatoxins, och-
ratoxin A, patulin, fumonisins, deoxynivalenol, other
trichothecenes and zearalenone. TLC procedures are
described below for these mycotoxins.
Ochratoxin A
Ochratoxin A (Figure 3) is a metabolite of some As-
pergillus and Penicillium species. It is found as a con-
taminant of barley, corn, wheat, oats and coffee.
It has also been found in meat, human blood and
human milk. Ochratoxin A causes porcine neph-
ropathy, notably in some Scandinavian countries
when contaminated barley is fed to swine. Och-
ratoxin A is extracted from samples with chloroform
in the presence of phosphoric acid and cleaned up
using partition into sodium bicarbonate and
C18 solid-phase extraction. In a similar manner to the
TLC of aSatoxin discussed above, ochratoxin A is
spotted on a plate pre-coated with a 0.25 mm layer of
silica gel 60 (E. Merck, Darmstadt) and developed
with benzene}methanol}acetic acid (18 : 1 : 1 v/v) or
toluene}ethyl acetate}formic acid (5 : 4 : 1 v/v). After
drying, the plate is examined under long and short
wave ultraviolet light (365 and 254 nm). Ochratoxin
A (RF"0.65) Suoresces brightest under long wave
ultraviolet light and is usually accompanied by the
less toxic ochratoxin B (RF"0.5) which Suoresces
brightest under short wave ultraviolet light. The Su-
orescence of the ochratoxins can be enhanced by
spraying the plate with alcoholic sodium bicarbonate
solution which changes them to their more Suor-
escent salt forms. The ochratoxins can be conRrmed
by ester formation using boron triSuoride in ethanol
and re-chromatographing using the same conditions
as above. The ethyl esters appear at lower RF values
than the parent compounds under long and short
wave ultraviolet light.
Patulin
Patulin (Figure 3) is a lactone metabolite of several
moulds, including Penicillium expansum, which
causes brown rot in apples. Patulin is often found in
apple juice, especially juice from fallen apples. Patulin
can be extracted from apple juice with ethyl acetate
and cleaned up using silica gel column chromatogra-
phy. After evaporation, the extract is dissolved in
chloroform and spotted on silica gel plates and de-
veloped with toluene}ethyl acetate}formic acid
(5 : 4 : 1 v/v). After drying, the plate is sprayed
 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography 1893

Figure 3 Structures of ochratoxin A, patulin, fumonisin B1, deoxynivalenol, T-2 toxin, diacetoxyscirpenol, satratoxin H and
zearalenone.

with 3-methyl-2-benzothiazolinone hydrazone}HCl chloroform}acetone (9 : 1 v/v) can be used to conRrm

(MBTH) solution and heated for 15 min in an oven at
1303. Under ultraviolet light at 365 nm, patulin
(RF"0.5) appears as a yellow-brown Suorescent
spot. The amount of patulin in the sample can be
determined by comparing the intensity of Suores-
cence of the standard and sample spots. Other
TLC developers, such as hexane}anhydrous ether
(1 : 3 v/v), chloroform}methanol (95 : 5 v/v), and
the identity of the patulin. After development, plates
are sprayed with MBTH to reveal the patulin.
Fumonisins
Fumonisins B1 (Figure 2) and B2 are metabolites of
Fusarium moniliforme and F. proliferatum. They are
common natural contaminants of corn and have
1894 III / AFLATOXINS AND MYCOTOXINS / Thin-Layer (Planar) Chromatography
caused deaths in horses and swine. Small amounts
have been found in cornmeal and breakfast cereals. In
order to ensure that they are not present in food in
excessive amounts, methods of analysis have been
developed. Most methods use LC after formation of
derivatives of the primary amine function with re-
agents such as o-phthaldialdehyde. However, a rever-
sed-phase TLC determination has been devised.
The fumonisins, dissolved in acetonitrile}water
(1 : 1 v/v), are spotted at the origin of a 10;10 cm
C18 plate and developed with methanol}1% KCl in
water (3 : 1 v/v). After drying, the plates are sprayed
with 0.1 mol L\1 sodium borate (pH 8}9) followed
by Suorescamine (0.4 mg mL\1 in acetonitrile). After
a 1 min delay, further spraying with 0.01 mol L\1
boric acid}acetonitrile (2 : 3 v/v) is carried out. Ex-
amination under 365 nm ultraviolet light reveals Su-
orescent yellow spots of fumonisin B1 (RF"0.5) and
fumonisin B2 (RF"0.1).
Deoxynivalenol
Deoxynivalenol (Figure 2), also called vomitoxin, is
a trichothecene metabolite of F. graminearum, an
organism which causes a disease in barley and wheat
called head blight or scab. Deoxynivalenol is found as
a contaminant of barley, wheat, corn and rye and
causes adverse health effects in animals and hu-
mans, including feed refusal and vomiting in swine.
An advisory level of 1
g g\1 has been set for Rnished
wheat products. Methods for analysis have been de-
vised using LC and TLC. The TLC method uses
acetonitrile}water (84 : 16 v/v) extraction and clean-
up using a charcoal}alumina}Celite (7 : 5 : 3 v/v) col-
umn. The extracts and standard deoxynivalenol are
dissolved in methanol and spotted near the 20 cm
edge of a 20;10 cm Linear-K High Performance
(Whatman, Clifton NJ) or equivalent silica gel plate
and developed with chloroform}acetone}2-propanol
(8 : 1 : 1 v/v). After drying the plate is sprayed with
aluminium chloride solution (20 g AlCl3.6H2O in
100 mL methanol}water: 1 : 1 v/v) and then heated
in an oven at 1203 for 7 min. Under 365 nm ultra-
violet light, deoxynivalenol appears as a blue Suor-
escent spot at RF"0.78. Spots may be scanned with
a densitometer.
Other Trichothecenes
The trichothecenes are a large group of fungal meta-
bolites produced by various species of Fusarium,
Myrothecium, Stachybotrys, Verticimonosporium,
Cylindrocarpon, Trichoderma and Tricothecium.
They have been implicated in numerous farm-animal
poisonings. A human disease called alimentary toxic
aleukia occurred in the former Soviet Union during
World War II when grains were eaten after they had
lain out in the Reld under snow during the winter.
Fusarium moulds isolated from these grains were
shown to produce large amounts of T-2 toxin (Fig-
ure 3) and related derivatives. T-2 toxin and related
compounds can be analysed by silica gel TLC using
chloroform}methanol (9 : 1 v/v) as the developer.
Since the trichothecenes are colourless and do not
Suoresce, it is necessary to spray the developed plate
with sulfuric acid}methanol (1 : 1 v/v), heat for
10 min at 1003C and examine the plate under 365 nm
ultraviolet light. Trichothecenes of the T-2 group will
appear as blue Suorescent spots, T-2 at RF"
0.64, diacetoxyscirpenol (Figure 3) at RF"0.60,
neosolaniol at RF"0.39, dihydroxydiacetoxy scir-
penol at RF"0.32 and HT-2 toxin at RF"0.31.
Other trichothecenes of the nivalenol group do not
form Suorescent derivatives with sulfuric acid but
instead give a dark pink to brown spot when the plate
is examined under visible light. A more useful detec-
tion procedure for these compounds is spraying with
4-(p-nitrobenzyl)-pyridine (NBP: 1% in chloroform),
heating for 30 min at 1503C and spraying with tet-
raethylenepentamine (TEPA). Under visible light
a plate developed with chloroform}methanol
(9 : 1 v/v) will have blue spots of fusaranon-X at
RF"0.36, and dihydronivalenol at RF"0.07.
A plate developed with benzene}acetone (1 : 1 v/v)
will have tetraacetylnivalenol at RF"0.62, crotocin
at RF"0.59, dihydronivalenol at RF"0.07 and
nivalenol at RF"0.09.
Another type of trichothecene is a series of macro-
cyclic di- and trilactonic derivatives of verrucarol.
These have been implicated in a disease of horses and
other farm animals called stachybotryotoxicosis. Re-
cently they are suspected of contributing to the death
of some infants exposed to the air in mouldy houses.
Since they contain an ultraviolet-absorbing func-
tional group, they can be detected by using silica gel
plates which Suoresce under 254 nm ultraviolet light.
The satratoxins appear as dark spots on a white
background. If developed with chloroform}2-pro-
panol (99 : 1 v/v) roridin E will appear at RF"0.85,
verrucarin J at RF"0.45, satratoxin F at RF"0.40,
satratoxin G at RF"0.20 and satratoxin H (Fig-
ure 3) at RF"0.15.
Zearalenone
Zearalenone (Figure 3) is a metabolite of the mould
F. graminearum also called by its perfect name Gib-
berella zeae. Zearalenone is found in barley, wheat
and corn and causes hyperoestrogenism in swine,
resulting in infertility and spontaneous abortions. It

sometimes co-occurs with deoxynivalenol. Zeara-
lenone can be extracted from grain with chloroform
and cleaned up using a silica gel column, followed by
defatting by partitioning between hexane and
acetonitrile. For TLC the samples and standards are
dissolved in benzene and spotted on a silica gel and
developed with ethanol}chloroform (5 : 95 v/v) or
acetic acid}benzene (5 : 95 v/v). Under 254 nm ultra-
violet light, zearalenone appears as a greenish-
blue Suorescent spot at RF"0.5. If the plate is
sprayed with an aluminium chloride solution and
heated for 5 min at 1303C, zearalenone will appear
under 365 nm ultraviolet light as a blue Suorescent
spot.
Summary
TLC methods have been developed to analyse for
a variety of mycotoxins in the commodities in which
they occur. TLC densitometric determinations pro-
vide precise quantitative data at ng g\1 to
g g\1
levels. The major advantages of TLC over LC are
its low cost and its use as a screening tool. The
commercial availability of precoated TLC plates,
including silica gel, reversed-phase and high perfor-
mance plates has resulted in expanded applications
in the mycotoxin Reld. The role of TLC in the analysis
of mycotoxins will continue for the foreseeable
future.
See also: II/Chromatography: Thin-Layer (Planar):
Historical Development; Preparative Thin-Layer (Planar)
AIR LIQUEFACTION: DISTILLATION
R. Agrawal and D. M. Herron, Air Products
and Chemicals, Hamilton Boulevard, Allentown,
PA, USA
Copyright ^ 2000 Air Products and Chemicals, Inc
Oxygen, nitrogen and argon, the major components
of air, have been separated by distillation at cryogenic
temperatures for nearly a century. Air was commer-
cially liqueRed as early as 1895 by Carl von Linde and
also by William Hampson. Linde separated oxygen
from air by distillation in a single column in 1902.
A commercial plant producing pure nitrogen was
already in operation by 1904. The Rrst double-col-
III / AIR LIQUEFACTION: DISTILLATION 1895
Chromatography. III/Aflatoxins and Mycotoxins: Chrom-
atography. Immunoaffinity Extraction.
Further Reading
Bullerman LB and Draughon FA (eds) (1994) Fusarium
moniliforme and Fumonisin symposium. Journal of
Food Protection 57: 513}546.
Cole RJ and Cox RH (eds) (1981) Handbook of Toxic
Fungal Metabolites. New York: Academic Press.
Eaton DE and Groopman JD (eds) (1994) The Toxicology
of AUatoxins. San Diego: Academic Press.
Purchase IFH (ed.) (1974) Mycotoxins. Amsterdam: Elsevier.
Rodricks JV (ed.) (1976) Mycotoxins and Other Fungal
Related Food Problems. Advances in Chemistry Series
149. Washington, DC: American Chemical Society.
Rodricks JV, Hesseltine CW and Mehlman MA (eds)
(1977) Mycotoxins in Human and Animal Health. Park
Forest South, IL: Pathotox.
Scott PM (ed.) (1995) Chapter 49, Natural toxins. In:
Cunniff P (ed.) OfTcial Methods of Analysis of
AOAC International, 16th edn., Gaithersburg. MD:
AOAC International.
Stack, ME (1996) Toxins. In: Sherma J and Fried B (eds)
Handbook of Thin-layer Chromatography. New York:
Marcel Decker.
Steyn PS (ed.) (1980) The Biosynthesis of Myco-
toxins. New York: Academic Press.
Touchstone JC (ed.) (1982) Advances in Thin Layer
Chromatography. New York: Wiley.
Whitaker TB, Springer J, DeRze PR et al. (1995) Evaluation
of sampling plans used in the United States, United
Kingdom, and the Netherlands to test raw shelled pea-
nuts for aSatoxin. Journal of AOAC International 78:
1010}1018.
umn distillation system, the predecessor to current
double-column processes, was commissioned in 1910
by Linde. Argon was produced on an industrial scale
by 1913. Today the major industrial companies sup-
plying products from air distillation and liquefaction
and also the equipment for this purpose are: AGA,
Air Liquide, Air Products and Chemicals, the BOC
Group, Linde, Messer Group, Nippon Sanso and
Praxair.
The composition of dry and impurities-free air is
given in Table 1. The critical temperature and normal
boiling point (at 101.3 kPa) for each component is
also listed. In this table, and in the rest of this chapter,
concentration in p.p.m. refers to parts per million on
a volume basis. The gases listed in Table 1 are used in
1896 III / AIR LIQUEFACTION: DISTILLATION

Table 1 Composition of air and thermodynamic properties of its constituent gases

Constituent gas Concentration Boiling temperature Critical temperature

(mol%) (3C)a (3C)

Nitrogen 78.12 !195.8 !146.9

Oxygen 20.95 !182.9 !118.8
Argon 0.93 !185.9 !122.4
Neon 18 p.p.m. !246.1 !228.8
Helium 5.3 p.p.m. !268.9 !267.9
Krypton 1.1 p.p.m. !153.4 !63.8
Xenon 0.08 p.p.m. !108.1 !16.6

a
Boiling temperature at 101.3 kPa.

a wide range of industrial and medical applications.
Typical industries using these gases include: ferrous
and nonferrous metals, chemicals, petroleum, food,
paper, glass, textile and electronics. Oxygen is gener-
ally used as an oxidant while nitrogen and argon are
used to provide inert atmospheres. Krypton is used in
light bulbs, lasers, sputtering of electronic compo-
nents and high energy physics. Neon is used in Suor-
escent lighting, infrared detection equipment and ex-
perimental physics at cryogenic temperatures. Xenon
is used in electronic Sashlights, as an anaesthetic and
in a new application where an on-board xenon ion
propulsion system is used for positioning a satellite.
Helium is not generally recovered from air due to its
low concentration.
The history of air distillation started with oxygen
production followed by recovery of other constitu-
ents. Therefore, the distillation processes to produce
oxygen are described Rrst here followed by argon and
then nitrogen. These topics are followed by liquefac-
tion processes and Rnally a brief description is given
of the major equipment used in cryogenic air separ-
ation and liquefaction processes.
Distillation
Distillation of Air to Recover Oxygen
The basic steps of any cryogenic air distillation pro-
cess are shown in Figure 1.

Figure 1 Basic steps in a cryogenic air distillation plant.

 III / AIR LIQUEFACTION: DISTILLATION 1897

Air is Rrst compressed in a multistage compressor A simpliRed sketch of this single column arrangement
and cooled to near ambient temperature. Given the is shown in Figure 2.
boiling point temperatures of nitrogen and oxygen in Compressed, cleaned and cooled air that is near its
Table 1, it is clear that air has to be cooled to ex- dewpoint temperature is Rrst condensed in a reboiler-
tremely low temperatures before it can be distilled. It condenser located in the bottom (sump) of the distil-
follows that a number of impurities present in air and lation column. The condensed liquid air is then re-
which would freeze at such cryogenic temperatures duced in pressure across a valve and fed to the top of
must be removed to avoid plugging of heat exchange the distillation column. The stream provides both the
and separation equipment. Typical impurities that are feed and the liquid reSux. From Table 1 it can be seen
not listed in Table 1 but are present in air include: that, of the three major components in air, nitrogen is
water (after compression air is saturated); carbon the most volatile, oxygen the least volatile and argon
dioxide (about 375 p.p.m.); hydrocarbons such as is of intermediate volatility. As a result, the liquid
acetylene (0.1}1 p.p.m.), methane (2}10 p.p.m.) and descending the distillation column becomes enriched
some higher hydrocarbons in varying concentrations in oxygen. The vapour needed for distillation is pro-
(ethylene, propylene, ethane, etc.); carbon monoxide; vided by boiling the oxygen-enriched liquid in the
nitrogen oxides and sulfur compounds. Therefore, in sump by heat exchange against the condensing air in
the second step, compressed air is sent through a puri- the reboiler-condenser. A portion of the vapour rising
Rcation system at least to remove impurities such as from the sump is collected as gaseous oxygen product
water, carbon dioxide, acetylene, nitrogen oxides and while the rest is allowed to travel up the distillation
sulfur compounds. column. As vapour ascends the distillation column, it
In the third step, the compressed and cleaned air is becomes enriched in nitrogen and Rnally leaves from
cooled to near its dewpoint by heat exchange. Finally, the top of the column as a nitrogen-enriched waste
the cooled air is sent to an appropriate distillation stream. If needed, a liquid oxygen product stream is
column system. Here air is distilled into at least two collected from the sump of the distillation column.
product steams } one stream is enriched in oxygen Even when liquid oxygen is not a desired product,
and the other is enriched in nitrogen. Both of these a very small quantity of liquid oxygen is continuously
streams are then warmed to near ambient temper- withdrawn from the bottom of the distillation col-
ature by countercurrent heat exchange with the in- umn to avoid accumulation of hydrocarbons in the
coming air. When a product stream is required at sump. Both the gaseous oxygen product stream and
a higher pressure, it is further compressed. Liquid nitrogen-rich vapour stream are then warmed to near
products such as liquid oxygen, liquid nitrogen or
liquid argon can also be produced from the distilla-
tion column system and sent to liquid storage for later
distribution.
The heat exchanger and the distillation system are
enclosed in a well-insulated enclosure called the cold
box. Despite the insulation, there is heat leakage and
therefore, refrigeration is provided to the cold box to
keep the inside equipment cold. For this purpose,
modern cryogenic plants employ turbo-expanders that
are also enclosed in the cold box. These turbo-expan-
ders produce work out of the cold box and keep all
the equipment at the desired cryogenic temperatures.
Now that the basic steps that are common to all the
air distillation plants have been described, attention
will be paid to the distillation of air after it is cooled
to near its dewpoint temperature. The distillation of
air is at the heart of an air separation plant and its
arrangement varies with the number, quantity and
purity of products being produced.
The early developments in air distillation to pro-
duce oxygen were propelled by the invention of the
oxyacetylene blow torch for welding and steel cut-
ting. In 1902, Carl von Linde introduced the Rrst air
distillation process using a single distillation column. Figure 2 A single column to produce oxygen.
1898 III / AIR LIQUEFACTION: DISTILLATION

ambient temperature by heat exchange against the

incoming air stream (Figure 1). While the early plants
produced oxygen at a purity of 80}90%, the single-
column process can provide oxygen at any desired
higher purity. Generally, the purity of oxygen used in
metal welding and cutting is 99.5% or greater.
The problem with the single distillation column
process shown in Figure 2 is that the recovery of
oxygen is low. The reason is that the minimum con-
centration of oxygen in the nitrogen-rich vapour
stream leaving the top of the distillation column is
limited to that value which is in equilibrium with the
liquid air that is fed at the top. Since the concentra-
tion of oxygen in air is fairly high, a sizeable fraction
of the oxygen in the feed air leaves in the nitrogen-
rich vapour stream. To illustrate this for a distillation
column operating at 1.4 atm and producing 99.5%
oxygen, the pressure of feed air is about 5 atm and
a vapour stream in thermodynamic equilibrium with
the liquid air stream (at 1.4 atm) will be 6.9% oxy-
gen. A typical oxygen recovery from such a distilla-
tion column would only be in the neighbourhood of
14 mol of oxygen per 100 mol of feed air.
It is clear that, for higher recoveries of oxygen, the
concentration of oxygen in the nitrogen-rich vapour
stream leaving the top of the distillation column must
be low. In other words, both product streams should
be relatively pure. This requires that the liquid reSux
to the top of the distillation column should also be
relatively pure. It seems that Georges Claude was the
Figure 3 A double-column arrangement to produce oxygen.
Rrst to provide the solution by using his dephlegma-
tion equipment in 1903. However, it was Carl von
Linde’s double distillation column of 1910 that revol- sure column, crude liquid oxygen is distilled to pro-
utionized the industry and is still the workhorse of the duce a nitrogen-rich vapour stream at the top and an
modern cryogenic oxygen plants. oxygen product stream at the bottom. The boil-up at
A typical double distillation column conRguration the bottom of the low pressure column is provided by
is shown in Figure 3. vaporizing the liquid oxygen stream in the sump by
In this arrangement, compressed, cleaned and heat exchange against the condensing nitrogen
cooled air is now sent to a high pressure distillation vapour stream from the top of the high pressure
column that operates at about 6 atm. As the vapour column. A portion of the vapour from the reboiler-
rises up this high pressure column, it is enriched in condenser is recovered as gaseous oxygen product
nitrogen and at the top of the column, the concentra- while the rest rises to perform distillation in the low
tion of oxygen has been reduced to an extremely low pressure column. When needed, some liquid oxygen
level. The nitrogen vapour is condensed by heat ex- can also be recovered from the sump as product.
change in a reboiler-condenser. Of this condensed In a double-column arrangement, the main purpose
nitrogen stream, about 60% is returned back to the of the high pressure column is to distil and provide
top of the high pressure column as liquid reSux; two saturated liquid streams from the feed air } a
approximately 40% of the Sow is sent to the top of liquid nitrogen reSux and a crude liquid oxygen feed.
a low pressure column that operates at around It is in the low pressure column that the crude liquid
1.4 atm. The liquid descending the high pressure col- oxygen is distilled to provide the needed oxygen
umn becomes enriched in oxygen to produce crude product stream. The liquid nitrogen stream provides
liquid oxygen leaving the bottom (typically around the much needed liquid reSux at the top of the
35% oxygen). This crude liquid oxygen is reduced in low pressure column. By using sufRcient stages of
pressure across a valve and fed to an intermediate separation in the high pressure column, the concen-
location in the low pressure column. In the low pres- tration of oxygen in the liquid nitrogen stream can

often be reduced to p.p.m. level. Therefore, the con-
centration of oxygen in the nitrogen-rich vapour
stream from the top of the low pressure column is
reduced to extremely low levels. This not only allows
the potential to recover the nitrogen-rich vapour
stream as a useful product stream, but also makes
very high recoveries of oxygen possible. For a double-
column process, production of 99.5% oxygen in ex-
cess of 20.5 mol per 100 mol of feed air (maximum
oxygen content being 20.95 mol) is quite common.
For most uses, gaseous oxygen is needed at pres-
sures greater than atmospheric pressure. This pres-
sure can range from about 2 atm absolute pressure
for glass-making to pressures in the range of
30}80 atm for the gasiRcation of hydrocarbons such
as coal and petroleum residuum. One obvious
method to deliver pressurized oxygen is to compress
gaseous oxygen to the desired pressure after it exits
the cold box. However, safety considerations tend to
make the equipment associated with oxygen com-
pressors expensive. In certain applications, where
both oxygen and nitrogen are needed at higher pres-
sures, one has the option of increasing the pressure of
the distillation columns and directly produce both
products at elevated pressures. Unfortunately, the
low pressure column is seldom operated at pressures
greater than 8 atm absolute. This is because the pres-
sure of the high pressure column is typically greater
than two to three times the pressure of the low pres-
sure column and distillation in the high pressure col-
Figure 4 Pumped liquid oxygen flowsheet.
III / AIR LIQUEFACTION: DISTILLATION 1899
umn must be conducted at a pressure that is sufR-
ciently lower than the critical pressures of nitrogen
and oxygen. A third method that is becoming more
popular is the use of a pumped liquid oxygen process.
This method is also sometimes referred as internal
oxygen compression. A schematic of such a plant is
shown in Figure 4.
In the pumped liquid oxygen process of Figure 4,
air is compressed in a multistage air compressor
(MAC) to about 6 atm absolute pressure, cooled to
near ambient temperature and then sent to a molecu-
lar sieve puriRer. About 70% of the cleaned air is
directly fed to the cold box for cooling in the main
heat exchanger. From an intermediate location of the
main heat exchanger, corresponding to 1003C to
!1303C, approximately 10}20% of this Sow is
withdrawn and expanded in a turbo-expander to
a pressure slightly above atmospheric pressure and
fed to an intermediate location of the low pressure
column. The work extracted from the turbo-expan-
der provides the needed refrigeration for the cold
box. The air that remains after the withdrawal of
expander feed is cooled to near the dewpoint temper-
ature and is fed to the bottom of the high pressure
column. The arrangement of the double distillation
column process is the same as discussed in Figure 3,
with only two differences. The Rrst is that the
liquid nitrogen stream from the top of the high
pressure column is cooled in a subcooler heat ex-
changer against the nitrogen-rich vapour streams.
1900 III / AIR LIQUEFACTION: DISTILLATION
This increases the fraction of liquid in this stream as
its pressure is reduced to that of the low pressure
column. This technique, which is commonly used in
any oxygen plant to increase liquid nitrogen reSux to
the low pressure column, has the beneRcial effect
of increasing the purity and recovery of products.
The second major difference is the withdrawal
of oxygen product from the low pressure column as
liquid and its subsequent vaporization. The liquid
oxygen is pumped in a liquid oxygen pump to the
desired oxygen product pressure. This pumped liquid
oxygen is then vaporized in the main heat exchanger.
In order to maintain refrigeration balance, it is essen-
tial that another stream be condensed through heat
exchange as the pumped liquid oxygen is being va-
porized. For this purpose, about 30% of the cleaned
air is further boosted to a higher pressure in a booster
compressor. The pressure of the boosted air is chosen
such that it would easily condense through heat ex-
change with the vaporizing oxygen stream. Generally
the pressure of the condensing air stream is much
greater than the oxygen stream. The condensed liquid
air from the main heat exchanger is appropriately fed
to either one or both of the distillation columns. The
warmed gaseous oxygen stream provides the desired
pressurized oxygen product.
While the early oxygen plants produced only a few
tons of oxygen per day, modern plants are capable of
producing in excess of 3000 tons per day of oxygen in
a single train.
Distillation of Air to Recover Argon
After nitrogen and oxygen, argon is the most
abundant component in air. Its inert property is
quite attractive for metals and several other material-
processing applications. Within a very short time of
the discovery of the double-column system, argon
was distilled from air in 1913. The distillation ar-
rangement to produce argon in modern plants was
generally described in a German patent by 1935.
The arrangement for argon production starts with
an examination of the argon concentration proRle in
the low pressure column of a double-column process.
From the normal boiling temperatures listed in
Table 1, it is readily observed that the volatility of
argon is between that of nitrogen and oxygen and,
furthermore, it is closer to oxygen than nitrogen. As
a result, the concentration of argon in the liquid
nitrogen stream from the high pressure column (Fig-
ures 3 and 4) is at p.p.m. level and virtually all the
argon is contained in the crude liquid oxygen. There-
fore, the bulk of the argon in air enters at an inter-
mediate point on the low pressure column. When
oxygen containing less than 0.5% argon is produced,
argon is forced to escape from the top of the low
pressure column in the nitrogen-rich vapour stream.
However, the liquid nitrogen reSux stream is virtually
free of argon and tends to drive argon down the low
pressure column. Consequently, the concentration of
argon in the vapour phase at an intermediate location
between the crude LOX feed and the oxygen product
withdrawal point reaches levels approaching 20%.
A typical vapour-phase argon concentration proRle in
the low pressure column is shown in Figure 5. The
build-up of argon provides an opportunity to with-
draw a side vapour stream from near the location
where a peak in argon concentration occurs and to
distil it further in a side distillation column to pro-
duce concentrated argon.
A typical argon recovery arrangement is shown in
Figure 6. An argon-rich vapour stream containing
between 10 and 25% argon, p.p.m. levels of nitrogen
and the rest oxygen is withdrawn from an intermedi-
ate location of the bottom section of the low pressure
column and is fed to the bottom of a side argon
column. The Sow of this vapour stream is about 20%
of the feed air. As vapour ascends the side argon
column, it is depleted in oxygen. The development of
structured packing for cryogenic distillation columns
has allowed the modern cryogenic plants to use in
excess of 175 theoretical stages of separation in the
side argon column. As a result, the vapour at the top
of this column can contain only p.p.m. levels of oxy-
gen. This vapour stream is condensed in a reboiler-
condenser; most of it is returned back as reSux to the
side argon column and a small portion is recovered as
a crude argon stream. The liquid to vapour Sow ratio
in this column is in the neighbourhood of 0.97. The
liquid from the bottom of the side argon column is
Figure 5 Vapour-phase composition in the low pressure
column.

Figure 6 Distillation arrangement for argon separation from air for the process shown in Figure 4.
pumped back to the argon-rich vapour draw location
of the low pressure column. Condensation at the top
of the side argon column is provided by boiling a por-
tion of the crude liquid oxygen at nearly the low
pressure column pressure, as shown in Figure 6. The
vaporized crude liquid oxygen stream is fed to the
low pressure column a few stages of separation below
the location where the unboiled crude liquid oxygen
is fed. The recovery of argon from cryogenic air is
easily in the range of 70}85% and occasionally, if
needed, it can be as high as 95% of the total argon
contained in the feed air.
While the concentration of oxygen in the crude
argon recovered from the arrangement in Figure 6 is
below 5}100 p.p.m., the nitrogen concentration is
much higher, as nearly all the nitrogen contained in
the argon-rich vapour stream shows up in the crude
argon. Generally, the argon product speciRcation re-
quires that the nitrogen concentration also be below
5 p.p.m. Therefore, the crude argon stream is sub-
sequently distilled in a pure argon column with separ-
ation stages both below and above the feed. Boil-up
and condensing duties for this column are extremely
III / AIR LIQUEFACTION: DISTILLATION 1901
low and are easily provided by using side streams that
are withdrawn from one or more appropriate process
streams, shown in Figure 6, such as crude liquid oxy-
gen, high pressure liquid air or high pressure nitrogen
vapour. A waste vapour stream containing all the
nitrogen is withdrawn from the top of the pure argon
column and pure liquid argon product is collected
and sent to a storage tank from the bottom of this
column.
Distillation of Air to Recover Nitrogen
In most industrial applications, nitrogen is used as an
inert gas. Cryogenic air separation can easily produce
nitrogen gas with concentrations of oxygen below
5 p.p.m. Until the 1950s, the demand for nitrogen
was low. Supply could easily be met by withdrawing
a portion of nitrogen vapour from the top of the high
pressure column as a co-product of a double-column
process for oxygen production (for example, see Fig-
ure 6). Generally, up to 30% of the feed air can be
recovered as high pressure nitrogen product from the
top of the high pressure column. When needed, a por-
tion of the nitrogen-rich vapour stream from the top
1902 III / AIR LIQUEFACTION: DISTILLATION
of the low pressure column can also be recovered as
a useful product.
In the 1960s industrial demand for nitrogen in-
creased, and this led to a need for plants that were
designed solely for nitrogen with no co-production of
oxygen. For most applications, nitrogen product is
required at a pressure between 6 and 10 atm. There
are two basic schemes for nitrogen separation from
air: one uses a single column while the other uses two
columns (similar to the double-column process
for oxygen production). The single-column process
is used for relatively small size nitrogen plants
(up to about 500 tons per day of nitrogen) and
two-column processes are used for larger size plants.
Cold boxes can now be designed to produce as much
as 10 000 tons per day of nitrogen in a single
train.
A single-column process for nitrogen separation is
shown in Figure 7.
Feed air is compressed to a pressure in excess of
about 5 atm absolute, cleaned of impurities in the
molecular sieve puriRer and cooled to near its dew-
point in the main heat exchanger by heat exchange
against the returning streams. The cooled air stream
is then fed to the bottom of a single column. SufR-
cient separation stages are used in this column to
attain the desired purity at the top of this column.
A portion of the nitrogen vapour from the top is
withdrawn and warmed in the main heat exchanger
to provide the desired nitrogen product. The rest of
Figure 7 A single distillation column process for nitrogen production.
the nitrogen vapour stream is condensed in a reboiler-
condenser and returned as reSux to the column.
The ratio of liquid to vapour Sow rates in the column
is in the neighbourhood of 0.6. The crude liquid
oxygen stream from the bottom of the column is
reduced in pressure and vaporized in the reboiler-
condenser. The vaporized stream is partially warmed
in the main heat exchanger and then expanded in
a turbo-expander to near atmospheric pressure to
provide the needed refrigeration for the plant. The
expanded stream is then warmed to near ambient
temperature in the main heat exchanger and eventual-
ly discharged as an oxygen-rich waste stream. The
concentration of oxygen in the waste stream is ap-
proximately 35%. The Sow rate of the nitrogen prod-
uct stream is about 40}50 mol per 100 mol of the
feed air.
The main problem with the single-column process
is that the crude liquid oxygen leaving from the bot-
tom of the column is at best, in thermodynamic equi-
librium with the feed air. This means that there is
a lower limit to the concentration of nitrogen in the
crude liquid oxygen stream. This limits the recovery
of nitrogen. For higher recoveries of nitrogen and
more efRcient processes, it is essential that the
crude liquid oxygen stream be further distilled to
recover the contained nitrogen. Figure 8 shows such
a two-column process.
The major difference between this process and
the single-column process of Figure 7 is that now

Figure 8 A two-column process fo nitrogen production.
crude liquid oxygen from the bottom of the high
pressure column is fed to an intermediate location of
a low pressure column for further distillation. A low
pressure nitrogen vapour stream is recovered from
the top of the low pressure column as a second prod-
uct stream. Another portion of this low pressure
nitrogen stream is condensed in the top reboiler-
condenser and returned as the major reSux stream to
the top of the low pressure column. Optionally, a mi-
nor nitrogen reSux stream can also be provided to the
low pressure column from the high pressure column.
An oxygen-rich liquid stream containing about 70%
oxygen is withdrawn from the bottom of the low
pressure column, reduced in pressure and vaporized
in the top reboiler-condenser. The vaporized stream is
then warmed in the main heat exchanger and event-
ually discarded as a waste stream. Note that the
refrigeration for the plant is met by expanding a por-
tion of the gaseous feed air stream to the low pressure
column in a turbo-expander. In this process, the feed
air is compressed to about 8}9 atm absolute and the
pressure of the low pressure column is about 3 atm
absolute. A nitrogen compressor is generally used to
compress the low pressure nitrogen product stream
and then it is combined with the high pressure nitro-
gen product stream. The typical Sow rate of the
III / AIR LIQUEFACTION: DISTILLATION 1903
combined nitrogen product stream is about 72 mol
per 100 mol of the feed air stream. This two-column
nitrogen generator and its variations are particularly
attractive for enhanced oil recovery where a very
large quantity of nitrogen is injected in the wells to
maintain pressure.
Ultrahigh Purity Nitrogen and Oxygen Production
for the Semiconductor Industry
The fast-growing semiconductor industry uses bulk
nitrogen and oxygen gases in the manufacturing of
computer chips. The acceptable level of impurities in
the supply of these bulk gases has been continually
declining over the last decade. Currently, impurities
are limited to few parts per billion (p.p.b.) and levels
are expected to drop down to parts per trillion levels
as wafer sizes increase. The cryogenic distillation pro-
cess is the only known method of present that, in
conjuction with other adsorption and catalytic pro-
cesses, can meet the stringent demands of ultrahigh
purity (UHP) gases.
In addition to the constituent components listed in
Table 1, air typically consists of several impurities
at p.p.m. levels. Hydrogen is a light impurity that is
in the range of 1}5 p.p.m.; carbon monoxide is
also present at the same levels. There are several
1904 III / AIR LIQUEFACTION: DISTILLATION
impurities that are heavier than oxygen } methane
and higher hydrocarbons and nitrogen oxides are all
present in p.p.m. concentrations. When nitrogen is
distilled in one of the typical processes discussed
earlier, it contains almost all the hydrogen, helium
and neon and a major fraction of the carbon monox-
ide contained in the air. Similarly, a typical high
purity oxygen contains all the unacceptable heavier
impurities } krypton, xenon, methane and higher
hydrocarbons, for example. Clearly the conventional
distillation methods need modiRcation to be able to
supply the UHP gases.
One early method used to produce UHP nitrogen
was to pass the nitrogen from a typical cryogenic
distillation process over a bed of a nickel supported
on silica to remove the trace levels of oxygen, hydro-
gen and carbon monoxide. This method is now used
for back-up systems since the regular supply of UHP
nitrogen is produced directly from the cold box. One
distillation scheme to produce UHP nitrogen is shown
in Figure 9.
The feed air is compressed to a pressure that is
slightly greater than the pressure at which UHP nitro-
gen is required. This is done to supply the UHP
nitrogen product directly from the cold box to the
semiconductor fabrication plant without any further
compression. The pressure of the supply UHP nitro-
gen is generally 8}10 atm absolute. The feed air after
compression is heated to a temperature of about
Figure 9 A UHP nitrogen scheme.
2003C and passed over a noble metal catalyst such as
platinum to oxidize all the carbon monoxide and
hydrogen. The exhaust gas is then cooled and passed
through the molecular sieve unit. In an alternative
process, a separate noble metal catalyst is not
used but instead, layers of adsorbent catalysts are
supplied within the molecular sieve unit to remove
hydrogen and carbon monoxide to p.p.b. level. The
impurities-free air is then cooled in the main heat
exchanger and distilled in a column similar to the one
shown in Figure 7. A large number of separation
stages (60}100) are used in this main distillation
column to reduce the oxygen concentration in the
resulting nitrogen product stream to a level of a few
p.p.b.
In addition to stage count, another major dif-
ference between the processes of Figures 7 and 9 is
the manner in which crude liquid oxygen from the
main distillation column is treated. In the UHP nitro-
gen process of Figure 9, crude liquid oxygen is fed to
the top of a short column containing three to six
stages of separation. The boil-up at the bottom of this
column is provided by condensing a portion of the
nitrogen vapour stream from the top of the main
(high pressure) distillation column. The nitrogen and
oxygen concentration in the vapour leaving the top of
the short column is similar to that in air } this stream
is called synthetic air. The pressure of the synthetic air
is generally greater than 4 atm absolute. To recover

the pressure energy, this stream is recycled after
further boosting its pressure to that of the main
distillation column. In Figure 9, synthetic air is fed to
an interstage of the main air compressor. This conRg-
uration saves the capital cost associated with a separ-
ate booster recycle compressor.
Not all the liquid at the bottom of the short column
is vaporized; instead, an oxygen-rich liquid contain-
ing about 60}70% oxygen is withdrawn. The pres-
sure of this liquid is reduced by 1}2 atm and then it is
vaporized in a separate reboiler-condenser to produce
another portion of the liquid nitrogen reSux for the
main distillation column. A liquid purge stream is
taken from this reboiler to prevent the accumulation
of hydrocarbons to unsafe levels. The vaporized oxy-
gen-rich stream is then expanded to provide the
needed refrigeration for the plant and is eventually
discharged as a waste stream.
The reason for modifying the distillation scheme of
Figure 7 to that of Figure 9 stems from the fact that,
in a UHP nitrogen process, the distillation pressure is
quite high. In Figure 7, high pressure distillation
causes the waste stream to vaporize in the reboiler-
condenser at a pressure greater than 4 atm. When
a large portion of the feed air (in this case 60%) is
sent to the turbo-expander at such a high pressure,
excess refrigeration is produced. The consequence is
that the pressure energy of the waste stream is not
utilized effectively and the process becomes inef-
Rcient. In constrast, the UHP nitrogen distillation
scheme in Figure 9 recovers nearly half of the crude
liquid oxygen stream as a recycle synthetic air stream,
thereby reducing the Sow to the turbo-expander. In
other words, the production of excess refrigeration is
avoided and efRcient operation is maintained by
sending only that portion of the crude liquid oxygen
stream needed for the refrigeration to the turbo-
expander.
There are several other distillation schemes
used for the production of UHP nitrogen. However,
all the efRcient schemes are based on modiRca-
tion of the scheme in Figure 7, such as is shown in
Figure 9.
Generally, UHP oxygen is required in much smaller
quantities (generally 1}5% of the UHP nitrogen pro-
duction rate). It is essential that not only the concen-
tration of a lighter impurity such as argon be in the
p.p.m. to p.p.b. level but also that the concentration
of heavier impurities, such as krypton, xenon and
hydrocarbons, be no more than a few p.p.b. in the
UHP oxygen stream. In contrast to this, a standard-
grade oxygen from the process shown in Figure 4
contains about 0.5% argon and all of the heavier
impurities such as methane, krypton and xenon are
contained in the feed air.
III / AIR LIQUEFACTION: DISTILLATION 1905
A UHP oxygen distillation scheme that is a modiR-
cation to the distillation conRguration of Figure 9 is
shown in Figure 10.
This modiRcation results from an observation that,
as the feed air ascends the main distillation column,
all the impurities that are heavier than oxygen are
rapidly reduced to nearly zero within a few separ-
ation stages. The concentration of oxygen, however,
is still at signiRcant levels. Thus, a heavies-free liquid
stream is withdrawn from about 10}15 separation
stages above the vapour feed air location of the main
distillation column. This heavies-free liquid stream
contains about 10}20% oxygen and, after pressure
reduction to near atmospheric pressure, is fed to the
top of the UHP oxygen column. Since the feed to this
column only contains components that are more vol-
atile than oxygen, the purpose of the column is essen-
tially to distil off these components from oxygen.
Depending on the desired purity of UHP oxygen,
60}100 separation stages are used. Since the amount
of UHP oxygen to be produced is low compared to
the amount of UHP nitrogen, the boil-up at the bot-
tom of the UHP oxygen column is met by cooling
crude liquid oxygen in the sump of this column. The
vapour from the top of this column is typically mixed
with the discharge stream from the turbo-expander.
About 25% of the heavies-free liquid feed to the
UHP oxygen column is recovered as UHP liquid
oxygen product from the bottom of the column. It
is a true credit to the cryogenic distillation
industry that the stringent purity demanded by the
semiconductor industry can be met without post-
treatment.
Liquefaction
Liquid nitrogen and liquid oxygen are produced and
stored in a back-up system to supply gases (after
vaporization of the stored liquid) in the event of the
cryogenic air separation plant shut-down. Liquid may
also be supplied in tankers from a central plant loca-
tion to an end-use site where the consumption of
nitrogen and oxygen is not high and economically it is
not justiRable to build a dedicated plant. Liquid nitro-
gen is also used as a source of refrigeration in such
applications as food freezing.
In 1895, Carl von Linde built the Rrst industrial-
scale air liqueRer. His liqueRer used a Joule}Thomp-
son (JT) valve to create refrigeration. His genius was
the realization that, for the same pressure ratio across
a JT valve, the amount of cooling (drop in temper-
ature) increases rapidly as the absolute pressure of
air is increased. Therefore, such a liqueRer operated
at about 125 atm while the pressure across the JT
valve dropped to approximately 5 atm. In 1902,
1906 III / AIR LIQUEFACTION: DISTILLATION
Figure 10 A distillation scheme for UHP oxygen production.
Georges Claude demonstrated that it was possible to
lubricate a piston expander with petroleum ether at
cryogenic temperatures. He then built an air liqueRer
using his piston expander. Since this liqueRer did not
rely on a JT valve to supply all the refrigeration, it
was much more efRcient than the liqueRer built
seven years earlier by Linde. In 1935, Kapitza built
a piston expansion engine with gas lubrication and
in 1939 he built an air liqueRer with an expansion
turbine. Most modern liqueRers use expansion turbines.
Although the early ‘masters’ of cryogenics were
interested in liquefying air, the focus of modern
liqueRers is mainly to liquefy nitrogen. This is due to
the dominant use of liquid nitrogen for refrigeration
supply. Liquid oxygen is generally produced by sup-
plying some liquid nitrogen as reSux to the low pres-
sure column of a double-column process and by with-
drawing an equivalent amount of liquid oxygen from
the bottom of this column. The gaseous nitrogen
needed for the nitrogen liqueRer is provided by any of
the suitable air distillation processes described earlier.
LiqueRers that are capable of producing in excess of
1000 tons per day of liquid nitrogen and oxygen are
now in operation.
A two-expander nitrogen liqueRer is shown in
Figure 11.
Make-up nitrogen from an air distillation cold box
is compressed to about 6 atm in a make-up compres-
sor and is further compressed to a pressure in excess
of 27 atm in a recycle compressor. The pressurized
nitrogen leaving the recycle compressor is further
boosted to a pressure in excess of 45 atm in compres-
sor 1 and compressor 2, and then fed to a heat
exchanger for cooling. A portion of the high pressure
nitrogen stream is withdrawn near the warm end of
the heat exchanger and expanded in a warm expan-
der to provide a portion of the refrigeration needed
for the liquefaction. A second portion of the high
pressure nitrogen stream is withdrawn from an inter-
mediate location of the heat exchanger and expanded
in a cold expander to provide the refrigeration in the
cold part of the heat exchanger. The remaining

Figure 11 A nitrogen liquefier.
portion of the high pressure nitrogen stream exits the
cold end of the heat exchanger at a temperature
below !1703C and is sent to an optional dense Suid
expander. The pressure drop across this dense Suid
expander is maximized, subject to the constraint that
very little vapour forms in the exhaust. The pressure
of this stream is further reduced to about 6 atm in
a JT value and the resulting two-phase stream is
separated in separator I. The vapour from this separ-
ator and the exhaust streams from the cold and warm
expanders are mixed at appropriate temperatures,
warmed and returned to the recycle compressor. The
liquid from separator I is further cooled and reduced
to near atmospheric pressure through another JT
valve. The resulting liquid is collected as liquid
nitrogen from separator II and the vapour is recycled
to the make-up compressor.
The liqueRer shown in Figure 11 is quite efR-
cient. The use of a dense Suid expander contributes to
increased efRciency but its use is optional. The
working pressure range of the modern brazed plate
and Rn aluminium heat exchangers now approaches
100 atm. For increased efRciency, the pressure of
the high pressure nitrogen steam is increased to max-
imum feasible values. Recently, processes using more
than two gaseous expanders have been suggested for
incrementally higher efRciencies.
III / AIR LIQUEFACTION: DISTILLATION 1907
In Figure 11, if none of the expanders are used then
the liquefaction process reduces to the one proposed
by Carl von Linde. On the other hand, if the warm
expander and the dense Suid expander are removed,
then the resulting process is similar to the one used by
Georges Claude.
Equipment
Machinery
The selection of a compressor depends on the volu-
metric Sow rate, the operating pressures, the com-
pressor efRciency, its capital cost and the cost of
energy. Because of their lower installation cost, cen-
trifugal compressors are chosen over reciprocating
compressors when volumetric Sow rates and pressure
allow their use. Axial compressors are used for large
volumetric Sow rates. Reciprocating machines are
used at very high pressures and small volumetric
Sows.
For most air plants in the size range of 30}3000 tons
per day of oxygen, centrifugal compressors contain-
ing three or four stages are used for compressing the
feed air. Interstage cooling is provided with cooling
water to approximate isothermal compression.
A large fraction of the water contained in the feed air
1908 III / AIR LIQUEFACTION: DISTILLATION
is condensed in these interstage coolers. Most of these
compressors are electrically driven; however, steam
or gas turbines are occasionally used when economi-
cally justiRed. Air is passed through one or two stages
of Rltration to remove particulates prior to entry in
a MAC. For plants that are larger than 3000 tons per
day of oxygen, a combination of axial and centrifugal
compressors is used. At the other end of the scale, for
small size plants in the size range of less than 30 tons
per day of oxygen, inexpensive screw compressors are
used. These guidelines for oxygen plants can easily be
translated to nitrogen plants, because for the same
production rates, a nitrogen plant requires only
30}50% of the feed air Sow required by an oxygen
plant.
When gaseous oxygen is to be compressed, a centri-
fugal compressor is used for low to moderate pres-
sures while a reciprocating compressor is used for
higher pressures. When oxygen is needed at fairly
high pressures, the initial stages of compression may
be centrifugal. The design of an oxygen compressor
requires careful selection of materials and seals, and
total prevention of rubbing contacts to avoid ignition
in the presence of high pressure oxygen. Furthermore,
an oxygen compressor is generally enclosed in
a building with an external barrier to increase the
safety of plant personnel. Special test and start-up
procedures are also used for oxygen compressors.
Expanders are used to provide refrigeration by ex-
tracting work from an expanding Suid. In the expan-
sion process, the temperature of the expanded Suid is
reduced. An air separation or a liquefaction plant
generally uses a single-stage radial inSow turbine as
a standard. For small plants, the work energy from
the expander is either dissipated in an oil brake or
through an ambient air blower. For medium to large
size separation plants and liqueRers, it is essential that
the work energy from an expander be recovered to
increase process efRciency. This is done by either
loading an expander with an electric generator or
a compressor for some other process Suid. When an
expander is directly coupled to a compressor, the
arrangement is called a compander. As seen from
Figure 11, companders are widely used in liqueRers.
Expanders used in the cryogenic air separation and
liquefaction industry typically have isentropic ef-
Rciencies in the range of 85}90%. The dense Suid
expanders are essentially reverse-running liquid
pumps (Figure 11).
Front-end Puri\cation
The compressed air from a main air compressor must
be cleaned of impurities such as water, carbon diox-
ide and some hydrocarbons to avoid plugging at
cryogenic temperatures. The original plants used re-
cuperative heat exchangers which were later replaced
by regenerators in 1930 upon their invention by Mat-
thias FraK nkl. Around the mid-1950s, owing to the
introduction of large brazed aluminium plate and Rn
heat exchangers, the regenerators were replaced by
reversing heat exchangers. Recuperators, regener-
ators and reversing heat exchangers all operated to
freeze the impurities within the device } complete
removal of these trace components was never
achieved. Beginning in the early 1980s, reversing ex-
changers were replaced with ambient temperature
adsorption beds. Today almost all cryogenic air sep-
aration plants use molecular sieve vessels to remove
impurities.
A typical two-bed adsorption system for air puriR-
cation is shown in Figure 12. Each vessel is Rlled with
13X (Na-X zeolite) molecular sieve. This sieve has an
excellent capacity for carbon dioxide and water re-
moval. Sometimes an additional layer of alumina is
used at the entrance for bulk water removal to de-
crease the energy demand during regeneration. While
one bed is on stream, the other bed is being regen-
erated. In Figure 12, bed A is on stream and bed B is
being regenerated. A bed is on stream for a pre-
speciRed period until carbon dioxide is about to break
through the bed. At this point the feed air is directed to
another bed. The pressure of the spent bed is reduced
to near atmospheric pressure and a hot dry gas in the
temperature range of 150}2003C is passed through the
bed to desorb the adsorbed impurities. After the impu-
rities have been removed, the bed is cooled by a Sow of
cool dry gas and it is then ready to be brought on
Figure 12 A front-end system for air purification.

stream. The dry gas used for regeneration is generally
a portion of the gas exiting the cold box. Regenera-
tion gas Sow rate is typically in the range of 10}20%
of the feed air Sow. For oxygen plants, the regenera-
tion gas is a portion of the nitrogen stream.
Heat Exchangers
Around the mid-1950s, large brazed aluminium plate
and Rn heat exchangers were commercially intro-
duced. They readily became the heat exchangers of
choice for cryogenic air separation and liquefaction
plants. In this type of heat exchanger, corrugated Rns
are sandwiched between plates to form a passage for
gas Sow. The use of Rns provides increased surface
area for heat transfer. Typical Rn heights range be-
tween 5 and 9 mm; Rn spacing can be as low as 1 Rn
per mm. A heat exchanger block is formed by stack-
ing passages. Generally, Sow through individual pas-
sages is countercurrent with a warming stream in one
passage and a cooling stream in the adjacent passage.
A heat exchanger block can easily handle multiple
warming and cooling streams. Plate and Rn heat ex-
changers are applied in virtually all the heat ex-
changer services of an air separation plant. They are
used as main heat exchangers, reboiler-condensers
and subcoolers. The maximum size and pressure rat-
ing of these heat exchangers depend on the manufac-
ture; however, heat exchangers 1200 mm wide by
1200 mm stack height by 6000 mm long with a pres-
sure rating up to 50 atm can easily be found. For large
size plants, multiple heat exchangers are used in par-
allel and careful attention is paid to the Sow distribu-
tion in the manifolds.
Distillation
Until the 1980s almost all air distillation was per-
formed in columns containing sieve trays. Due to the
close relative volatility between argon and oxygen
and the purity of the products, columns with over 100
separation stages are not uncommon. Therefore, tray
spacing is generally 150 mm or less. Pressure drop
across a large number of trays in the low pressure
column increases the pressure of the boiling Suid in
the low pressure column sump. This increases the
pressure of the condensing nitrogen at the top of the
high pressure column. In turn, the pressure of the air at
the discharge of the main air compressor is increased.
This leads to an increase in power consumption. As
a result, there is an incentive to use a liquid}vapour
contact device with lower pressure drop.
Today, most modern cryogenic air separation
plants use low pressure drop structured packing in
one or more of the distillation columns. The pressure
drop through structured packing is only one-Rfth to
one-tenth that of a trayed column. The use of struc-
III / AIR LIQUEFACTION: DISTILLATION 1909
tured packing has led to more than 3% power savings
for a typical oxygen plant. It has also allowed the use
of over 175 stages of separation in the side argon
column of Figure 6 for the production of argon with
less than 5 p.p.m. oxygen through distillation. This
eliminates the need for a second deoxidation (Deoxo)
process using hydrogen and a catalyst, thereby mak-
ing pure argon production much simpler.
Cold Boxes
The cryogenic equipment is enclosed in an insulated
enclosure termed a cold box. A rectangular cold box
consists of a steel frame with panels of sheet metal. It
also provides structural support for the equipment.
Cylindrical cans with insulation are also used in cer-
tain applications. Mineral wool was used for insula-
tion prior to the late 1940s. Starting around 1948,
a powdered insulation called perlite was increasingly
used. The advantage of perlite is that installation
costs are lower and cold boxes can be insulated with
greater uniformity, leading to reduced heat leak and
improved plant efRciency.
Materials of Construction
Early plants used copper or copper alloys to fabricate
vessels and piping. Austenitic stainless steels were
occasionally used. In the later 1950s, with the devel-
opment of welding techniques for aluminium, its use
became the most popular. This occurred because alu-
minium and aluminum alloys are easily available and
are low cost and light weight. Cryogenic liquid con-
tainers are also constructed from low carbon 9%
nickel steel.
Safety
Many materials react with pure oxygen, so great care
is taken in the selection and clean-up of materials that
come into contact with oxygen. Potential ignition
sources must be minimized. All impurities that come
into contact with oxygen, especially unsaturated hy-
drocarbons, must be reduced to safe levels. To avoid
hydrocarbon build-up, generally a small purge stream
is taken from the sumps where oxygen-rich liquids
are being boiled. The combustion hazard increases as
pure gaseous oxygen is compressed to higher pres-
sures and therefore, special care should be taken in
the compression and handling of high pressure oxy-
gen gas.
See also: II/Distillation: Energy Management; Historical
Development; Instrumentation Control Systems; Multi-
component Distillation; Theory of Distillation; VapourI
Liquid Equilibrium: Theory.
1910 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION
Further Reading
Agrawal R (1995) Production of ultra high purity oxygen:
a distillation method for the co-production of the heavy
key component stream free of heavier impurities. Indus-
trial Engineering Chemical and Research 34: 3947.
Agrawal R and Thorogood RM (1991) Production of me-
dium pressure nitrogen by cryogenic air separation. Gas
Separation and PuriTcation 5: 203.
Agrawal R and Woodward D (1991) EfRcient cryogenic
nitrogen generators } an exergy analysis. Gas Separation
and PuriTcation 5: 139.
Agrawal R, Woodward DW, Ludwig KA and Bennett DL
(1992) Impact of low pressure drop structure packing on
air distillation. In: Distillation and Absorption. IchemE
Symposium Series no. 128, A125.
Isalski WH (1989) Separation of Gases. Oxford: Oxford
Science Publications Clarendon Press.
Latimer RE (1967) Distillation of air. Chemical Engineer-
ing Progress 63: 35.
Linde W and Reider R (1997) How it all began. In: The
Invisible Industry. Cleveland, Ohio: The International
Oxygen Manufacturers Association.
AIRBORNE SAMPLES: SOLID PHASE
EXTRACTION
D. J. Eatough, Brigham Young University,
Provo, Utah, USA
Copyright ^ 2000 Academic Press
Introduction
Organic material in the atmosphere may exist in
either the gas phase or in particles. For the purposes
of this chapter, atmospheric organic material will be
divided into three classes, deRned by the phase distri-
bution of the organic material in the atmosphere. Gas
phase compounds will include those organic com-
pounds which are present only in the gas phase. This
will include essentially all non-aromatic organic ma-
terial with fewer than about 12}14 carbon atoms.
Nonvolatile organic material will include those com-
pounds which are present in particles and whose
concentrations in the gas phase are negligible com-
pared to the particulate material. Semi-volatile or-
ganic material includes those compounds which are
present in equilibrium between the gas and parti-
culate phases in the atmosphere and for whom the
concentrations in both phases are signiRcant. The
collection of gas phase and nonvolatile organic
material is relatively straightforward. However, the
McGuinness RM (1998) Oxygen Production. In: Baukal
CE (ed.) Oxygen-enhanced Combustion, Ch. 3. Boca
Raton: CRC Press.
Scott RB (1988) Cryogenic Engineering. Boulder, Colora-
do: Met-Chem Research.
Scurlock RG (1992) History and Origins of Cryogenics.
Oxford: Clarendon Press.
Springmann (1977) The planning of large oxygen plants for
steel works. Linde Report in Science and Technology 25:
28.
Thorogood RM (1986) Large gas separation and liquefac-
tion plants. In: Hands BA (ed.) Cryogenic Engineering,
Ch. 16. London: Academic Press.
Timmerhaus KD and Flynn TM (1989) CryogenicProcess
Engineering. New York: Plenum Press.
Venet FC, Dickson EM and Nagamura T (1993) Under-
stand the key issues for high purity nitrogen production.
Chemical Engineering Progress 89: 78.
Wilson KB, Smith AR and Theobald A (1984) Air puriRca-
tion for cryogenic air separation units. IOMA Broad-
caster January, pp. 15d20.
accurate determination of the phase distribution of
semi-volatile organic material requires the use of dif-
fusion denuder technology.
Correct assessment of the contribution of Rne
particulate carbonaceous material to various atmos-
pheric processes is dependent on the accurate deter-
mination and characterization of Rne particulate
organic material as a function of particle size. Several
studies have shown that about one-third of the mass
of Rne particulate matter (dia.(2.5
m) collected on
Rlters in remote desert regions of the Southwest U.S.
is organic compounds and elemental carbon. Similar
fractions of carbonaceous material are found in par-
ticles collected on Rlters in western urban areas. In the
eastern United States sulfate is the major component
of Rlter collected airborne Rne particles. However,
organic material comprises one-fourth or more of the
Rne particulate mass. In the Northwest, organic ma-
terial has been found to be the dominant Rne partic-
ulate component. However, unless proper sampling
procedures are used to collect particulate material,
the composition of organic material in Rne particles
will be signiRcantly underestimated due to losses
from the semi-volatile particulate organic fraction
during sample collection, i.e. a ‘negative’ sampling
artifact.
 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION
Several studies have also indicated the presence of
a ‘positive’ artifact in the determination of particulate
organic compounds collected on a quartz Rlter, due to
the adsorption of gas phase organic compounds by
the quartz Rlter during sampling. Data obtained using
sampling systems with two quartz Rlters in series
suggest that quartz Rlters collect at least some gas
phase organic compounds. In addition, particulate
material collected on a Rlter can also absorb some gas
phase organic compounds. The adsorption of organic
compounds by a second quartz Rlter has been shown
to be reduced, but not eliminated, in samples col-
lected in the Los Angeles Basin if a multi-channel
diffusion denuder with quartz Rlter material as the
denuder collection surface precedes the quartz Rlters.
This artifact can be eliminated by the use of activated
charcoal at the denuder surface. Recent experiments
have shown that the quartz Rlter artifact can result
both from the collection of gas phase organic com-
pounds and from the collection of semi-volatile or-
ganic compounds lost from particles during sampling.
Thus, results available to date suggest that both
a ‘positive’ and a ‘negative’ artifact can be present in
the determination of particulate phase organic com-
pounds using two tandem quartz Rlters.
Collection of Gas Phase Organic
Material
A well validated technique for the collection of gas
phase organic material for subsequent analysis is the
use of SUMMA stainless steel canisters. If the cani-
sters are properly cleaned before use and analysed
within a few weeks of sample collection, valid results
can be obtained for most gas phase compounds.
A second method which has frequently been used
to collect gas phase organic materials consists of the
use of a Rlter to remove particulate material, followed
by a sorbent bed to collect the gas phase organic
compounds. This approach is not valid if (1) the gas
phase organic material is oxygenated or polar and
therefore capable of being absorbed by a quartz Rlter
or by organic material colleted by the particle remov-
ing Rlter, or (2) the gas phase organic material is
semi-volatile and therefore, may be present on and
lost from particles during sampling (see following
section). The absorption of organic material by vari-
ous types of Rlters has been reviewed. TeSon has been
suggested to be relatively inert to absorption artifacts,
but this Rlter is not amenable to the determination of
total carbon. Glass Rbre and cellulose membrane Rl-
ters both absorb signiRcant quantities of gas phase
organic material. Quartz membrane Rlters are suit-
able for the determination of total carbon, but they
also can absorb signiRcant quantities of gas phase
1911
organic material. This is illustrated in Figure 1 which
shows the analysis of total carbon for a Rlter which
was preceded and not preceded by a charcoal based
diffusion denuder to remove gas phase material. The
large peak seen in the absence of a diffusion denuder
is gas phase organic material collected by the quartz
Rlter. A similar peak (plus some higher temperature
material) is seen on a second quartz Rlter which is not
preceded by a denuder.
Materials which have been validated as sorbents
for the removal of gas phase organic compounds
include polyurethane foam (PUF), poly(oxy-m-ter-
phenyl-2,5-ylene), Tenax, copolymers of styrene and
divinylbenzene (XAD), Chromosorb and charcoal.
Of these sorbents, Tenax is best suited for the collec-
tion of very low molecular weight organic material
and Chromosorb or XAD are effective for collection
over a wide range of molecular weights. A caution is
that many of the sorbents can produce spurious re-
sults due to reactions during sample collection and
each of the sorbents can be difRcult to clean for the
detection of trace substances. Thus, for example,
a PUF cartridge produces mutagenic compounds
upon extraction with methanol and Tenax forms de-
composition products during sampling.
Collection of Non-Volatile Organic
Material
Compounds which are sufRciently volatile that they
exist essentially only in the gas phase can be collected
on any Rlter suitable for total particle collection, such
as quartz or TeSon Rlters. Quartz Rlter are usually
used when the determination of total carbonaceous
material in addition to the identiRcation of speciRc
compounds is desired. However, if only speciRc com-
pound identiRcation is desired, the use of TeSon
Rlters avoids the complication associated with the
absorption of gas phase material by the Rlter. How-
ever, if the target species include compounds which
are reactive or unstable, they may be altered by chem-
ical reactions associated with the sampling process.
Examples of potential problems are given in the fol-
lowing sections.
Collection of Semi-Volatile Organic
Material
To address the issues of both ‘positive’ and ‘negative’
artifacts in the sampling of particulate phase organic
compounds, several groups have constructed and tes-
ted sampling systems employing diffusion denuders,
Rlters and sorbent Rlters. The data obtained to date
with these sampling systems show that particulate
1912 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION

Figure 1 Temperature-programmed volatilization analysis of quartz filters (A) not preceded and (B) preceded by a denuder with
charcoal impregnated filter surfaces. The large initial peak seen in (A) but not in (B) is due to the absorption of gas phase organic
material by the quartz filter not preceded by a diffusion denuder to remove gas phase organic compounds.

phase organic compounds have been signiRcantly
underestimated by the collection of particles with
only a Rlter. The collection of gas phase compounds
by a quartz Rlter may produce a signiRcant ‘positive’
artifact (Figure 1), but a much larger negative error
usually results from the loss of 20}80% of the partic-
ulate phase semi-volatile organic material during
sampling. This sampling artifact must be considered
in the collection of semi-volatile particulate organic
compounds. Accurate collection procedures for semi-
volatile organic compounds must meet the following
two criteria:
1. Organic compounds initially present in the gas
phase which can be adsorbed onto particles or the
Rlter must be distinguished from semi-volatile or-
ganic compounds lost from particles during samp-
ling.
2. Organic compounds initially present in the parti-
culate phase and lost from particles during samp-
ling must be captured during sampling separate
from compounds which are present in the gas
phase in the atmosphere.
These two criteria cannot be met by any sampling
procedure in which the particulate phase organic
compounds are collected before the collection or sep-
aration of gas phase organic compounds because the
gas phase organic compounds and organic com-
pounds volatilized from particles become indistin-
guishable. Thus, it is necessary Rrst to remove the gas
phase organic compounds and then to collect the
particulate phase organic compounds with a sampler
which will collect all organic material, gas and par-
ticle. This can be accomplished using diffusion de-
nuder sampling technology.
The BOSS and BIG BOSS Diffusion Denuder
Samplers
Diffusion denuder sampling systems for the deter-
mination of total Rne particulate organic material
have been developed at Brigham Young University.
 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION 1913

The objectives which guided the development of these

sampling systems were:

1. The sampling system should have a Sow rate

sufRcient to enable measurement of low concen-
trations of particulate carbonaceous material
and to allow the detailed chemical characteriza-
tion of particulate organic material, e.g. Sow rates
of from 30 to 300 L min\ were considered desir-
able.
2. The sampler should have a diffusion denuder ca-
pable of removing all gas phase semi-volatile or-
ganic compounds which are in equilibria with
compounds in the particulate phase in the atmos-
phere.
3. The diffusion denuder of the sampler should be
effective in removing all gas phase compounds
which can be adsorbed by a quartz Rlter or by
collected particles during sampling.
4. The capacity of the diffusion denuder for the re-
moval of gas phase organic compounds should be
high enough that samples can be collected at the
target Sow rates over sampling periods of several
days to weeks.
5. Particle losses during the passage of sampled
air through the diffusion denuder should be small.
6. The sampler after the diffusion denuder should
collect both particles and any semi-volatile or-
ganic material lost from particles during sampling
with high efRciency.
7. The collection materials used in the sampler
should be compatible both with the determination
of total carbonaceous material and with the de-
tailed chemical characterization of particulate or-
ganic material.

The BOSS (BYU Organic Sampling System) re-
quires two different samplers as shown schematically
in Figure 2:
1. A charcoal impregnated Rlter (CIF), multi-chan-
nel, parallel plate diffusion denuder followed by
a Rlter pack containing quartz and CIF Rlters. The
denuder removes gas phase organic compounds.
The quartz Rlter after the denuder collects Rne
((2.5
m) particles. The organic compounds col-
lected by the CIF sorbent Rlter in this sampler are
semi-volatile organic compounds lost from the
particles during sampling and a small fraction
(about 5%) of the gas phase organic material not
collected by the diffusion denuder.
2. A quartz Rlter followed by a CIF diffusion denuder
and a CIF sorption Rlter. The quartz Rlter collects
particles and any gas phase organic compounds
which can be absorbed by quartz, both those ori-
Figure 2 Schematic of the BOSS. Non-volatile particulate car-
bonaceous material is determined from analysis of T1,1. Semi-
volatile carbonaceous material lost from particles is determined
from analysis of CIF1,1, corrected for the denuder inefficiency
determined from analysis of CIF2,1.
ginally in the gas phase and those lost from the
particles during sampling. The denuder then re-
moves gas phase compounds passing the quartz
Rlter. Any gas phase compounds not removed by
the denuder are then collected by the CIF sorbent
Rlter. This system is used to determine indepen-
dently the gas phase organic compounds not col-
lected by the denuder to correct the data obtained
with the CIF Rlter of Sampler 1.
The various 47 mm diameter Rlters of the BOSS are
contained in TeSon Rlter packs (University Research
Glass, Model 2000-30F) with the Rlter packs holding
1914 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION
the quartz Rlter in Sampler 2, Figure 2, being modi-
Red so that the outlet is identical to the inlet to allow
for convenient connection to the diffusion denuder
(University Research Glass, Model 2000-30FB). The
diffusion denuder is based on a design originally
reported by Fitz (1990). Each denuder is comprised of
17 (4.5;58 cm) strips of Schleicher and Schuell char-
coal impregnated Rlter paper which are separated at
the long edges by 2-mm rods. The multi-parallel plate
array of Rlter strips is contained within a (5;5 cm)
square aluminium tube. The entire assembly is nom-
inally 90 cm in length to accommodate 58 cm sorbent
Rlter strips and two nominally 15 cm long Sow
straightening sections ahead of and behind the denud-
ing section. The multi-channel diffusion denuder is
designed to have acceptable efRciency for the removal
of gas phase organic material in the denuder, negli-
gible loss of particles to the denuder during sampling,
and high capacity for the collection of gas phase
organic material. The total capacity of the CIF multi-
channel denuder has not been directly measured.
However, no degradation of the efRciency of the
denuder for the collection of gas phase organic com-
pounds was seen during continuous operation at
40 L min\ for over two months or for sampling at
180 L min\ for continuous periods equivalent to
seven and fourteen days in the Los Angeles Basin, for
ten days in the Mohave Desert, or for twelve days at
Research Triangle Park, NC.
The CIF (Schleicher and Schuell, Inc.) strips in the
diffusion denuder are used as received from the
manufacturer. The 47 mm CIF (Schleicher and
Schuell, Inc., No. 508) Rlters are cleaned with dich-
loromethane and dried at 2003C before use. Alter-
nately, a 47 mm Carbon EMPORE (3M) Rlter may be
used. The Carbon EMPORE Rlters may be used as
received from the manufacturer, however, Sow
through these Rlters is limited to about 7 L min\.
The 47 mm quartz Rlters (PallSex, 2500 QAT-UP) are
pretreated by Rring at 8003C for four hours prior to
sample collection. The Sow through the two samplers
of the BOSS, Figure 2, is controlled at about 40 L min\.
A version of the BOSS using a shortened denuder
(27 cm CIF strips) with a Sow of from 4 to 20 L min\
has also been described. The CIF or Carbon EM-
PORE Rlters may also be replaced with an XAD
sorbent bed. The XAD (Rohn & Haas) is cleaned by
Rrst sonicating 10 times with CH3OH to remove very
Rne particles and then Soxhlet extracting for 24 hours
sequentially with CH3OH, CH2Cl2 and C2H5OC2H5.
The efRciency of removal of gas phase organic
compounds by the CIF denuder (or by an annular
denuder conRguration) is described by eqn [1]:
C/Co"0.819e\22.5(DjLW/4Fd) [1]
where Co and C are the concentrations of organic
compounds entering and exiting a section of the de-
nuder, respectively, Dj is the diffusion coefRcient of
the gas phase organic compound(s) at the experi-
mental conditions, L and W are the length and effec-
tive width of the denuder section, F is the Sow and
d is the space between the denuder surfaces. A plot of
the log of the amount collected in equal length sec-
tions of a denuder versus the distance from the start
of the denuder through the section should be linear
with a slope of!22.5 DjW/4Fd. The expected depos-
ition gradient was observed for organic material col-
lected by a CIF based denuder containing two parallel
sheets of the charcoal impregnated Rlter material. The
slope of the line describing the deposition pattern for
the collection of ambient gas phase organic com-
pounds gives an average diffusion coefRcient for the
collected gases of 0.052$ 0.008 cm2 s\. This diffu-
sion coefRcient gives a calculated effective average
molecular weight of 160$25. This average molecu-
lar weight is consistent with the majority of the or-
ganic material which has been shown to be collected
by the diffusion denuder. The deposition pattern was
also consistent with the measured efRciency of the
CIF denuder for the removal of gas phase organic
compounds.
The importance of the particulate organic com-
pounds which have not been identiRed in past studies
where particles are collected on a Rlter will be depen-
dent on the chemical composition and the size distri-
bution of the particulate organic compounds, both
those lost from the particles during sampling and
those remaining on the particles after sampling.
A high-volume, multi-component diffusion denuder
sampling system (BIG BOSS) for the determination of
the size distribution and chemical composition of Rne
particulate organic compounds using diffusion de-
nuder sampling technology has been developed and
tested.
The BIG BOSS uses a variety of size selective virtual
impactor inlets to control the particle size of the
particles introduced to the diffusion denuder sampler.
The inlet system is a modiRcation of a high-volume,
multi-jet virtual impactor. The nominal total Sow
through all systems of the BIG BOSS is 0.9 m3 min\
inlet Sow. This Sow is divided among four systems,
each with a coarse particle minor Sow stream and
a Rne particle major Sow stream. Two of the four
systems have an inlet cut of 2.5
m. The other two
systems are designed to operate with an inlet cut of
0.8 and 0.4
m (see Tang, 1994).
The PC-BOSS Denuder Sampler
The combination of the technology used in the pre-
viously described BIG BOSS sampling system and the
 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION 1915

Figure 3 Schematic of the PC-BOSS. The composition of fine particulate matter is determined from analysis of the two filter packs
after the denuder. The efficiency and losses of the fine particle concentrator is determined by comparison of sulfate on Q2 with that on
Q1 or T1.

Harvard particle concentrator results in the Particle
Concentrator-Brigham Young University Organic
Sample System (PC-BOSS) shown schematically in
Figure 3. The system has been optimized to meet the
following criteria: (1) removal of at least 75% of the
gas phase material before the sampled aerosol is
passed through the diffusion denuder, (2) efRciency,
'99% for the removal of SO2, HNO3 and gas phase
semi-volatile organic material, (3) determination of
particle mass, carbonaceous material and nitrate with
a diffusion denuder sampler, (4) operation on less
than 20 amps of 110 V power.
The inlet to the sampler is a Bendix cyclone with
a particle cut of 2.3
m aerodynamic diameter at an
inlet Sow of 150 L min\. Following the inlet,
20 L min\ is diverted to a Rlter pack to provide data
for calculating the efRciency of and losses in the
PC-BOSS particle concentrator. The remaining Sow
enters the virtual impactor particle concentrator. The
design and evaluation of the particle concentrator has
1916 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION
been previously described. The particle concentrator
separates most of the gas phase material into the
major Sow and leaves particles larger than the cut
point (about 0.1
m) along with a signiRcantly re-
duced fraction of the gas phase material in the minor
Sow. The performance of the particle concentrator
for collection of ambient samples with the PC-BOSS
was evaluated as a function of the minor to major
Sow ratio, and the distance between the accelerator
and receiver slits of the virtual impactor. The opti-
mum design uses a single particle concentrator with
a 9.5 cm long slit and a distance between the acceler-
ator and receiver slits 1.5 times the slit width of
0.32 mm. The minor Sow (25% of the total
150 L min\ Sow) containing concentrated particles
enters the BOSS diffusion denuder. The denuder is
followed by two parallel Rlter packs (Figure 3). The
Rlter pack containing a 47 mm quartz Rlter (PallSex,
preRred) followed by a 47 mm charcoal impregnated
Rlter is used to determine Rne particulate carbon-
aceous material, including semi-volatile organic ma-
terial lost from the particles during sampling. The
second Rlter pack contains 47 mm TeSon (Gelman
ZeSuor) and nylon (Gelman Nylasorb) Rlters to de-
termine mass, sulfate and nitrate, including any ni-
trate lost from particles during sampling.
The IOVPS and IOGAPS Denuder Samplers
Researchers at Lawrence Berkeley Laboratories have
developed an annular denuder sampling system, the
Integrated Organic Vapour/Particle Sampler (IOVPS)
with an XAD-IV based diffusion denuder for the
measurement of SVOC. This diffusion denuder sam-
pler is similar in design and operation to the BOSS
systems described above. The IOVPS is shown sche-
matically in Figure 4. An advantage of the IOVPS
sampler is that the gas phase material collected by the
denuder can be easily recovered for organic com-
pound chemical characterization and quantitation.
Current disadvantages of the sampler are the total
carbonaceous material is not determinable in the de-
nuder or post-Rlter XAD sorbent beds (Figure 4) and
the capacity of the denuder limits the length of time
over which the denuder may be used from hours to
days.
The denuder of the IOVPS system is prepared by
adhering very Rne mesh XAD to a glass multi-annular
denuder surface. The adhesion of the Rnely ground
XAD to the sandblasted glass is strong enough that
the coating is resistant to removal by handling, sol-
vent washing and air sampling. Quantitation of gas
phase organic compounds removed by the IOVPS
denuder is accomplished by extraction with a suitable
solvent and analysis by GC or GC/MS. The collection
efRciency of these denuders for various gas phase
organic compounds has been shown to be close to
that predicted by eqn [1]. A 5-channel denuder with
1 mm spacing in the annulus and a coating length of
38 cm has been used for most applications of the
IOVPS denuder.
The capacity of the IOVPS XAD based denuder is
dependent on two factors: (1) the capacity of the
XAD surface for a given compound and (2) the time
required to elute a dilute concentration of a given gas
down the XAD column length. The dominant factor
appears to be the movement of collected gas phase
material down the XAD column. As a result, studies
using the IOVPS denuder have generally been limited
to chamber studies where the sampling period is short
or to ambient studies where the sample collection
occurred only over a few hours. By increasing the
length and surface area of the denuder (including
using parallel denuders) prototype systems have been
developed by Lawrence Livermore Laboratory and
the Atmospheric Environment Service of Environ-
ment Canada (IOGAPS, Integrated Organic Gas and
Particle Sampler) which are capable of sample collec-
tion for up to 48 hours. Comparisons of results ob-
tained from 24 hour IOGAPS and sequential 4 hour
IOVPS data where the annulus width of the IOGAPS
was 1.5}3.0 mm with a residence time of 2.6 s in-
dicated there was about 10% breakthrough of naph-
thalene in the IOGAPS. A redesign with an annulus
width of 1.0}1.4 mm is expected to eliminate this
problem.
Particle losses to the wall of the IOVPS denuder has
been evaluated in several studies. The results are
essentially identical to those reported above for the
BOSS and BIGBOSS samplers. With face velocities of
around 20 cm s\ through the denuder, losses are less
than 2%. At higher face velocities of 35 to 50 cm s\,
the losses increase to about 5}7%. These losses are
comparable to that seen for conventional annular
denuders.
Other Diffusion Denuder and Related Samplers
Diffusion denuder sampling techniques have also
been developed and used by several other investiga-
tors to determine Rne particulate organic material.
The focus of these studies has been on the determina-
tion of speciRc organic compounds. Krieger and Hites
have used short sections of capillary gas chromato-
graphic columns as a diffusion denuder and deter-
mined concentrations of gas and particulate phase
polychlorinated biphenyl (PCB) and polyaromatic
hydrocarbon (PAH) compounds. Coutant et al. have
described the development of a circular multi-channel
diffusion denuder for the study of PAH in ambient
air. However, results on Reld studies using the
sampling system have not yet been published. The
 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION 1917

Figure 4 Schematic of the IOVAPS (from Gundel, 1999). The denuders contain XAD as the gas phase organic sorbent. Non-volatile
particulate carbonaceous material is determined from analysis of the filters in either of the filter packs. Semi-volatile carbonaceous
material lost from particles is determined from analysis of the denuder d3.

Atmospheric Environment Service of Environment
Canada has been involved since 1984 in the develop-
ment and use of a diffusion denuder sampler for the
determination of PCBs and chlorinated hydrocar-
bons. The instrument uses a silicone gum/Tenax-
coated, multi-tube, annular, diffusion denuder to re-
move the target organic compounds. Turpin et al.
have developed a sampling system which corrects for
the loss of semi-volatile organic compounds during
sampling by removal of most of the gas phase mate-
rial from the particles in a diffusion separator samp-
ling system. The system has been evaluated for the
collection of PAH. All of the systems which have been
described by other research groups collect samples at
a Sow rate of a few L min\. One advantage of the
use of the diffusion denuder sampling systems de-
scribed above is that the attainable high Sow rate,
200 L min\, allows for more collected material and
a wider range of analyses on the collected samples.
Residence Time in the Denuder
The efRciency of a diffusion denuder sampler for the
removal of gas phase material can be improved by
1918 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION
increasing the residence time of the sampled aerosol
in the denuder. However, the residence time can only
be increased within limits. Since the diffusion de-
nuder reduces the concentration of gas phase semi-
volatile organic material, semi-volatile organic mater-
ial present in the particles passing through the
denuder will be in a thermodynamically unstable
environment and will tend to outgas SVOC during
passage through the denuder. The residence time of
the aerosol in the denuder should be short enough to
prevent signiRcant loss of particulate phase SVOC to
the denuder. Various studies have suggested that the
residence time in the denuder should be less than
about 2 s. The residence times in the various denuder
designs described above are about 1.5, 0.2, 0.2 and
1.4 s for the BOSS (or PC-BOSS), BIG BOSS, IOVPS
and IOGAPS denuders, respectively.
Changes in Chemical Composition
during Sampling
The preceding sections have outlined sampling sys-
tems designed to identify correctly the atmospheric
gas and particulate phase distribution of collected
organic material. An additional sampling artifact
which has been little considered in the collection of
atmospheric sampling is the potential alteration of
organic compounds as a result of the sampling pro-
cess. These alterations appear to result from the
movement of ambient air containing oxidants and
other reactive compounds past the collected particles.
The addition of NO2 ((1 p.p.m.) or O3 ((200 p.p.b.)
to the sampled air stream (0 to 53C) for a high volume
sampler reduced the concentrations of benzo(a)
pyrene and benzo(a)anthracene from a few up to
38%, with the observed reduction increasing with
increased concentration of the added gases. Spiking
a Rlter with an amine resulted in an increase in mea-
sured concentrations of nitrosoamines in both the
Rlter and a following XAD sorbent bed for a mid-
volume sampler. Similar results have been obtained
for the exposure of a deuterated amine on a Rlter to
NOx. When Tenax columns spiked with deuterated
styrene and cyclohexane were exposed to p.p.m. con-
centrations of ozone or halogens, oxygenated and
halogenated compounds were shown to be formed.
Similar oxidation of aldehydes and PAN during
sampling has been observed. Collected PAH com-
pounds can be oxygenated and/or nitrated on a Rlter
but 1-nitropyrene has been shown to be resistant to
additional nitration. These various chemical trans-
formations of collected organic compounds can be
eliminated by removal of the gas phase oxidants,
NOx, HNO3, etc., prior to collection of the particles.
The PC-BOSS denuder described above should be
effective in eliminating most of chemical transforma-
tion artifacts since reactive gases are removed by the
charcoal denuder which precedes the particle collec-
tion Rlter.
Application of Diffusion Denuder
Samplers to the Determination
of Semi-Volatile Organic Material
The application of diffusion denuder samplers to the
determination of gas and particulate phase semi-vol-
atile organic material is illustrated with results from
three different studies, one each using the BIG BOSS,
PC-BOSS and IOVPS samplers.
Semi-volatile organic compounds lost from par-
ticles during sampling and subsequently collected by
an XAD-II trap and semi-volatile organic compounds
retained by the quartz Rlters during sampling have
been chemically characterized for(2.5
m particles
in BIG BOSS samples collected at Azusa in the Los
Angeles Basin. The XAD-II sorbent beds included
signiRcant concentrations of aliphatic, acidic and aro-
matic organic compounds. Similar compounds were
also detected in the GC}MS analysis of the Rlter
extracts. However, the compounds retained by the
Rlter were of higher molecular weight. The distribu-
tion of compounds lost from particles during samp-
ling and remaining on the particles during sampling is
illustrated by the GC/MS results for parafRnic com-
pounds (Figure 5).
The pattern seen in Figure 5 is typical of results
obtained for all classes of compounds and all samples
studied to date. For those compounds which have
been characterized, the envelopes of each class of
compounds remaining in the particles and lost from
the particles overlap. For each compound class, the
more volatile compounds predominate in the material
lost from the particles and collected in the XAD-II
bed during sampling. In contrast, the higher molecu-
lar weight organic compounds are retained by the
particles during sampling. For example, particulate
n-tetradecane and n-pentadecane are found only in
the XAD-II bed and not in the particles after sampling
(Figure 5). Hydrocarbons lower in molecular weight
than these two compounds are found in comparable
concentrations in the XAD-II beds of both Samplers
1 and 2 of the BOSS (Figure 2) indicating they orig-
inate mainly from the breakthrough of some fraction
of the gas phase component of these species. In con-
trast, n-tetracosane and higher molecular weight
aliphatic hydrocarbons are retained by the particles
during sampling and are not found in the XAD-II
sorbent beds (Figure 5). Compounds of intermediate
molecular weight, e.g. n-decosane, are partially lost
and partially retained by the particles. Also illustrated
 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION 1919

Figure 5 GC/MS data (m/z"85) for paraffinic compounds; (A) retained by particles and (B) lost from particles during collection on
a filter (from Tang, 1994).

by the GC-MS data is the increased tendency for
lower molecular weight semi-volatile organic com-
pounds to be retained by the particles during sample
collection as the polarity of a given molecular weight
compound increases. For example, n-heptadecane
(MW 226) is largely lost from particles during samp-
ling (Figure 5). However, lauric acid (MW 214) and
Suoranthene (MW 202) are largely retained by the
particles during sampling.
Results for the determination of PAH compounds
in indoor air obtained with the IOVPS and with
a conventional Rlter-sorbent sampler are given in
Figure 6. As indicated in Figure 6(A), about 90% of
the phenanthrene, pyrene and chrysene are present in
the gas phase. However, about 60% of the more
volatile phenanthrene (MW 178) and pyrene (MW
202) are lost from the Rlter of the Rlter pack during
sample collection. In contrast, the loss of the less
volatile chrysene (MW 228) was negligible. These
results are comparable to those given above for the
Azusa study with the BIG BOSS.
Recent studies have indicated that the U.S. Envir-
onmental Protection Agency (EPA) PM10 air quality
standard does not provide adequate human health
protection because the Rne particle (PM2.5) compon-
ent of PM10 is related to observed health effects at
concentrations substantially below the PM10 stan-
dard. As a result, EPA has promulgated a PM2.5 air
quality standard. In order to implement the new
PM2.5 standard, a Federal Reference Method (FRM)
for Rne particulate monitoring has been proposed (see
Schaefer, 1997). The PM2.5 FRM is a single Rlter pack
sampling method with gravimetric determination of
the collected mass.
For the reasons outlined above, the FRM will tend
to not measure semi-volatile Rne particulate constitu-
ents. The amount of semi-volatile material is ex-
pected to be a substantial fraction of the total PM2.5

Figure 6 Retention and loss of particulate PAH compounds during sampling. (A) The lower concentrations determined by a filter pad,
compared to IOVPS, is due to losses from particles during sampling. (B) Concentrations of both particle and gas phase PAH with the
IOVPS.
1920 III / AIRBORNE SAMPLES: SOLID PHASE EXTRACTION
Figure 7 Average composition of PM2.5 in Riverside CA, including semi-volatile ammonium nitrate and organic material lost during
sampling from particles collected on a filter.
mass observed in many urban areas. As a result, the
proposed Federal Reference Method may under-
determine Rne particulate mass. A comprehensive
Reld study to evaluate the PC-BOSS and compare
with results obtained by other PM2.5 sampling
methods, including the FRM has been conducted in
Riverside, California. Riverside was chosen for the
study because high particulate pollution resulting
from summer inversions is expected. Both annual and
24 hour maximum concentration of PM10 exceeded
the federal standards in 1995 and high concentrations
of particulate semi-volatile ammonium nitrate and
organic materials are expected to be present in this area.
The average result for the determination of the
composition of Rne particulate matter in Riverside
during August and September 1997 are given in
Figure 7. Substantial amounts of both ammonium
nitrate and semi-volatile organic material were lost
from the Rlters of both the PC-BOSS and the PM2.5
FRM. The average loss of ammonium nitrate (34%,
1.7
g m\3) was smaller than that for the semi-vol-
atile organic material (54% of total Rne particulate
organic material, 9.5
g m\3). As a result of the loss
of these species, the PM2.5 FRM lost an average of
39% of the Rne particulate material during the collec-
tion of the sample.
The results obtained in these three examples illus-
trate the importance of correctly sampling for semi-
volatile particulate organic material.
See also: II/Extraction: Solid-Phase Extraction. Mem-
brane Separations: Filtration. III/Atmospheric Analy-
sis: Gas Chromatography: Supercritical Fluid
Chromatography. Solid-Phase Extraction with Discs.
Further Reading
Chow JC (1995) Measurement methods to determine com-
pliance with ambient air quality standards for suspended
particles. J. Air & Waste Management Assoc. 45:
320}382.
Cui W, Eatough DJ and Eatough N (1998) Fine particulate
organic material in the Los Angeles Basin } I: Assess-
ment of the high-volume Brigham Young University
Organic Sampling System, BIG BOSS. J. Air & Waste
Manage. Assoc. 48: 1024}1037.
Ding Y, Lee ML and Eatough DJ (1998) The deter-
mination of total nitrite and n-nitroso compounds in
atmospheric samples. J. Environ. Anal. Chem. 69:
243}255.
Eatough DJ (1999) BOSS, the Brigham Young University
Organic Sampling System: Determination of particulate
carbonaceous material using diffusion denuder sampling
technology. In: Douglas Lane (ed.) Gas and Particle
Phase Partition Measurements of Atmospheric Com-
pounds, Vol. 2, 233}285. Gordon and Breach Science
Publishers.
Eatough DJ, Obeidi F, Pang Y et al. (1999) Integrated and
real-time diffusion denuder samplers for PM2.5 based on
BOSS, PC and TEOM technology. Atmospheric Envi-
ronment 33: 2835}2844.
Eatough DJ, Tang H, Cui W and Machir J (1995) Deter-
mination of the size distribution and chemical composi-
tion of Rne particulate semivolatile organic material in
urban environments using diffusion denuder technology.
Inhal. Toxicol. 7: 691}710.
Fitz DR (1990) Reduction of the positive organic artifact
on quartz Rlters. Aerosol Sci. Technol. 12: 142}148.
Fraser MP, Cass GR, Simoneit BRT and Rasmussen RA
(1998) Air quality model evaluation data for organics. 5.
C6}C22 nonpolar and semipolar aromatic compounds.
Environ. Sci. Tech. 32: 1760}1770.
 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY
Gundel LA and Lane DA (1998) Direct determination
of semi-volatile organic compounds with sorbent
coated diffusion denuders. J. Aerosol Sci. 29:
S341}S342.
Gundel LA, Lee VC, Mahanama KRR, Stevens RK and
Daisey JM (1995) Direct determination of the phase
distributions of semi-volatile polycyclic aromatic hydro-
carbons using annular denuders. Atmos. Environ. 29:
1719}1733.
Hart KM and Pankow JF (1994) High-volume air sampler
for particle and gas sampling. 2. Use of backup Rlters to
correct for the adsorption of gas-phase polycyclic aro-
matic hydrocarbons to the front Rlter. Environ. Sci.
Technol. 28: 655}661.
Kamens RM, Odum J and Fan Z-H (1995) Some observa-
tions on times to equilibrium for semivolatile polycyclic
aromatic hydrocarbons. Environ. Sci. Technol. 29: 43}50.
Lane DA and Johnson ND (1993) Vapor and particle phase
measurements of polycyclic aromatic compounds (PAC)
in ambient air. Poly. Arom. Comp. 13 (Supplement):
511}518.
McDow SR and Huntzicker JJ (1990) Vapor adsorption
artifact in the sampling of organic aerosol: face velocity
effects. Atmos. Environ. 24: 2563}2571.
ALCOHOL AND BIOLOGICAL MARKERS
OF ALCOHOL ABUSE:
GAS CHROMATOGRAPHY
F. Musshoff, Institute of Legal Medicine, Bonn,
Germany
Copyright ^ 2000 Academic Press
The use of alcoholic beverages is probably the most
ancient social habit worldwide, but alcohol abuse
has generated severe problems. Chronic and/or
acute alcohol intoxication has been demonstrated to
be connected with serious pathologies, suicides,
homicides, fatal road and industrial accidents and
many criminal offences. Alcoholism is a widespread
social, medical and economic problem in a large
section of the population of nearly all ethnic
groups. Therefore, it is of great importance to
have diagnostic tools (biological markers) to detect
excessive alcohol consumption and alcoholism. This
article deals with gas chromatographic techniques to
determine excessive alcohol consumption. The fol-
lowing parameters are described: ethyl alcohol and
congeners, ketone bodies, ethyl glucuronide,
fatty acid ethyl esters and condensation products like
salsolinol.
1921
Pankow JF (1989) Overview of the gas phase retention
volume behavior of organic compounds on poly-
urethane foam. Atmos. Environ. 23: 1107}1111.
Pankow JF (1988) Gas phase retention volume behavior
of organic compounds on the sorbent poly(oxy-m-ter-
phenyl-2
,5
-ylene). Anal. Chem. 60: 950}958.
Pellizzari ED and Krost KJ (1984) Chemical transforma-
tions during ambient air sampling for organic vapors.
Anal. Chem. 56: 1813}1819.
Schaefer G, Hamilton W and Mathai CV (1997) Implemen-
ting the NAAQS and FACA subcommittee for ozone,
particulate matter and regional haze. Environ. Man. Oct
1997: 22}28.
Tang H, Lewis EA, Eatough DJ, Burton RM and Farber RJ
(1994) Determination of the particle size distribution
and chemical composition of semi-volatile organic com-
pounds in atmospheric Rne particles with a diffusion
denuder sampling system. Atmos. Environ. 28: 939}947.
Turpin BJ and Huntzicker JJ (1994) Investigation of or-
ganic aerosol sampling artifacts in the Los Angeles
basin. Atmos. Environ. 28: 3061}3071.
Williams EL and Grosjean D (1990) Removal of atmo-
spheric oxidants with annular denuders. Environ. Sci.
Technol. 24: 811}814.
Ethyl Alcohol
The most obvious and speciRc test for heavy drinking
is the measurement of blood, breath or urine alcohol
(ethyl alcohol). However, this simple test cannot dis-
tinguish between acute and chronic alcohol consump-
tion, unless it can be related to an increased tolerance
of alcohol. According to the American National
Council on Alcoholism (NCA), the Rrst-level criteria
for the diagnosis of alcoholism are blood alcohol
exceeding 1.5 g L\1 without gross evidence of intoxi-
cation, over 3 g L\1 at any time, or over 1 g L\1 in
routine examination. The determination of alcohol
has already been the subject of many reviews. The
most important facts are summarized here.
As a Rrst step, various pitfalls and analytical prob-
lems such as interference in alcohol analysis induced
by cleaning the skin with ethanol or isopropanol
before expert venepuncture should be borne in mind.
The stability of ethanol during storage is a problem.
The main factors affecting alcohol determination in
stored blood are the duration and temperature of
storage, with negligible losses in the frozen state, and
the presence of a preservative. Three mechanisms
accounting for these changes are: oxidation (highly
1922 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY

temperature-dependent, needing oxygen from oxy-
haemoglobin), the growth of microorganisms meta-
bolizing ethanol (inhibited by sodium Suoride at
50.5%, w/v) and diffusion from containers owing
to closure failure. A further potentially interfering
factor, especially in autopsy cases, is ethanol produc-
tion in (postmortem) tissues by bacteria and yeasts.
Freezing seems to be the best precaution in order to
maintain the original alcohol levels.
Gas chromatography (GC) is par excellence the
all-purpose technique for the determination of
volatile molecules, such as alcohols and related
compounds. Almost all GC methods for ethanol de-
termination allow the simultaneous measurement of
a wide range of other volatile analytes (alcohols,
aldehydes, ketones, glycols, etc.). Although some of
the earlier techniques have become obsolete, the
incorporation of advances such as headspace
chromatography have extended the popularity of
chromatography. The analytical conditions of the
most interesting methods are summarized in Tables 1
and 2. The following classiRcation has been used.
Direct Injection
Methods using direct injection of whole blood suffer
from the adsorption of undesirable compounds (pro-
teins and other macromolecules) on the column and,
consequently, in most procedures prior dilution or
centrifugation have been used.
With Extraction
For a prior extraction step organic solvents such as
n-propyl acetate, n-butanol or dioxan are used.
With Distillation
Sample and internal standard in sodium tungstate/
sulfuric acid are subjected to distillation. The distil-
late is injected into the column and detection is per-
formed by thermal conductivity or Same ionization.

Table 1 Direct injection gas chromatography. Representative overview of standard procedures for the determination
of ethyl alcohol

Specimen Diluent Column Packing Oven Carrier gas Detection Internal

(mL or g)a (mL) (m;mm I.D.) (mesh) temperature (mL min\1) standard
(3C)

Blood (0.5) Int. standard solution (0.5) 1.8;6 30% Carbowax 20M 100 Nitrogen FID Isobutanol
on Chromosorb W (35)
(60}80)
Blood (0.01) Int. standard solution (0.1) 1.5;4.8 10% Carbowax 400 75 Nitrogen FID n -Propanol
on Chromosorb W (80}100) (75)
Blood Int. standard solution 2;3 (1) 0.2% Carbowax 1500 120 FID n -Propanol
Urine (0.5
L) on Carbopack C
Serum (80}100)
Plasma (2) 30% Carbowax 20M 100 (20) Isobutanol
(0.5
L) on Chromosorb W HP
(60}80)
Serum (0.1) Int. standard solution# 3;3.2 Porapak Q (80}100) 155 Nitrogen FID Acetonitrile
Triton-X-100 (0.1) (18)
Serum (0.2) Int. standard solution (0.2) 30;0.25 Methylsilicone-bonded 35 Helium FID n -Propanol
Sodium tungstate phase (0.25
m)
0.2 mol L\1 (0.2)
Copper (II) sulfate
0.2 mol L\1 (0.2)
Blood Water (50-fold sample vol.) 15;0.53 Polyethylene glycol (1.0
m) 40 Helium FID
(25)
Blood Int. standard solution 1.8;2 Porapak S (80}100) 165 Nitrogen FID Acetonitrile
Urine (twofold) (45)
Serum
Plasma
Blood Sodium tungstate 2;3 Porapak Q (80}100) 180 Nitrogen FID Isopropanol
(0.1}0.3) 12.5% (0.2) (30)
Sulfuric acid
0.33 mol L\1 (0.2)
Blood (0.2) Int. standard solution (0.8) 1.2;4 5% Carbowax 20M 100 Helium FID n -Butanol
on Supelcoport (100}120) (30)

a
mL for serum/plasma/urine or g for blood.
Selection according to Tagliaro et al. (1992) Chromatographic methods for blood alcohol determination. Journal of Chromatography 580: 161.
 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY 1923

Table 2 Headspace gas chromatography. Representative overview of standard procedures for determination of ethyl alcohol

Specimen Incubation Packing Oven Carrier gas Detection Internal

(mL or g) a (mesh) temperature (mL min\1) standard
Temperature Time (3C)
(3C) (min)

Blood (0.02) 60 3 Porapack Q (80}100) 150 Nitrogen FID n -Propanol

(30)
Blood 60 30 5% Carbowax 20M 65}110 Nitrogen n -Propanol
Serum (0.5) on Carbopack B (60}80) (30)
Blood (0.2) 60 20 (1) 0.2% Carbowax 1540 85}100 FID tert -Butanol
on Carbopack C (60}80)
(2) 15% Polyethylene glycol
on Celite (60}100)
Blood (0.2) 20}40 30 0.2% Carbowax 1500 125 Nitrogen FID n -Propanol
on Carbopack C (80}100) (20)
Blood 20}40 30 Methylsilicone 35}40 Helium FID n -Propanol
(25)
Blood (0.5) 55 12 (1) Methylsilicone (megabore) 45 Helium FID n -Propanol
(2) DB-wax (megabore) (7.5)
Blood 40 18 (1) 0.2% Carbowax 1500 100 Nitrogen FID n -Propanol
Urine (0.1) on Carbopack C (80}100) (20)
(2) 5% Carbowax 20M
on Carbopack B (60}80)
(3) 15% Carbowax 20M
on Chromosorb W
Plasma 25 N Porapak S (80}100) 165 Nitrogen FID
(45)

a
mL for serum/plasma/urine or g for blood.
Selection according to Tagliaro et al. (1992) Chromatographic methods for blood alcohol determination. Journal of Chromatography
580: 161.

Headspace Methods the development of this technique. Chromatograms

The most important advantage is the prevention of
column contamination.
Methods requiring solvent extraction or distillation
should be considered obsolete mainly because they
are time- and sample-consuming and not susceptible
to automation. Direct injection and headspace GC
are the only techniques in general use that can be fully
and easily automated. The description of direct injec-
tion technique is mostly connected with the dilution
of the sample (mostly with aqueous solutions con-
taining the internal standard) and with the injection
of small volumes. Additional protection from con-
tamination can be obtained with a glass sleeve in-
serted in the injection port or with a pre-column glass
insert Rlled with a silanized glass wool plug. Triton
X-100 has been reported to improve the performance
of the direct injection of serum by acting as a protein-
dispersing agent. Protein precipitation, which can be
carried out in conjunction with the addition of the
sample with the internal standard, has been proposed
as a simple means of overcoming the problems related
to the injection of whole blood. Headspace GC for
blood alcohol analysis was the subject of a review in
1975 by Machata who made many contributions to
are shown in Figure 1. Headspace analysis prevents
any contamination of the column and injector with
involatile material and is preferred in routine labora-
tories. Also, reproducibility is often better than in
direct injection (typical within and between-run coef-
Rcients of variation(1.5% and (2.5%, respective-
ly). Analytical problems arise concerning the choice
of the sample equilibration temperature; oxidation of
ethanol takes place at temperatures exceeding 403C,
but higher temperatures increase the air}blood parti-
tion coefRcient and, consequently, the sensitivity. The
conversion of ethanol into acetaldehyde is reportedly
inhibited by the addition of sodium nitrite or sodium
dithionite. Increased sensitivity due to a salting-out
effect is obtained using sodium chloride, sodium
nitrite, potassium carbonate, sodium Suoride and
ammonium sulfate. In such non-ideal solutions, the
vapour pressures of volatile components at a Rxed
temperature have been reported to depend on the
water content of the sample.
An additional advantage of headspace technique
is the complete elimination of matrix-related effects,
which prompted its use for the analysis of tissues,
stool samples or other biological material. A new pro-
cedure is the headspace}solid-phase microextraction
1924 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY
Figure 1 Representative headspace gas chromatograms de-
termining alcohol concentrations in human serum samples. A,
Blank (serum); B, 0.48 g L\1; C, 1.95 g L\1. Retention times:
EtOH, 1.65 min; t-butanol, 2.2 min.
(HS-SPME) technique, based on the adsorption of
analytes directly from the headspace on to a coated
fused silica Rbre. Various Rbres for different analytes
are available and a 65
m Carbowax/divinylbenzene
coating is used for alcohols.
Alcohols can be efRciently separated with different
GC columns and the choice is often only based on
practical considerations such as total analysis time,
cost, column life and the possibility of using the same
column for different analyses. Carbopack B coated
with Carbowax 20M is superior to Carbopack
C coated with Carbowax 1500 for the determination
of acetaldehyde and methanol and is also superior to
adsorption chromatography on Porapak Q and Chro-
mosorb 102. Separation is generally carried out under
constant temperature conditions; temperature pro-
gramming has been used for the simultaneous deter-
mination of less volatile compounds. Detection is
universally carried out by a Same ionization detector
(FID). Capillary chromatography (Carbowax 20M)
allows a higher separation performance and easier
coupling with mass spectrometry, which is preferred
for the determination of lower volatile alcohols.
Congeners
Besides ethanol, alcoholic beverages contain up to
800 Savour compounds and some of these congeners
can be found in sufRcient quantities to allow their
detection in the blood of the consumer. There are
characteristic differences in the congener content of
alcoholic beverages. A close correlation between the
consumed amount of a congener alcohol and the
resulting blood level can be helpful for the evaluation
of allegations concerning alcohol intake in forensic
cases, especially when determining types of drinks
and when estimating the time of drinking (Figure 2).
The sensitivity of conventional headspace GC is sufR-
cient for blood ethanol determinations down to
0.01 g L\1, but for the detection of congener alcohols
the limits of detection had to be improved to
0.01 mg L\1. Some procedures contain special
sample preparation steps, which include homogeniz-
ation by ultrasound and/or ultraRltration. As the long
chain alcohols are partly or completely bound to
glucuronic acid, incubation with -glucuronidase is
necessary. Using a temperature programme and capil-
lary columns the loading capacity can be enhanced by
a cryofocusing technique.
Methanol is an important congener of most alco-
holic beverages. Metabolism of methanol via liver
alcohol dehydrogenase is inhibited by ethanol levels
exceeding 0.4 g L\1. Consequently, excessive and
prolonged drinking results in high blood methanol
levels. Increased blood methanol levels are frequently
found in drunken drivers and alcoholics. On the basis
of these Rndings, blood methanol levels exceeding
10 mg L\1 have been suggested to be an indicator of
alcoholism. Additionally higher concentrations of
acetone and propanol-2 have been proposed as an
indication of drinking behaviour. This phenomenon
is caused by reciprocal formation through the alcohol
dehydrogenase system. If the sum of the concentra-
tions exceeds 9 mg L\1, heavy drinking is suspected.
However, due to the effects of metabolic disorders
(ketosis, diabetes, hunger, physical stress), the signiR-
cance has been regarded as very low.
Ketone Bodies
In many forensic cases alcohol abusers have been
found dead and the cause of death cannot be ascer-
tained. In order to examine the possible role of
ketoacidosis as the cause of death the concentrations
of ketone bodies (acetone, acetoacetate, D--hy-
droxybutyrate) have to be determined in postmortem
blood specimens. The phenomenon of ketoacidosis is
often seen as typical in periods of abstinence with low
intake of food. It is due to the accumulation of D--
hydroxybutyrate and acetoacetic acid. The accumula-
tion is probably the result of various factors such as
volume depletion and starvation, which have a
lipolytic effect.
A routine procedure is a coupled enzymatic head-
space GC method (Figure 3). This procedure is based
 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY 1925

Figure 2 Total ion chromatograms ((A) selected ion monitoring and (B) full scan mode) of a standard solution of 28 substances
relevant in congener analysis in concentrations of 2 mg L\1 (methanol 10 mg L\1, acetaldehyde 0.5 mg L\1). 1, Acetaldehyde; 2,
methanol; 3, ethanol; 4, propionaldehyde; 5, acetone; 6, propanol-2; 7, methyl acetate; 8, t -butanol (internal standard); 9, i -
butyraldehyde; 10, propanol-1; 11, n -butyraldehyde; 12, methyl ethyl ketone; 13, ethyl acetate; 14, butanol-2; 15, i -butanol; 16,
i -valeraldehyde; 17, 2-methylbutyraldehyde; 18, butanol-1; 19, n -valeraldehyde; 20, 1,1-diethoxyethane; 21, 3-hydroxybutanone-2; 22,
3-methylbutanol-1; 23, 2-methylbutanol-1; 24, i -butyl acetate; 25, pentanol-1; 26, butyl acetate; 27, ethyl lactate; 28, hexanol-1. GC
parameter: HP 5890 II GC with HP MSD 5972, equipped with a DB 624 column (60 m;0.32 mm, df"1.8
m); helium flow 1 mL min\1;
injector 1503C; detector 2003C; oven initially 303C for 8 min, 33C min\1 up to 1803C. (Reproduced with permission from Roemhild
W (1998) Congener analysis by means of ‘headspace’I GC/MS. Blutalkohol 35: 10.)

on enzymatic dehydrogenation of D--hydroxy- portions are taken from each sample. One portion is
butyrate into acetoacetate and subsequent decar- heated to 603C in a headspace sampler, which gives
boxylation of this compound into acetone. Three the free acetone. Acetoacetate is converted into

Figure 3 Schematic presentation of a standard procedure for determination of ketone bodies in blood specimens. Three portions are
taken from each sample to determine free acetone and the sums of acetone#acetoacetate and acetone#acetoacetate#D--
hydroxybutyrate.
1926 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY
acetone by decarboxylation at 1003C in the second
portion. This part gives the combined amount of
acetone and acetoacetate. In the third portion,
D--hydroxybutyrate is Rrst enzymatically dehyd-
rogenized into acetoacetate by D--hydroxybutyrate
dehydrogenase and then decarboxylated into acetone.
QuantiRcation of acetone then yields the molar equiv-
alent of the total ketone bodies. Omission of the
enzymatic stage of the analysis allows quantiRcation
of the molar equivalent of acetone and acetoacetate
present, and the substraction of this value from total
ketone quantitation allows calculation of the D--
hydroxybutyrate concentration.
The reported ketone body concentrations vary
a lot. It was held that if the ketone body concentra-
tion of the blood exceeds 531
mol L\1 and if there is
no other plausible cause of death in a group of alco-
hol abusers, the term ketoalcoholic death should be
used. In another study it was pointed out that very
hight levels, above 10 mmol L\1, are indicative of
profound alcoholic ketoacidosis.
Ethyl Glucuronide
Ethyl glucuronide (EtG) is a minor metabolite of
ethanol and is formed from ethanol by conjugation
with uridine diphosphate (UDP)-glucuronic acid. EtG
has been detected in human urine, serum and clipped
hair samples of ethanol consumers. The formation of
EtG depends on the serum ethanol concentration. It
was shown that serum EtG concentrations higher
than 5 mg L\1 may indicate alcohol misuse, espe-
cially if the serum ethanol concentration is less than
1 g L\1. The EtG concentration declines exponenti-
ally with a half-life of 2}3 h and testing for EtG is
restricted to a period of about 6}18 h after drinking,
depending on the ethanol dose and individual meta-
bolism. In forensic cases testing is predominantly
indicated when the ethanol determination gives nega-
Figure 4 Mass spectrum of the triacetyl derivative of ethyl glucuronide.
tive results and consumption is denied. For retrospec-
tive studies detection of EtG in hair samples also
seems to be possible. However, if excessive ethanol
consumption over a period of months or years pro-
vokes a stimulation of glucuronyltransferase in the
liver, the extent of the EtG formation might be an
indicator of ethanol abuse.
For the determination of EtG in serum the sample
is precipitated with acetone or methanol and the
dried supernatant is derivatized by addition of acetic
anhydride and pyridine. A mass spectrum of the
triacetyl derivative is shown in Figure 4. Hair samples
are extracted with methanol, including treatment by
ultrasound prior to derivatization. On an OV-1 capil-
lary column, the retention index is 1920. Gas
chromatography}mass spectrometry (GC-MS) was
performed with an electron energy of 70 eV and gave
the following m/z values (intensities higher than 20%
in parentheses): 85 (53), 88 (41), 101 (38), 113 (66),
114 (42), 115 (100), 117 (47), 130 (25), 157 (73),
142 (25) and 143 (28); there is no parent peak. An
m/z value of 303 (1%, M-45) indicates that EtG is
decarboxylated.
Fatty Acid Ethyl Esters
Fatty acid ethyl esters (FAEE) are formed by an
enzyme-mediate esteriRcation of ethanol with fatty
acids or fatty acyl-coenzyme A. It has been shown
that FAEE and the FAEE synthase are predominantly
present in those organs most often damaged by
ethanol abuse, notably the pancreas and liver. This
has led to speculation that FAEE, lipids more hydro-
phobic than triglycerides, are mediators of ethanol-
induced organ damage. Following ethanol consump-
tion by humans, FAEE have also been found in serum
lipoproteins. Recently it was reported that the con-
centration of FAEE in the blood closely parallels the
concentration of blood ethanol. In serum samples of
 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY 1927

subjects who had blood ethanol concentrations

'1.5 g L\1, FAEE concentrations ranged up to
2500 nmol L\1 and were still detectable 24 h after
ethanol ingestion. However, serum FAEE may evolve
into both a short-term and long-term marker of
ethanol ingestion. In forensic cases the determination
of a recent intake of ethanol may be necessary.
A negative blood ethanol with a positive FAEE test is
consistent with ethanol intake 4}24 h before blood
collection. Additionally it has been reported that
FAEE are present in signiRcantly higher amounts in
postmortem adipose tissues obtained from indi-
viduals with a history of chronic alcohol abuse, with
ethanol-induced organ damage at autopsy and zero
blood ethanol at the time of death (mean$SEM
equals 300$46 nmol g\1) compared to those from
a control group without a history of chronic ethanol
ingestion, without ethanol-related organ damage and
with zero blood ethanol at the time of death
(43$13 nmol g\1; Figure 5).
Studies on FAEE frequently involve isolating the
compounds by liquid}liquid extraction and thin-layer
chromatography (TLC) prior to identiRcation and
quantiRcation. The isolation of FAEE by these
methods is especially suitable for adipose tissue. Figure 5 Analysis of FAEE from human plasma. Lipids from
sera of patients with markedly elevated blood ethanol levels were
Sample material (1}2 g) is extracted in acetone
extracted into hexane and applied to an aminopropyl silica col-
(10% w/v) and the lipids are separated by TLC on umn. Lipids eluted from the column were dried under nitrogen to
silica gel using a petroleum ether/diethyl ether/acetic a small volume and an aliquot injected into a gas chromatograph
acid (75:5:1) solvent system. Fatty acid ethyl esters, } mass spectrometer (WCOT Supelcowax capillary column). The
RF"0.5, are identiRed by comparison with stan- peaks identified as FAEEs are labelled: E 16:0, ethyl palmitate;
E 17:0, ethyl heptadecanoate; E 18:0, ethyl stearate; E 18:1, ethyl
dards and eluted from the silica gel with acetone. The
oleate; E 18:2, ethyl linoleate; E 20:4, ethyl arachidonate. (Repro-
reproducibility of this procedure is sometimes a prob- duced from Bernhardt TG et al. (1995) Purification of fatty acid
lem and the method often results in low yields. The ethyl esters by solid-phase extraction and high-performance
small amounts of the very hydrophobic FAEE present liquid chromatography. Journal of Chromatography B 675: 189,
in human plasma after ethanol ingestion are com- with permission from Elsevier Science.)
monly lost during extraction. As with fatty acids,
FAEE moieties which contain two or more double
bonds can be oxidized within minutes on a dried TLC edly caused by the accumulation of the congener
plate and are thereby lost prior to quantiRcation. To alcohol, methanol, during chronic alcohol abuse.
enhance the recovery of the relatively small amounts The GC analysis of FAME after esteriRcation of
of FAEE, an effective solid-phase extraction (SPE) lipids was the subject of an excellent review by Eder
method for isolation is preferred. Extraction of FAEE in 1995 and the comments are applicable to FAEE.
from serum is initiated by the addition of The most critical step in the GC analysis of FAME is
acetone/hexane solution. After being dried and re- sample introduction. The classical split injection tech-
suspended in hexane the extract is applied to an nique, which is the most widely used procedure, has
aminopropyl silica SPE column with simultaneous the potential disadvantage of boiling-point discrim-
elution of FAEE and cholesteryl esters from the col- ination. Cold injection of the sample, either on-col-
umn with hexane. The FAEE can then be separated umn or by programmed-temperature vaporization,
from cholesteryl esters, if necessary, by chromatogra- does not present this problem and is therefore prefer-
phy on an octadecylsilyl (ODS) SPE column and red. Separation of FAME can be carried out with
elution with isopropanol/water (5:1, v/v). Recently nonpolar, polar and very polar stationary phases. The
a relationship between various levels of alcohol con- polarity inSuences the retention times, especially
sumption and the appearance of fatty acid methyl those of polyunsaturated FAME. The resolution ca-
esters (FAME) in postmortem tissue samples have pability is highest in columns with very polar phases.
been reported. In addition, this connection is suppos- However, very polar phases have a shorter lifetime
1928 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY

Table 3 Selection of procedures for determination of tetrahydroisoquinolines (TIQ) and tetrahydro--carbolines (THBC)

Sample material Analytes Work-up procedure Packing (mesh)/column Limit of detection

(m;mm I.D.)

Tissue and body Various TIQs and Al2O3 extraction; fluor- 3}5% OV-17, SE-30, SE-54, 0.2}50 pg per sample
fluids catecholamines acylation; GC with electro- XE-60 or GE
chemical detection XF-1150 on Gas
(GC-ECD) Chrom Q (80}100) (6 ft;2)
Brain Salsolinol Liquid}liquid reextraction; 3% OV-1 on Gas Chrom 10 pg per sample
fluoracylation; GC-ECD Q (100}120) (6 or 8 ft;2)
Urine TIQs Liquid}liquid reextraction; 3% OV-1 on Gas Chrom Q
trimethylsilyl (TMS) (100}120) (6 ft;2)
derivatives; GC with mass
spectrometry (MS)
Blood, platelets, Various THBCs Liquid}liquid reextraction; 2% SP-2250 (4 ft;2) or SE-30 1 pmol per sample
plasma and brain heptafluorobuturyl (HBF) (15;0.3) on Chromosorb
derivatives; GC-MS W-HP (100}120)
Brain and bio- Salsolinol and others Al2O3 extraction with 1% OV-17 (2.5;2) or 1 pmol per sample
logical fluids deuterated standards; SE-54 (25;0.2)
fluoracylation; GC-MS
Biological fluids THBCs Liquid}liquid extraction with SE 52 W COT (20;0.25) 0.3 pmol mL\1
and foods deuterated standards;
fluoracylation; GC-MS
Brain Nor salsolinol Amberlite extraction; 2% SP}2250 on Chromosorb 1 ng g\1
propionyl derivatives; W-HP
GC-MS (100}120) (4 ft;2)
Brain and foods TIQ and N- Liquid}liquid extraction; 3% OV-17 on Shimalite 0.25 ng per sample
methyl-TIQ HFB derivatives; GC-MS (80}100) (2;2.5) or OV-1
or OV-101 or DB-17 (25;0.2 mm)
Brain and foods Various THBCs Liquid}liquid extraction; SE 52 WCOT (20;0.35 mm) 0.1}0.5 ng per sample
pentafluorobenzyl
(PFP) derivatives; GC-MS
Brain and foods 1-methyl-THBC Liquid}liquid extraction; OV-1701 (25;0.25) 10 fg per sample
TFA derivative; GC-MS
with negative CI
Urine Various THBCs Combined liquid}liquid OV-1 (12;0.2 mm) 100 pg mL\1
and TIQs and solid-phase extraction;
carbomethoxy/propionyl
derivatives; GC-MS
Urine 1-methyl-THBC Liquid}liquid extraction; RTX-cross-bonded SE-30
derivatization with (R )-(!)- (30;0.25 mm)
2-phenylbutyryl chloride
(PBC)-enantiomeric composi-
tion; GC-NICI-MS
Plasma and urine Salsolinol and others Solid-phase extraction over BGB-silaren (30;0.32 mm) 100 pg mL\1
phenylboronic acid (PBA)
cartridges; two-step derivati-
zation (TMS-PBC)-
enantiomeric composition;
GC-MS
Urine Salsolinol Extraction and derivatization DB-5 (30;0.25) 10 fmol mL\1
in one step by Schotten}
Baumann two-phase reaction
utilizing pentafluorbenzoyl-
chloride
Brain THBC and 1-methyl- Liquid}liquid extraction; RTX-cross-bonded SE-30 20 pg per sample
THBC TFA derivatives; (30;0.25 mm)
GC-NCI-MS
Urine Salsolinol and Solid-phase extraction OV-1 (12;0.2 mm) 100 pg per sample
norsalsinol (PBA); propionyl derivative;
GC-MS

TFA, trifluoroacetyl; CI, chemical ionization; NCI, negative chemical ionization; TMS, trimethylsilyl.
 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY
than nonpolar phases and, in many cases, nonpolar
phases provide adequate separation. The most impor-
tant very polar phases are composed of 100%
cyanoethylsilicone oil (SP-2340, OV-275), 100%
cyanopropylsilicone (CP-Sil 88) or 68% biscyano-
propyl/32% dimethylsiloxane (SP-2330).
The most important stationary phases of inter-
mediate polarity are polyethylene glycol (DB-Wax,
Supelcowax 10, Carbowax 20M), acidiRed polyethy-
lene glycol (FFAP), 86% dimethyl/14% cyanopropyl-
phenylpolysiloxane (DB-1701), and methylsilicone
polymer, 25% cyanopropyl/25% phenyl/50%
methyl (OV-225, DB-225, SP-2300). Intermediate-
polarity columns allow acceptable separation of
FAME from biological samples such as plasma or
adipose tissue and combine the advantages of a rela-
tively high resolution capability with those of a rela-
tively high thermal stability. The most important
nonpolar stationary phases are based on methyl-
silicones (SPB-1, SPB-5), 95% dimethyl/5%
diphenylpolysiloxane (DB-5, CP-Sil 8CB) or 100%
dimethylpolysiloxane (DB-1, Rt-1, SP-2100, OV-1,
OV-101, CP-Sil 5CB). FAEE are eluted according to
Figure 6 Electron impact mass spectra of (A) salsolinol and (B) norsalsolinol after derivatization with N-methyl-N-trimethylsilyltri-
fluoracetamide (MSTFA) and (R )-(!)-2-phenylbutyrylchloride.
1929
their boiling points. Therefore, unsaturated com-
pounds elute before being saturated. This elution
order is the reverse of that on very polar and polar
columns. The main disadvantage of nonpolar col-
umns is partial overlapping of some unsaturated
FAME. Advantages are high thermal stability, a wide
range of operating temperatures and chemical inert-
ness.
In summary, FAEE detection can lead to a major
improvement in the monitoring of ethanol ingestion
and the treatment of ethanol-induced organ damage.
Condensation Products
During the past decades research in the aetiology of
alcoholism has focused on the hypothesis that con-
densation products formed endogenously by the reac-
tion of indolalkylamines and catecholamines with
aldehydes or pyruvic acid might be implicated in
neurochemical mechanisms underlying addictive al-
cohol drinking. The formation of 1,2,3,4-tetrahydro-
-carbolines (THBC) and 1,2,3,4-tetrahydroiso-
quinolines (TIQ) via the Pictet}Spengler reaction is
1930 III / ALCOHOL AND BIOLOGICAL MARKERS OF ALCOHOL ABUSE: GAS CHROMATOGRAPHY
Figure 7 Identification of dopamine, (R )-(#)- and (S )-(!)-salsolinol and norsalsolinol in an authentic urine sample of a chronic
alcoholic.
extensively documented. Salsolinol, which might be
formed in vivo by ring cyclization of dopamine with
acetaldehyde, is one of the most discussed tetrahyd-
roisoquinolines. Several studies have been done to
improve analytical techniques for identiRcation in
human urine, plasma, brain and cerebrospinal Suid
samples. Poor assay speciRcity and possible artefact
formation of the alkaloids during work-up and stor-
age have been suggested to be responsible for contro-
versial reports on the detection of these compounds in
mammalian tissues and Suids after alcohol intake.
The variability of reported levels of Salsolinol might
also be a result of variables, including dietary condi-
tions during the experiments or the duration of
ethanol ingestion and analytical problems associated
with the detectability of the analytes. The presence of
TIQ and THBC compounds has been established us-
ing (radioenzymatic) TLC methods, high perfor-
mance liquid chromatography coupled with electro-
chemical or Suorescence detection, or GC procedures
mostly combined with mass spectrometry (Table 3).
Recently, it has been considered that the (R)-(#)-
and (S)-(!)- enantiomers of salsolinol do not exert
identical biological activities. Thus, methods for the
determination of the enantiomeric composition of
endogenous salsolinol have been developed (Fig-
ures 6 and 7). More experimental work is necessary
to determine whether alcohol really has an inSuence
on the biosynthesis of salsolinol or other condensa-
tion products and if the (S)-(!)-salsolinol enantiomer
is a sufRcient clinical marker to distinguish between
alcoholics and nonalcoholics.
Conclusion
Several chemical abnormalities associated with ex-
cessive alcohol consumption are useful in the diag-
nosis of alcoholism. Additionally, in forensic cases
information can be helpful to evaluate allegations
concerning alcohol intake, especially when determin-
ing the types of drinks and estimating the time of
drinking. In these problems GC procedures measur-
ing the concentration of ethyl alcohol and congeners
or EtG can be helpful. The determination of ketone
bodies is a diagnostic tool in a prospective postmor-
tem toxicology analysis in alcoholics for consider-
ing a ketoalcoholic death. Further studies are
necessary to determine the connection between alco-
hol abuse and the formation of FAEE and condensa-
tion products. Further investigations could lead to
important pathopysiological bases of alcohol drink-
ing behaviour and ethanol-induced organ damage
and ultimately to better forms of prevention and
therapy.
See also: II/Chromatography: Gas: Headspace Gas
Chromatography. Detectors: Mass Spectrometry.
III/Clinical Diagnosis: Chromatography. Forensic
Sciences: Liquid Chromatography.
Further Reading
Bonte W (1987) Begleitstoffe alkoholischer Getra( nke.
LuK beck: Schmidt-RoK mhild.
Bonte W (1990) Contributions to congener research. Jour-
nal of TrafTc Medicine 18: 5.
Eder K (1995) Gas chromatographic analysis of fatty acid
methyl esters (review). Journal of Chromatography
B 671: 113.
Laposata M (1997) Fatty acid ethyl esters: short-term and
long-term serum markers of ethanol. Clinical Chemistry
43: 1527.
Machata G (1975) The advantage of automated blood
alcohol determination by head space analysis (review).
Zeitschrift fu( r Rechtsmedizin 75: 229.
 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY
Musshoff F and Daldrup T (1998) Determination of biolo-
gical markers for alcohol abuse (review). Journal of
Chromatography B 713: 245.
Pounder DJ, Stevenson RJ and Taylor KK (1998) Alcoholic
ketoacidosis at autopsy. Journal of Forensic Sciences 43:
812.
Ruz J, Fernandez A, De Castro MDL and Valcarcel
M (1986) Determination of ethanol in human Suids I.
Determination of ethanol in blood, II. Determination of
thanol in urine, breath and saliva (reviews). Journal of
Pharmaceutical and Biomedical Analysis 4: 545.
ALCOHOLIC BEVERAGES: DISTILLATION
See III / WHISKY: DISTILLATION
ALDEHYDES AND KETONES:
GAS CHROMATOGRAPHY
H. Nishikawa, Gifu Prefectural Institute of
Health and Environmental Sciences, Gifu, Japan
Copyright ^ 2000 Academic Press
Introduction
Simple aldehydes, such as formaldehyde, acetal-
dehyde and acrolein, are known to be hazardous
air pollutants. Aldehydes are emitted from incom-
plete burning of various organic compounds
and from various chemicals, and are formed by
photochemical reaction with hydrocarbons in the
atmosphere.
Volatile ketones are used as solvents in various
chemical plants and laboratories and are emitted
into the atmosphere. The toxicity of ketones is, in
general, not as high as that of aldehydes. Carbonyl
compounds are signiRcant in environmental chem-
istry, i.e. in rainwater and as a photochemical
oxidant.
Separation of aldehydes and ketones is very impor-
tant for the determination of volatile aldehydes.
Usually, analysis of aldehydes is performed by de-
rivatization and gas chromatography (GC) or high
performance liquid chromatography (HPLC). Selec-
tive and sensitive gas chromatographic methods for
separation of aldehydes and ketones are described
below.
1931
Schmitt G, Aderjan R, Keller T and Wu M (1995) Ethyl
glucuronide: an unusual ethanol metabolite in humans.
Synthesis, analytical data, and determination in serum
and urine. Journal of Analytical Toxicology 19: 91.
Tagliaro F, Lubli G, Ghielmi S, Franchi D and Marigo
M (1992) Chromatographic methods for blood alcohol
determination (review). Journal of Chromatography
580: 161.
Thomsen JL, Felby S, Theilade P and Nielsen E (1995)
Alcoholic ketoacidosis as a cause of death in forensic
cases. Forensic Science International 75: 163.
2,4-Dinitrophenylhydrazone
Derivatization
2,4-Dinitrophenylhydrazone (DNPH) derivatives of
aldehydes and ketones have been used in gas
chromatography for many years. The reaction pro-
cedure of aldehyde or ketone is as follows:
Kallio et al. analysed 15 carbonyl compounds (al-
dehydes and ketones) known to be Savour compo-
nents by derivatization/GC with DNPH. The DNPHs
of the carbonyl compounds were prepared by shaking
100
L of each compound with 100 mL of a
saturated solution of DNPH in aqueous 2 mol L\1
hydrochloric acid and allowing the mixture to stand
at room temperature overnight. The precipitated
DNPHs were dissolved in ethyl acetate, then analysed
by GC-FID or dissolved in benzene and analysed by
GC-ECD (electron-capture detector). Packed col-
umns with silicone stationary phases were used. Rela-
tive retention times of DNPHs of aldehydes and
ketones on one of these columns are listed in Table 1
1932 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY

Table 1 Relative retention times (R ) of 2,4-dinitrophenylhyd- detectable concentrations of formaldehyde, acetal-

razones of carbonyl compounds dehyde and acrolein were 85, 140 and 190 ppb,
respectively, for a 10 L gas sample with a relative
Carbonyl compound 2% SE-30 column
200I270 3C a standard deviation of less than 8%. Using this method,
R formaldehyde, acetaldehyde, propionaldehyde, iso-
butyraldehyde, acetone, acrolein, methyl ethyl ketone
Formaldehyde 0.52 and methyl isopropyl ketone (or butyraldehyde) were
Acetaldehyde 0.71
measured in gasoline engine exhaust gas.
Propenal 0.84
Acetone 0.84
Propanal 0.86
2-Methylpropanal 0.94
Benzyloxime Derivatization
2-Butanone 1.00 Derivatization of simple aldehydes and ketones to
3-Methylbutanal 1.13
benzyloximes was developed by Magin. The reaction
2-Butenal 1.16
Hexanal 1.42 proceeds as follows:
Furfural 1.62
1.58b
Heptanal 1.63
Octanal 1.83
Benzaldehyde 1.97
Nonanal 2.03

a
Perkin-Elmer F-11 gas chromatograph equipped with a coiled The GC used for the analysis of these derivatives
glass column, 6 ft long and 1/8 in i.d. was equipped with a nitrogen-selective detector and
b
Relative retention times of secondary peaks.
Reproduced with permission from Kallio H et al. (1972) GasIliquid
chromatographic analysis of 2,4-dinitrophenylhydrazones of car-
bonyl compounds. Journal of Chromatography 65: 355.

and a typical chromatogram of DNPHs of carbonyl

compounds is shown in Figure 1. The peaks of the
DNPHs of propanal, propenal and acetone are not
separated under these conditions. The sensitivity was
of the order of nanograms of carbonyl compounds
when FID was used and Rve-hundred times greater
with an ECD.
Saito et al. reported an improved GC method with
DNPH for the determination of trace low molecular
weight aliphatic carbonyl compounds in auto ex-
haust. They used a glass capillary column, 30 m;
0.27 mm i.d., coated with OV-17 and operated iso-
thermally at 2103C. Aldehydes and ketones in the
exhaust were selectively collected by passing
600 ml min\1 of exhaust gas through two impingers
which were connected in series and which contained
hydrochloric acid saturated with DNPH. The deriva-
tives were extracted twice with chloroform in a separ- Figure 1 Chromatogram of a mixture of 2,4-dinitrophenyl-
hydrazones of carbonyl compounds on a 2% SE-30 column. 1,
ating funnel. After concentration by evaporation un-
Formaldehyde; 2, acetaldehyde; 3, acetone; 4, 2-butanone; 5,
der a stream of nitrogen, anthracene was added as 2-butenal; 6, hexanal; 7, heptanal; 8, octanal; 9, nonanal. Pro-
internal standard, and 1
L of the solution was injec- grammed temperature analysis from 200 to 2703C (43C min\1) on
ted into a GC equipped with an FID. Six aliphatic a Perkin-Elmer F-11 chromatograph equipped with a hydrogen
aldehydes and three aliphatic ketones were analysed flame ionization detector. Injected sample: 1.0
L of ethyl acetate
containing about 1000 ng of each of the derivatives. Attenuation
(Figure 2). The derivatives of C3 carbonyl com-
128, range 1. (Reproduced with permission from Kallio H et al.
pounds, propionaldehyde, acetone and acrolein, were (1972). GasIliquid chromatographic analysis of 2,4-dinitrophenyl-
completely separated, and simultaneously determined hydrazones of carbonyl compounds. Journal of Chromatography
with formaldehyde and acetaldehyde. The minimum 65: 355.)
 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY 1933

Figure 2 Chromatogram of the 2,4-dinitrophenylhydrazones of carbonyl compounds. 1, Formaldehyde; 2, acetaldehyde; 3, acetone;

4, propionaldehyde; 5, isobutyraldehyde; 6, acrolein; 7, methyl ethyl ketone; 8, butryraldehyde; 9, methyl isopropyl ketone; 10, diethyl
ketone; 11, methyl t -butyl ketone; 12, isovaleraldehyde; 13, methyl propyl ketone; 14, methyl s -butyl ketone; 15, methyl isobutyl ketone;
16, valeraldehyde; 17, crotonaldehyde, 18, methyl butyl ketone; 19, capronaldehyde. (Reproduced with permission from Saito T et al.
(1983) Determination of trace low molecular weight aliphatic carbonyl compounds in auto exhaust by gas chromatography with a glass
capillary column. Bunseki Kagaku 32: 33.)

a glass capillary column (12 m;0.4 mm i.d.) coated aldehyde appeared in the reagent blanks. Mass
with a 0.4
m Rlm of free fatty acid phase (FFAP); spectra of these peaks were identical to the spectra of
helium was used as the carrier gas. The column tem- the genuine acetaldehyde derivatives, but the source
perature programme was 100}1803C at 23C min\1;
the temperature was kept at 1803C to the end of
Table 2 Adjusted retention times of O-benzyloxime derivatives
the analysis. The retention times of the ben-
zyloximes of a number of aldehydes and ketones Aldehydes Retention Ketones Retention
(1d7 carbon atoms) are shown in Table 2. The re- time (min) time (min)
sults show almost complete separation under these
conditions. Formaldehyde 6.5 Acetone 12.6
Acetaldehyde 9.9 2-Butanone 14.2
Magin also reported the application of the ben-
Propanal 11.7 2-Pentanone 16.1
zyloxime-GC analysis of aldehydes and ketones to the Butanal 17.6
semi-quantitative analysis of simple monocarbonyls Pentanal 22.5 3-Pentanone 16.8
in cigarette smoke. The smoke was passed through Hexanal 30.0 4-Heptanone 22.1
a silica gel column to trap the carbonyls, followed by Heptanal 36.5
Octanal 43.5 Cyclopentanone 34.5
elution with water. About 15 mL of eluted solution
Cyclohexanone 38.4
was collected in a screw-capped bottle, and the ben- Isobutanal 13.9 Cycloheptanone 45.0
zyloximes of the carbonyls were prepared. Separation Isopentanal 20.0
was accomplished by a temperature-programmed 3-Methyl-2-butanone 14.4
12 m glass capillary FFAP column. An internal stan- Propenal 15.9 3-Methyl-2-pentanone 15.4
2-Butenal 27.9
dard (hexanal) was added both as a reference for
2-Hexenal 44.0 4-Methyl-2-pentanone 15.2
retention time determination, and as an aid in estima- Methacrolein 17.9 5-Hexen-2-one 25.9
ting the amounts of the individual carbonyls in the Benzaldehyde 67.7
smoke samples. Levels of some carbonyls in the ciga-
rette whole smoke samples are shown in Table 3. Perkin-Elmer Model 3920 gas chromatograph equipped with
glass capillary column, 12 m long and 0.4 mm i.d.
One problem with this method was that acetal-
Reproduced with permission from Magin DF (1979) Preparation
dehyde, one of the major carbonyl compounds in and gas chromatographic characterization of benzyloximes and
cigarette smoke, could not be determined since peaks p-nitrobenzyloximes of short-chain (C1IC7) carbonyls. Journal of
corresponding to the benzyloxime derivative of acet- Chromatrography 178: 219.
1934 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY
Table 3 Levels of selected carbonyls in the whole smoke of some cigarettes
Carbonyl Cigarette A
(filter)
Formaldehyde 31 (10}50)
Acetone 400 (325}475)
Propanal 61 (37}100)
Acrolein 23 (13}37)
Methacrolein 17 (14}38)
Butanal 20 (9}29)
Levels are tabulated as average; values in parentheses indicate the range. The values are
given in
g per cigarette.
Reproduced with permission from Magin DF (1980) Gas chromatography of simple
monocarbonyls in cigarette whole smoke as the benzyloxime derivatives. Journal of
Chromatography 202: 255.
in the blanks was unknown. Another problem was
the lack of reproducibility from run to run. For
this reason, the method was described as semi-quant-
itative.
Penta]uorophenylhydrazone
Derivatization
A sensitive and selective GC analysis of lower
aliphatic carbonyl compounds as their penta-
Suorophenylhydrazones was described by Hoshika
and Muto.
The general procedure for the preparation of the
pentaSuorophenylhydrazone derivatives was as fol-
lows: 0.5;10\3 mol of lower aliphatic carbonyl
compound was added, using a 100
L microsyringe,
to 1 mL methanol containing 1.01;10\3 mol pen-
taSuorophenylhydrazine (PFPH). The mixture was
allowed to stand overnight at room temperature.
A 1
L volume of the solution was injected into the
GC. The acetone, acrolein and propionaldehyde de-
rivatives were separated on a 30 m;0.25 mm i.d.
glass capillary column coated with polyethylene
glycol 20 M at 1303C with an FID. This temperature
compares favourably with the 2003C column temper-
ature necessary with DNPH derivatives.
A procedure for the GC determination of low
molecular weight carbonyl compounds in aqueous
Cigarette B Cigarette C
(filter, low delivery) (non-filter)
10 (9}10) 21 (12}30)
137 (130}144) 330 (310}350)
37 (30}40) 50 (50}53)
3 (3}4) 22 (20}25)
18 (18}19) 27 (20}32)
13 (12}13) 18 (17}20)
solution as their pentaSuorophenylhydrazones and its
application to such compounds produced by photoly-
sis of -amino acids were reported by Kobayashi et al.
(Figure 3). The calibration graphs for formaldehyde,
acetaldehyde and isobutyraldehyde showed good lin-
earity in the range 10}40
g. Solutions of DL-alanine,
DL-valine, DL-leucine or DL-isoleucine in phosphate
buffer containing mercury (II) chloride were irra-
diated in a quartz vessel with a 20 W blacklight lamp
(300d400 nm) as a light source for 300 h and the
mixtures were measured by this method. Two peaks
corresponding to formaldehyde and acetaldehyde
were observed for valine.
Figure 3 Chromatogram of some carbonyl compounds as their
pentafluorophenylhydrazones. Conditions: 3% XE-60, 2.0 m
glass column, temperature-programmed from 105 to 1303C at
23C min\1, FID. 1, Formaldehyde; 2, acetaldehyde; 3, acetone;
4, isobutyraldehyde; 5, diethyl ketone; IS, p -xylylene dichloride;
6, methyl isobutyl ketone. (Reproduced with permission from
Kobayashi K et al. (1979) Gas chromatographic determination of
low-molecular-weight carbonyl compounds in aqueous solution
as their pentafluorophenylhydrazones. Journal of Chromatogra-
phy 176:118.)
 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY 1935

Penta]uorobenzyloxime
Derivatization
PentaSuorobenzyloxyamine (PFBOA) was synthesized
by Nambara et al. as a new derivatizing agent for
GC of ketones using electron-capture detection.
Kobayashi et al. reported a GC analysis of low
molecular weight carbonyl compounds in aqueous
solution as their O-pentaSuorobenzyloximes. The
reaction procedure of aldehyde or ketone is as
follows:

Good linearity of response was obtained for for-

The retention times of 13 O-pentaSuoroben-
zyloxime (O-PFBO) derivatives of carbonyl com-
pounds relative to the internal standard obtained on
XE-60 and other columns are shown in Table 4.
When some carbonyl compounds react with PFBOA,
two peaks are obtained, corresponding to syn- and
anti-isomers formed by the condensation reaction
with PFBOA:
maldehyde, acetaldehyde, isobutyraldehyde and di-
ethyl ketone in the range 1}50
g in 0.5 mL of
aqueous solution. The utility of PFBOA was com-
pared with that of PFPH. The formation of the O-
PFBO derivatives of aldehydes was easily achieved
with a much lower concentration of PFBOA than that
required for PFPH. The derivatization of aldehydes
with PFBOA is complete in 20 min at room tem-
perature, but the reaction with ketones pro-
ceeds slowly (within 24 h). The PFBO derivatives
are much more volatile than the corresponding pen-
taSuorophenylhydrazones and therefore the separ-
ation can be carried out at lower temperatures
(70}1003C). It was shown that the PFBO derivatives
are stable in ethyl acetate at room temperature for
several days.

Table 4 Relative retention times of the O -pentafluorebenzyloximes of carbonyl compounds

Parent compound Stationary phase and column temperature (each column: 2 m;3 mm i.d.)

3 % XE-60 (90 3C) 3 % XF-1105 (100 3C) 3%SE-30 (80 3C) 2 % OV-17 (70 3C)

PFBOA 0.62 0.69 0.88 0.65

HCHO 0.19 0.27 0.29 0.19
CH3CHO 0.32, 0.34a 0.48, 0.51a 0.57, 0.59a 0.43
C2H5CHO 0.47, 0.51a 0.80 1.05 0.72
n-C3H7CHO 0.80 1.41 1.83, 1.92a 1.30, 1.38a
iso-C3H7CHO 0.57 1.01 1.32 0.88
n-C4H9CHO 2.72 1.91, 1.99a 2.55, 2.72a 1.76, 1.94a
CH3COCH3 0.40 0.70 0.88 0.65
CH3COC2H5 0.61 1.13 1.50 1.04
CH3CO-iso-C3H7 0.73 1.48 2.08 1.24, 1.43a
CH3CO-iso-C4H9 1.16 2.36 3.38 2.08
C2H5COC2H5 0.85 1.74 2.43 1.61
C2H5CO-n-C3H7 1.27 2.70 3.95 2.61
CH2ClC6H4CH2Clb 1.00 1.00 1.00 1.00

a
Double peaks.
b
Internal standard (
,
-dichloro-p -xylene). PFBOA, Pentafluorobenzyloxyamine. Shimadzu Model GC-4APF gas chromatograph
equipped with an FID.
Reproduced with permission from Kobayashi K et al . (1980) Gas chromatographic determination of low-molecular-weight carbonyl
compounds in aqueous solution as their O-(2,3,4,5,6-pentafluorobenzyl) oximes. Journal of Chromatography 187: 413.
1936 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY
Aldehydes in air are important as pollutants
and photochemical products because they are irri-
tants to the skin, eyes and nasopharyngeal mem-
branes. Trace amounts of formaldehyde in air was
determined as the PFBO derivative using GC-ECD by
Nishikawa et al. Sub-ppb levels of formaldehyde in
air could be determined in this way. Although the
FID or ECD can be used with PFBO derivatives,
the FID has poor sensitivity and in some cases the
peaks of the derivatives suffer interference from co-
existing materials with both detectors. In the case of
analysis of samples with a complex matrix such as
exhaust gas or emission gas, the thermionic detector
(TID) is more effective. The TID is very sensitive and
selective to nitrogen compounds such as oximes and
can therefore be used with advantage for the analysis
of aldehydes as PFBOA derivatives in automobile
exhaust and emission gas. Gas (2}30 L) was collected
by bubbling the exhaust through two impingers con-
nected in series. Each impinger contained 10 mL of
the absorption solution (300 mg L\1 PFBOA ) HCl in
ethanol). The sample gas was drawn at a rate of
0.5 L min\1. After sampling, the absorbed solution
was made up to 20 mL with ethanol and was allowed
to stand for 80 min at room temperature. A 10 mL
portion of the solution and 20 mL of distilled water
were mixed well and the mixture was passed through
a Sep-Pak C18 cartridge. The cartridge was eluted
with 1.5 mL hexane and the eluate was analysed with
an SE-52 fused silica column (25 m;0.25 mm i.d.) at
1303C with an FTD. The relative retention times for
aldehyde and ketone derivatives are shown in
Table 5. Except for formaldehyde, acetone and di-
Table 5 Relative retention times of O-pentafluorobenzyloxime
derivatives
Parent compound Relative retention time
Formaldehyde 1.00
Acetaldehyde 1.30, 1.34a
Propionaldehyde 1.75, 1.79a
Butyraldehyde 2.52, 2.58a
Acrolein 1.76, 1.88a
Crotonaldehyde Ib
Acetone 1.63
Methyl ethyl ketone 2.20, 2.24a
Methyl isopropyl ketone 2.69, 2.75a
Diethyl ketone 3.00
Methyl propyl ketone 3.03, 3.14a
GC conditions: SE-52 capillary column (25 m;0.25 mm i.d.),
temperature 1303C.
a
Double peaks; b no peaks appeared.
Reproduced with permission from Nishikawa H et al. (1987)
Determination of aldehydes in exhaust gas and thermal degrada-
tion emission using volatile derivatization and capillary GC with
flame thermionic detector. Bunseki Kagaku 36: 381.
ethyl ketone, each derivative has syn- and anti-
isomers and the double peaks appear for the isomers.
The peaks of the formaldehyde, acetaldehyde,
propionaldehyde and butyraldehyde derivatives were
completely separated from those of the ketones. Al-
though the Rrst peak of propionaldehyde was not
separated from the Rrst peak of acrolein, the second
peak of propionaldehyde was separated from the sec-
ond peak of acrolein. Therefore, it can be judged
whether the Rrst peak of propionaldehyde contained
the peak of acrolein or not. The calibration graphs
obtained show good linearity over the range 0}5
g
for formaldehyde, 0}4
g for acetaldehyde, 0}15
g
for propionaldehyde and 0}13
g for butyraldehyde
in 1 mL of the absorption solution. The determina-
tion limits ranged from 14 ppb (v/v) for formalde-
hyde to 67 ppb for propionaldehyde when a 30 L gas
sample was used.
This method is very selective, without inter-
ference from ketones, and is well-suited to analyse
simple aldehydes in various exhaust gas samples.
The PFBO method was applied to analysis of carbon-
yl compounds in clothes, river water, seawater,
tap water, indoor air samples and environmental air
samples.
Thiazolidine Derivatization
Hayashi et al. developed a method to determine
formaldehyde and methyl glyoxal in foods and bever-
ages. This method is known as the cysteamine
method. Volatile aliphatic aldehydes and ketones re-
act with 2-aminoethanethiol (cysteamine) to form
thiazolidine compound as follows:
Only one derivative is formed from each aldehyde
or ketone; the reaction proceeds rapidly under mild
conditions and the derivatives can be readily separ-
ated on a capillary column. The derivatives can be
detected selectively with a nitrogen}phosphorus de-
tector (NPD). Carbonyl compounds in the headspace
gases above heated pork fat were analysed as
thiazolidine derivatives by Yasuhara and Shibamoto
(Figure 4). A GC with a 30 m;0.25 mm i.d. DB-
Wax fused silica capillary column and an NPD was
used for the qualitative and quantitative analysis of
the carbonyl compounds in the samples. The column
temperature was held at 503C for 2 min and then
programmed to 1903C at 33C min\1. Nine aldehydes
and four ketones were analysed as thiazolidine deriv-
atives in traps containing aqueous cysteamine or
 III / ALDEHYDES AND KETONES: GAS CHROMATOGRAPHY 1937

Figure 4 Chromatogram of the thiazolidine derivatives of carbonyl compounds produced from heated pork fat. a, Acetaldehyde;
b, formaldehyde; c, propionaldehyde; d, 2-pentanone; e, butyraldehyde; f, 2-hexanone; g, valeraldehyde; h, 2-heptanone; i, hexylal-
dehyde; j, 2-octanone; k, heptaldehyde; l, octaldehyde; m, nonylaldehyde; IS, N-methyl acetamide. (Reproduced with permission from
Yasuhara A (1991) Analysis of lower aldehydes in air. Journal of Environmental Analytical Chemistry 1: 253.)

aqueous sodium bisulRte. The major compounds pro-
duced from the samples were pentanal, hexanal and
heptanal. Generally, aqueous cysteamine was more
efRcient at trapping carbonyl compounds than was
aqueous sodium bisulRte but formaldehyde and acet-
aldehyde were trapped better by sodium bisulRte.
Acrolein and malonaldehyde were analysed as 1-
methyl-2-pyrazoline and 1-methyl-pyrazole, respec-
tively, from the trap containing methylhydrazine.
Good separation of volatile aldehydes and ketones
was obtained by this method.
The cysteamine derivatization method has been
applied to the determination of trace amounts of
carbonyl compounds in automobile exhaust, air sam-
ples, the headspace of heated food oils, foods and
beverages.
Conclusion
The DNPH-GC and PFBO-GC methods are now
widely used for the derivatization and gas chrom-
atographic analysis of aldehydes and ketones.
The PFBO-GC and cysteamine-GC methods will
become more common. These reactions proceed
under mild conditions and the derivatives are well
separated with a capillary column at comparatively
low temperatures. The GC-MS measurement of these
derivatives will be utilized for analysis of trace
amounts of aldehydes and ketones in a variety of
samples.
See also: II/Chromatography: Gas: Derivatization; De-
tectors: Selective.
Further Reading
Afgran BK, Chau ASY (1989) Analysis of Trace Organics in
the Aquatic Environment. Florida: CRC Press.
Kallio H, Linko RR and Kaitaranta J (1972) Gas}liquid
chromatographic analysis of 2,4-dinitrophenylhyd-
razones of carbonyl compounds. Journal of Chromatog-
raphy 65: 355.
Katz M (ed.) (1977) Methods of Air Sampling and Analysis.
Washington, DC: American Public Health Association.
Izard C and Libermann C (1978) Acrolein. Mutation Re-
search 47: 115.
Nishikawa H and Sakai T (1995) Derivatization and
chromatographic determination of aldehydes in gaseous
and air samples. Journal of Chromatography A 710:
159.
Yasuhara A (1991) Analysis of lower aldehydes in air.
Journal of Environmental Chemistry 1: 253.
1938 III / ALKALOIDS / Gas Chromatography
ALKALOIDS
Gas Chromatography
M. Muzquiz, SGIT-INIA, Madrid, Spain
Copyright ^ 2000 Academic Press
Introduction
Alkaloids are an important class of compounds that
have pharmacological effects on the human body.
These compounds can be found in natural products
such as plants, and the type and amount of these
alkaloids varies greatly, depending on the portion
of plant analysed and the stage of maturation. Al-
though alkaloids have traditionally been isolated
from plants, an increasing number are to be found
in animals, insects, marine invertebrates and micro-
organisms.
There is no clear deRnition of what constitutes an
alkaloid, but these compounds do share the following
characteristics: they are basic components that con-
tain nitrogen; they are mostly complex components,
derived biosynthetically from various amino acids;
and they show pronounced pharmacological effects
on various tissues and organs of humans and other
animal species.
Pelletier deRnes an alkaloid as ‘a cyclic compound
containing nitrogen in a negative oxidation state
which is of limited distribution in living organisms’.
This deRnition includes both alkaloids with nitrogen
as part of a heterocyclic system as well as the many
exceptions with extracyclic bound nitrogen (Figure 1).
Although a wealth of information is available on
the pharmacological effects of these compounds, little
is known about how plants synthesize these substan-
ces or about how this synthesis is regulated. Alkaloids
belong to the broad category of secondary metab-
olites. This class of molecule has historically been
deRned as a naturally occurring substance that is not
vital to the organism that produces them. Alkaloids
have traditionally been of interest only due to their
pronounced and various physiological activities in
animals and humans. A picture has now begun to
emerge that alkaloids do have important ecochemical
functions in the defence of the plant against patho-
genic organisms and herbivores and are found to play
an important role in plant interactions with animals
and higher and lower plants. Alkaloids are now gen-
erally considered to be part of an elaborate system of
chemical defence in plants; indeed, the same seems to
be true in vertebrates, invertebrates and microorgan-
isms. Alkaloids have now been isolated from such
diverse organisms as animals, insects, marine organ-
isms, microorganisms and lower plants, although it is
not yet clear whether de novo alkaloid biosynthesis
occurs in each organism.
In the past ten years there has been an increasing
interest in the isolation and determination of alkal-
oids in plant materials, in pharmaceutical products,
and in other samples of biological interest. In addi-
tion, numerous alkaloids have been synthesized and
chemically characterized. The active agents of around
13 000 plant species are known to have been used as
drugs throughout the world. Some are used as pure
compounds for therapeutic purposes (such as the
narcotic and analgesic, morphine; the analgesic and
antitussive, codeine; and the chemotherapeutic
agents, vincristine and vinblastine) or as teas and
extracts. Plant constituents have also served as mod-
els for modern synthetic drugs, such as atropine for
tropicamide, quinine for chloroquine, and cocaine for
procaine and tetracaine. Alkaloids can also be found
in the stimulants caffeine in coffee and tea and nic-
otine in cigarettes. Currently, much work is being
done to discover new alkaloid molecules for different
applications such as new antiviral and tumour treat-
ments.
However, many alkaloids are toxic substances and
it is important to evaluate these. The vegetables
Solanaceae, which contain steroidal glycoalkaloids,
and Leguminosae, which contain quinolizidine alkal-
oids, are the principal food crops that contain alkal-
oids. Grain legumes are extremely important owing
to their signiRcance in human and animal nutrition.
They also conserve the soil and Rx nitrogen, and are
used as sources of timber, fuel oils, etc. Plants of the
Leguminosae rank second in economic importance
only to those of the Gramineae, and the demand for
legumes is likely to escalate as humans begin to utilize
more marginal agricultural lands to provide food for
the increased population. The largest legume subfam-
ily is the Papilionaceae, which embraces approxim-
ately 440 genera and 12 000 species in 32 tribes, as
recently reclassiRed by Polhill. Over 450 alkaloids
have been reported to occur in plants of the Legumin-
osae, with the majority of such compounds occurring
in papilionaceous species. Quinolizidine alkaloids
(QA), contained in lupins, are the largest single
group of legume alkaloids. Since lupin seeds contain
 III / ALKALOIDS / Gas Chromatography 1939

up to 50% protein and up to 20% lipids, they are of been consumed for centuries in European countries,
interest in terms of animal and human nutrition. while L. mutabilis (tarwi) is an important component
Lupinus luteus, L. albus and L. angustifolius have of the South American diet.

Figure 1 Chemistry classification of alkaloids.

1940 III / ALKALOIDS / Gas Chromatography
Figure 1 Continued
This article aims to provide an overview of various
aspects of separation of alkaloids by gas chromatog-
raphy (GC). Although a number of phytochemical
methods have been developed for the qualitative and
quantitative determination of alkaloids, one of the
most popular methods for the evaluation of complex
alkaloid mixtures is capillary gas}liquid chromatog-
raphy combined with mass spectrometry (MS). De-
pending on the task high performance liquid chromato-
graphy (HPLC), thin-layer chromatography (TLC),
colorimetry, NMR, radioimmunoassay, capillary elec-
trophoresis and enzyme-linked immunosorbent assay
(ELISA) are additional helpful analytical techniques.
GC-MS Method for Analysis of Alkaloids
Capillary gas chromatography (CGC) analysis has
been described for several classes of alkaloids. A ma-
jor advantage of GC over other methods is its en-
hanced sensitivity and high resolution. Another ad-
vantage is its easy coupling to a mass spectrometer,
which allows the identiRcation of new and minor
compounds of a mixture without laborious isola-
tion procedures. This makes it a particularly attract-
ive method for thermally stable mixtures. The
analysis of pyrrolizidine alkaloids, tropane alkaloids,
steroidal alkaloids, quinazoline alkaloids, quino-
lizidine alkaloids, diterpenoid alkaloids and lyco-
podium alkaloids has been described by a number of
authors.
Capillary gas chromatography was the method of
choice and was ofRcially accepted at the 6th Interna-
tional Lupin Conference in Chile (1990) as a method
of determination of quinolizidine alkaloids in lupins.
As an example, we will describe the methodology for
the analysis of these compounds.

Sample Preparation for Chemical Analysis of
Alkaloids
Successful chemical analysis of alkaloids depends on
the sampling method and pretreatment of the sample.
It is therefore important to know the chemistry of the
compounds to be analysed. As described by Roberts
and Wink, a basic character is no longer a prerequi-
site for an alkaloid and the chemistry of the nitrogen
atom allows for at least four groups of nitrogenous
compounds:
1. Secondary and tertiary amines, which are more or
less protonated and therefore hydrophilic at
pH(7.0, or the more general case where they are
lipophilic and unprotonated at pH'8.0. This is
the classical alkaloid type.
2. Quaternary amino compounds, which are very
polar, charged at all pH values, and have to be
isolated as salts, e.g. berberine and sanguinarine.
3. Neutral amino compounds, which include the am-
ide-type alkaloids such as colchicine, capsaicin,
and most lactams, e.g. ricinine.
4. N-oxides, which are generally highly water sol-
uble, are frequently found in many alkaloid
classes. The pyrrolizidine group of alkaloids is rich
in this particular alkaloid type.
A conventional alkaloid extraction process in-
volves successive removal or nonalkaloids and alkal-
oids by organic solvents from acidiRed and basiRed
aqueous solutions of an ethanol extract. The extrac-
tion of alkaloids is generally based on the fact that
they normally occur in the plant as salts and on their
basicity, in other words on the differential solubility
of the bases and salts in water and organic solvents
(Figure 2).
The techniques used for sample preparation are
liquid}liquid extraction, solid-phase extraction and,
more recently, supercritical Suid extraction.
Liquid}liquid solvent extraction This technique is
the most commonly used method for sample treat-
ment and is based on the observation that alkaloids
can usually be removed from the sample by extracting
them into a water-immiscible solvent. The method
relies on the relative solubility of alkaloids in the
extracting solvent and the sample matrix.
Although such techniques are usually satisfactory,
difRculties can be found when they are applied to
chromatography where the limits of quantiRcation
are often in the ppb range. This is principally caused
by the solvents being nonselective and therefore
tending to extract endogenous material from the
matrix, which results in spurious peaks in the
chromatogram.
III / ALKALOIDS / Gas Chromatography 1941
An example of quinolizidine alkaloid liquid}liquid
extraction is provided in Figure 3.
Solid-phase extraction for sample preparation
Sample clean-up is required when impurities in the
sample matrix interfere with analyte measurement.
The interest in this technique led to the commercial
introduction of small disposable cartridges packed
with relatively large particles of various bonded sil-
icas. The particle size allows the use of minimum
pressure to force the sample and wash solutions
through the column. Indeed, it is common practice to
suck the solution through the packings rather than to
use pressure.
There are many advantages of solid-phase extrac-
tion including: (1) the possible use of large sample
sizes in pretreatment; (2) the technique is quick and
automated; (3) the low consumption of solvents used;
(4) the use of selective sorbents and solvents; (5) the
possible achievement of a high pre-concentration of
the component of interest, enabling high sensitivities
to be obtained; (6) there is good reproducibility in
GC; and (7) the technique is inexpensive.
In developing assays using solid-phase extraction,
it is necessary to take into account several factors
when deciding on the choice of sorbent to be used in
a particular assay for alkaloid analysis.
The most important consideration of the technique
is that the compounds of interest must be capable of
being readily absorbed from the matrix. In some
cases, pretreatment of the sample is necessary, espe-
cially in cases of protein binding. This can usually be
solved by the addition of perchloric or trichloroacetic
acid to denature the proteins. In addition, it may be
necessary to adjust the pH of the sample to ensure
that the compound is in the correct ionic form to
achieve efRcient retention by the packing. Proteins
can also be removed by the addition of organic sol-
vents such as acetonitrile or methanol.
After removing the majority of the interfering sub-
stances, the Rnal step of the technique is efRcient
elution from the bonded silica. This step must ensure
that the compounds of interest are desorbed in the
least volume of eluent, since it is usual to evaporate
the solution to dryness and reconstitute the residue in
a small volume prior to chromatographic analysis.
The evaporation step generally precludes the use
of inorganic salts in the Rnal wash solution, with
the exception of those compounds that are readily
volatile.
Quinolizidine alkaloid solid-phase extraction is
illustrated in Figure 4.
Analytical supercritical Wuid extraction At present,
and in view of increasing environmental concerns of
1942 III / ALKALOIDS / Gas Chromatography
Figure 2 Extraction of alkaloids.
the use of liquid solvents in the extraction of natural
products, there has been growing interest in alterna-
tive and reliable sample extraction techniques using
supercritical Suids. Supercritical Suids have been
widely used for the extraction of alkaloids on both
analytical and industrial scales and for many years for
the selective extraction of selected compounds from
bulk samples. The extraction of caffeine from coffee
is a well-known process performed on an industrial
scale. The aim here is to remove a speciRc component
(i.e. caffeine) from large quantities of the bulk matrix
in order to increase its commercial value.
Analytical-scale supercritical Suid extraction (SFE)
is concerned more with extraction of analytes of
interest from a bulk matrix as a sample preparation
step prior to characterization by other analytical
methods such as GC. It is therefore potentially very
useful for the extraction of natural products prior to
structural characterization. SFE is gaining acceptance
as an alternative to Soxhlet extraction. Much of the

Figure 3 Extraction of quinolizidine alkaloids (Muzquiz et al. 1994).
current interest in using analytical-scale SFE stems
from the need to replace conventional liquid extrac-
tion methods with sample preparation methods that
are more efRcient, easier to automate, faster and safer
to use. Many of the properties of supercritical Suids
such as carbon dioxide have facilitated advances in
these areas. Thermally labile compounds can be ex-
tracted at low temperatures and greatly reduced ex-
traction times. Extracts can also be analysed online
by coupling the SFE directly with a gas chromato-
graph (SFE-GC).
III / ALKALOIDS / Gas Chromatography 1943
Determination of Alkaloids by Gas+Liquid
Chromatography+Mass Spectrometry
Methods for the unequivocal identiRcation and quan-
tiRcation of alkaloids in various, often complex,
matrices are of great interest. For this purpose
chromatography is widely used. Originally thin-layer
chromatography (TLC) was the major method ap-
plied for both qualitative and quantitative analysis of
alkaloids. Although TLC is still a major tool in alkal-
oid analysis, in recent years high performance liquid
1944 III / ALKALOIDS / Gas Chromatography
Figure 4 Extraction of quinolizidine alkaloids (Wink et al. 1995).
chromatography (HPLC) has also developed as an
important method for the quantitation of alkaloids.
However, more and more applications of capillary
(CGC) for complex alkaloids have been reported
recently.
Combined gas chromatography}mass spectro-
metry (GC-MS) has been increasingly used over the
last decade for the convenient analysis of alkaloids.
This sensitive technique is applicable to the qualitat-
ive analysis of individual components of crude alkal-
oid fractions and is normally able to resolve alkaloid
diastereoisomeric pairs. GC-MS is particularly suit-
able for work of a chemotaxonomic nature, since in
such studies it is desirable to identify all the alkaloids

that may have accumulated at a speciRc time and site
in a speciRc part of a species, rather than only the
most abundant compounds present. Also, the use of
GC-MS may enable the experimenter to rule out the
presence of a particular alkaloid group in the plant
material being examined. Impressive separations of
alkaloids have been obtained using the high column
efRciencies achieved in CGC.
Until 1980 most GLC applications for separating
alkaloid mixtures involved packed columns. How-
ever, better results can be obtained using the new
generation of fused silica capillary columns with
bonded phases. The advantage of using small internal
diameter columns is not only the higher plate number
per unit length, but also the improved lower level of
detection due to reduced band broadening. Much
more important than efRciency however, is the selec-
tivity that can be introduced into the chromato-
graphic system. The reason for this is that even the
best capillary column still has a limited peak capacity
(maximum 1000), which is certainly insufRcient for
unravelling the complex proRles that have to be dealt
with in natural product research.
Some GC systems used for the analysis of alkaloids
are indicated in Table 1. The column selectivity can
be adapted to the speciRc problem by selecting the
most suitable stationary phase. Stationary phase se-
lection, however, has no inSuence on the peak capa-
city. In addition to universal inlets such a split, split-
less, cool on-column and temperature-programmed
vaporization, a number of selective inlets are avail-
able in CGC.
In the case of QA the capillary columns used (di-
mensions 15 m;0.23 mm to 30 m;0.32 mm) have
a high number of theoretical plates ('70 000),
which allow the separation of complex mixtures and
even of enantiomers, epimeric at C11 or C6, such as
sparteine and
-isosparteine, lupanine and
-
Table 1 Some GC systems used for the analysis of alkaloids
Alkaloid Column type, GC conditions
length (m);i.d. (mm)
Injector
Pyrrolizidine WCOT DB-1 2503C
25 m;0.25 mm
Quinolizidine SPB-1 30 m;0.25 mm 2403C
Tropane DB-1 15 m;0.25 mm 2503C
Morphinan DB-5 15 m;0.25 mm 2803C
Aconitum DB-5 15 m;0.25 mm 3203C
Amaryllidaceae DB-1 15 m;0.25 mm 2603C
Solanum RTx-1 15 m;0.53 mm 2703C
Ephedra HP-5 25 m;0.20 mm 2203C
WCOT, Wall coated open tubular column.
III / ALKALOIDS / Gas Chromatography 1945
isolupanine, 13-hydroxylupanine and 23-epihyd-
roxylupanine, anagyrine and thermopsine, 13
-
tigloyloxylupanine and 13-tigloyloxylupanine, and
of cis and trans isomers, such as 13
-angeloyl-
oxylupanine and 13
-tigloyloxylupanine, as well as
the trans- and cis-cinnamic acid esters.
As a liquid phase several silicone derivatives
(0.1
m or 1
m Rlms) are employed; good resolu-
tions have been obtained using DB-1 or DB-5
columns, but equivalent products of other manu-
facturers also work. Split injection techniques are
usually appropriate. On-column injection does
not provide signiRcant advantages for most ap-
plications.
Helium is routinely used as carrier gas, but hydro-
gen or nitrogen will also work. The injector temper-
ature is usually set at 2503C, that of the detectors at
3003C. Furthermore, even nanogram amounts of al-
kaloids can be detected by the FID (Same ionization
detector) or more sensitively and speciRcally by a
nitrogen-speciRc detector (NPD).
Hydroxylated QA, such as 13-hydroxylupanine or
3-hydroxylupanine, may be derivatized by trimethyl-
silyl prior to injection to avoid tailing and to achieve
better quantiRcation. Care should be taken not to use
the NPD for these derivatives, since the detector
would soon be destroyed.
Some authors give relative retention indices for
QA. However, KovaH ts retention indices (RI) give bet-
ter comparative information and are helpful in identi-
fying individual alkaloids in a GC proRle.
Additionally, since this method can be combined
with mass spectrometry (GC-MS) it is easy to identify
the individual compounds present. Among the spec-
troscopic methods, mass spectroscopy is deRnitely the
most powerful technique and should therefore take
an important place in any laboratory. The problems
of interfacing both techniques have been completely
Carrier
gas
Temperature program (3C) Detector
120}290 (83C min\1) FID, NPD He
150}235 (53C min\1) FID, NPD He
150}270 (63C min\1) FID, NPD He
180}270 MSD He
(303C min\1)}3203
(403C min\1)
250}320 (163C min\1) MSD He
200}250 (43C min\1) FID, NPD He
210}260 (13C min\1) FID He
90}124 (33C min\1)}280 (203C min\1) NPD He
1946 III / ALKALOIDS / Gas Chromatography

overcome by direct coupling (no interface) or the use or chemical ionization and can be operated in the full
of an open split interface. Low and high resolution scan mode for identiRcation of unknowns or in the
mass spectrometers are the most universal detection ion-monitoring mode for quantiRcation of target
devices for CGC. They are capable of electron impact compounds.

Figure 5 Separation of an alkaloid extract from L. angustifolius (A) and L. mutabilis (B) bitter seeds by capillary GC. Injector, 2403C;
detector 3003C; oven 150}2353C, 53C min\1; carrier gas, helium; detection of alkaloids by nitrogen-specific detector (NPD) and mass-
selective detector.
 III / ALKALOIDS / Gas Chromatography 1947

Mass spectrometry is widely used today, since QA ionization (CI-MS), Reld desorption (FD-MS) and
usually provide distinctive fragmentation patterns fast atom bombardment (FAB-MS) are suitable for
in the electron impact mode (EI-MS). Chemical identifying molecular ions of QA esters and of

Figure 6 Separation of an alkaloid extract from L. luteus (A) and L. hispanicus (B) bitter seeds by capillary GC. Injector, 2403C;
detector 3003C; oven 150}2353C, 53C min\1 ; carrier gas, helium; detection of alkaloids by nitrogen-specific detector (NPD) and mass-
selective detector.
1948 III / ALKALOIDS / Gas Chromatography
tricyclic alkaloids, whose molecular ions are usually
obscure or absent in EI-MS spectra. A major advant-
age of MS is the possibility it gives of combining the
high resolution power of capillary GC with the sensi-
tivity of and information provided by EI- or CI-MS.
Work using GC-MS was very much facilitated after
1980 by the development of new GC capillary col-
umns, the development of new methods to position
the GC column exit near the MS ion source and, most
importantly, by improved data processing.
Alkaloid extracts of many legumes contain
piperidine alkaloids such as ammodendrine, N-
methylammodendrine, hystrine or smipine. These
alkaloids also derive biogenically from lysine via ca-
daverine. Simple indole and quinolizidine alkaloids,
such as gramine and lupinine may also be encoun-
tered. Even combinations of both indole and
quinolizidine units are possible, as in the case of
Lupinus hispanicus.
Separation and identiRcation of QA by GC-MS is
shown in Figures 5}7.
Conclusion and Future Developments
Gas chromatography is a versatile tool in the analysis
of natural products with a wide area of application. It
Figure 7 Separation of alkaloid extracts from Chamaecytisus proliferus by capillary GC. Injector, 2403C; detector 3003C; oven
150}2353C, 53C min\1; carrier gas, helium; detection of alkaloids by nitrogen-specific detector (NPD) and mass-selective detector.
is capable of extracting a wide range of diverse com-
pounds from a variety of sample matrices.
Clear advantages of GC are the high sensitivity of
the most common detection method, the FID, and the
fact that the detector response of similar compounds
will be about the same (i.e. peak areas may be directly
compared for quantiRcation). By using a nitrogen-
speciRc detector (NPD) sensitivity for alkaloids can
be even further improved while at the same time
introducing selectivity.
No systematic studies to determine which column
is best suited for alkaloid analysis have been reported,
but from the methods described to date it is clear that
thinly coated apolar columns are preferred for the
analysis of underivatized alkaloids. The length of the
columns used varies considerably and it is advisable
to test the stability of a compound under GC condi-
tions with a short column. A longer column may be
used later if the desired chromatographic resolution
has not been achieved.
A wealth of information can be obtained by the
analysis of alkaloids by GC coupling to MS. Coupled
techniques (GC-MS) have demonstrated their analyti-
cal potential. The large amounts of data produced
by capillary GC, especially when coupled to a mass
spectrometer, can now be handled by a personal

computer. The data can be acquired, manipulated
and displayed in real time and can be stored for
record purposes.
Looking to the future, it is reasonable to expect
continued evolutionary development: new selective
detectors, more complex analysers for automated
sample processing, increasing use of coupled tech-
niques, columns with immobilized phases of a wider
range of selectivity, etc. It is hoped that further re-
search and development will encourage the use of
GC-MS in the areas of alkaloid analysis that still
await investigation.
Acknowledgements
The author gratefully acknowledges M. Martin-Ped-
rosa, T. Ortega, C. Cuadrado and C. Burbano for
their helpful comments.
See also: II/Chromatography: Gas: Detectors: General
(Flame Ionization Detectors and Thermal Conductivity
Detectors); Detectors: Mass Spectrometry; Detectors:
Selective. III/Alkaloids: Liquid Chromatography; Solid-
Phase Extraction; Solid-Phase Microextaction; Supercriti-
cal Fluid Extraction; Thin-Layer (Planar) Chromatography.
Extraction: Analytical Extractions.
Further Reading
Bruneton J (ed.) (1995) Pharmacognosy Phytochemistry
Medicinal Plants. Paris: Technique and Documentation
Lavoiser.
Cheeke PR (ed.) (1989) Toxicants of Plant Origin. vol. 1:
Alkaloids. Boca Raton, FL: CRC Press.
High Speed Countercurrent Chromatography
See III / MEDICINAL HERB COMPOUNDS: HIGH SPEED COUNTERCURRENT
CHROMATOGRAPHY
Liquid Chromatography
R. Verpoorte, Leiden/Amsterdam Center
for Drug Research, Leiden, The Netherlands
Copyright ^ 2000 Academic Press
De\nition and Classi\cation
of Alkaloids
Alkaloids represent a wide variety of chemical struc-
tures (Figure 1). More than 16 000 are known and most
III / ALKALOIDS / Liquid Chromatography 1949
Dagnino D and Verpoorte R (1994) Gas chromatography
in the analysis of alkaloids. In: Linskens HF and Jackson
JF (eds) Modern Methods of Plant Analysis. Berlin:
Springer-Verlag.
David F and Sandra P (1992) Capillary gas chromtogra-
phy}spectroscopic techniques in natural product
analysis (Review paper). Phytochemical Analysis 3:
145}152.
D’Mello JPF, Duffus CM and Duffus JH (eds) (1991) Toxic
Substances in Crop Plants. Cambridge: Royal Society of
Chemsitry.
Kutchan TM (1995) Alkaloid biosynthesis. The basis for
metabolic engineering of medicinal plants. Plant Cell 7:
1059}1070.
Modey WK, Mulholland DA and Raynor MW (1996) Ana-
lytical supercritical Suid extraction of natural products
(Review paper). Phytochemical Analysis 7: 1}15.
Papadoyannis IN and von Baer D (1993) Analytical tech-
niques used for alkaloid analysis in legume seeds. In:
Van der Poel AFB, Huisman J and Saini HS (eds) Recent
Advances of Research in Antinutritional Factors in
Legume Seeds. Wageningen: EAAP.
Pelletier SW (ed.) (1984) Alkaloids: Chemical and Bio-
logical Perspectives. New York: John Wiley and Sons.
Roberts MF and Wink M (eds) (1998) Alkaloids. Biochem-
istry, Ecology, and Medicinal Applications. New York
and London: Plenum Press.
TomaH s-BarberaH n FA (1995) Capillary electrophoresis:
a new technique in the analysis of plant secondary
metabolites (Review paper). Phytochemical Analysis 6:
177}192.
Verpoorte R and Niessen WMA (1994) Liquid chromatog-
raphy coupled with mass spectrometry in the analysis of
alkaloids. Phytochemical Analysis 5: 217}232.
Wink M (1993) Quinolizidine alkaloids. Methods in Plant
Biochemistry.
are derived from higher plants. Alkaloids have also been
isolated from microorganisms, marine organisms like
algae, dinoSagellates and puffer Rsh and terrestrial
animals like insects, salamanders and toads.
An alkaloid has been deRned by Pelletier as a cyclic
organic compound containing nitrogen in a negative
oxidation state which is of limited distribution among
living organisms. From the analytical chemical point
of view, the most important trait of alkaloids is their
1950 III / ALKALOIDS / Liquid Chromatography

basicity arising from a heterocyclic tertiary nitrogen amino acids, such as phenylalanine, tyrosine, tryp-
atom. Notable exceptions are colchicine and the tophan, ornithine and lysine. The biogenesis of alka-
xanthines (e.g. caffeine), with pKa values between loids is used for their classiRcation, as this is directly
1 and 2. Alkaloids are biosynthetically derived from linked with their molecular skeleton, e.g. the two

Figure 1 Structures of alkaloids. (A) L-hyoscyamine; (B) taxol; (C) quinine; (D) nicotine; (E) caffeine; (F) colchicine; (G) vincamine;
(H) R"H ellipticine; R"OH 10-hydroxyellipticine; (I) camptothecin; (J) strychnine; (K) reserpine; (L) H-20
tetrahydro alstonine;
H-20 ajmalicine; (M) emetine; (N) berberine; (O) galanthamine; (P) sanguinarine; (Q) R"H morphine; R"CH3 codeine; (R) R"H
d -tubocurarine; R"CH3 d -chondrocurarine; (S) solasodine.
 III / ALKALOIDS / Liquid Chromatography 1951

Figure 1 Continued
1952 III / ALKALOIDS / Liquid Chromatography

Table 1 Alkaloids of pharmaceutical interest diseases. For example, for some types of alkaloids,
insect antifeedant activity has been established.
Indole alkaloids Tropane alkaloids Thus, many alkaloids have strong biological activ-
Ajmalicine Cocaine
Ajmaline Scopolamine ities. Their effect in humans can be explained
Camptothecin Atropine (d / l -hyoscyamine) by structural relationships with important signal
Ergocornine compounds (neuro-transmitters) like dopamine,
Ergocristine Terpenoid alkaloids acetylcholine, noradrenaline and serotonin. Conse-
Ergocryptine Aconitine
quently, some alkaloids are used as medicines or in
Ergonovine Solasodine
Ergosine Taxol pharmacological studies (Table 1). In addition to
Ergotamine Tomatidine pure compounds, crude plant extracts containing
Harmane Veratrine alkaloids are used (phytotherapy). Another area
9-Hydroxyellipticine where alkaloids play a major role is in drugs of
Lysergic acid Miscellaneous abuse, e.g. mescaline, cocaine, morphine and its
Physostigmine Caffeine
Psilocybin Colchicine semisynthetic derivative, heroin. Alkaloids are also
Reserpine Ephedrine of interest in the analysis of doping (e.g. strychnine,
Rescinnamine Lobeline ephedrine, caffeine) and poisons (e.g. strychnine,
Serotonin Mescaline pyrrolizidine alkaloids, coniine, nicotine, aconitine,
Strychnine Muscarine
tetrodotoxin).
Yohimbine Nicotine
Vincamine Pilocarpine Due to their various applications, the analysis of
Vinblastine Saxitoxin alkaloids is of great importance. The very different
Vincristine Sparteine types of (ab)use of the alkaloids mean that the type of
Tetrodotoxin analyses also varies. Alkaloids must be analysed in
Quinoline alkaloids Theobromine
a broad variety of matrices, such as plant material,
Quinine Theophylline
Quinidine tablets, drug seizures, urine and blood. Each requires
different sample clean-up methods and chromato-
Isoquinoline alkaloids graphic selectivities. Liquid chromatography is the
Apomorphine most commonly used method since the instability and
Berberine
low volatility of alkaloids mean that gas chromato-
Boldine
Chelerythrine graphy has a limited applicability. Because the ex-
Codeine tracts are often complex and ‘dirty’, thin-layer
Emetine chromatography is useful in analysing alkaloid-con-
Galanthamine taining plant extracts.
Heroin
Morphine
Narceine Chemical Properties and
Noscapine
Papaverine
Artefact Formation
Sanguinarine Most alkaloids have basic properties with pKa values
Thebaine
of about 6 to 12, but usually 7}9. The free base is
Tubocurarine
soluble in organic solvents and not in water. Protona-
tion of the nitrogen in the free base usually results in
a water-soluble compound. This behaviour is the
largest groups are indole alkaloids (more than 4100 basis of the selective isolation of alkaloids by
compounds) and isoquinoline alkaloids (more than liquid/liquid partitioning processes. Quaternary al-
4000 compounds). Other important groups are kaloids are poorly soluble in organic solvents but
tropane alkaloids (c. 300 compounds), steroidal al- soluble in water at any pH.
kaloids (c. 450 compounds), pyridine and pyr- Many alkaloids are difRcult to crystallize in the
rolizidine alkaloids (about 250 and 570 compounds, form of the free base, but do crystallize as a salt.
respectively). Alkaloids are usually colourless; only some highly
The botanical origin of the alkaloids is also used conjugated compounds are coloured or show strong
as a classiRcation method, e.g. Papaver (opium) Suorescence (e.g. berberine and serpentine).
alkaloids, Cinchona alkaloids, RauvolTa alkaloids, Alkaloids are not very stable; in particular, N-
Catharanthus alkaloids, Strychnos alkaloids, ergot oxidation is quite common. Stability is inSuenced by
alkaloids, cactus alkaloids and Solanum alkaloids. solvents, as well as heat and light. Halogen-contain-
As secondary metabolites, alkaloids probably play ing organic solvents such as chloroform and di-
a role in defending organisms against pests and chloromethane are widely used in alkaloid research.

Chloroform in particular is a very suitable solvent,
because of its relatively strong proton donor charac-
ter. However, this solvent easily causes the forma-
tion of artefacts, e.g. (N-)oxidation occurs easily.
Dichloromethane may result in the formation of
quaternary N-dichlorometho-compounds. Similar
compounds are formed with the minor impurities
present in chloroform. Peroxides in ethers may also
cause N-oxidation.
Alkaloids are more stable in toluene, ethyl acetate
and alcoholic solutions. Carbinolamine functions are
often found in alkaloids, either formed during the
coupling of a carbonyl group and an amine in the
biosynthesis, or as products formed from rearrange-
ments of N-oxides. Carbinolamines readily react with
alcohols (e.g. O-methyl pseudostrychnine formed
from pseudostrychnine with methanol). Ketones such
as acetone and methylethylketone are well-known
artefact formers. Berberine, for example, may react
with acetone. Ammonia and acetone may react dur-
ing column chromatography, yielding condensates
that give a Dragendorff-positive reaction. Ammonia
may also react with aldehydes present in plant mater-
ials, giving rise to artiRcial alkaloids, e.g. the py-
ridine-type alkaloid gentianine is formed from swero-
side during extraction.
Extraction
Due to the more lipophylic character of alkaloids as
free bases, they can be extracted under neutral or
basic conditions (e.g. after basiRcation of the plant
material or bioSuid to pH 7}9 with ammonia, so-
dium carbonate or sodium bicarbonate) with organic
solvents (such as dichloromethane, chloroform,
ethers, ethyl acetate and alcohols). Strongly basic
alkaloids can only be completely extracted at higher
pH ('10), e.g. tryptamine. As a general rule of
thumb, for the extraction of an alkaloid one should
choose a pH of pKa#2. On the other hand, alkaloids
containing phenolic groups are protonated at higher
pH, and thus not extracted by organic solvents under
such conditions (e.g. morphine).
Alkaloids can be extracted in protonated form
(after acidiRcation to pH 2}4 with diluted acids like
phosphoric acid, sulfuric acid, citric acid) with water
or alcohols (e.g. methanol).
Alkaloids can be further puriRed by liquid}liquid
extraction or liquid/solid extraction. In liquid}liquid
extraction the alkaloids are, after basiRcation, extrac-
ted form an aqueous solution with an immiscible
organic solvent or from an organic solvent with an
aqueous acid solution. To avoid the formation of
lipophylic ion pairs, phosphoric acid, sulfuric acid
and citric acid are preferred over acetic acid and
III / ALKALOIDS / Liquid Chromatography 1953
hydrochloric acid. By using a back-extraction from
aqueous solution to organic and back to aqueous, or
from organic to aqueous and back to organic solu-
tion, alkaloids can easily be separated from neutral
and acidic compounds.
Alkaloids can be extracted from acidic aqueous
solutions with organic solvents by using ion-pairing
reagents (e.g. alkylsulfonic acids). It should be noted
that common anions such as Cl\, Br\, I\ and acetate
also form ion pairs which are readily soluble in or-
ganic solvents.
Solid-phase extraction using adsorption or ion
exchange can also be used. For adsorption of the
alkaloids in the free form, reversed-phase materials,
such as chemically bonded C8 and C18 on silica,
are widely used. A suitable solvent system is a mix-
ture of methanol and water; the crude extract is
fractionated by stepwise elution of the adsorbent
with a solvent of decreasing polarity. XAD-2 is
also used for the concentration of alkaloids, e.g.
from biological Suids. Various cation exchange ma-
terials can be used for the selective extraction of
alkaloids.
For preparative purposes puriRcations based on the
precipitation of alkaloids are employed. A crude ex-
tract of the alkaloids is made with aqueous acid; sub-
sequently the alkaloids are precipitated with reagents
such as Mayer’s reagent (1 mol L\1 mercury chloride
in 5% aqueous potassium iodide) or Reinecke’s salt
(5% ammonium reineckate in 30% acetic acid) at pH
2, or picric acid (saturated aqueous solution) at pH
5}6. After collection by Rltration or centrifugation,
the precipitate is dissolved in an organic solvent
(acetone : methanol : water; 6 : 2 : 1). The complex-
ing group is then removed by means of an anion
exchanger. Quaternary alkaloids cannot be puriRed
by means of liquid}liquid extraction, therefore
precipitation is particularly suited for their puri-
Rcation.
Thin-layer Chromatography
Thin-layer chromatography (TLC) is widely used as
a versatile method in the analysis of alkaloids. It
offers the advantage of a broad range of polar-
ities being separated in one single analysis, which is of
interest in plant materials and metabolism studies.
The most widely used stationary phase is silica;
alumina plates are rarely employed nowadays. Rever-
sed-phase materials, such as chemically bonded
C18 on silica, are also applied but silica is still used
most widely.
Strongly basic alkaloids will show severe tail-
ing on silica gel plates, due to the acidic properties
of silica. The use of mobile phases which contain
1954 III / ALKALOIDS / Liquid Chromatography

Table 2 Some common thin-layer chromatography systems for the analysis of alkaloids

Solvent system (all with silica plates) Commonly used ratios Polarity range

Cyclohexane}chloroform}diethylamine 5 : 4 : 1}(0) : 9 : 1 lp-mp

Chloroform}acetone}diethylamine 5:4:1 mp
Chloroform}methanol}ammonia 8:1:1 mp
Chloroform}methanol/ethanol 99 : 1}1 : 1 lp-mp, wb
Ethyl acetate}isopropanol}25% ammonia 100 : 2 : 1, 80 : 15 : 5, 45 : 35 : 5 lp-mp
Ethyl acetate}methanol 9 : 1}1 : 1 lp-mp, wb
Toluene}ethyl acetate}diethylamine 7:2:1 lp-mp
Toluene}acetone}ethanol}25% ammonia 20 : 20 : 3 : 1 mp
Dicholoromethane}diethyl ether}diethylamine 20 : 15 : 5 mp
Acetone}methanol}25% ammonia 40 : 10 : 2, 95 : (0) : 5 mp-p
Methanol}25% ammonia 95 : 5 lp-p
n -Butanol}acetic acid}water 4:1:1 lp-p
Methanol}1 mol L\1 aq. M NH4NO3}2 mol L\1 aq. ammonia 7:1:2 lp-p
Methanol}0.2 mol L\1 aq. M NH4NO3 3:2 lp-p

lp, Low polarity compounds; mp, medium polarity compounds; p, polar compounds; wb, weakly basic compounds.

a base such as ammonia or diethylamine will over-
come this problem. A more elaborate method is
the use of TLC plates impregnated with a basic
solution.
For the analysis of highly polar quaternary alka-
loids and N-oxides, solvent systems consisting of
methanol and aqueous salt solutions are useful. In
Table 2 some widely used TLC systems are sum-
marized. For the detection of alkaloids a large num-
ber of methods have been reported. Besides quench-
ing ultraviolet (UV) light on Suorescent plates and
Suorescence, general reagents for selectively detecting
alkaloids are Dragendorff’s reagent (orange-brown
spots) and potassium iodoplatinate (brown-violet-
purple spots; Table 3). Dragendorff’s reagent may
cause false-positive reactions with, for example, com-
pounds containing conjugated carbonyl or lactone
functions. The iodoplatinate reagent has less risk of
false-positive reactions and is more selective due to
a broader spectrum of colours observed for individual
alkaloids.
Highly selective reagents have been reported for the
visualization of various classes alkaloids (Table 4).
These are based on different colorations under
strongly oxidative conditions.
Liquid Chromatography
High performance liquid chromatography (HPLC) is
a major tool for the analysis of alkaloids. Most separ-
ations are done on reversed-phase (RP) materials
(C8-, C18- and phenyl-bonded phases on silica). Al-
though extensive tailing due to the interaction of the
basic nitrogen and residual acidic silanol groups may
occur on the RP materials. In particular, strong bases
show this problem. Several solutions have been found
to circumvent this. First, special RP materials have
been developed for basic compounds. These materials
have an altered silica surface, a high load of the alkyl
groups or they have undergone a rigorous endcapp-
ing treatment to reduce the number of free silanol
groups. Often the plate numbers observed for alkal-
oids on an HPLC column are considerably lower
than those measured with the usual neutral test
compounds. Polymeric (e.g. polystyrene-based) sta-
tionary phases do not have the problem of residual
silanol groups; however, plate numbers observed
with such columns are not usually better than those
found with specially made RP silica materials.
Phenyl-type RP columns are also successful in the
separation of alkaloids.
Another way of reducing the tailing is through
modiRcation of the eluent. By adding long chain al-
kylamines (e.g. hexylamine) in low concentrations
Table 3 Detection reagents for alkaloids on thin-layer
chromatography plates
Dragendorff’s reagent (modification according to Munier )
(A) 1.7% bismuth subnitrate in 20% aq. tartaric acid solution
(B) 40% potassium iodide in water
A and B are mixed (5 : 2) and the spray reagent is prepared by
mixing 50 mL of the stock solution with 100 g tartaric acid and
500 mL water.
Colours observed after spraying: orange-brown spots for alka-
loids
Potassium iodoplatinate reagent
The reagent is prepared freshly by mixing 3 mL of 10% aq.
hexachloroplatinic acid solution with 97 mL water and 100 mL of
6% aqueous potassium iodide solution.
Colours observed after spraying: brown-violet-purple spots for
alkaloids

Table 4 Selective colour reactions for the detection of alkaloids on thin-layer chromatography plates
Spray reagent
0.2 mol L\1 ferric chloride in 35% perchloric acid and heat
1% ceric sulfate in 10% sulfuric acid
1% p -dimethylaminobenzaldehyde in ethanol, followed by exposure
to hydrochloric acid vapour
Sulfuric acid and heat
(typically 1 mmol L\1) to the mobile phase, tailing
can be considerably reduced. Also, addition of amines
like triethylamine or tetramethylammonium can be
helpful in reducing tailing. Moreover, alkaloids have
been analysed on aminopropyl- and cyanopropyl-
type of columns, in both normal and reversed-phase
modes.
In liquid chromatography of alkaloids, the pH of the
mobile phase must be strictly controlled, as stationary
phases are unstable at a pH above 8, usually a pH
between 2 and 4 is used, i.e. the alkaloids are present
in the protonated form. Ion suppression systems are
quite common. Because of the preference for the
lower pH range of the eluent, ion pairing is used
with C4}C8 alkylsulfonates at a concentration of
25}100 mmol L\1 for the analysis of alkaloids. Increas-
ing length of the alkyl chain causes longer retention.
Some general features of RP HPLC systems for the
analysis of alkaloids are given in Table 5.
The number of applications of ion exchange
chromatography for the separation of alkaloids is
limited. In general, cation exchange columns will also
affect the selectivity of the separation through
nonionic interactions, e.g. through the stationary
phase matrix. Usually an elevated temperature is used
to improve peak shape.
A large number of liquid}solid separations on silica
have been reported (Table 6). The systems applied
are similar to those reported for TLC.
UV is most widely used for detection. Particularly
for the groups of indole and isoquinoline alkaloids,
Table 5 General outline of reversed-phase high performance liquid chromatography
systems for the separation of alkaloids
Stationary phase
C8, C18 or phenyl-bonded phase
with low percentage of free silanol groups
III / ALKALOIDS / Liquid Chromatography 1955
Commonly used for the detection of
Indole alkaloids, isoquinoline alkaloids
Indole alkaloids
Ergot alkaloids
Various alkaloids
strong and speciRc UV chromophores are found.
These can greatly assist in identifying compounds,
e.g. in using HPLC with diode array detection. The
pH of the solvent as well as the solvent itself may have
an effect on the UV spectra, e.g. causing shifts of
maxima and minima. Some alkaloids can be detected
by means of their Suorescence. Some type of alkaloids
have poor UV absorption properties, e.g. tropane
alkaloids, pyrrolizidine alkaloids and steroidal
alkaloids require detection at lower wavelengths
(200}220 nm). Electrochemical detection has been
applied, enabling the selective attenuation of interfer-
ing compounds. Mass spectrometry is a major tool in
the identiRcation and structure elucidation of alkal-
oids. In combination with gas chromatography and
liquid chromatography, it is very useful in the quali-
tative and quantitative analysis of complex mixtures
of alkaloids. Solvent systems suited for liquid
chromatography}mass spectrometry should only
contain volatile compounds (e.g. ammonium acetate,
ammonium formate).
Countercurrent Chromatography
The preparative isolation of alkaloids can be achieved
by means of modern countercurrent chromatogra-
phy. Because of the ionic nature of the alkaloid sys-
tems with a controlled pH are preferred for the
separation. Improved efRciency can be obtained by
using ion pair gradients, e.g. solvent two-phase sys-
tems consisting of chloroform}methanol}aqueous
Mobile phase
Ion supression mode
Methanol (acetonitrile)}water containing
c. 0.01}0.1 mol L\1 phosphate buffer,
ammonium carbonate or sodium acetate
(pH 4}7)
Ion pair mode
Methanol (acetonitrile)}water containing
c. 25}100 mmol L\1 alkylsulfonate and
1% acid (e.g. acetic acid), pH 2}4
1956 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
Table 6 General outlines of normal-phase high performance liquid chromatography
systems for the separation of alkaloids
Stationary phase Mobile phase
Silica gel Dichloromethane,
Chloroform,
Diethyl/isopropyl ether, or
Tetrahydrofuran, or
Ethyl acetate
phosphate or citrate buffer, pH c. 4, containing per-
chlorate, acetate or chloride as the ion pairing agent.
High loadability and different selectivity compared
with column chromatography are important features
of countercurrent chromatography.
See also: III/Alkaloids: Gas Chromatography; Thin Layer
(Planar) Chromatography. Natural Products: High-
Speed Countercurrent Chromatography.
Further Reading
Baerheim Svendsen A and Verpoorte R (1983) Chromatog-
raphy of alkaloids. Part A: Thin-layer chromatography.
Amsterdam: Elsevier Science Publishers.
Manske RHF and Holmes HL (eds) The Alkaloids, Volume
1}5 (1950}1995), Manske RHF (ed.) Volume 6}16
(1955}1977), Manske RHF and Rodrigo R (eds) Vol-
ume 17}20 (1979}1981), Brossi A (ed.) Volume 21}40
(1983}1992), Cordell GA (ed.) Volume 40} (1992})
New York: Academic Press.
Cordell GA (1981) Introduction to Alkaloids. A Biogenetic
Approach. New York: John Wiley.
Thin-Layer (Planar) Chromatography
J. Flieger, Medical Academy, Lublin, Poland
Copyright ^ 2000 Academic Press
Introduction
In 1938, Izmailow and Schraiber pioneered the thin-
layer chromatography (TLC) method for the analysis
of plant material containing alkaloids. The subject
matter of their scientiRc research was an extract of a
plant rich in tropane alkaloids. Later on, the method
was developed by Bekesy, who applied it to the separ-
ation of ergot alkaloids. Since then, numerous papers
exploring the detection, isolation and quantitative
determination of alkaloids by TLC have been pub-
Methanol Ammonia,
Diethylamine or
Isopropanol Triethylamine
(c. 1% of the mobile phase)
Glasby JS (1975) Encyclopedia of Alkaloids, vols 1 and 2.
New York: Plenum Press.
Hesse M (1974) Progress in Mass Spectrometry, vol. 1,
parts 1 and 2. Mass Spectrometry of Indole Alkaloids.
Weinheim, Verlag Chemie.
Hesse M and Bernhard HO (1975) Progress in Mass Spec-
trometry, vol. 3. Mass Spectrometry of Alkaloids. Wein-
heim: Verlag Chemie.
Pelletier SW (ed.) (1983) Alkaloids: Chemical and Biolo-
gical Perspectives, vols 1}6. New York: John Wiley.
Popl M, FaK hnrich J and Tatar V (1990) Chromatographic
Analysis of Alkaloids. New York: Marcel Dekker.
Sangster AW and Stuart KL (1965) Ultra-violet spectra of
alkaloids. Chemical Reviews 65: 69}130.
Southon IW and Buckingham J (1989). Dictionary of Al-
kaloids. London: Chapman & Hall.
Verpoorte R and Baerheim Svendsen A (1984) Chromatog-
raphy of alkaloids. Part B: Gas-liquid chromatography
and high-performance liquid chromatography. Journal
of Chromatography Library. Volume 23B. Amsterdam:
Elsevier Science Publishers.
Verpoorte R (1986) Methods for the structure elu-
cidation of alkaloids. Journal of Natural Products 49:
1}25.
lished. It has been stated that no other method has
delivered so much information on alkaloids.
From the chemical point of view, alkaloids form
a very diverse group of organic nitrogen compounds
of a basic character (with the exception of some
derivatives of purine and colchicine). They have terti-
ary or quaternary amino groups in their molecules
and only a few contain secondary amino groups.
Considering the fact that analytical problems connec-
ted with alkaloids are mostly concerned with their
physicochemical properties, they are commonly
divided according to the type of chemical struc-
ture into tropane, quinoline, indole, diterpene and
others. Another useful classiRcation is based on
botanical groups (e.g. tobacco, lupine, ergot, strych-
nos, vinca and catharanthus alkaloids), and this is
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
especially valuable as far as chemotaxonomical
studies are concerned.
In early work, alkaloids were predominantly iso-
lated from the natural plant material. TLC was
then used for qualitative and quantitative analysis
of plants and the study of the biosynthesis of alkal-
oids. Because of their powerful physiological proper-
ties alkaloids have become important therapeutic
compounds and many of them have been prepared
synthetically or by partial synthesis. As a conse-
quence, many derivatives have been formed that
do not occur in nature. TLC is particularly well suited
for checking the processes of synthesis as well as
for establishing the progress of reactions and Rnally
testing of products in pharmaceutical preparations.
The importance of alkaloids is also fundamental in
toxicological analysis; many are used as narcotics
and hallucinogenic drugs, as doping substances and
as poisons. The presence of alkaloids in drugs of abuse
and their metabolites in biological Suids such as urine
and blood has also been tested by means of TLC.
Preparation of Samples
Various sample preparation procedures have been
developed for pharmaceutical formulations, plant
and biological materials. Due to the fact that, in most
of them, alkaloids occur as salts together with com-
plex mixtures of nonalkaloid compounds such as
inorganic salts or substances of lipophilic character,
their pre-separation by a suitable extraction proced-
ure is necessary.
While in the case of the analysis of solutions, alkali-
Red (or acidiRed) samples and extraction with an
organic solvent such as chloroform or diethyl ether is
usually sufRcient, isolating alkaloids from a plant
material is a multistage process and may be conduc-
ted using several methods.
Most often preparative isolation is carried out by
liquid}liquid extraction. Plant material with a high
liquid content should be initially extracted with light
petroleum or water containing diluted hydrochloric
acid to remove lipids. The release of alkaloidal bases
occurs under the inSuence of the addition of a mineral
base, commonly ammonia. Then they are extracted
by means of water-immiscible organic solvents or
water}alcohol mixtures.
For efRcient extraction in the above cases de-
scribed, alkaloids should be present in the extractable
form in at least 95%, so pH adjustment of the sample
to pH"pKa#2 is sufRcient.
Further puriRcation is achieved by re-extracting
alkaloids from organic solvents into an aqueous
phase of the opposite pH, where the alkaloids are
present as salts.
1957
Figure 1 Scheme for the back-extraction procedure of a basic
drug (B) (after Adamovics JA (1990) Chromatographic Analysis of
Pharmaceuticals. New York and Basel: Marcel Dekker, Inc.)
This back-extraction procedure for basic com-
pounds (B) is shown schematically in Figure 1.
A liquid extraction technique used to increase ex-
traction efRciency and selectivity is an ion pair extrac-
tion originally used to extract strychnine from syrup.
PuriRcation of crude plant extracts from non alkal-
oidal compounds may be carried out by precipitating
the alkaloids with picric acid, Reinecke’s salt or
Mayer’s reagent or by using ion exchange or a small
adsorption column. Solid-phase extraction (SPE) is
gaining in popularity. SpeciRc sorption conditions
under which alkaloids are strongly retained lead to
preconcentration of free bases (on aluminium oxide),
their salts (on phosphoric acid impregnated silica) or
as an ionic form (on ion exchangers).
It should be emphasized that, in the case of silica
gel, quaternary alkaloids are more strongly retained
than ternary ones with an aqueous buffer}methanol
mobile phase. Such differences also create the possi-
bility of separating these two groups of alkaloids.
One of the latest methods of isolating groups of
alkaloids from solid samples is supercritical Suid ex-
traction (SFE). The method increases the efRciency
of extraction and shortens the overall time of
analysis.
While considering the problems of extraction, iso-
lation and puriRcation of alkaloids, one should be
cautious about the possibility of undesirable reactions
and artefact formation. One reason may be impurities
present in the solvents applied. Thus, peroxides (in
ethers) cause oxidation, ethyl chloroformate (in
chloroform) forms ethylcarbamates of alkaloids;
halogen-containing compounds; bromochloro-
methane and dichloromethane (in chloroform) cause
quaternization of tertiary alkaloids, while cyanogen
1958 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
chloride (in dichloromethane) is the cause of nitrila-
tion of primary and secondary amines. Decomposi-
tion may also be caused by a photochemical reaction,
especially in chloroform solutions. Finally there may
be a reaction with a solvent itself, mainly with chloro-
form, but also with ketones or strong alkali. The fact
that the chloroform used as a component of the mo-
bile phase may present a quenching effect should also
be emphasized.
Development Techniques
Adsorbents used in TLC may be either commercial
products or home-made plates (now seldom em-
ployed). High quality chromatograms can be
achieved with HPTLC plates which were introduced
in the 1980s.
Plates may be developed in a linear, circular or
anticircular mode. The most common technique in
TLC of alkaloids is ascending, single, one-dimen-
sional development in tanks saturated with the va-
pour of the solvent system.
Preconditioning the plate with the vapours, thus
preventing demixing of the mobile-phase components,
can also be performed in sandwich-type chambers
produced by Camag (Vario-KS) and Chromdes (DS).
In some cases, especially where compounds differ in
polarity, repeated development of the plate with the
same solvent or solvents of increasing strength or the
continuous development technique has some advant-
ages. In other cases, programmed multiple develop-
ment with the same solvent may be successfully
applied. Also useful is two-dimensional develop-
ment, which is especially valuable for separating
a greater number of alkaloids in a given section of the
plant.
Great differences in the polarity of alkaloid mol-
ecules make gradient elution advantageous. This
technique may be developed in both glass chambers
and in horizontal chambers as well as with overpres-
sured layer chromatography.
Worth noticing is one technique related to TLC
} thin-layer electrophoresis, which has been used as
a two-dimensional combination with TLC for the
separation of ergot alkaloids.
Separation Methods
It is obvious that the kind of adsorbent used and
solvent system composition determine the separation
mechanism occurring in the chromatographic pro-
cess. The adsorbent also determines the method of
sample preparation. Thus, for adsorption and parti-
tion chromatography, alkaloids are mostly applied as
bases in organic polar solvents; for ion exchange
sorbents they are applied in the form of salts in
aqueous solution.
Choosing the optimal chemical character of the
stationary and mobile phase is especially important in
the case of alkaloids because of the ionization ability
of their molecules. Dissociation of bases in aqueous
solution can be expressed by the following equation:
B#H2O 0 BH##OH\
or, in the case of the conjugated acid BH#, by:
BH#0 B#H#
with a dissociation constant (acidic) Ka.
The dependence of the molar ratio of nondis-
sociated molecules [B] to the total concentration of an
alkaloid [B]#[BH#] on the pH of the mobile phase
is shown in the curves presented in Figure 2. The pKa
values of chosen alkaloids are summarized in Table 1.
For TLC of alkaloids, numerous chromatographic
systems have been reported. Some are presented in
Table 2, together with their practical applications.
Adsorption Chromatography
Silica gel is the most frequently used solid-phase in
adsorption chromatography. The weakly acidic prop-
erties of its surface may be the reason for the
chemisorption of alkaloids, especially when neutral
nonpolar solvents are used.
Tailing of spots may occur and the danger in using
a neutral mobile phase is the formation of double
spots, resulting from partial deprotonation of mol-
ecules if alkaloids are applied as salts. This is why
Figure 2 Dependence of the degree of dissociation of an alkal-
oid (B) on pH of buffer in mixed solvent (
pKa(B)(0 and
pHm'
0). 1, solution in water; 2, solution in mixed solvent pKHa(B)"pKa in
50% (w/w) methanol (after Popl M, FaK hnrich J and Tatar V (1990)
Chromatographic Analysis of Alkaloids. Chromatographic
Science. New York and Basel: Marcel Dekker, Inc.).
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography 1959

Table 1 Values of pKa for the dissociation of alkaloids in water

Alkaloid pKa Alkaloid pKa

Aconitine 8.35 Methylecgonine 9.16

Arecaidine 9.07 Morphine 8.21
Arecoline 7.41 Narceine 3.30
Atropine 9.85 -Narcotine 6.37
Benzoylecgonine 11.80 Nicotine 8.02
Berberine 11.73 (pKa2"3.12)
Brucine 8.16 Nicotyrine 4.76
(pKa2"2.50) Papaverine 6.40
Caffeine 1.00 d,l-Pelletierine 9.40
Cinchonidine 8.40 Pilocarpine 6.87
(pKa2"4.17) Piperine 1.98
Cinchonine 8.35 Protopine 5.99
(pKa2"4.28) Quinidine 8.77
Cocaine 8.39 (pKa2"4.20)
Codeine 8.21 Quinine 8.34
Colchicine 1.85 (pKa2"4.30)
d-Coniine 10.90 Retronecine 8.88
Cytisine 8.12 1-Scopolamine 7.55
(pKa2"1.20) Solanine 7.54
Emetine 8.43 Sparteine 11.96
(pKa2"7.56) (pKa2"4.80)
Ergometrine 6.73 Strychnine 8.26
Harmine 7.61 (pKa2"2.50)
Heliotridine 10.55 Thebaine 8.15
Heroine 7.60 Theobromine 1.00
1-Hyoscyamine 9.65 Theophylline 1.00
Tropacocaine 9.88
Isopilocarpine 7.18 Tropine 10.33
Yohimbine 7.45
(pKa2"3.00)

Reproduced with permission from Popl et al. (1990).

silica gel is most often used in combination with
basic mobile phases or the gel is impregnated with
basic buffers or basic compounds (KOH, NaOH,
NaHCO3). Colchicine is the exception to these rules
and, because of its neutral character, can be analysed
in neutral solvent systems in combination with silica
gel plates.
There are fewer applications using alumina. Basic
alumina is most often used. The weakly basic charac-
ter of the surface allows the use of neutral solvent
systems as mobile phases. Depending on the nature of
the alkaloids examined, neutral or acidic alumina
may sometimes be more suitable.
As presented in detail in Table 2, solvent systems
used in adsorption chromatography are either binary
or ternary mixtures of chloroform, benzene, ethyl
acetate and others. AlkaliRcation of the mobile phase
is achieved by the addition of ammonia, di-
ethylamine, triethylamine or triethanolamine. Very
interesting methods for choosing a suitable solvent
were proposed in the late 1960s, and were based on
the weighted average values of dielectric constants,
and by the introduction of homogenous azeotropic
mixtures (methanol}chloroform}methyl acetate,
methanol}acetone-chloroform, methanol}benzene).
When choosing the proper solvent strength, especially
in complex eluent mixtures used for the analysis of
alkaloids, the xe, xd, xn parameters developed by
Snyder are useful. They refer to the possibility of
a solvent acting as a proton acceptor, proton donor or
the one exhibiting strong dipole interaction. All pos-
sible compositions of quaternary, ternary and binary
solvent mixtures have been described by the Prisma
model. It may be applied either in normal or reversed-
phase systems with the aim of optimizing the condi-
tions of separation.
Pseudo-reversed-phase
Chromatography
Chromatographic systems composed of silica gel and
buffered aqueous organic mobile phases have been
successfully used in recent years to isolate and separ-
ate alkaloids. The retention mechanism occurring
here, described as pseudo-reversed phase, is fairly com-
plex. An important role is played by the hydrophobic
1960 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
Table 2 Examples of the most popular chromatographic systems for TLC of the main alkaloid groups
Compounds separated Applications
Phenylethylamine derivatives
Ephedrine and its Qualitative identification of
derivatives ephedrine
Determination of ephedrine
in Herba Ephedrae
Determination of ephedrine
in bulk drugs
Colchicine and related Determination of colchicine
compounds in bulk drugs, dragees (BP)
Analysis of colchicine in tablets Silica gel
and plant material: Colchicum
autumnale seeds, Iphigena indica
Separation of colchicine and
3-demethylocolchicine,
demecolcine in Turkish
Colchicum and Merendera
species
Imidazole alkaloids
Pilocarpine Qualitative identification of
pilocarpine
Determination of pilocarpine
in ocular system
Determination of pilocarpine
nitrate in bulk drugs (EP)
Separation of pilocarpine,
isopilocarpine, pilocarpic acid
and isopilocarpic acid in
eye drops
Indole alkaloids
Strychnos alkaloids Determination of strychnine in
biological specimens
TLC analysis of strychnine
and brucine in plant extract
from Strychnos nux vomica
Separation of strychnine
and brucine
Yohimbine type Determination of reserpine
Rauwolfia alkaloids in Rauwolfia serpentina and
and related bases R. cubana stem bark
Isolation of alkaloids from
Mitragyna speciosa
TLC analysis of extract from
Uncaria
Determination of serpentine
and ajmalicine in
Catharanthus roseus
TLC analysis of ajmaline
steroisomers, vincine,
vincamine
Adsorbent Solvent system
Silica gel Butanol}acetic acid}water (6 : 3 : 1)
Silanized silica gel 1 mol L\1 acetic acid}3% potassium
chloride
Silanized silica gel 1 mol L\1 acetic acid}methanol (80 : 20)
impregnated with anionic
and cationic detergents
Silica gel Isopropyl ether}acetone}
tetrahydrofuran (15 : 3 : 2)
Silica gel n-Butanol}water}formic acid (7 : 2 : 1)
Silica gel Chloroform}acetone}ammonium
Aluminium oxide hydroxide (5 : 4 : 0.2 or 25 : 20 : 0.4)
Chloroform}methanol (80 : 0.5)
Silica gel Benzene}ethyl acetate}butylamine
(5 : 4 : 1 or 7 : 2 : 1)
Silica gel Chloroform}acetone}diethylamine
(5 : 4 : 1)
Chloroform}acetone}water (5 : 4 : 1)
Aluminium oxide Chloroform
Silica gel Methanol}1% potassium dihydrogen
phosphate (pH 6 : 9 : 1)
Silica gel Chloroform}methanol}ammonium
hydroxide (85 : 14 : 1)
Silica gel Ethanol}chloroform}28% ammonium
hydroxide (53 : 30 : 17)
Silica gel Dichloromethane}methanol}water}
formic acid}diethanolamine
(72.3 : 25 : 2.5 : 0.1 : 0.1)
Silica gel Methanol}4 mol L\1 ammonium
hydroxide (9 : 1)
Aluminium oxide Benzene-ethanol (9 : 1 or 8 : 2)
Silica gel Chloroform}methanol (19 : 1 or 9 : 1)
Cellulose Ethyl acetate}cyclohexane}diethylamine
(210 : 90 : 1)
Butanol}acetic acid}water (60 : 15 : 25)
RP-18 Methanol}water (4 : 2)
Silica gel Chloroform}acetone (5 : 4)
Ethyl acetate}isopropanol}ammonium
hydroxide (100 : 2 : 1)
Silica gel Chloroform}methanol (9 : 1)
Ethyl acetate}methanol (3 : 1)
Chloroform}acetone}diethylamine
(5 : 4 : 1)
Silica gel Acetone}petrol ether}diethylamine
(2 : 7 : 1)
Hexane}chloroform}methanol (5 : 1 : 1)

Table 2 Continued
Compounds separated Applications
Ergot alkaloids Determination of ergotamine
tartrate in bulk drugs and
dihydroergotamine mesylate
(USPXXI, EP,BP)
Qualitative identification of
hallucinogen ergot alkaloids
from Ipomoea Tricolor Cav
Quantitative analysis of ergot
alkaloids: lysergol, ergometrine,
agroclavine, ergotamine,
ergocristine, ergotaminine,
ergocristinine
Qualitative identification of
ergot alkaloids
Pyridine and piperidine alkaloids
Tobacco alkaloids Rapid TLC identification of
cytisine and nicotine
Determination of nicotine,
nornicotine, anabasine,
nicotyrine, 2,2-dipiridyl
Tropane alkaloids Quantitative determination of
atropine in Chinese medicine
Determination of atropine in
pharmaceutical preparations:
bulk drugs and injections
(USPXXI)
Qualitative identification of
atropine, scopolamine,
tubocurarine in African arrow
poison
TLC analysis of Belladonna
tinctura (atropine, scopolamine)
Analysis of Hyoscyamus extract
Pseudotropine alkaloids Determination of cocaine and
local anaesthetics
Identification of alkaloids in
Erythroxylium hypericifolium
leaves
Quinoline alkaloids
Cinchona alkaloids Quantitative analysis of
17 cinchona alkaloids
TLC analysis of cinchona
alkaloids as pure substances
III / ALKALOIDS / Thin^Layer (Planar) Chromatography 1961
Adsorbent Solvent system
Silica gel Dimethylformamide}ether}chloroform}
ethanol (15 : 70 : 10 : 5)
Silica gel Chloroform}ethanol (9 : 1)
Silica gel Ethanol}tetrahydrofuran}ethyl acetate
(1 : 1 : 8)
Water}ethanol}ether (5 : 35 : 60)
Acetonitrile}ethanol}toluene (85 : 10 : 5)
Silica gel (circular U-RPC) Water}ethanol}ether (1 : 7 : 12)
Acetonitrile}ethanol}toulene (17 : 2 : 1)
Silica gel Stepwise gradient elution:
1 Chloroform}diethylamine (12 stages,
7 steps)
2 Chloroform}acetone}diethylamine
(11 stages, 5 steps)
Silica gel Dichloromethane}methanol}10%
ammonium hydroxide (83 : 15 : 2)
Silica gel (OPLC) Ethyl acetate}methanol}water
(12 : 35 : 3)
Silica gel Chloroform}acetone}methanol}
ammonium hydroxide (70 : 10 : 15 : 1)
Silica gel Chloroform}diethylamine (9 : 1)
Silica gel Chloroform}cyclohexane}diethylamine
(3 : 6 : 1)
Silica gel with micro- Chloroform}acetone}methanol}
crystalline cellulose (5 : 2) ammonium hydroxide (73 : 10 : 15 : 2)
Silica gel Methanol-ammonium hydroxide (98 : 2)
Chloroform}butylamine (9 : 1)
Ethyl acetate}formic acid}ammonium
hydroxide (10% : 83 : 15 : 2)
Water}methanol}sodium acetate buffer
(0.2 mol L\1 aqueous: 28 : 12 : 60 : 1)
Silica gel Two-dimensional:
1 Cyclohexane}benzene}diethylamine
(75 : 15 : 10)
2 Chloroform}methanol (8 : 1)
Aluminium oxide Chloroform}ethanol (1 : 1)
Butanol}ethanol (95 : 1)
Silica gel Chloroform}acetone}methanol}25%
ammonium hydroxide (60 : 20 : 20 : 1)
Silica gel Chloroform}diethylamine (9 : 1)
Chloroform}methanol}ammonium
hydroxide (25% : 85 : 14 : 1)
Kerosene}acetone}diethylamine
(23 : 9 : 9)
1962 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
Table 2 Continued
Compounds separated Applications
Determination of quinidine
and dihydroquinidine in serum
Preparative TLC quinoline
alkaloids from Orixa japonica
stems
Determination of quinine
hydrochloride, quinidine
sulfate in bulk drugs (EP, BP)
Determination of cinchonine
in bulk drugs
Isoquinoline alkaloids
Protoberberine and Determination of berberine in
protopine alkaloids biological matrix
Separation of berberine in
presence of quaternary
alkaloids in plant extracts
Quantitative analysis and
qualitative identification of
protoberberine alkaloids
Quantitative analysis of
berberine in capsule
TLC analysis of protopine
and allocryptopine from
Turkish Papaver curviscapum
Determination of berberine in
bulk drugs
Determination of sanguinarine,
chelidonine, protopine,
allocryptopine in Chelidonium
maius
Morphine alkaloids Analysis of morphine alkaloids
in opium
Determination of morphine
and semisynthetic derivatives
Adsorbent Solvent system
Toluene}diethyl ether}diethylamine
(20 : 12 : 5)
Silica gel Ethyl acetate}ethanol}n-butanol}
ammonium hydroxide (56 : 28 : 4 : 0.5)
RP-18 Methanol}water (2 : 1)
Silica gel Benzene}ethyl acetate (4 : 1)
Silica gel Diethylamine}ether}toluene
(10 : 24 : 40)
Silica gel sprayed Ammonium hydroxide}methanol
with 0.1 mol L\1 (1.5 : 100)
methanolic potassium
hydroxide
Silica gel Ethyl acetate}methyl
acetate}methanol}water (27 : 23 : 6 : 5)
Silica gel (OPLC) Ethyl acetate}tetrahydrofuran}acetic
acid (6 : 2 : 2)
Silica gel Two-step development in twin trough
chamber:
1 Ethyl acetate-methanol}ammonium
hydroxide (10 : 10 : 1)
2 Benzene}ethyl
acetate}isopropanol}methanol}water
(20 : 10 : 5 : 5 : 1)
Second trough containing 5 mL conc.
NH3
Silica gel Ethyl acetate}acetone}formic acid}water
(20 : 17 : 4 : 2)
Silica gel Benzene}ethanol}ammonium hydroxide
(8 : 2 : 0.03)
Benzene}acetone}methanol (7 : 2 : 1)
Toluene}acetone}methanol}ammonium
hydroxide (45 : 45 : 7 : 3)
Silica gel sprayed Ammonium hydroxide}methanol
with 0.1 mol L\1 (1.5 : 100)
methanolic potassium
hydroxide
Silica gel Toluene}methanol}diethylamine
(60 : 5 : 2) saturated with formamide
Silica gel Benzene}ethanol (17 : 1 or 9 : 1)
Benzene}dioxane}ethanol}ammonium
hydroxide (50 : 40 : 5 : 5)
Toluene}acetone}ethanol
(96%)}ammonium hydroxide (25%)
(45 : 45 : 7 : 3)
Hexane}chloroform}diethylamine
(50 : 30 : 7)
Ethyl acetate}methanol}ammonium
hydroxide (85 : 10 : 5 or 75 : 20 : 5)
Silica gel Chloroform}triethanolamine (95 : 5)
Chloroform}methanol}water (7 : 5 : 1)
Butanol}ammonium hydroxide}
water}methanol (20 : 1 : 4 : 2)
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography 1963

Table 2 Continued

Compounds separated Applications Adsorbent Solvent system

Determination of Dabsyl Silica gel Chloroloform}ethanol}triethanolamine

derivatives of morphine
in urine
Isoquinoline bases Determination of papaverine,
codeine, eupaverine
Determination of emetine
and tubocurarine
TLC analysis of emetine
hydrochloride in bulk drugs
(USPXXI, BP)
TLC analysis of codeine
in bulk drugs (EP)
TLC analysis of papaverine
hydrochloride in bulk drugs
(EP)
Determination of codeine,
chlorpheniramine,
phenylephrine, paracetamol
(acetaminophen) in syrup
and tablets
Benzylisoquinoline Determination of alkaloids
alkaloids in Anisocycla cymosa roots
and plant extract
Determination of
bisbenzylisoquinoline alkaloids
in A. jollyana leaves
Aporphine alkaloids Analysis in plant material
Various isoquionoline Determination of cocaine,
alkaloids heroin and local anaesthetics
in street drugs
Analysis of major drugs
of abuse in urine
(30 : 2 : 0.05)
RP-18 (IP-TLC) Water}acetone (20 : 80, 100 : 0) with
0.1 mol L\1 of ion reagent}sodium
alkylsulfonate
Silica gel Ethyl acetate}isopropanol}ammonium
hydroxide (25% : 9 : 7 : 2)
Silica gel Chloroform}diethylamine (9 : 1)
Silica gel Ammonium hydroxide}cyclohexane}
ethanol (6 : 30 : 72)
Silica gel Diethylamine}ethyl acetate}toluene
(1 : 2 : 7)
Silica gel Butanol}methanol}toluene}water}acetic
acid (3 : 4 : 1 : 2 : 0.1)
Silica gel Chloroform}methanol}diethylamine}
ammonium hydroxide (8 : 2 : 2 : 0.5)
Benzene}acetone}ammonium hydroxide
(15 : 15 : 1)
Silica gel Chloroform}toluene}methanol}acetone}
ethyl acetate}ammonium hydroxide
(270 : 30 : 80 : 30 : 3)
Aluminium oxide Toluene}chloroform}methanol}ammonium
hydroxide (100 : 150 : 40 : 3)
Silica gel Cyclohexane}ethyl acetate (3 : 2)
Cyclohexane}acetone (9 : 1)
Petrol ether}acetone (7 : 3)
Chloroform}methanol (9 : 1)
Silica gel Benzene}chloroform}triethanolamine
(9 : 9 : 4)
Ethyl acetate}isopropanol}28%
ammonium hydroxide (40 : 30 : 3)
Silica gel Ethyl acetate}cyclohexane}
methanol}ammonium hydroxide
(conc.)}water (70 : 15 : 8 : 2 : 0.5)
Ethyl acetate}cyclohexane (50 : 60)

Diterpene and steroidal alkaloids

Diterpene Determination of aconitine
nitrate in bulk drugs
Determination of aconitine,
3-deoxyaconitine, mesaconitine
in Wutou and Aconitum
Isolation of norditerpenoid
alkaloids from extract of roots
of Delphinum tatsienense
TLC of 8 diterpenoid
alkaloids from Aconitum
septentrionale
Silica gel spray 0.1 mol L\1
potassium hydroxide
methanol
Silica gel
Aluminium oxide (neutral)
Silica gel (centrifugal TLC)
Silica gel (preparative TLC)
Silica gel
Aluminium oxide
(centrifugal TLC)
Ammonium hydroxide}
methanol (1.5 : 100)
Cyclohexane}ethyl
acetate}ethylenediamine (8 : 1 : 1)
Gradient elution: hexane, hexane}diethyl
ether (25 : 75), diethyl ether, diethyl
ether}methanol
Diethyl ether}75% methanol}0.3%
diethylamine
Diethyl ether}5% methanol
Hexane}chloroform (6 : 4)
Chloroform}methanol (8 : 2 or 97 : 3)
Gradient of hexane, ether and methanol
1964 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
Table 2 Continued
Compounds separated Applications
Steroidal alkaloids Isolate ecdysteroids from
the herba of Siline tatarica
Veratrum alkaloids Determination of veratrum
alkaloids jervine,
veratroylzygadenine,
rubijervine, isorubijervine,
veromine in Veratrum root
and tincture
Solanum alkaloids Determination of solanum
alkaloids (solanidine)
from spiked milk and
-solasonine,
, -solamargine from
Solanum ptycanthum
Miscellaneous heterocyclic systems
Pyrrolizidine alkaloids TLC analysis in plant material
Lupin alkaloids TLC of lupanine and
7-hydroxylupanine from
Cytisophyllum seccilifolium
Carbazole alkaloids
Xanthine alkaloids Qualitative identification and
preparative TLC of alkaloids
from Bosistoa floydi leafs
Purine bases Determination of purine bases
in urine
Determination of caffeine,
theophylline and 15 drugs
in Chinese herbal preparations
Determination of caffeine
and theobromine in bulk
drugs (EP)
Determination of theophylline
in capsules (USPXXI)
in tablets with ephedrine
hydrochloride and phenobarbital
(USPXXI, EP)
Quinolizidine Qualitative identification
BP, British Pharmacopoeia; EP, European Pharmacopoeia, USPXXI, The United States Pharmacopeia, Twenty-first Revision.
interactions of siloxane groups with the non-polar
fragments of the separated alkaloids, as well as by ion
exchange interactions. In the retention of alkaloids
a dominant role is played by the ion exchange mecha-
nism which is due to the weak cation exchange prop-
Adsorbent Solvent system
Silica gel (droplet Ethyl acetate}methanol}ammonium
countercurrent hydroxide (17 : 5 : 3)
chromatography) Dichloromethane}ethanol (17 : 3)
Chloroform}methanol}acetone (6 : 2 : 1)
Methanol}water (13 : 7)
Silica gel Benzene}ethanol}diethylamine
(80 : 16 : 4)
Aluminium oxide Benzene}ethanol (95 : 5)
Silica gel Methanol}chloroform}1% ammonium
hydroxide (2 : 2 : 1)
Silica gel Dichloromethane}methanol}ammonium
hydroxide (85 : 15 : 2 or 75 : 23 : 2 or
79 : 20 : 1)
Chloroform}methanol (4 : 1)
Silica gel impregnated Chloroform}methanol}ammonium
with 0.1 mol L\1 NaOH hydroxide (60 : 10 : 1 or 17 : 38 : 0.25)
Silica gel Chloroform}methanol}28% ammonium
hydroxide (85 : 14 : 1)
Silica gel Benzene}chloroform (1 : 1)
Silica gel Chloroform}ethyl acetate (3 : 2)
Silica gel Two-dimensional:
1 Chloroform}methanol (4 : 1)
2 Butanol}chloroform}acetone}
ammonium hydroxide (4 : 3 : 3 : 1)
Silica gel Dichloromethane}methanol}water
(183 : 27 : 5)
Ethyl acetate}toluene}
dimethylformamide}formic acid
(75 : 70 : 4 : 2)
Dichloromethane}methanol (183 : 27)
Silica gel Ammonium hydroxide}acetone}
chloroform}butanol (1 : 3 : 3 : 4)
Cellulose Methanol}water
Silica gel Chloroform}acetone}methanol}
ammonium hydroxide (50 : 10 : 10 : 1)
Silica gel Chloroform}cyclohexane}butylamine
(5 : 4 : 1)
Aluminium oxide 1.5% Methanol in chloroform
erties of silica gel at pH"2}8 and the fact that
aromatic amines chromatographed in an aqueous
mobile phase are weakly protonized at pH"
pKa!1. The selectivity of such systems depends
then, primarily, on the pH of the mobile phase but
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
also on the kind of organic modiRer, which is usually
methanol or acetonitrile.
Reversed-phase Chromatography
Nonpolar adsorbents have rarely been applied in
TLC of alkaloids, perhaps because of the low efRcien-
cy of such systems, which is caused by the interaction
of alkaloid molecules with silanol groups present
on the adsorbent surface in addition to the hydrocar-
bon chains. In reversed-phase chromatography on
silanized silica gel, alkaloids as easily ionized com-
pounds require speciRc conditions of analysis such as
suppression of dissociation, ion suppression or the
application of speciRc ion pair reagents.
The suppression of dissociation is achieved with
a mobile phase of a suitable pH (pH5pKa) for the
speciRc solvent, in accordance with the curve shown
in Figure 2.
Reversed-phase conditions may also be obtained by
impregnating silica gel with parafRn or silicone oil.
Additionally, chemically bonded reversed phases
with polar groups (cyano- and amino-layers) have
been employed. Their properties depend on the kind
of compounds to be separated and on the composi-
tion of the mobile phase.
Ion Pair Chromatography
The use of ion pair chromatography (IP-TLC) of
alkaloids may be carried out on normal and reversed
phases. This technique has been applied to analyse
basic drugs, including alkaloids, on silica gel using
normal-phase systems. The best results are obtained
by applying chloride and bromide as counterions of at
least 0.1 mol L\1 concentration in the spreading
slurry or in the solvent.
Reversed-phase IP-TLC is far more widely used.
The counterion reagents which may be present in the
mobile phase and serve for impregnation in the non-
polar stationary phase may be di-(2-ethylhexyl) or-
thophosphoric acid (HDEHP), camphoric and cam-
phorosulfonic acids, sodium dodecylsulfate and
simple hydrophilic anionic reagents such as per-
chloric acid, oxalic acid and trichloroacetic acid. The
acidic environment of the mobile phase ensures ioniz-
ation of the acidic counterions and enables the cre-
ation of an ion pair with the alkaloids protonized
under these conditions. The behaviour of some
isoquinoline bases using RP-18 plates and alkylsul-
fonates as counterions has also been investigated.
Although retention and separation selectivity in
IP-TLC depend on many factors, optimization of
such chromatographic systems is basically concerned
with pH changes, concentration and the chain length
1965
of the counterion or the concentration of organic
modiRer in the mobile phase.
Partition Chromatography
In the past, partition chromatography conducted on
paper was a perfect model for establishing optimum
extraction systems for alkaloid isolation. In paper
chromatography, the system allowing partition con-
ditions is mainly composed of cellulose with an aque-
ous solvent or an aqueous buffer solution of pH 3}7,
depending on the nature of the alkaloids. Silica gel
combined with an aqueous phase or a water-
saturated organic solvent also allows for the domina-
tion of the partition mechanism, thanks to deactiva-
tion of the surface silanol groups. The aqueous phases
in such systems are often alkalized with aqueous am-
monium hydroxide or acidiRed with hydrochloric acid.
Partition conditions, similar to paper chromatogra-
phy, may be obtained by impregnating cellulose or
silica gel with a solution of formamide in ethanol and
using mobile phases immiscible with the stationary
phase, such as chloroform, benzene, cyclohexane or
their mixtures.
Ion Exchange Chromatography
Ion exchange techniques are applied for both crude
fractionation and separation and determination of
alkaloids.
The typical ion exchange sorbents used for TLC of
alkaloids have been as follows: anion exchangers AG
1-X4 and Cellex D, and cation exchangers with cellu-
lose (alginic acid and sodium carboxymethylcel-
lulose), parafRn (Rexyn 102) and polystyrene
(Dowex 50-X4) matrices.
While choosing the best eluent for ion exchange
chromatography, pH values should be carefully con-
sidered. They are closely correlated with the number
of charges in the alkaloid molecules and at the same
time decide the retention values. The trends for most
alkaloids Rt the type of curves shown in Figure 3.
One of the popular adsorbents which may function
as an ion exchanger is aluminium oxide (AI2O3) with
an aqueous mobile phase. Depending on the kind of
aluminium oxide used, a cation- or anion-exchanging
mechanism may occur. Thus, in aqueous alcoholic
solution basic alumina functions as a cation exchanger
(I), but acidic alumina acts as an anion exchanger (II).
With neutral alumina, both types of reactions may
take place depending on the conditions used:
(I) Al}O}Na#(BH)#Cl\
PAl}OH#B#Na##Cl\
(II) Al}Cl#BH#OH\PAl}OH#(BH)#Cl\
1966 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
Figure 3 RM versus pH curves for some alkaloids on alginic
acid thin layers (after Lepri L, Desideri PG and Lepori M (1976)
Chromatographic Behaviour of Alkaloids of Thin Layer of Cation
Exchangers. Journal of Chromatography 123, 175. Amsterdam:
Elsevier).
Adsorbents Impregnated with
Metal Salts
The use of silica gel and aluminium oxide impreg-
nated with metal salts (cadmium and zinc nitrate) for
the separation of some alkaloids has been studied.
For steroid alkaloids, the impregnation of the sta-
tionary phase with silver salts } so-called argentation
TLC } has been applied. This technique is based on
the formation of -complexes with the separated
compounds during the chromatographic process.
Detection of Alkaloids
Only a few alkaloids are directly visible on the
chromatogram as coloured spots and visualization
methods have to be applied to detect them. In order to
detect the compounds under UV light, Suorescing
indicators are added to the adsorbent.
Alkaloids become visible in short wavelength UV
light ("254 nm), where they appear as dark zones
on a Suorescent background. This is considered to be
a nonselective method of detection because, on the
layer containing a Suorescent indicator, the emission
is quenched in regions where all aromatic organic
compounds absorb the UV light with which the plates
are irradiated.
Some alkaloids, such as indoles, quinolines,
isoquinolines and purines, cause pronounced quench-
ing of Suorescence, but some (e.g. tropine alkaloids)
only weakly quench UV light. Sometimes compounds
can be detected under a UV lamp due to their own
luminescence. Excitation is usually performed using
long wavelength radiation of "365 nm. Alkaloids
absorb radiation and then usually emit it in the visible
region of the spectrum, where they appear as bright-
coloured luminous zones of blue, blue-green or violet,
for example, RauwolRae radix, Chinae cortex,
Ipecacuanhae radix, Boldo folium, and of yellow, e.g.
colchicine, sanguinarinae, berberine.
Other methods of physical detection make the most
of the chemical properties of alkaloids. As basic com-
pounds, these properties can be detected using pH
indicators applied to the chromatogram by dipping it
or spraying it with 0.01}1% indicator solutions.
Bromocresol Green with pH transition from 3.8 to
5.4 is applied for many alkaloids; Bromocresol Purple
(pH"5.2}6.8) is predominantly applied for ephed-
rine.
Another nonselective detection method for alkal-
oids as lipophilic substances is the treatment of
a chromatogram with iodine vapour or dipping
into or spraying with 0.5}1% iodine solutions.
Molecular iodine is enriched in the chromatogram
zones and colours them brown. Emetine and cephae-
line, the two major alkaloids of ipecacuanha, begin to
glow after treatment with iodine vapour. In this case,
the molecular iodine which acts as a quencher must
be removed by heating, before the yellow (emetine)
and blue (cephaeline) Suorescent zones become
visible.
Although the methods described are usually fairly
sensitive and allow a detection limit of less than 1
g,
sometimes they are insufRcient. That is why they have
to be supplemented by speciRc reactions with a num-
ber of detection reagents (Table 3).
The most popular reagents which react with terti-
ary and quaternary nitrogen atoms present in alkal-
oid molecules are Dragendorff’s reagent and potassi-
um iodoplatinate. Alkaloids containing primary and
secondary amino groups treated with dimethyl sulfate
give quaternary nitrogen atoms, permitting effective
detection with these reagents too.
Dragendorff ’s and iodoplatinate reagents exists in
various modiRcations. The replacement of water in
these reagents by acetic acid or ethyl acetate, diethyl
ether}methanol or hydrochloric acid increases the
sensitivity of the reaction and signiRcantly improves
the sharpness of spots. Spraying 10% sodium nitrate
solution after the use of Dragendorff’s reagent causes
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography 1967

Table 3 Selection of detection reagents for postchromatographic derivatization of alkaloids

Reagent Substances Reaction Method Result
detected

Ammonia vapour Alkaloids, e.g. Morphine and heroin Heat the chromatogram Morphine, 6-monoacetylmorphine
morphine, heroin, form fluorescent in the drying cupboard and heroin appear as blue
6-mono- oxidation products to 110}1203C for 25 min fluorescent zones on a dark
acetylmorphine and place it for 15 min background under UV light
in a twin-trough chamber, ("365 nm). In each case the
whose second trough detection limits are 2 ng of
contains 10 mL of 25% substance per chromatographic
ammonia solution. Then zone. The fluorimetric
immerse for 2 s in a determination is carried out in UV
solution of liquid light exc"313 nm, fl"390 nm
paraffin}n-hexane (1 : 2)
Formaldehyde Alkaloids, e.g. The reaction mechanism Dry the chromatogram Morphine alkaloids yield blue
reagent (1,2- codeine, morphine, has not been elucidated. in a stream of warm air chromatogram zones on a pale
naphthoquinone- heroin, 6-mono- It is possible that for 5 min, immerse in the blue background. The detection
4-sulfonic acid)} acetylmorphine formaldehyde reacts by reagent solution for 4 s limits are 10}20 ng of substance
perchloric acid oxidation, as in Marquis and heat to 703C for c. per chromatogram zone. The
reaction 10 min absorption photometric analysis
can be performed at reflectance
"610 nm
2-Methoxy-2, Alkaloids from MDPF reacts directly Free the chromatogram Colchicine appears as a yellow
4-diphenyl-3(2H)- Colchicum autum- with primary amines to from mobile phase in a fluorescent zone on a dark
furanone (MDPF) nale (Colchicine) form fluorescent products stream of warm air (45 min), background in UV light (365 nm).
immerse in the reagent The detection limit is 10 ng per
solution for 4 s and then heat chromatogram zone. The
to 1103C for 20 min fluorimetric analysis is carried out
with excitation at exc"313 nm,
and evaluation at fl'390 nm
2,4-Dinitrophenyl- Alkaloids Reagent reacts with Immerse the chromatogram Substances yield yellow to
hydrazine carbonyl groups with in the dipping solution for orange-yellow chromatogram
the elimination of water 2 s or spray and then dry zones on an almost colourless
to yield hydrazone and in a stream of warm background
with aldoses or ketoses air (10}20 min at 1103C)
to yield coloured
osazones
2,6-Dichloro- Isoquinoline Reagent reacts with Dry the chromatogram Cephealine produces a blue
quinone}4- alkaloids phenols or anilines for 5 min in a stream of colour immediately on immersion,
chloroimide which are not substituted warm air, immerse in while emetine only does so on
in the p-position the dipping solution for heating. On storage this colour
5 s and then heat to 1103C slowly changes to brown
for 2 min (background light brown). The
detection limits are c. 10 ng per
chromatogram zone. The
absorption photometric analysis
was made at "550 nm
o-Phthal- Ergot alkaloids In the presence of Immerse the dried Substance zones are produced
aldehyde 2-mercaptoethanol, chromatogram for 1 s in that mainly yield blue (or yellow)
(OPT, OPA) o-phthalaldehyde the reagent solution and fluorescence under long
reacts with primary then heat to 40}503C in wavelength light ("365 nm)
amines to yield the dry cupboard for 10 min
fluorescent
isoindole derivatives
Phosphomolybdic Morphine Morphine can be Dry the chromatogram Blue zones appear on a yellow
acid oxidized with in a stream of warm air background immediately or after
phosphomolybdic and immerse for 2}3 s a few minutes
acid, whereby a portion in the reagent solution,
of the Mo(VI) is reduced or spray the layer with it
to Mo(IV) which forms
blue-grey oxides with
the remaining Mo(VI)
1968 III / ALKALOIDS / Thin^Layer (Planar) Chromatography

Table 3 Continued

Reagent Substances Reaction Method Result

detected

Trichloroacetic Alkaloids from, The reaction mechanism Dry the chromatogram in Light blue fluorescent zones
acid e.g. Veratrum has not yet been a stream of cold air and appear mainly under long
colchicum elucidated immerse for 1 s in the wavelength UV light ("365 nm).
reagent solution or spray The fluorescence can be
with it and then heat at stabilized and intensified by
1203C for 10 min dipping the plate into
a solution of liquid
paraffin}n-hexane (1 : 2)
Sulfuric acid Alkaloids The reaction mechanism Dry the chromatogram Under long wavelength UV light
has not yet been in a stream of warm air ("365 nm) characteristic
elucidated for 10 min, immerse in the substance-specific yellow, green,
dipping solution for 1}2 s red and blue fluorescent
or spray with the spray chromatogram zones usually
solution, dry in a stream appear, and are often
of warm air and then heat recognizable in visible light
to 95}1403C for 1}20 min
7-Chloro-4- Alkaloids NBD reacts with Dry the chromatograms. Under UV light ("365 nm) the
nitrobenzo-2- nucleophilic compounds Immerse in dipping solution chromatogram zones fluoresce
oxa-1,3-diazole to yield the of sodium acetate in greenish-yellow, olive brown or
(NBD-chloride corresponding methanol}water for 1 s. violet. The plate background also
reagent) 7-substituted Dry in a stream of warm air fluoresces, but appreciably less.
4-nitrobenzofurazan and dip after cooling in The detection limits are
derivatives NBD-chloride reagent 100}800 ng substance per
in ethanol and then heat chromatogram zone
to 1003C for 2}3 min.
Alternatively the
chromatogram can be
sprayed with the
appropriate spray solutions
tert-Butyl Alkaloids The reaction mechanism Dry the chromatogram, The analysed compounds appear
hypochlorite has not yet been immerse in dipping solution in long wavelength UV light
elucidated of reagent in carbon tetra- (365 nm), yellow to violet
chloride or cyclohexane for fluorescent zones, on a dark
1 s (or spray or expose to its background. The detection limit
vapours) then immerse in for morphine is 10 ng and for
dipping solution of chloro- papaverine 1 ng per
form, paraffin oil and chromatogram zone
triethanolamine (6 : 1 : 1)
for 1 s and dry in hot air
Formaldehyde- Alkaloids, e.g. Morphine reacts with Dry the chromatogram Morphine alkaloids yield reddish
sulfuric acid morphine, codeine, formaldehyde in in a stream of warm air chromatogram zones (codeine
(Marquis reagent) heroin, acidic solution to yield for 5 min, immerse in yielded blue on a pale pink
6-monoacetyl- a cyclic ketoalcohol, the dipping solution for 6 s background). If a quantitative
morphine which is transformed and heat to 1103C fluorimetric analysis is to follow,
into the coloured for 20 min the chromatogram is exposed to
oxonium or carbenium ammonia vapour for 20 min and
ion in acidic conditions immersed for 2 s in 20% dioctyl
sulfosuccinate in chloroform. After
drying, morphine alkaloids
appear as pink to red flourescent
zones on a blue fluorescent
background under UV light
("365 nm). The fluorimetric
analysis is carried out at
exc"313 nm, fl"560 nm
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography 1969

Table 3 Continued

Reagent Substances Reaction Method Result

detected

Iron (III) chloride- Indole alkaloids, The reaction mechanism Free the chromatogram Variously coloured chromatogram
perchloric acid e.g. from has not yet been from mobile phase in a zones are produced on
(FCPA reagent) Rauwolfia, elucidated stream of warm air a colourless background. For
Tabernaemontana, (45 min), immerse in the instance, strychnine appears as
Mitragyna, dipping solution for 4 s. a red and brucine as a yellow
Strychnos, Dry and heat to 1103C chromatogram zone on
Synclisia, Cinchona for 20 min a colourless background. The
detection limit for both is 10 ng per
chromatogram zone. The light
absorption in reflectance was
measured at "450 nm
Hydrochloric Alkaloids, The reaction Free the chromatogram Alkaloids are visible after
acid vapour e.g. mechanism has not yet from mobile phase (first irradiation with unfiltered UV light
papaverubines been elucidated in a stream of cold air for a from a mercury lamp
few minutes, than at 1003C
for 5 min), place in the
free trough of the
prepared twin-trough
chamber for 5 min and
then evaluate

Figure 4 (See Colour Plate 54). The chromatograms of the separated alkaloids developed on silica gel or alumina in solvent systems
1}4, detected with different reagents. Solvent systems: 1, toluene}ethyl acetate}diethylamine (70 : 20 : 10); 2, chloroform}diethyl-
amine (90 : 10); 3, toluene}chloroform}ethanol (28.5 : 57 : 14.5); 4, 1-propanol}water}formic acid (90 : 9 : 1). For identification of
compounds, reagents used and obtained results, see Table 4. (Reproduced with permission from Wagner H and Bladt S (1996) Plant
Drug Analysis. Thin-layer Chromatography Atlas. Berlin: Springer.)
1970 III / ALKALOIDS / Thin^Layer (Planar) Chromatography

Table 4 Symbols used in Figure 4

Symbol Detection Solvent Reference compounds Result

system

A Marquis reagentPvis
B Natural products,
polyethylene glycol
reagent (NP/PEG)P
UV 365 nm
C Sulfuric acid
reagentPUV 365 nm
D Dragendorff reagentPvis
E Dragendorff reagent
followed by sodium
nitritePvis
F Iodine/CHCI3 reagentP
UV 365 nm
G Pvis
H 10% H2SO4 followed
by iodoplatinate
reagentPvis
I van URK reagentPvis
J UV 254 nm
K UV 365 nm
1 Morphine (1), codeine (2),
papaverine (3), noscapine (4),
opium extract (5)
1 Morphine, papaverine, noscapine give a blue
1 Serpentine (1), quinine (2),
cinchonine (3), quinidine (4),
cinchonidine (5), cephaaeline (6),
emetine (7), yohimbine (8),
noscapine (9), hydrastine (10),
berberine (11), sanguinarine (12)
Strychnine (1), yohimbine (2),
physostigmine (3), nicotine (4),
1 veratrine (5), emetine (6),
papaverine (7), lobeline (8),
aconitine (9), narcotine (10)
1 Cephaelis accuminata (1),
cephaeline: R f&0.2; emetine:
1 R f&0.4 (2).
Cephaelis ipecacuanha (3)
2 China alkaloid mixture (1) Cinchona
succirubra (2)
3 Ergocristine (1), Secale cornutum (2),
ergotamine (3), ergometrine (4)
1 Strychnine (1), Strychni semen (2),
Ignatii semen (3), brucine (4)
4 Chelidonii herba different trade
samples (1}3), sanguinarine (4)
Morphine and codeine are immediately
stained violet; papaverine: weak violet;
noscapine: weak yellow brown
fluorescence in UV 365 nm; codeine does
not fluoresce
The fluorescence of quinine and quinidine is
a radial blue; cinchonine and cinchonidine:
deep violet, berberine and sanguinarine:
bright yellow
Alkaloids give orange-brown, stable colours
The zones become dark brown
Cephaeline fluoresces bright blue and
emetine: yellow-white
Cephaeline gives red and emetine weak
yellow zones
The violet-brown zone of quinine is followed
by the grey-violet zone of cinchonidine,
a weak red-violet zone of quinidine and
brown-red cinchonine (1)
In Cinchona succirubra extract additionally
three red-violet zones appear in the R f range
0.4}0.6 (2)
Secale alkaloids appear as blue zones in the
R f range of 0.05}0.4
Strychnine and brucine are characterized in
UV 254 nm by their strong quenching zones
The major alkaloid coptisin at R f&0.15
(bright-yellow) is followed by berberine,
chelerythrine, sanguinarine (broad yellow)
and chelidonine (weak yellow-green) in the
R f range of 0.75}0.85

the colour of alkaloid zones to be intensiRed or stabil- reagent, also causes an increase in the sensitivity of
ized and increases the sensitivity to 0.01}0.1
g. the reaction. Potassium iodoplatinate reagent gives
ModiRcation, where a chromatogram is sprayed preliminary identiRcation, due to the fact that differ-
with 10% sulfuric acid after the use of Dragendorff’s ent colours are obtained with different alkaloids.

Table 5 Examples of prechromatographic derivatization of alkaloids

Prechromatographic Reagent used Special applications

derivatization

Oxidation 10% Chromic acid in glacial acetic acid Strychnine and brucine
Potassium dichromate
Dehydration by heating the applied sample on silica layer
Reduction Sodium borohydride solution Not specified
Iodination Iodine vapour saturated chamber (18 h) Quinoline, isoquinoline, indole alkaloids
Nitration Concentrated nitric acid Brucine
Dansylation Dansyl chloride and twice bigger volume of 8% Morphine, 6-monoacetylmorphine,
sodium bicarbonate solution morphine-6-nicotinate
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography 1971

Table 6 Systematic analysis of alkaloids on TLC plates

Chemical Plant drug Botanical Major alkaloid Fluorescence Colour with hRF values
skeleton origin in UV light iodoplatinate
(366 nm) reagent S1 S2 S3 S4 S5 S6 S7 S8

Tropane Fol. Belladonnae Atropa

Rad. Belladonnae belladonna L,
Solanaceae Atropine Violet-blue 38 40 16 5 12 0 10 17
Fol. Hyoscyami Hyoscyamus Homatropine Violet-blue 37 45 15 5 23 4 24 15
niger L,
Solanaceae
Fol. Stramonii Datura Apoatropine Violet-blue 54 67 40 20 26 15 40 16
stramonium L,
Solanaceae
Rad. Scopoliae Scopolia Scopolamine Violet 56 60 19 3 34 30 0 52
carniolica Jacq. Scopoline White 60 90 44 20 44 46 50 37
Solanaceae
Fol. Duboisiae Duboisia Tropacocaine Violet 65 90 56 34 45 58 78 35
myoporoides R.
Br., Solanaceae
Fol. Cocae Erythroxylon coca Cocaine Violet 73 90 65 36 58 84 77 62
Lamarck
Erythroxylaceae
Indole Semen Calabaris Physostigma Physostigmine Pink 65 '90 32 4 44 59 50 46
venenosum
Balfour
Papilionaceae
Rad. Rauwolfiae Rauwolfia Reserpine Green-yellow White 72 80 20 0 46 63 35 69
serpentina
Rad. Serpentinae Bentham, Serpentinine Dark brown Red-brown 24 15 0 0 4 0 0 0
Apocynaceae
Semen Strychni Strychnos nux Serpentine Yellow-green Yellow-brown 53 56 8 0 10 0 3 12
vomica L, Ajmaline Blue Beige 47 42 12 3 30 6 13 56
Loganiaceae Strychnine Yellow 53 76 28 5 38 57 60 22
Cortex Pausinystalia Brucine Violet-brown 42 63 18 0 19 50 54 12
Yohimbehe Yohimbe Pierre, Yohimbine Green-blue Light yellow 63 62 18 3 37 33 15 60
Rubiaceae Ergocristinine Violet-blue Light brown 61 57 13 0 20 0 27 70
Secole cornutum Claviceps Ergotamine Violet-blue Pink 24 16 0 0 3 10 5 59
purpurea Tulasne Ergometrine Violet-blue White 14 6 0 0 2 3 0 64
Clavicipitaceae Ergometrinine Violet-blue Violet-blue 42 25 3 0 8 12 10 62
Ergocristine Violet-blue Beige-light 51 38 14 5 13 46 15 70
brown
Ergotaminine Violet-blue Pink 24 51 0 0 14 42 15 68
Dihydroergotamine Violet-blue Brownish 21 12 0 0 3 7 0 61
Dihydroergocristine Violet-blue Brownish 12 30 3 0 7 15 7 69
Isoquinoline Opium Papaver Thebaine Red-brown 65 90 51 16 50 71 76 40
somniferum L, Narceine Deep-blue 3 0 0 0 3 0 0 0
Papaveraceae Morphine Deep-blue 10 8 0 0 3 3 0 34
Papaverine Yellowish Yellow 67 90 42 3 47 85 84 70
Codeine Pink-violet 38 53 16 4 26 12 27 35
Noscapine Blue Light-yellow 72 90 51 10 57 81 79 72
Hydrastinine Steel blue Violet-blue 66 90 58 41 50 0 25 0
Dihydromorphinone Brownish 24 23 8 1 11 5 8 16
yellow
Dihydrocodeine Blue Violet-blue 38 54 18 6 28 10 30 25
Dihydrocodeinone Violet 51 65 21 4 30 48 43 18
Fol. Boldo Peumus boldus Boldine Violet Beige 16 16 3 0 5 24 6 58
Monimiaceae
Quinoline Cortex Chinae Cinchona Quinidine Blue Light yellow 34 40 15 0 25 12 18 50
Succirubra, Quinine Blue Yellow-white 19 26 7 0 17 9 18 43
Pavon, Rubiceae Cinchonine Beige-brown 38 44 17 7 27 0 22 40
Imidazole Fol. Jaborandi Pilocarpus Pilocarpine Light brown 41 52 9 0 13 32 25 55
microphyllus
Stapf e.a.;
Rutaceae
1972 III / ALKALOIDS / Thin^Layer (Planar) Chromatography

Table 6 Continued

Chemical Plant drug Botanical Major alkaloid Fluorescence Colour with hRF values
skeleton origin in UV light iodoplatinate
(366 nm) reagent S1 S2 S3 S4 S5 S6 S7 S8

Pyridine Semen Arecae Areca catechu L., Arecoline Violet 66 90 56 34 48 0 0 0

Herba Lobeliae Palmae
Lobelia inflata L., Lobeline Red-brown 68 90 48 14 48 55 60 55
Lobeliaceae
Quinolizidine Sarothamnus Sparteine Violet 70 90 68 68 55 0 55 5
Scoparius;
Leguminosae
Dihydroindole Fol. Catharanti Catharantus roseus Aspidospermine White 65 90 54 20 49 50 60 65
Apocynaceae
Aporphine Rhizoma Corydalidis Corydalis cava L. Bulbocapnine Blue White 65 '90 35 7 54 78 70 48
Schweigg et
Koerte
Papaveracae,
Fumariaceae
Isoquinoline Rad. Ipecacuanhae Cephaelis Emetine Blue Red-brown 67 90 40 6 45 38 58 50
ipecacuanha Cephaeline Violet-blue White 56 63 19 2 23 25 17 37
Rubiaceae
Miscellaneous
alkaloids
Derivatives of Aconiti Tuber Aconitum Aconitine Red-browm 68 '90 35 3 49 36 60 65
diterpene napellus L.,
Ranunculaceae
Xanthine Herba Ephedrae Ephedra sinica Ephedrine Light-grey 47 41 4 0 4 11 0 57
Stapf.
Ephedraceae
Colchicine Semen Colchici Colchicum Colchicine Light brown
autumnale L,
Liliaceae

TLC systems
S1, Silica gel G, activated: chloroform}acetone-diethylamine (5 : 4 : 1).
S2, Silica gel G, activated: chloroform}diethylamine (9 : 1).
S3, Silica gel G, activated: cyclohexane}chloroform}diethylamine (5 : 4 : 1).
S4, Silica gel G, activated: cyclohexane}diethylamine (9 : 1).
S5, Silica gel G, activated: benzene}ethyl acetate}diethylamine (7 : 2 : 1).
S6, Aluminium oxide G, activated: chloroform.
S7, Aluminium oxide G, activated: cyclohexane}chloroform (3 : 7)#0.05 diethylamine.
S8, Silica gel G, impregnated with 0.1 mol L\1 sodium hydroxide, activated: methanol.
(Reproduced with permission from Svendsen AB and Verpoorte R (1983) Chromatography of Alkaloids. Journal of Chromatography Library.
Amsterdam: Elsevier.)

For particular alkaloids, speciRc reagents can
be used; for instance, Marqui’s reagent (formalde-
hyde}sulfuric acid) or FroK hde’s reagent (sul-
fomolybdic acid}sulfuric acid) for morphine. KoK nig’s
reaction can be used to detect nicotine and related
alkaloids; Wachtmeister’s reagent (bis-diazatized
benzidine-sulfuric acid) is applied for alkaloids
belonging to the protoberberine and protopine
group.
The Vitaly reaction is speciRc for the tropane alkal-
oids, and reaction with 4-dimethylaminobenzal-
dehyde for indole alkaloids. Some examples of
applications of different reagents are illustrated in
Figure 4 and Table 4.
The use of -acceptor reagents producing colour
spots (TCNQ: 7,7,8,8-tetracyano-quinodimenthane;
TNF: 2,4,7-trinitroSuorenone; TetNF: 2,4,5,7-
tetranitro-9-Suorenone; DDQ: 2,3-dichloro-5,6-di-
cyanoquinone; DNFB: 2,4-dinitroSuorobenzene) for
the detection of alkaloids has been employed.
Initial derivatization during sample preparation or
in situ on the layer after the application of the sample
is called prechromatographic derivatization and com-
prises oxidation, reduction, iodination, nitration and
dansylation (Table 5).
Starting chromatographic separation with sample
derivatization allows better-quality results to be
obtained, especially as far as reproducibility and
 III / ALKALOIDS / Thin^Layer (Planar) Chromatography
lowering the detection limits are concerned. Mor-
phine as a dansyl derivative is an example of Suores-
cence stabilization and intensity augmentation as
a result of treatment of the chromatogram with
a 20% solution of liquid parafRn in n-hexane.
A similar phenomenon is observed for codeine,
morphine, monoacetylmorphine and heroin with the
aid of hydrophilic liquids, such as a 20% solution of
dioctyl sulfasuccinate in ethanol as a Suorescence
intensiRer.
Enhanced sensitivity can be achieved by impregnat-
ing the layer, by adding the reagent to the solvent or
by spraying the plate after development. In addition
to the reagents mentioned above, Suorescence intensi-
Rers such as triethanolamine, glycerol and Triton
X-100 are quite popular.
Identi\cation and Quanti\cation
The forte of TLC is qualitative analysis. It is possible
to identify unknown alkaloids owing to the large
amount of RF data available from the literature and
the ability to perform a chemical reaction using
a wide spectrum of different reagents in situ. For
some alkaloid drugs, a compilation of TLC data has
been elaborated and stored in computer-based in-
formation systems.
Many authors make an identiRcation based on
RF values in a number of chromatographic systems.
One scheme has been described in which the analysis
of a series of alkaloids by eight TLC systems, com-
bined with observations under UV light ("366 nm)
and colour reactions with iodoplatinate reagent
(Table 6) is used.
For precise identiRcation, UV or infrared spectra
after elution have become indispensable. Together
with the melting point and optical rotation values, they
are sufRcient for the identiRcation and comparison of
isolated pure substances. Other spectral methods such
as nuclear magnetic resonance or mass spectrometry
have frequently been used to identify alkaloids.
Although quantitative determination in TLC is
more difRcult and requires more effort, it is becoming
increasingly important nowadays. There exist two
forms of quantitative analysis: direct and indirect.
The Rrst method is based on the elution of spots with
a suitable solvent and determination in solution, fol-
lowed by spectrophotometric, Suorometric or
acid}base potentiometric titration. The second possi-
bility utilizes adsorption of UV and visible radiation
or luminescence of alkaloids, and is performed by the
means of photodensitometry, densitometry and
Suorimetry in situ. This latter technique requires the
use of an optical scanner, which is a relatively expen-
sive piece of equipment.
1973
See Colour Plate 54.
See also: II/Chromatography: Thin-Layer (Planar):
Layers; Modes of Development: Forced Flow, Overpres-
sured Layer Chromatography and Centrifugal; Spray Re-
agents. III/Alkaloids: Gas Chromatography; Impregna-
tion Techniques: Thin-Layer (Planar) Chromatography;
Liquid Chromatography.
Further Reading
Adamovics JA (1990) Chromatographic Analysis of Phar-
maceuticals. New York: Marcel Dekker.
Bieganowska ML and Petruczynik A (1994) Thin-layer
reversed-phase chromatography of some alkaloids
in ion-association systems. Chemia Analityczna 39:
139.
Camag Bibliography Service Thin-layer Chromatography.
Cumulative CD, Version 1.00. Camag 1997.
Deyl A, Macek K and Janak J (1975) Liquid Column
Chromatography. A Survey of Modern Techniques and
Applications. Amsterdam: Elsevier.
Golkiewicz W, Gadzikowska M, KuczynH ski J and Jusiak L
(1993) Micropreparative chromatography of some quat-
ernary alkaloids from the roots of Chelidonium majus L.
Journal of Planar Chromatography 6: 382.
Jork H, Funk W, Fischer W and Wimmer H (1990)
Thin-layer Chromatography, Reagents and Detection
Methods. Weinheim: VCH.
Lepri L, Desideri PG and Lepori M (1976) Chromato-
graphic behaviour of alkaloids of thin layer of cation
exchangers. Journal of Chromatography 123: 175.
Niederwiesser A and Pataki G (1972) Progress in Thin
Layer Chromatography and Related Methods. Michi-
gan: Ann Arbor Science.
Popl M, FaK hnrich J and Tatar V (1990) Chromatograpic
Analysis of Alkaloids. Chromatographic Science. New
York: Marcel Dekker.
RoK nsch H and Schreiber K (1967) Analytische und
praK parative DuK nnschichtchromatographische Trennung
von 5-gesaK ttigten BZW.
5-ungesaK ttigten Steroidal-
kaloiden und}sapogeninen an silbernitrat-haltigen Ad-
sorptionsschichten. Journal of Chromatography 30:
149.
Smith RM (1996) Supercritical Suid extraction of natural
products. LC/GC International, the Magazine of Separ-
ation Science, Vol. 9, p. 8. Chester, UK: Advanstar
Communications.
SoczewinH ski E and Flieger J (1996) Thin Layer Chromatog-
raphy of Alkaloids. Journal of Planar Chromatography.
9, 107.
Svendsen AB (1989) Thin layer chromatography of alkal-
oids. Journal of Planar Chromatography 2: 8.
Svendsen AB and Verpoorte R (1983) Chromatography of
Alkaloids. Amsterdam: Elsevier.
Touchstone JC (1992) Practice of Thin Layer Chromatog-
raphy. New York: John Wiley.
Wagner H and Bladt S (1996) Plant Drug Analysis. Thin
Layer Chromatography Atlas. Berlin: Springer.
1974 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY

ALLERGENS IN PERFUMES:
GAS CHROMATOGRAPHY^
MASS SPECTROMETRY
S. C. Rastogi, National Environmental deodorants, creams, lotions, shampoos and other per-
Research Institute, Rokilde, Denmark fumed consumer products which may contain
Copyright ^ 2000 Academic Press both natural as well as synthetic fragrance materials.
The method was later modiRed slightly so that quant-
itative analysis of many more fragrance substances in
Perfumes (fragrance substances) are used in the for- perfumes or in perfumed products could be per-
mulation of consumer products to provide pleasure to formed. This method, described in the present
the user and/or to mask malodours of some other article, has been applied to the analysis of perfumes
ingredients in the products. Perfumes are also used in in various consumer products. To demonstrate the
aromatherapy. A typical perfume may be composed potential of the method for perfume analysis,
of 10}300 substances selected from a battery of over example of analysis of fragrance substances in a de-
3000 synthetic and natural fragrance materials. It has odorant and in an eau de toilette are presented here.
been shown that approximately 2% of the general Sample preparation methods for the GC analysis of
population is allergic to perfumes. Furthermore, per- fragrances in various types of consumer products is
fumes have also been shown to be one of the major also described. The quantitative data on fragrance
cause of allergic contact dermatitis from the use of substances in various consumer products are reported
cosmetics and toiletries. Besides cosmetics, the use of in the publications described in the Further Reading
many other consumer products such as perfumed section.
laundry detergents and dishwashers have also been
implicated as the cause of perfume allergy in contact
Target Fragrance Substances
eczema patients.
Perfume allergy in contact eczema patients is The analytical method has been developed for the
diagnosed by patch-testing with a fragrance mix con- quantiRcation of 21 fragrance substances which in
taining 1% each of geraniol, eugenol, isoeugenol, relatively high concentrations are commonly used in
cinnamic alcohol, cinnamic aldehyde,
-amylcin- the composition of perfumes, or which are estab-
namic aldehyde, hydroxycitronellal and an extract lished contact allergens:
from oakmoss } oakmoss absolute. However, only
50}80% of perfume allergy cases are diagnosed by 1 geraniol: CAS registration number 106-24-1;
this test. For the management of allergy, it is impor- 2 eugenol: 97-53-0;
tant to identify the fragrance allergen responsible for 3 isoeugenol: 97-54-1;
contact eczema in a patient, as this makes it possible 4 linalool: 78-70-6;
for the patient to avoid the use of products containing 5 linalyl acetate: 115-95-7;
the sensitizing allergen(s). To establish the identity of 6 citronellol: 106-22-9;
the fragrance substance responsible for perfume al- 7 cinnamic alcohol: 104-54-1;
lergy in a contact eczema patient, it is recommended 8 cinnamic aldehyde: 104-55-2;
that the product(s) used by a patient should be ana- 9 hydroxycitronellal: 107-75-5;
lysed for the contents of fragrance allergens followed 10
-amylcinnamic aldehyde: 122-40-7;
by patch-testing the patient with the relevant fra- 11
-hexylcinnamic aldehyde: 101-86-0;
grance allergens present in the product. 12
-isomethylionone: 127-51-5;
Gas chromatography}mass spectrometry (GC-MS) 13 coumarin: 91-64-5;
is frequently used for the analysis of fragrance sub- 14 piperonal: 120-50-7;
stances in essential oils. This approach is used for the 15 benzyl alcohol: 100-51-6;
identiRcation and semiquantitative determination of 16 benzyl acetate: 140-11-4;
the fragrance substances of interest in essential oils. 17 benzyl benzoate: 121-51-4;
In 1995, GC-MS was used for the identiRcation 18 benzyl salicylate: 118-51-8;
and quantiRcation of 10 selected fragrance substances 19 Lilial威: 80-54-6;
including the seven chemically deRned substances 20 Lyral威: 31906-04-4;
of fragrance mix in perfumes, eau de toilette, 21 Hedione威: 24851-98-7.
 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY
Approximately 1.0% (w/v) solutions of all of the
substances in ethanol served as stock solutions. The
stock solutions were stored in closed vials at 43C and
were used within 1 month.
Sample Preparation
Perfumes, Eau de Toilette, Aftershave and
Deodorant Sprays
These products were approximately diluted in
ethanol so that the concentrations of target fragrance
substances were 40.1%. Depending on the concen-
trations of the target fragrance substances in
a sample, it may be necessary to analyse several dilu-
tions of the sample.
Shampoos, Creams, Lotions, Lipsticks,
Face Powders and Deodorant Sticks
Perfumes from 1 g sample were extracted in 10 ml
methanol at 603C (to facilitate the extraction) fol-
lowed by removal of matrix components by silica gel
column chromatography. The extract was loaded on
a 7;1.8 cm silica gel column, and the fragrance
fraction was eluted with methanol. The perfume ex-
tract was stored at 43C and analysed within 24 h.
Soap Bar and Laundry Detergents
Perfumes from 1 g sample dissolved in 50 ml water
were extracted in 10 ml ethyl acetate employing
liquid}liquid extraction. The perfume extract in ethyl
acetate was centrifuged to remove any solid or aque-
ous contamination. The perfume extract was stored
at 43C and analysed within 24 h.
Dishwashing liquid
The method used for the extraction of perfumes from
dishwashing liquids was the same as for shampoo.
GC-MS Analysis
MS Conditions
Electron impact ionization at 70 eV was used, scann-
ing m/z 29}250 in 0.6 min.
Results
The method described here has been applied to the
determination of 21 target fragrance substances in
consumer products. The chromatographic separation
of these 21 compounds employing GC is shown in
Figure 1. Day-to-day variation of retention times of
the fragrance substances is (0.5%. The detection
limits of all of the target substances are 41 p.p.m.,
1975
the calibration curves for all of the target substances
are linear (coefRcient of correlation '0.995) in
the tested concentration range 10}2000 p.p.m.,
the relative standard deviations of the determina-
tion of all of the substances are (11%. The recov-
ery of all of the target substances from the spiked
samples is 82}116%, and day-to-day variations of
quantitative analysis for all of the substances are
within 5%.
The reconstructed ion chromatogram obtained by
GC-MS analysis of fragrance substances in a deodor-
ant (undiluted) is shown in Figure 2. The fragrance
substances in the product were identiRed by compar-
ing the retention times of the GC peaks with those of
the reference materials, as well as by comparing the
spectra of the GC peaks with the reference spectra of
standard compounds in the mass spectrum library.
Followed by GC-MS identiRcation, quantiRcation of
target fragrance substances in the sample is carried
out with external standards.
Most consumer products contain many more
fragrance substances other than the target com-
pounds. The identiRcation of these substances was
only performed by comparing the mass spectrum of
a GC peak with the mass spectra of reference com-
pounds in the MS library. In this case, both the
spectrum Rt and spectrum purity of match of the
unknown spectrum with those of library spectra were
'900. An example of identiRcation of fragrance
substances in an eau de toilette is shown in Fig-
ure 3A}E, where the results are divided in six win-
dows for the clarity of peak identiRcation. ConRrma-
tion of the identiRcation of these substances and their
quantiRcation were performed where a reference
material was available.
In some cases it is not possible to identify all the
peaks because of the absence of mass spectra of the
compounds in the mass spectral library.
Discussion
For the analysis of perfumes on a routine basis,
GC-MS identiRcation of the fragrance substances
followed by quantiRcation employing GC-Same ion-
ization detection (FID) was found to be a more suit-
able approach. The main reason for this is that the use
of GC-FID allows relatively rapid production of vali-
dated data. Thus, several relevant analysis recom-
mended by quality assurance/quality control
(QA/QC) protocol for a set of samples can be easily
performed by GC-FID. FulRlling the requirements of
QA/QC protocol for the analysis by GC-MS is time-
consuming, because it requires tuning and calibration
of the MS at regular intervals and frequent cleaning
of the ion source. The detection limits of the target
1976 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY

Figure 1 GC-MS analysis of a mixture containing 83}117 p.p.m. of the 21 target fragrance substances. 50 m;0.32 mm, 1.2
m film
thickness Chrompack fused silica capillary columns coated with CP-Sil-5CB, were used. 1
L split injection; helium carrier gas flow
30 ml min\1, column-head pressure 20 psi; injector temperature 3003C; column temperature program: 40}1403C in 4 min, thereafter
53C min\1 to 2803C, 5 min at 2803C. 2
L injection volume was used when the content of perfume in a sample was relatively low.
 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY 1977

Figure 2 GC-MS analysis of the target fragrance in an undiluted deodorant. The following were present among the target fragrance
substances: 102 p.p.m. benzyl alcohol, 1028 p.p.m. linalool, 141 p.p.m. citronellol, 136 p.p.m. geraniol, 614 p.p.m. linalyl acetate,
205 p.p.m. hydroxycitronellal, 183 p.p.m. cinnamic alcohol, 408 p.p.m. eugenol, 1051 p.p.m. coumarin, 7 p.p.m. isoeugenol, 319 p.p.m.

-isomethylionone, 291 p.p.m. Lilial威, 199 p.p.m. Hedion威, 68 p.p.m.
-amylcinnamic aldehyde, 101 p.p.m. benzyl benzoate and
112 p.p.m. benzyl salicylate. Quantification of Hedion威 was performed by the analysis of 1 : 10 dilution of the sample, where no
interference by the sesquiterpene alcohol present in the sample was observed.
1978 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY

Figure 3 GC-MS analysis of an eau de toilette, diluted 1 : 10 in ethanol. The reconstituted ion chromatogram is divided in six windows
(A}F) for the clarity of the compounds identified in the sample. Peaks with no name could not be identified.
 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY 1979

Figure 3 Continued
1980 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY

Figure 3 Continued
 III / ALLERGENS IN PERFUMES: GAS CHROMATOGRAPHY^MASS SPECTROMETRY
substances by GC-FID, however, are 2}5 p.p.m. So,
unless the quantiRcation was required at 1 p.p.m.
level, GC-FID was chosen for the determination of
fragrance substances after prior identiRcation by
GC-MS.
Most of the fragrance substances in use, includ-
ing the target fragrance substances, have a molecu-
lar weight (250 Da. Therefore, the MS scan was
performed only up to m/z 250. Occasionally, for
example in the identiRcation of musk ketone, it is
necessary to scan masses up to 300.
Not all the fragrance ingredients in all tested prod-
ucts could be identiRed or quantiRed, in some cases
due to interferences. Occasionally the GC peak of
a relatively high amount of dipropylene glycol present
in a sample overlapped the peak by benzyl alcohol;
a C11-alkyne interfered with the analysis of Lilial威;
high amounts of triethyl citrate and/or a sesquiter-
pene alcohol (C15H26O) interfered with the analysis of
Hedione威 and relatively high amounts of Hedione威
interfered with the analysis of
-amylcinnamic al-
dehyde. An unidentiRed compound was found to in-
terfere with the analysis is benzyl salicylate. In most
cases these problems could be solved by analysing
diluted samples.
By using GC-MS, identiRcation of 226 substances
in deodorants has recently been reported. A struc-
ture}activity relationship (SAR) analysis of contact
allergens revealed that 84 of the identiRed com-
pounds possess at least one structural alert (chemical
group) having sensitizing potential, and 70 belong to,
or are susceptible to metabolize into, the chemical
groups having sensitizing properties: aldehydes,
ketones and
,-unsaturated aldehydes, ketones or
esters. The combination of GC-MS and SAR analysis
could be helpful in the selection of substances for
supplementary investigations regarding sensitizing
properties.
Analysis of as many fragrance ingredients as
possible in a perfumed product is of great import-
ance for clinicians to establish the identity of
contact allergens in each case. This information
is also important for clinical research to investi-
gate cross-reactions of fragrance allergens. The
quantitative data on the fragrance ingredients in
consumer products make a basis for exposure as-
sessment that is a help for establishing threshold
concentrations of fragrances for the elicitation of
contact allergy.
Conclusions
Chemical analysis of perfumes and perfumed prod-
ucts is of great importance for the diagnosis and
management of perfume allergy. The GC-MS/GC-
1981
FID method described here for the analysis of fra-
grance substances in consumer products has proved
to be valuable to identify allergens in patients with
contact eczema from the use of perfumes and per-
fumed products. Using GC-MS in combination with
SAR it has been possible to identify several fragrance
substances in perfumes which possess sensitizing po-
tential.
See also: II/Chromatography: Gas: Column Techno-
logy; Derivatization; Detectors: General (Flame Ionization
Detectors and Thermal Conductivity Detectors); Detectors:
Mass Spectrometry; Detectors: Selective; Headspace
Gas Chromatography; Theory of Gas Chromatography.
III/Flavours: Gas Chromatography: Sulphur Com-
pounds: Gas Chromatography.
Further Reading
Calkin RR and Jellinek JS (1994) Perfumery Practice and
Principles. New York: Wiley.
De Groot AC and Frosch P (1997) Adverse reactions to
fragrance. A clinical review. Contact Dermatitis 36:
57}86.
Frosch PJ, Pliz B, Andersen KE et al. (1995) Patch testing
with fragrances: results of a multicenter study of the
European Environmental and Contact Dermatitis Re-
search Group with 48 frequently used constituents of
perfumes. Contact Dermatitis 33: 333}342.
Frosch PJ, Johansen JD and White IR (eds) (1998) Fragran-
ces: BeneTcial and Adverse Effects. Berlin: Springer-
Verlag.
Johansen JD, Rastogi SC and MenneH T (1996) Contact
allergy to popular perfumes; assessed by patch test, use
test and chemical analysis. British Journal of Dermatol-
ogy 135: 419}422.
Larsen W, Nakayama H, Lindberg M et al. (1996) Fra-
grance contact dermatitis: a worldwide multicenter
investigation (part I). American Journal of Contact Der-
matitis 7: 77}83.
Pybus DH and Sell CS (eds) (1999) The Chem-
istry of Fragrances. Cambridge: Royal Society of
Chemistry.
Rastogi SC (1995) Analysis of fragrances in cosmetics by
gas chromatography-mass spectrometry. Journal of
High Resolution Chromatography 18: 653}658.
Rastogi SC, Johansen JD and MenneH T (1996) Natural
ingredients based cosmetics: content of selected fra-
grance sensitizers. Contact Dermatitis 34: 423}426.
Rastogi SC, Johansen JD, Frosch P et al. (1998) Deodor-
ants on the European market: quantitative chem-
ical analysis of 21 fragrances. Contact Dermatitis 38:
29}35.
Rastogi SC, Leppoitevin JP, Johansen JD et al. (1998)
Fragrances and other materials in deodorants: search for
potentially sensitizing molecules using combined GC-
MS and structure activity relationship (SAR) analysis.
Contact Dermatitis 39: 293}303.
1982 III / AMINES: GAS CHROMATOGRAPHY
AMINES: GAS CHROMATOGRAPHY
H. Kataoka, S. Yamamoto and
S. Narimatsu, Okayama University, Tsushima,
Okayama, Japan
Copyright ^ 2000 Academic Press
Aliphatic and aromatic mono-, di- and polyamines
are naturally occurring compounds formed as meta-
bolic products in microorganisms, plants and ani-
mals, in which the principal routes of amine forma-
tion include the decarboxylation of amino acids,
amination of carbonyl compounds and degradation
of nitrogen-containing compounds. Accordingly,
amines are important indicators of a wide variety
of biochemical, clinical, toxicological and fermenta-
tion processes. Amines are also widely used as
raw materials or as intermediates in the manufacture
of industrial chemicals, e.g. pesticides, medicines,
dyestuffs, rubbers, polymers, surfactants, cosmetics
and corrosion inhibitors. Many of them are dis-
charged into the atmosphere and water from anthro-
pogenic sources such as foods, cattle feeds, livestock
buildings, waste incineration, sewage treatment,
automobile exhaust, cigarette smoke and various
industries. Furthermore, many amines have an
unpleasant smell and are hazardous to health as
sensitizers and irritants to the skin, eye, mucous
membranes and respiratory tract. Some amines are
also suspected to be allergenic, mutagenic or carcino-
genic substances due to their adsorption in living
tissue. Amines are not only toxic of themselves but
can also become toxic N-nitrosamines through chem-
ical reactions with nitrosating agents such as nitrite
or nitrate.
Gas chromatography (GC) has been widely used
for amine analysis because of its inherent advantages
of simplicity, high resolving power, high sensitivity,
short analysis time and low cost. In addition, a wide
variety of detectors can be used: nitrogen}phos-
phorus (NPD), electrolytic conductivity (ELCD) and
chemiluminescent (CLD) detectors offer increased
selectivity for speciRc amines. Furthermore, the com-
bined technique of GC-mass spectrometry (MS) can
provide structural information for the unequivocal
identiRcation of amines. Sub-nanogram detection
limits can be achieved using these detectors.
However, GC separation of free amines at very low
concentrations generally has inherent problems
related to the difRculty in handling low molecular
mass amines because of their high water solubility,
high volatility and ready oxidation under chromato-
graphic conditions. Furthermore, amines tend to be
strongly adsorbed and decomposed on the columns
and give tailing peaks, ghosting phenomena and low
detector response. The adsorption tendency in the
analytical system, i.e. in sample vessels, injector, glass
wool and GC column, is in the order primary'
secondary'tertiary amines, and tailing becomes in-
creasingly severe as the basicity of the amines in-
creases. In addition, it is generally more difRcult to
chromatograph aliphatic than aromatic amines.
A common method of overcoming these problems
is to convert such polar compounds to relatively non-
polar derivatives more suitable for GC analysis.
A number of derivatives such as acyl, silyl, dinitro-
phenyl, permethyl, Schiff base, carbamate, sulfon-
amide and phosphonamide compounds have been
used for this purpose.
Another successful approach has been to employ
less reactive column packing materials to reduce the
interaction with solutes, for example, the use of por-
ous polymers and the deactivation of supports by
treatment with alkali. Wall-coated (WCOT), sup-
port-coated (SCOT) and porous-layer (PLOT) open
tubular capillary columns, which minimize col-
umn}solute interactions, have also been used for this
purpose. Free amines can be analysed after addition
of alkali, either by direct injection or by headspace
sampling, or they can be extracted into an organic
solvent before analysis. Direct or headspace analysis
of samples minimizes sample preparation, thereby
reducing the possibility of contamination. Solid-
phase microextraction (SPME), with integrated
sampling, extraction, concentration and sample in-
troduction in a single step, has recently been used for
amine analysis by coupling with GC.
This article is concerned with the general aspects
of direct GC separation of underivatized aliphatic
and aromatic amines, and various characteristics with
respect to columns are considered in more detail
below.
Column Development
Packed-column GC is generally simpler to set up than
capillary GC, because of the ability to apply the
stationary phase easily to the solid support and
modify it appropriately to the particular analysis
 III / AMINES: GAS CHROMATOGRAPHY 1983

required. Deactivation of the glass surface can be
effected using a suitable silylating reagent to limit the
effect of adsorption on the wall of the column. How-
ever, the general difRculty in the chromatography lies
in absorptivity on the solid support leading to tailing.
The adsorption of amines by the support material has
been attributed to the presence of free silanol groups
on the silica surface participating in hydrogen bond-
ing with the free electron pair of the nitrogen atom of
the amine. Simple treatment with KOH reduces the
adsorption to a minimum, allowing good peak shape
and optimum performance.
Glass and fused silica capillary columns have also
been used for the analysis of free amines. The in-
herent strength and Sexibility of fused silica make it
easier to use and less fragile than glass capillary col-
umns. Furthermore, fused silica provides a more inert
surface for improved performance and less adsorp-
tion. The analysis of free amines on packed columns
has now largely been replaced by analysis on fused
silica capillary columns that are commercially avail-
able with a range of stationary phases. The packed
and capillary GC columns reported in the past 30
years for amine analysis are summarized in Tables 1
and 2.
Packed Columns
Three types of packing can be used to separate
amines: graphitized carbon coated with a stationary
phase and deactivated, coated and uncoated porous
polymers, and conventional columns packed with
a deactivated diatomaceous earth coated with a sta-
tionary phase. The columns are usually deactivated
with KOH, trimethylchlorosilane (TMCS) or ammo-
nia in the carrier gas. Carbopack graphitized carbon
and porous polymer packings are well suited for sep-
arating C1}C10 compounds, but retention times for
larger molecules are excessive; deactivated and
coated conventional packings are better suited to the
analysis of higher molecular weight amines.
Graphitized carbon packings Graphitized carbon
packings are generally used for free amine analysis
after coating with a stationary phase. Sterling FT-G
and Vulcan, sold by Supelco as Carbopack A and
Carbopack B, respectively, have been used for the
analysis of C1}C16 aliphatic amines with suitable
amounts of KOH and polyethylene glycol (PEG), e.g.
PEG 20M and PEG-1500. A 4.8% PEG 20M/0.3%
KOH on Carbopack B column is recommended for
the analysis of C1}C4 aliphatic amines in aqueous
solution at nanogram level. This column offers com-
plete separation of the C2}C3 amine isomers and is
less affected by water than the other packed columns.
However, the preparation of the column seems to be
difRcult for routine analysis. A 1.5% UCON 50-HB-
2000 on Carbopack B packing deactivated with 0.8%
KOH has also been used to separate a mixture of
aliphatic, aromatic and cyclic amines, and rapid sep-
aration of nine amines without ghosting was obtained
by temperature programming and treatment of glass
wool in the column ends with dimethylchlorosilane
(DMCS). On the other hand, 4% Carbowax 20M on
0.8% KOH-deactivated Carbopack B packing has

Table 1 Packed columns for analysis of free amines

Column packing Type Length (m) Amine Detection

1.3% PEG 20M/0.3% KOH on Sterling FT-G GC 2.0 AL FID

0.5% PEG-1500/0.2% KOH on Sterling FT-G GC 1.4 AL FID
4% PEG 20M/0.8% KOH on Vulcan GC 1.4 AL FID
4.8% PEG 20M/0.3% KOH on Carbopack B GC 1.8 AL FID
1.5% UCON 50-HP/0.8% KOH on Carbopack B GC 1.83 AL, AR FID
4% Carbowax 20M/0.8% KOH on Carbopack B GC 1.7}3.75 AL FID, NPD
Tenax GC PP 1.52 AL FID
Chromosorb 103 PP 1.5}3.3 AL FID, NPD
Chromosorb 102/5% TMCS/5% KOH PP 2.0 AL NPD
5% Squalene/2% KOH on Chromosorb 103 or 104 PP 3.0 AL CLD
4% Carbowax 20M/1% KOH on Corning glass PC 1.8 PO FID
10% Carbowax 20M/2% KOH on Chromosorb W AW PC 1.5}1.9 AL FID, NPD
5% PEG-1000/0.5% Na3PO4 on Chromosorb G PC 2.0 AL SID
5% PEG-HT/1% KOH on Umiport HP PC 2.0 AR FID
3% SP-2250 on Supelcoport PC 1.83 AR NPD
5% SP-2401-DB on Supelcoport PC 1.83 AR NPD
1.5% SP-2250/1.95% SP-2401 on Supelcoport PC 1.83 AR NPD
3% Silar 5CP on Supelcoport PC 1.83 AR NPD

GC, Graphitized carbon; PP, porous polymer; PC, partition column; AL, aliphatic amine; AR, aromatic amine; PO, polyamine; FID,
flame ionization detection; NPD, nitrogen}phosphorus detection; surface ionization detection; CLD, chemiluminescence detection.
1984 III / AMINES: GAS CHROMATOGRAPHY

Table 2 Capillary columns for analysis of free amines

Column Amine Detection

Stationary phase Type Length (m)

10% PEG 400 WCOT/G 99 AR FID

5% PEG 400/2% KOH WCOT/G 40 AR FID
Supelcowax 10 WCOT/G 10 AL, AR FID
SP-2250 WCOT/G 30 AR NPD
SP-2100 WCOT/G 30 AR NPD
Carbowax 20M WCOT/G,F 25}37 AR FID, NPD, MSD
SE-54 WCOT/G,F 30 AR, DR FID, NPD
SE-52 WCOT/G,F 30 AR NPD, ELCD, PID
SE-30 WCOT/G,F 30 AR FID, NPD
CAM WCOT/F 30 AL FID
HP-20M WCOT/F 25 AR FID
Carbowax Amine WCOT/F 30 AL, AR FID
PoraPLOT Amines PLOT/F 25 AL FID, ELCD
CP-Sil-19CB WCOT/F 10 AL, AR FID
DB-35ms WCOT/F 25 AR FID
DB-5ms WCOT/F 30 AL, AR FID, MSD
HP-5 WCOT/F 25}30 DR FID, MSD
HP-101 WCOT/F 25 AL, AR FID
HP-1 WCOT/F 10}30 AL, AR FID, NPD
DB-1 WCOT/F 30 AL, DR FID, MSD
OV-1 WCOT/F 25 DR MSD
SBP-1 WCOT/F 30 AL, AR FID
CBJ-17 WCOT/F 30 DR NPD

WCOT, Wall-coated; PLOT, porous layer; G, glass; F, fused silica; AL, aliphatic amine; AR, aromatic amine; DR, basic drug; FID, flame
ionization detection; NPD, nitrogen}phosphorus detection; ELCD, electrolytic conductivity detection; MSD, mass selective detection;
PID, photoionization detection.

been speciRcally developed for monitoring low mo-
lecular weight aliphatic amines at p.p.m. levels in
water. Heterocyclic amines can also be separated on
this packing, but aromatic amines exhibit excessively
long retention times. By using this column and an
NPD, low molecular weight amines in sea water were
determined. In order to reduce the appearance of
ghost peaks, 15% ammonia solution was injected on
to the hot (150}2003C) column after each sample
run. As shown in Figure 1, 23 amines were separated
within 25 min (Figure 1A), and 12 amines were selec-
tively detected in sea water (Figure 1B).
A general characteristic of Carbopack-based col-
umns is that sample components are separated by
carbon number and are eluted in the order C1P
C2PC3, and so forth. This is seen in the separation of
methylamine, dimethylamine, trimethylamine and
ethylamine. Both C2 amines (dimethylamine and
ethylamine) are eluted before the C3 amine (tri-
methylamine). On these packings, it is necessary to
use small samples to prevent tailing due to overload-
ing. Furthermore, the column must be conditioned by
injecting a number of relatively large amounts of
water when analysing amines in aqueous solution.
This treatment converts any K2CO3 in the column to
KOH, making the column more basic and improving
its inertness for amines. Acidic compounds in the
sample are irreversibly adsorbed by the KOH. In
addition, a certain amount of stationary phase is
hydrolytically decomposed and appears as a water
peak in the chromatogram when water passes
through the column. Therefore, conditioning is
needed to clean the column and minimize the water
peak when standards and sample are subsequently
injected. These packed columns should not be ex-
posed to air, since the packing will adsorb carbon
dioxide and lose its deactivation. Furthermore, the
water used should be distilled or deionized, and fresh-
ly boiled to remove CO2.
Porous polymer packings Porous polymers possess-
ing large surface area are often used as column pack-
ing in GC without coating with a stationary phase.
Tenax GC and Chromosorb 103 were speciRcally
developed to separate low molecular weight aliphatic
amines. Although Chromosorb 103 proved inconsist-
ent and difRcult to handle, and tended to expand on
heating, leaving gaps in the column upon cooling,
these effects could be minimized by paying scrupu-
lous attention to packing. By using Chromosorb 103,
 III / AMINES: GAS CHROMATOGRAPHY 1985

Figure 1 (A) Amine standards and (B) a sea water sample. GC conditions: packed column, 4% Carbowax 20 M and 0.8% KOH on
Carbopack B (2 m;2.5 mm i.d. glass); column temperature, initially hold at 853C for 2.5 min, increase to 1503C at 323C min\1 for 6 min
and then to 2203C at 103C min\1; injector and detector temperatures, 200 and 2203C, respectively; He carrier gas flow rate,
22 mL min\1; detector, NPD. Peaks: 1, ammonia; 2, monomethylamine; 3, dimethylamine; 4, ethylamine; 5, trimethylamine; 6,
2-propylamine; 7, 1-propylamine; 8, tert-butylamine; 9, diethylamine; 10, sec-butylamine; 11, 2-butylamine; 12, pyrrolidine; 13,
1-butylamine; 14, piperidine; 15, triethylamine; 16, pyridine; 17, 2-amylamine; 18, 1-amylamine; 19, pyrrole; 20, dipropylamine; 21,
cyclohexylamine; 22, tripropylamine; 23, dibutylamine. (Reproduced with permission from Yang et al. (1993) Analytical Chemistry
65: 572.)

11 aliphatic amines were isothermally separated
without the ghosting observed with alkali-washed
support packings. The use of longer columns resulted
in increased analysis time that could not be reduced
with a higher Rnal temperature owing to excessive
column bleed. Amines tail on other porous polymers,
but performance can be improved by coating them
with a stationary phase and TMCS. By using 5%
squalene/2% KOH on Chromosorb 103 or 104 and
GC-CLD, low molecular weight aliphatic amines
have been determined. A column packed with Chro-
mosorb 102 treated with 5% TMCS and coated with
KOH has been used to determine methylamines in
biological materials at low concentration by head-
space GC-NPD. Use of headspace sampling avoids
the possibility of interference from other water-
soluble biological substances.
Partition columns A partition column consists of
a support, generally a diatomaceous material, coated
with a stationary phase. However, the support tends
to interact with active analytes, such as amines, caus-
ing the peaks to tail, unless it is made strongly basic
by adding KOH or an amine or by using an amine as
the stationary phase. This alkaline deactivation of
diatomaceous supports appears to be more effective
than silanization for the analysis of amines. The ma-
jor disadvantage of alkali-washed packings lies in the
thermal instability of the liquid phases that prevents
temperature-programmed analysis. Generally, Chro-
mosorb W (white and light weight) and Chromosorb
P (pink) supports are used as support materials; poly-
tetraSuoroethylene supports are widely regarded as
very inert, but they do not appear to be especially
inert to amines. The stationary phase must be com-
patible with the basic material. Polyglycols, such as
Carbowax and certain hydrocarbons, have been used
successfully with basic materials.
Although a 10% Carbowax 20M/2% KOH on
Chromosorb W AW packing was used for separating
aliphatic mono- and diamines, many of the higher
boiling anilines apparently did not elute from the
column. Other PEG packings, such as PEG-1000 on
Chromosorb G and PEG-HT on Uniport HP, have
1986 III / AMINES: GAS CHROMATOGRAPHY

also been used for the analysis of aliphatic and aro-

matic amines. Di- and polyamines in tissue samples
were analysed using Corning glass beads coated with
4% Carbowax and 1% KOH. This column gave
a good separation and a nearly complete recovery of
these amines.
Aromatic amines are generally less basic than
aliphatic amines and consequently present less of an
adsorption problem. Although they can be separated
on the highly basic columns described above, the
analysis is generally carried out with a silicone sta-
tionary phase on an acid-washed, dimethyl-
chlorosilane (AW-DMCS) treated support. The
SP-2250 and SP-2401-DB packings were specially
developed for separating amines at low concentra-
tions. When analyte concentrations are very low,
even AW-DMCS treatment is inadequate, and it is
necessary to use a specially deactivated column or
derivatize the analytes. The 3% SP-2250 column par-
tially resolved all the anilines, but some peak tailing
was evident, especially for aniline. The 5% SP-2401-
DB (containing KOH) and 1.5% SP-2250/1.95% SP-
2401 gave no improvement in peak shape, in spite of
the greater polarity, compared to SP-2250. In addi-
tion, several compounds were not resolved on these
columns. The Silar 5CP column partially resolved all
the anilines except for 3- and 4-chloroanilines, and
gave good peak shape for all the compounds. As
shown in Figure 2, the best separation of 19 anilines
was obtained using a 3% SP-2250 on Supelcoport.
Capillary Columns
Capillary columns offer a signiRcant improvement in
separation, in comparison to conventional packed
columns, and have been used for the separation of
complex mixtures and components closely related
chemically and physiologically. As shown in Table 2,
various glass and fused silica capillary columns have
been used for free amine analysis. Fused silica capil-
lary columns provide strength, Sexibility and a more
inert surface for improved performance and less ad-
sorption. Cross-linked or bonded-phase columns can
be washed with solvents, prolonging their lifetime.
Furthermore, the advance of commercially available
cross-linked and bonded-phase capillary columns and
precise temperature-controlled GC ovens has meant
that the retention times are extremely reproducible.
This is critical when using automated date-handling
equipment for identiRcation and quantiRcation. Typi-
cally, 10}30 m long columns, coated with either non-
polar or polar stationary phases, have been used for
amine analysis. Many of the phases used today are
speciRcally manufactured by the column supplier,
and give excellent performance, low bleed and high
efRciency. However, there is the drawback that a col-
Figure 2 Aniline compounds on packed column. GC condi-
tions: column, 3% SP-2250 on Supelcoport (1.83 m;2 mm i.d.
glass); column temperature, 43C min\1 from 80 to 2303C; injector
and detector temperatures, 250 and 3003C, respectively; He
carrier gas flow rate, 30 mL min\1; detector, NPD. Peaks: 1,
aniline; 2, 2-chloroaniline; 3, 3-chloroaniline; 4,4-chloroaniline; 5,
4-bromoaniline; 6, 3,4-dichloroaniline; 7, 2,4,6-trichloroaniline; 8,
2-nitroaniline; 9, 2,4,5-trichloroaniline; 10, 3-nitroaniline; 11, 4-
chloro-2-nitroaniline; 12, 4-nitroaniline; 13, 2,6-dichloro-2-
nitroaniline; 14, 2-chloro-4-nitroaniline; 15, 2-bromo-6-chloro-4-
nitroaniline; 16, 2,6-dibromo-4-nitroaniline; 17, 2-chloro-4,6-di-
nitroaniline; 18, 2,4-dinitroaniline; 19, 2-bromo-4,6-dinitroaniline.
(Reproduced with permission from Riggin et al. (1983) Analytical
Chemistry 55: 1862.)
umn from one supplier may not give the same separ-
ation as the nominally equivalent column from an-
other supplier.
Glass capillary columns In early work, glass capil-
laries were employed for the separation of aromatic
amines using alkaline PEG as the stationary phase.
Although a disadvantage of this phase is its tendency
to deteriorate at temperatures slightly above 2003C, it
has been used for the separation of methylanilines
and methylpyridines in coal-tar light oil. Carbowax
20M columns have been used for the determination
of airborne aromatic amines with an NPD. The neces-
sary inertness of glass capillary columns may be
achieved by deactivation with octamethylcyclo-
tetrasiloxane (OMCTS). The glass or fused silica
columns were silanized using OMCTS and triSuoro-
propyl(methyl)cyclosiloxane, and coated with vari-
ous phases (SE-30, SE-52, SE-54). Test mixtures
containing about 1 ng of such difRcult substances as
primary mono- and diaminoalkanes gave symmetri-
cal peaks on some of these phases. As shown in
 III / AMINES: GAS CHROMATOGRAPHY 1987

Figure 3 Drug standard mixtures on (A) AR glass and (B) fused silica capillary columns coated with SE-54 and with flame ionization
detector. Temperature programmes are shown within the figure. Peaks: 1, amphetamine; 2, phentermine; 3, propylhexedrine; 4,
methamphetamine; 5, ethylamphetamine; 6, propylamphetamine; 7, ephedrine; 8, phenmetrazine; 9, phendimetrazine; 10, amfep-
ramone; 11, benzocaine; 12, phenacetin; 13, methyl phenidate; 14, procaine; 15, methaqualone; 16, cocaine; 17, codeine; 18,
ethylmorphine; 19, morphine. (Reproduced with permission from Blomberg et al. Journal of Chromatography 239 (1982) 51).

Figure 3, the separation of some underivatized drugs
is equally good on alkali-resistant (AR) glass and
fused silica capillaries, although alkali-resistant (AR)
glass has a basic character that can be reduced by
careful leaching.
On the other hand, interesting results dealing with
the separation of free amines and other nitrogen com-
pounds were reported in glass capillary columns with
stationary phases polymerized in situ.
Fused silica capillary columns For the analysis of
amines, capillary columns with a nonimmobilized
PEG-type stationary phase have been specially pre-
pared and are commercially available. For the analy-
sis of volatile amines, aromatic and heterocyclic
amines and other amino substances, CAM, CP-Wax,
HP-20M, Carbowax 20M and Carbowax Amine
capillary columns have been recommended. These
columns are generally deactivated with KOH to elute
basic compounds with good peak shapes and re-
sponses. Three types of fused silica capillary columns,
Supelcowax 10 (PEG), CP-Sil-19CB (methylphenyl-
cyanopropylsilicone) and HP-1 (methylsilicone) have
also been used for the separation of aliphatic and
aromatic amines. Ammonia as a carrier gas can dras-
tically affect the retention factors and improve the
peak symmetry for aliphatic amines. A porous poly-
mer fused silica capillary column, PoraPLOT Amine,
has been used to separate very volatile amines. By
using this column and ELCD, C1}C6 amines in aque-
ous and methanolic solution were analysed. The sep-
aration of aniline and its halogen and nitrogen deriva-
tives in waste water were evaluated using several glass
and fused silica capillary columns of polysiloxane
type (SE-30, SE-52, SE-54, SP-2100) and NPD. Each
of the capillary columns gave excellent peak shape for
all the anilines, but failed to resolve at least one
compound pair (e.g. the SE-30 completely resolved
3- and 4-chloroaniline that co-eluted on SE-54, but
failed to resolve the 2,6-dibromo-4-nitroaniline and
2,4-dinitroaniline which were completely resolved on
SE-54). Figure 4 shows a chromatogram of an aniline
mixture on an SE-54 fused silica column. The NPD
sensitivities for many anilines are substantially better
with the SE-54 capillary column (Figure 4) than with
the 3% SP-2250 packed column (Figure 2), primarily
because less peak tailing is observed at low concentra-
tion. Interestingly, the fused silica and glass capillary
SE-54 columns gave different elution patterns for
the various anilines. Using both SE-54 and SE-30
1988 III / AMINES: GAS CHROMATOGRAPHY

Figure 4 Aniline compounds on fused silica capillary column. GC column, SE-54 (30 m;0.25 mm i.d.); He carrier gas flow rate,
30 cm s\1; split ratio, 10 : 1. Other conditions and peak numbers are the same as Figure 2. (Reproduced with permission from Yang et
al. (1993) Analytical Chemistry 65: 572.)

Figure 5 Aromatic amines on fused silica capillary column. GC conditions: column, DB-35ms (25 m;0.20 mm i.d. glass); column
temperature, initially hold at 503C for 2 min, increase to 3403C at 203C min\1 and then hold at 3403C for 10 min; injector and detector
temperatures, 280 and 3203C, respectively; He carrier gas flow rate, 35 cm\1; splitless injection; detector, NPD. Peaks: 1, o -toluidine;
2, 4-chloroaniline; 3, 2-methoxy-5-methylaniline; 4, 2,4,5-trimethylaniline; 5, 4-chloro-2-methylaniline; 6, 2,4-diaminotoluene; 7, 2,4-
diaminoanisole; 8, 2-aminonaphthalene; 9, 2-methyl-5-nitroaniline; 10, 4,4
-oxydianiline; 11, 4,4
-methylenedianiline; 12, benzidine; 13,
2-aminoazotoluene; 14, o -tolidine; 15, 4,4
-thiodianiline; 16, 3,3
-dimethoxybenzidine; 17, 3,3
-dichlorobenzidine. (Reproduced with
permission from Catalog and Technical Reference, C407, J & W Scientific, California.)

Figure 6 Chromatograms obtained from hair samples. (A) Normal hair; (B) normal hair with standard amphetamines added;
(C) abuser’s hair. GC conditions: column, CBJ-17 (30 m;0.53 mm i.d. fused-silica, Shimadzu); column temperature, initially hold at
1003C for 5 min, increase to 2203C at 103C min\1 and then hold at 2203C for 3 min; injector and detector temperatures, 2203C; He
carrier gas flow rate, 4 mL min\1; split ratio, 2 : 1; detector, NPD. Peaks: 1, -phenethylamine (internal standard); 2, amphetamine; 3,
methamphetamine; 4, N-propyl--phenethylamine (internal standard).
fused silica capillary columns, all 19 anilines can be
resolved.
A polysiloxane capillary column specially designed
for the analysis of basic compounds using new deacti-
vation technologies has been developed. This propri-
etary deactivation provides both the inertness (basic-
ity) and surface energies required to coat a 5%
diphenyl/95% dimethylpolysiloxane stationary phase
successfully. Using this column, C3}C10 primary
amines can be separated as symmetrical peaks. This
column allows lower limits of detection for basic
compounds such as substituted anilines and ben-
zidines. Since the column is virtually identical in po-
larity to the widely used ordinary columns with the
same stationary phase, it can be directly substituted
and run under the same temperature conditions. DB-
5ms and DB-35ms columns certiRed for use with MS
have been developed for the analysis of aliphatic and
aromatic amines. These columns have very low bleed
characteristics and excellent inertness. As shown in
Figure 5, 17 aromatic amines were completely separ-
ated using a DB-35ms column.
Lower aliphatic tertiary amines in environmental
samples were analysed by headspace GC with a mass
selective detector (MSD) using a polymethylsiloxane
column. The SPME method has gained popularity as
a solvent-free, reliable and Sexible tool for sampling
a variety of volatile and semi-volatile compounds. By
combining SPME with GC, these compounds can be
simply and rapidly extracted, concentrated and intro-
duced into the GC system. Using headspace SPME
and GC-MSD on polysiloxane-type fused silica capil-
lary columns such as DB-1, OV-1, SPB-1, HP-1 and
HP-5, amphetamine, methamphetamine and related
III / AMINES: GAS CHROMATOGRAPHY 1989
stimulants in urine can be analysed at the ng mL\1
level. Recently, headspace SPME and GC-NPD using
a slightly polar capillary column CBJ-17 (Figure 6)
has developed as a method for determining ampheta-
mines in human hair.
Future Prospects
Much of the early work on the separation of free
amines was done with columns packed with PEG and
KOH on diatomaceous earths. Although this ap-
proach was reasonably successful, the analysis of free
amines on packed columns has now largely been
replaced by analysis on fused silica capillary columns.
Application of capillary columns is expected to in-
crease as further developments in these columns, e.g.
shorter inactive columns with smaller internal dia-
meters giving ultra-high column efRciency and speed,
higher temperature phases and exterior coating for
the fused silica tubing, permit the analysis of both
high temperature and highly volatile amines. Further-
more, simple, rapid and automatic analysis of free
amines in various samples will be achieved by combi-
nation with convenient sample preparation tech-
niques such as SPME.
See also: II/Chromatography: Gas: Column Chromatog-
raphy; Detectors: Selective.
Further Reading
Clement RE (ed.) (1990) Gas Chromatography. Biochemi-
cal, Biomedical, and Clinical Applications. New York:
John Wiley.
1990 III / AMINO ACIDS / Gas Chromatography
Grant DW (ed.) (1996) Capillary Gas Chromatography.
New York: John Wiley.
Grob RL (ed.) (1995) Modern Practice of Gas Chromatog-
raphy, 3rd edn. New York: John Wiley.
Heftmann E (ed.) (1992) Chromatography, 5th edn. Part B:
Applications. (Journal of Chromatography Library, Vol.
51B). Amsterdam: Elsevier.
Hyver KJ and Sandra P (eds) (1989) High Resolution Gas
Chromatography, 3rd edn. Delaware: Hewlett-Packard.
Kataoka H (1996) Derivatization reactions for the deter-
mination of amines by gas chromatography and their
applications in environmental analysis. Journal of
Chromatography A 733: 19.
AMINO ACIDS
Gas Chromatography
S. L. MacKenzie, Plant Biotechnology Institute,
Saskatoon, Saskatchewan, Canada
Copyright ^ 2000 Academic Press
During the 1950s and 1960s, signiRcant progress was
made in the development of automated amino acid
analysers based on separation by ion exchange. How-
ever, such instruments were dedicated to the task of
amino acid analysis and were of limited application
to the analysis of other types of compounds. Further-
more, they were expensive. During the same period,
gas chromatography (GC) was being rapidly de-
veloped following the demonstration in 1952 by
James and Martin that fatty acids could be assayed by
GC. There followed a vast expansion in the applica-
tion of GC to the analysis of other types of com-
pounds. Amino acids were a logical target. In the
intervening years, methods have been developed for
assaying amino acids in protein hydrolysates and
physiological Suids, and for determining the propor-
tions of amino acid enantiomers in racemic mixtures.
Some landmark developments are listed in Table 1.
Proteic and Physiological Amino Acids
Derivative Development
Amino acids are not sufRciently volatile or stable at
the temperatures required for analysis by GC. Thus,
they must be converted to derivatives having the de-
sired characteristics. It was to be no simple task to
derivatize or mask the several functional groups in
even the 20 proteic amino acids. Carboxy, amino,
Kataoka H (1997) Methods for the determination of
mutagenic heterocyclic amines and their applications in
environmental analysis. Journal of Chromatography
A 774: 121.
Pawliszyn J (1997) Solid Phase Microextraction: Theory
and Practice. New York: Wiley VCH.
Riggin RM, Cole TF and Billets S (1983) Determination of
aniline and substituted derivatives in waste water by gas and
liquid chromatography. Analytical Chemistry 55: 1862.
Yang X-H, Lee C and Scranton MI (1993) Determination
of nanomolar concentrations of individual dissolved low
molecular weight amines and organic acids in seawater.
Analytical Chemistry 65: 572.
hydroxy and sulfhydryl groups all need to be con-
verted to eliminate internal zwitterionic charges and
hydrogen bonding, and thus increase the volatility of
the derivatives. It was thought in those early years that
the molecular mass also required to be reduced but it
was later realized that this was not an absolute require-
ment. As new reagents became available, it was found
that volatility could be signiRcantly increased while
increasing the derivative mass. Apart from the multi-
plicity of functional groups, it is also necessary that
each group should be quantitatively converted.
The Rrst report of amino acid analysis by gas liquid
chromatography was published in 1956. Hunter,
Dimick and Corse oxidized isoleucine and leucine
with ninhydrin to form volatile aldehydes. These
were resolved using a 10 ft long silicone oil}celite
column operated isothermally at 693C. Peaks were
detected at about 44 and 48 min (Figure 1). The al-
dehydes were generated using 2}5 mg of each amino
acid. Either of the leucines could be assayed in the
presence of 10-fold quantities of the other. However,
only about eight simple amino acids yield volatile
aldehydes.
From this simple but momentous beginning, there
followed, in the next two decades, a proliferation of
reaction schemes to prepare stable, volatile amino
acid derivatives. Various oxidation, hydrocracking,
pyrolysis and reduction reactions were explored but
signiRcant progress was to evolve from those proced-
ures which focused on substituting the exchangeable
protons of the reactive groups. In 1957, Bayer,
Reuther and Born separated glutamic acid, leucine,
methionine, norleucine, norvaline, phenylalanine,
sarcosine and valine methyl esters on a silicone
oil}sodium caproate packing. The use of an acyl ester
constituted the Rrst report of a key component in

Table 1 Advances in gas chromatography of amino acids
1956 First GC analysis (Hunter, Dimick and Corse)
1959 Acyl amino acid alkyl esters separated (Youngs)
1965 Resolution of alanine, leucine and valine enantiomers
(Gil-Av, Charles and Fischer)
1962}79 Development of derivatization and separation
procedures for the proteic amino acids (Gehrke)
1971 First single column separation of all proteic amino
acids (Moss, Lambert and Diaz)
1971}76 Further improvements in resolution
1977 Development of Chirasil Val威 (Frank, Nicholson and
Bayer)
1989 Use of cyclodextrins for enantiomer resolution (KoK nig,
Krebber and Mischnick)
1991 4 min analysis of proteic amino acids (Hus\ ek)
a derivatization strategy which would eventually
prove to be successful. One year later, Bayer reported
that good resolution could be achieved using N-tri-
Suoroacetyl (TFA) amino acid esters. This work
represented the Rrst use of N-TFA derivatives, repre-
sentatives of a class of compounds which would
feature strongly in later developments.
In the next decade, N-formyl and -acetyl deriva-
tives were combined with a variety of alkyl esters
such as methyl, ethyl, propyl, isopropyl, isobutyl,
amyl and isoamyl. The work of Youngs in 1959 was
the Rrst in which N-acyl derivatives were combined
with alkyl amino acid esters. N-acetyl ethyl and butyl
esters of six simple amino acids were separated on
Figure 1 Separation of 3-methylbutanal and 2-methylbutanal
using a 10ft column filled with a silicone}celite mixture. (Repro-
duced with permission from Hunter IR, Dimick KP and JW Corse
(1956) Determination of amino acids by ninhydrin oxidation and
gas chromatography. Chemistry and Industry 294}295.)
III / AMINO ACIDS / Gas Chromatography 1991
hydrogenated vegetable oil. This approach was to
provide the foundation for developments leading,
over the next several years, to the quantitative resolu-
tion of all the amino acids in a protein hydrolysate. In
1964 Karmen and Saroff showed that excellent yields
of N-TFA amino acid methyl esters were obtained
when the esters were Rrst prepared and then acylated.
This general protocol remains in use.
The use of N-TFA derivatives in combination with
amino acid alkyl esters was Rrst reported by Ettre in
1962. Starting in the same year, Gehrke and his
colleagues systematically studied the derivatization
and chromatography of the N-TFA n-butyl amino
acid esters. TFA derivatives were used in amino acid
chemistry by Weygand as early as 1952 but were Rrst
applied in the context of GC analysis in 1960. In the
Rrst report, 22 naturally occurring amino acid deriva-
tives were resolved in less than 45 min using a 2 m
column packed with Gas Chrom A coated with 1%
neopentyl succinate. Subsequently, the esteriRcation
reaction was simpliRed by using direct esteriRcation
instead of methylation followed by interesteriRcation.
Direct on-column injection and an all-glass system
were demonstrated to avoid degradation of some
derivatives. Rigorous exclusion of water is necessary
both for complete derivatization and to prevent
hydrolysis of derivatives once formed. These and
other procedures developed by Gehrke formed a solid
quantitative foundation for subsequent studies by
others.
Continued reRnement of both the reaction chem-
istry and the columns culminated in the complete
separation of the 20 proteic amino acids in 1971.
Seventeen amino acids were resolved using a 4 mm
i.d.;1.5 m glass column packed with 0.65% ethy-
lene glycol adipate (EGA) on 80}100 mesh Chromo-
sorb W}AW. The derivatives of arginine, histidine,
tryptophan and cystine were separated from those of
the other amino acids on a 4 mm i.d.;1.5 m glass
column packed with a mixed stationary phase of 2%
OV-17 and 1% OV-210 coated on 100}200 mesh
Supelcoport. In particular, histidine could be directly
assayed. The two columns were operated simulta-
neously, resulting in an analysis time of 15}30 min. In
1979, the same derivatives were separated on a single
EGA liquid phase but no signiRcant improvement
over other available procedures was obtained.
Gehrke also conducted a thorough assessment of
possible sources of contamination. As detection sensi-
tivity increased, contamination became a signiRcant
problem. At the nanogram level, contamination was
shown to derive from laboratory reagents such as
butanol, methylene chloride and water, and from
human sources such as dandruff, Rngerprints, hair,
saliva and skin fragments.
1992 III / AMINO ACIDS / Gas Chromatography
The stationary phases used during the early years of
development fell into three main classes: silicones,
polyglycols and polyesters. Because it was difRcult to
separate even the proteic amino acids on a single
phase, mixed phases were common. Eventually, how-
ever, the silicone phases, in nonpolar or slightly polar
forms, became favoured and were essential for quant-
itative elution of arginine, cystine and histidine deriv-
atives. Dual- and triple-column procedures were to
give way in the search for a single column separation
of the proteic amino acids. The Rrst such resolution
was achieved in 1971 by Moss, who prepared the
N-heptaSuorobutyryl (HFB) n-propyl esters. These
were resolved on a 10;1/4 in glass column packed
with 3% OV-1 coated on 80}100 mesh HP Chromo-
sorb W (Figure 2). No quantitative data were pro-
vided. There followed other variations on the same
theme. The N-HFB isoamyl (1973), isobutyl (1974)
and isopropyl (1979) esters provided similar resolu-
tions but with subtle separatory advantages depend-
ing on the relative proportions of speciRc amino acids
present. Resolution was primarily a function of the
ester, while the acyl group mainly moderated the
volatility.
The search for a single-column resolution of the
proteic amino acids was paralleled by a search for
a single reaction which would derivatize all the func-
tional groups present in amino acids. Trimethyl-
silylation was introduced as early as 1961 by RuK hl-
man and Giesecke who reacted trimethylchlorosilane
with amino acid salts. Six amino acids were separated
in less than 30 min. A fuller account in 1963 reported
that tyrosine and histidine derivatives tended to de-
compose in the presence of moisture or oxygen. The
early reagents were generally silylated amines or
monosubstituted amides and double derivative for-
mation was a signiRcant problem. However, newer
Figure 2 Separation of N(O, S )-heptafluorobutyryl n-propyl amino acids. (Reproduced with permission from Moss CW, Lambert MA
and Diaz FJ (1971) Gas-liquid chromatography of twenty protein amino acids on a single column. Journal of Chromatography 60:
134}136.)
reagents, for example bis-(trimethylsilyl)triSuoroacet-
amide, were considerably more potent and derivatiz-
ation became quantitative. In more recent work
(1993), all 22 proteic amino acids, including
glutamine and asparagine, which would not be pres-
ent in protein hydrolysates, have been quantitatively
resolved as the N(O)-tert-butyldimethylsilyl deriva-
tives in 41 min on a DB-1 column. The derivatives are
formed in 30 min at 753C.
Other approaches have also been used in the search
for the simplest derivatization commensurate with
reproducibility and stability, and with good
chromatographic characteristics. Reaction with di-
chlorotetraSuoroacetone forms stable 2,2-bis(chloro-
diSuoromethyl)-4-subst-1,3-oxazolidine-5-one
derivatives. All the proteic amino acids and more
than 30 other
-amino acids have been studied. How-
ever, a second reaction with HFB anhydride is
required and analysis of the diaminodicarboxylic
acids histidine and tryptophan required a second
column.
Alkoxycarbonyl alkyl esters, speciRcally the iso-
butoxycarbonyl methyl esters, were Rrst prepared by
Makita in 1976. Twenty proteic amino acid deriva-
tives were separated using a dual-column system but
the derivatization procedure involves multiple extrac-
tion. Arginine was Rrst converted to ornithine. At
that time, this procedure offered no signiRcant ad-
vantage over the other protocols available. However,
the method was subsequently improved so that, in
1996, all the proteic amino acid derivatives were
resolved as single peaks in 9 min using a DB-17 capil-
lary column. Serum amino acids could be assayed
without any prior clean-up except for deproteiniz-
ation. The isobutoxycarbonyl derivatives have also
been effectively combined with tert-butyldimethyl-
silyl esters.

Figure 3 Analysis of N(O, S )-ethoxycarbonyl amino acid ethyl esters on a 10 m;0.25 mm i.d. capillary column coated with OV1701.
(Reproduced with permission from Hus\ ek P and Sweeley CC (1991) Gas chromatographic separation of protein amino acids in four
minutes. Journal of High Resolution Chromatography 14: 751}753.)
In 1991, Hus\ ek prepared derivatives in the same
general class but using ethyl chloroformate which
reacts with all the functional groups found in amino
acids. The N(O, S)-ethoxycarbonyl ethyl esters are
formed in seconds in an aqueous medium. The deriv-
atives were resolved in less than 5 min using a moder-
ately polar OV1701 capillary column (Figure 3).
A variety of derivatization options are now avail-
able. The N-HFB isoamyl, isobutyl or isopropyl esters
are equally effective for the relatively simple task of
assaying the standard proteic amino acids. However,
procedures requiring only a single derivatization step
are more convenient and are preferred. Either the
isobutoxycarbonyl methyl esters or the ethoxycar-
bonyl ethyl esters can be quickly prepared and re-
solved in less than 10 min using moderately polar
capillary columns.
The several hundred amino acids which are present
in physiological Suids cannot be resolved by any
single method. Each method has advantages in a spe-
ciRc context. Frequently, however, the target is
a single or a few structurally related amino acids. In
such a context, any of the methods cited above may
be appropriate, depending on the speciRc separations
required. However, methods based on alkoxycar-
bonyl alkyl esters are more convenient to implement.
Furthermore, some physiological samples, such as
sera, can be assayed directly after deproteinization.
Very few amino acids are not amenable to being
analysed by GC. Furthermore, the resolving power of
capillary column chromatography cannot be matched
by any other separatory medium. GC remains the
III / AMINO ACIDS / Gas Chromatography 1993
method of choice for assaying amino acids in com-
plex physiological samples.
Resolution of Optical Isomers
The determination of the conRguration of amino
acids and the relative proportions of the D and L forms
is important in both natural and synthetic contexts.
Proteins in living organisms commonly contain only
the L-amino acids but D-amino acids occur in anti-
biotics (e.g. antiamoebin, gramicidin, valinomycin),
bacterial cell wall peptidoglycans and in animals and
insects. They have also been detected in human urine
and blood. On death, the L-amino acids racemize, but
so slowly that a racemic mixture is only produced
over a geological time scale. The racemization rate is
a function of temperature and the structure of each
amino acid. Aspartic acid, which has a racemization
half-life of about 15 000 years at 203C, is most com-
monly used for archaelogocial dating, but there is
considerable controversy over the results obtained.
Animal bones and shells and certain sediments con-
tain proteins, for example, collagen and conchiolin.
Extraction of the residual protein and determination
of the enantiomer ratio of aspartic acid following
hydrolysis can, when combined with knowledge of
the thermal history of the sample, be used to deter-
mine the age of the fossil. Recemization age dating is
generally more sensitive and less expensive than the
radiocarbon method. Typical examples of the use of
this technique have been analysis of Apollo 12 lunar
material and dating of the Dead Sea scrolls.
1994 III / AMINO ACIDS / Gas Chromatography
Amino acids also racemize under various condi-
tions such as prolonged acid hydrolysis and during
solid-phase peptide synthesis. Chemical procedures
such as asymmetric synthesis require proof of enan-
tiomeric purity, especially if the product is to be used
for pharmaceutical purposes. ConRgurational analy-
sis of peptide antibiotics and establishing retention of
conRguration during peptide synthesis are other con-
texts in which it is important to determine the enan-
tiomeric composition of amino acid samples.
Enantiomeric amino acid mixtures are resolved us-
ing two approaches. The Rrst is to derivatize (acylate
or esterify) with optically active reagents to form
diastereoisomers or diastereomers which are resolved
on an optically inactive stationary phase. The re-
agents must be of high optical purity and conversion
must be quantitative. The second approach is to de-
rivatize with optically inactive reagents, for example
the N-TFA isopropyl esters, and then conduct the
separation on columns containing optically active
stationary phases.
Most amino acid optical isomers result from asym-
metry at the
-carbon atom and depend on the pres-
ence of an
-hydrogen atom. However, some contain
two optically active centres. Thus, the threo and
erythro forms of the hydroxy amino acids and
isoleucine and allo-isoleucine can be resolved on con-
ventional columns. Similarly, isovaline, which con-
tains one asymmetric centre but no
-hydrogen, has
also been resolved. The mechanism has been postu-
lated to depend on the formation of transient dia-
stereomeric hydrogen-bonded association complexes
but other factors such as dipole}dipole interactions
and dispersion forces may also play a role.
Figure 4 Resolution of diastereomers of N-trifluoroacetyl amino acid (N-TFA) 2-octyl esters. (Reproduced with permission from
Gil-Av E, Charles R and Fischer G (1965) Resolution of amino acids by gas chromatography. Journal of Chromatography 17: 408}410.)
Resolution of Diastereomers
All four diastereomers cannot be resolved using
optically inactive stationary phases: the DD#LL and
DL#LD enantiomer pairs usually coelute. The elution
order depends on the speciRc derivatives. A diastereo-
mer can be formed by esteriRcation or by acylation.
Optically Active Esteri\cation Reagents
Initial studies focused on forming active esters of
N(O)-acyl amino acids and these were subsequently
to be the most widely used derivatives. In 1965,
Gil-Av reported the Rrst resolution of amino acid
diastereomers by GC (Figure 4). The 2-butyl and 2-
octyl amino acid esters of alanine, glutamic acid,
leucine, phenylalanine, proline and valine were re-
solved on capillary columns coated with either
poly(triSuoropropylmethylsiloxane) or poly(propy-
lene glycol) operated isothermally at 140 or 1803C. In
the same year, Pollock reported the resolution of the
N-TFA 2-butyl esters of 13 amino acids but those of
aspartic acid, serine and threonine were only partially
resolved. A study by Westley (1968) concluded that
the resolution was directly proportional to the size of
the groups attached to the alcoholic asymmetric car-
bon and to the proximity of the branching to the
asymmetric centre. Thus 3,3-dimethyl-2-butanol
gave superior resolution. In 1968, Pollock extended
his study to the resolution of all the proteic amino
acids except arginine, histidine and cystine. Three
years later, 37 amino acid diastereomers were re-
solved as the N-TFA 2-butyl esters.
In 1977, KoK nig separated the N-pentaSuoro-
propionyl (PFP) (#)-3-methyl-2-butyl esters of all the

common proteic amino acids, including arginine, his-
tidine and tryptophan. Excellent resolution was ob-
tained using a 25 m column coated with SE30 and
temperature programming from 85 to 2203C at
23C min\1.
Optically Active Acylation Reagents
A variety of carbonyl chlorides have been used to
generate optically active dipeptides. N-TFA-L(!)-
prolyl chloride was Rrst used in 1965 by Halpern and
Westley who separated the isomers of alanine,
leucine, methionine, phenylalanine, proline and va-
line. The reagent was chosen because it was thought
that the cyclic nature of the derivative would preclude
recemization via an oxazolinone mechanism. This
reasoning was later shown by Bonner to be incorrect
but the problem was overcome by modifying the
derivatization procedure. The reagent subsequently
came into fairly common use. It was extensively
studied by Iwase and Murai who combined TFA, PFP
and HFB forms with methyl, n-propyl, n-butyl, tert-
butyl and cyclopentyl esters. By assessing the resolu-
tion of alanine, valine, leucine and isoleucine, they
concluded that the esters of n-alkyl alcohols gave
better resolution than branched or cyclic chain
alcohols.
KoK nig introduced a second asymmetric centre into
amino acid methyl esters using the chiral reagent
L-
-chloroisovaleryl chloride. Formation of the 3-
methyl-2-butyl esters enabled resolution of all the
proteic amino acid diastereomers, including arginine,
on an SE-30 capillary column in less than 1 h. A sep-
arate analysis was required for the basic amino acids.
Nevertheless, the diastereomer approach was to be
overtaken by the more direct and absolute method of
enantiomer resolution on chiral phases.
Resolution of Enantiomers on
Optically Active Columns
In 1966, Gil-Av demonstrated the Rrst resolution of
amino acid enantiomers on an optically active sta-
tionary phase. The N-TFA-2-butyl esters of alanine,
valine and leucine were resolved on an N-TFA-L-
isoleucine lauryl ester phase coated on a capillary
column. However, phases of this type quickly gave
way to dipeptide phases such as N-acyl-L,L-dipeptide
alkyl esters which were Rrst introduced by Feibush
and Gil-Av in 1967 and which produced better
resolution.
In 1970 Nakaparskin and colleagues separated 17
amino acid enantiomers on an N-TFA-L-val-L-val-
cyclohexyl ester phase (val-val). In earlier studies,
stainless-steel columns up to 500 ft long were used.
III / AMINO ACIDS / Gas Chromatography 1995
Consequently, analysis times were prolonged and the
cystine, serine and threonine derivatives were de-
graded. In addition, dipeptide stationary phases such
as val-val were functional over a limited temperature
range or a limited maximum operating temperature.
Columns were usually operated in isothermal mode.
KoK nig addressed the problem of temperature stabil-
ity by introducing the N-TFA-L-phenylalanyl-L-
leucine cyclohexyl ester which could be operated
at 1403C. A later modiRcation, the N-TFA-L-
phenylalanyl-L-aspartic acid bis-(cyclohexyl) ester,
was stable over the range 96}1653C and allowed the
use of temperature programming. In addition, the
introduction of glass capillary columns reduced
degradation of the amino acid derivatives. The
high boiling N-PFP isopropyl esters of aspartic
acid, methionine, phenylalanine, glutamic acid,
tyrosine, ornithine and lysine were eluted using
a 20 m column. However, the diamide phases still left
room for improvement in thermal stability and in
peak resolution.
Another generation of phases was introduced by
Frank, Nicholson and Bayer who linked the diamide
moiety, L-valine tert-butylamide, to a polysiloxane
backbone. Later termed Chirasil Val威, phases of this
general type became predominant and are still in use.
Early versions of this phase resulted in the overlap of
D- and L-proline, D-isoleucine and L-allo-isoleucine,
and L-threonine and D-allo-isoleucine. Nevertheless,
the enantiomers of all the other proteic amino acids
were resolved as the N-PFP n- or isopropyl esters in
about 30 min by temperature programming from 90
to 1903C (Figure 5). Acid treatment of the glass capil-
lary followed by methanol washing was necessary
rigorously to exclude basic sites and thus to obtain
satisfactory elution of cysteine, serine, threonine and
tyrosine and to obtain a sharp peak for arginine. The
relative retention times of the amino acids can be
manipulated by including polar modiRers such as
cyanopropyl and phenyl groups but the effect varies
with speciRc amino acids. The L-valine tert-butyl moi-
ety was subsequently grafted to chloropropionyl-
methyl phenylmethyl silicone, a modiRed OV}225,
and to Silar 10C, but no overall improvement was
achieved.
Chirasil-Val威 was further improved by the incor-
poration of 15% phenyl groups substituted for
methyl groups in the dimethylsiloxane units and the
introduction of fused silica capillary columns. Ther-
mal stability, ease of handling and separation efRcien-
cy were improved. The product is commercially
marketed as HeliSexTM Chirasil-Val威.
Later improvements included the enhancement
of enantioselectivity and thermal stability by immo-
bilization of the Chirasil-Val威 by radical or thermal
1996 III / AMINO ACIDS / Gas Chromatography
Figure 5 Resolution of a racemic mixture of proteic amino acids as the N-(O, S )-pentafluoropropionyl n-propyl esters. (Reproduced
with permission from Bayer E, Nicholson G and Frank H (1987) Separation of amino acid enantiomers using chiral polysiloxanes:
quantitative analysis by enantiomer labeling. In Gehrke CC, Kuo KCT and Zumwalt RW (eds) Amino Acid Analysis by Gas
Chromatography, Volume II, pp. 35}53. Boca Raton, FL: CRC Press.)
reactions. Chiral polysiloxanes with regular repeat
units, e.g. triSuoroethyl ester-functionalized poly-
siloxanes supporting L-val-tert-butylamide or L-
-
naphthylethylamine liquid phases, have shown im-
proved enantioselectivity. Backbone modiRcation
achieved by replacing one methyl group per dialkyl-
siloxy unit with a pentyl or hexyl group improved
resolution of arginine and tryptophan N-TFA n-
propyl esters but the overall separation of the other
amino acids was not signiRcantly affected. However,
satisfactory results have been obtained by varying the
proportion of L-val-tert-butylamide on the poly-
siloxane backbone. A ratio of about 6}7 dimethyl-
siloxane units per chirally modiRed dialkyl siloxane
unit is effective for the complete resolution of all
components present in a chiral mixture of the 20
proteic amino acids in about 35 min on
a 20 m;0.3 mm glass capillary column.
Most studies on amino acid enantiomer resolution
on Chirasil-Val威 type columns have used N-per-
Suoroacyl alkyl ester derivatives. However, other
derivatives may present advantages in speciRc con-
texts. For example, the N-alkyloxycarbonyl alkylam-
ide derivatives of proline are completely resolved on
a Chirasil-Val威 column. Similarly, KoK nig demon-
strated the utility of isocyanate derivatives for resolv-
ing the enantiomers of N-methyl and -hydroxy
amino acids.
A radically different approach to enantiomer res-
olution has become possible with the development
of cyclodextrins as stationary phases. Although suit-
able for liquid chromatography, the high melting
point of cyclodextrins rendered them unsuitable for
GC without further modiRcation. KoK nig reduced the
melting point and increased stability by introducing
hydrophobic moieties by both partial and complete
alkylation and acylation of the hydroxy groups. -
Cyclodextrin substituted with 3-O-butyl and 2,6-di-
O-pentyl residues was found to resolve most of the
common amino acid enantiomers as the N-TFA
methyl esters. Histidine enantiomers were only par-
tially separated and arginine did not elute from the
column. However, proline, 3,4-dihydroxyproline and
pipecolic acid enantiomers were resolved, strongly
suggesting that hydrogen bonding is not involved
in the separatory mechanism. Atypical amino acids
such as N-methyl and -amino acids were also re-
solved.
More recently (1994), Abe explored capillary col-
umns coated with four types of cyclodextrin deriva-
tives of 6-O-tert-butyldimethylsilyl-2,3-di-O-acetyl
or n-butyl-- and -cyclodextrin. Depending on the
phase, all proteic amino acid enantiomers except for
those of tryptophan were resolved as the N-TFA
isopropyl esters. Variants such as 2,6-di-O-pentyl-3-
O-propionyl--cyclodextrin have also been used to
 III / AMINO ACIDS / Gas Chromatography 1997

separate a number of amino acid enantiomers. Mo-
lecular modelling has positively correlated the GC
elution order of proline derivatives on 2,6-di-O-
methyl-3-O-triSuoroacetyl--cyclodextrin with the
energies of the host}guest complexes.
Several satisfactory methods now exist for the res-
olution of amino acid enantiomers. Typically, 0.1%
of a minor enantiomer can be precisely determined
and, depending on the context and the speciRc
method used, it is possible to assay as little as 0.01%
or less.
Detectors
By far the majority of GC amino acid analyses have
been conducted using Same ionization detectors
(FID). These have the advantage of being sensitive
and economical, but are nonspeciRc and provide no
structural information.
Selective detectors confer distinct analytical ad-
vantages but are most often used to address special,
nonroutine analytical problems. For example, the
ability to detect a speciRc atom or molecular property
can simplify sample preparation. Thus, by using a ni-
trogen/phosphorus-selective detector, contaminating
compounds not containing nitrogen or phosphorus
are simply not detected in most samples. Further-
more, it can reasonably be assumed that those
peaks which have been detected contain nitrogen. In
addition, problems caused by overlapping peaks are
reduced. Selective detectors can also provide addi-
tional sensitivity. The ultimate detector is a mass
spectrometer which can, depending on the context,
provide all of these advantages and also provide de-
tailed structural information.
A number of selective detectors have been used to
assay amino acids. Their use will be illustrated using
some examples.
A nitrogen/phosphorus-speciRc detector, operated
in the nitrogen mode, has been used to assay free
amino acids in conifer tissues. All the proteic amino
acids and several biologically important nonproteic
amino acids were assayed at the low picomole level as
the N-HFB isobutyl esters. Comparison of the FID
chromatogram with the NPD chromatogram enabled
identiRcation of those compounds which did not con-
tain nitrogen (Figure 6). Similarly, 1-aminocyclo-
propane-1-carboxylic acid, a precursor of the plant
hormone ethylene, has been assayed as the N-benzoyl
n-propyl derivative in the leaves and xylem sap of
tomato plants. More recently (1997), 21 proteic and
33 nonproteic amino acids have been resolved in less
than 30 min as the N-isobutoxycarbonyl methyl es-
ters at a detection limit of 6}150 pg per injection.
Small urine samples were analysed without prior
clean-up and with no detectable inSuence from any
non-nitrogen-containing compounds present.
Flame photometric detection (FPD) is useful for
analysing sulphur-containing amino acids but has
rarely been used in that context. Amino acid phos-
phorylation is an important biochemical regulatory
mechanism and is also important for correlating pro-
tein structure and function. The O-phosphoamino
acids, speciRcally O-phospho serine, threonine and
tyrosine, have been assayed as the N-isobutoxycar-
bonyl methyl esters using FPD. The detection limits

Figure 6 Resolution of White Spruce leaf-free amino acids as the N-(O, S )-heptafluorobutyryl isobutyl esters using a flame ionization
detector. Peaks marked by asterisks were shown not to contain nitrogen by comparison with an analysis of the same sample using
a nitrogen-selective detector. (Reproduced with permission from MacKenzie SL (1986) Amino acid analysis by gas-liquid chromatogra-
phy using a nitrogen-selective detector. Journal of Chromatography 358: 219}230.)
1998 III / AMINO ACIDS / Gas Chromatography
ranged from 0.18 to 0.3 pmol, reSecting a sensitivity
about 200 times greater than FID detection. The
method has been applied to the determination of
O-phosphoamino acids phosphorylated by protein
kinase both in vitro and in vivo without radiolabell-
ing. Other amino acids did not interfere. The second-
ary amino acids, proline, pipecolic, thioproline, hy-
droxyproline and hydroxypipecolic acids, have also
been assayed using FPD. Detection limits for the
N-dimethylthiophosphoryl methyl esters were
0.1}0.7 pmol per injection.
Electron capture detectors are particularly useful
for detection of the strongly electronegative per-
Suoroacyl derivatives of amino acids, but few studies
have been conducted. Typically, as little as 1.4 pmol
of tyrosine has been detected in a standard amino acid
mixture. -Aminobutyric acid and Rve other aliphatic
acids have been assayed in small volumes of super-
natants from brain homogenates following sequential
reaction with isobutyl chloroformate and penta-
Suorophenol.
Mass spectrometric detection provides structural as
well as quantitative information. It is most frequently
used either to conRrm the structure of derivatives
during the development of new protocols or to ident-
ify unknown compounds. Detection limits are fre-
quently in the femtomole range. Electron impact (EI)
ionization is most commonly used but both positive
and negative ion chemical ionization have also been
applied. Selected ion monitoring (SIM) of diagnostic
ions is often used to increase sensitivity.
Typical examples of the structural information role
of a mass spectrometric detector are the identiRcation
of O-phosphoamino acids in urine hydrolysates, the
identiRcation of amino acid ethyl esters in wines, the
determination of amino acid composition in small
peptides, and the assaying of -aminobutyric acid in
mouse brain synaptosomes following therapy with
the antiepileptic drug valproic acid. The versality of
GC-mass spectrometry (GC-MS) is further illustrated
by the identiRcation of 3-OH-4-methyldecanoic acid,
a fungal cyclodepsipeptide, and by the simultaneous
analysis of branched-chain carboxylic,
-oxo,
-hy-
droxy and
-amino acids in the urine of patients
suffering from maple syrup urine disease. GC-MS has
also been used to characterize binding media from
medieval polychrome sculptures. Animal glues,
casein, egg and drying oils were identiRed as compo-
nents of the binders of paint and ground layers.
The expense of mass spectrometers mitigates
against their use as routine analytical detectors and
many real sample analyses by GC-MS (as distinct
from the analysis of standard mixtures) have been
directed to addressing analytical problems which can-
not be resolved using other types of detectors.
The ratios of the stable isotopes of C and N are
used in the assessment of in vivo protein turnover
studies, and in identifying the sources and history of
organic matter. Both natural abundances and the
ratios obtained after enrichment with singly or multi-
ply labelled amino acids or other compounds such as
13
C-glucose, pyruvate or acetate have been deter-
mined. The ratios may be determined after online
combustion following GC and introduction of the
resultant gases into a conventional isotope ratio mass
spectrometer. This approach has been used to study
15
N: 14N isotopic ratios in plasma-free amino acids
and, by eliminating many preparative steps, requires
only about 500
L-of plasma, whereas preparatory
methods may require as much as 60 mL.
Alternatively, the intact labelled compounds can
be introduced directly into the mass spectrometer.
For example, by combining stable isotope dilution
with the use of EI and SIM to monitor the [M-57]#
peak, homocysteine sulRnic acid, homocysteic acid,
cystine sulRnic acid and cysteic acid have been
shown to be agonistic to N-methyl-D-aspartate recep-
tors in brain tissue. This approach also enabled
the identiRcation and quantitation of these com-
pounds in normal human serum. Similarly, endogen-
ous and newly synthesized concentrations of the
neurotransmitter amino acids -aminobutyric acid,
glutamate and aspartate have been assayed in brain
slices following incubation with 13C-labelled pre-
cursors.
Future Developments
The techniques for derivatizing and separating the
standard amino acids in protein hydrolysates are ma-
ture and there is no signiRcant room for improve-
ment. Given the existence of quantitative derivatiz-
ation protocols which proceed very rapidly, it is
doubtful whether the development of on-column de-
rivatization would constitute a signiRcant advantage.
Similarly, proteic amino acids can now be assayed in
less than 10 min, so, given the availability of reliable
automatic injectors, a reduction in analysis time is not
of signiRcant value.
Physiological samples may contain several hundred
amino acids and these cannot, at present, all be re-
solved on any one single column. Frequently, how-
ever, only a subset is of interest. Thus, although the
simultaneous derivatization and separation of all the
proteic amino acids and as many as 50 of the more
common nonproteic amino acids are now possible, it
is likely that procedures targeted at speciRc subsets
will remain important in speciRc contexts. The
sensitivity of the FID detector is adequate for most
analytical purposes but mass spectrometric detectors

will remain important for specialized applications
requiring femtomole sensitivity.
See also: III/Amino Acids: Liquid Chromatography; Thin-
Layer (Planar) Chromatography. Amino Acids and
Derivatives: Chiral Separations. Amino Acids and
Peptides: Capillary Electrophoresis.
Further Reading
Gehrke CW, Roach D, Zumwalt RW et al. (eds) (1968)
Quantitative Gas-liquid Chromatography of Amino
Acids in Proteins and Biological Substances: Macro,
Semimicro and Micro Methods. Columbia, MO: Ana-
lytical Biochemical Laboratories.
Liquid Chromatography
I. MolnaH r-Perl, L. Eo( tvo( s University,
Budapest, Hungary
Copyright ^ 2000 Academic Press
The Rrst approach to the automatic liquid chromato-
graphy (LC) of amino acids (AAs) } known today as
ion exchange chromatography (IEC) } was published
by Spackman et al. in 1958. In over 40 years later, it
now takes less than 5 min (Figure 1) to separate and
quantitate the essential protein AAs instead of 2 days.
Early separations were carried out by post-column
derivatization.
Over the last 20 years LC has offered unlimited
possibilities in both the preparative and analytical
scale. The wide choice and sophisticated columns,
detectors, derivatization procedures, development of
modern instrumentation and data-handling systems
reduce time and costs, and give versatility and auto-
mation in Good Laboratory Practice (GLP)- control-
led conditions for selectivity, sensitivity and repro-
ducibility. It is the responsibility of the researcher to
choose the most appropriate method for the given
task. The most popular LC method for analysis of
both free AAs (present in many natural matrices,
biological Suids and tissues, feed and foodstuffs) and
of those constituents of protein hydrolysates is now
reversed-phase (RP) chromatography after pre-col-
umn derivatization of the AAs.
Numerous methods for derivatization are available
in the literature. This article will discuss the advant-
ages and drawbacks of the commonly used deriva-
tives.
Current trends in AA analysis identify the best
conditions for enantiomer separation and the devel-
opment of LC-mass spectrometry (LC-MS).
III / AMINO ACIDS / Liquid Chromatography 1999
Hus\ ek P and Macek K (1975) Gas chromatography of
amino acids. Journal of Chromatography 113: 139.
KoK nig WA (1987) The Practice of Enantiomer Separation
by Capillary Gas Chromatography. Heidelberg: HuK thig.
MacKenzie SL (1981) Recent developments in amino acid
analysis by gas-liquid chromatography. In: Glick D (ed.)
Methods of Biochemical Analysis, vol. 27, p. 1. New
York: Interscience.
Weinstein B (1966) Separation and determination of amino
acids and peptides by gas liquid chromatography. In:
Glick D (ed.) Methods of Biochemical Analysis, vol. 14,
p. 203. New York: Interscience.
Zumwalt RW, Kuo KCT and Gehrke CW (eds) (1987)
Amino Acid Analysis by Gas Chromatography. Boca
Raton, FL: CRC Press.
LC of Underivatized AAs
To attain one of the main advantages of LC } separat-
ing the ‘classical 20’ as underivatized AAs } has ap-
pealed to chromatographers. In spite of a number of
efforts, the simultaneous LC of underivatized AAs
has remained of secondary importance. Determina-
tion of a few selected AAs, such as tryptophan or
sulfur-containing AAs, has proved to be fruitful for
special tasks.
The aim of various investigations was to render
unnecessary the time-consuming derivatization tech-
niques. However, the characteristics of the free
AAs are considerably different from each other and
their various structural properties did not permit their
easy resolution. Thus, in attempting to achieve better
separation of free AAs, further means of discrimina-
tion were needed. For this purpose special techniques
have been introduced, such as the use of various
phase systems, ion pair and ligand exchange
chromatography, column-switching techniques or
anion exchange chromatography with electrochemi-
cal detection.
The solvent-generated ion exchange phase system
ensured the gradient elution of 19 AAs (Figure 2A):
some, but not all, are baseline-separated. A simple
isocratic method using aqueous, copper acetate/alkyl-
sulfonate additives containing acetate buffer (pH 5.6)
as mobile phase, a conventional RP column and UV
detection (230}240 nm) at different temperatures
and varying the concentrations of additives was un-
able to separate the classical 20 protein amino acids.
SigniRcant improvement in the separation can be
obtained by column switching (Figure 2B), as well
as by using an anion exchange column, a quaternary
2000 III / AMINO ACIDS / Liquid Chromatography
Figure 1 Separation of the phenylthiocarbamyl AAs separated on TSKgel Super-ODS (for details see Table 3). Peaks: 1, ASP;
2, Glu; 3, Ser; 4, Gly; 5, His; 6, Arg; 7, Thr; 8, Ala; 9, Pro; 10, NH#
4 ; 11, Tyr; 12, Val; 13, Met; 14, Cys; 15, Ile, 16, Leu; 17, Phe; 18, Lys.
(Reproduced with permission from Tosohaas, The Bioseparation Specialist, (1995) Catalogue p. 157, Figure 9/2.)
gradient mobile phase and pulsed amperometric
detection (Figure 2C). For tryptophan and the
sulfur-containing AAs (cysteine/cystine, methionine,
glutathione, etc.), the fast isocratic elution of the
underivatized samples has gained wide acceptance
and is a powerful tool in their quantitation.
Tryptophan can be measured directly, within
8 min, in neutralized alkaline hydrolysates of feed
and foodstuffs, using an RP column, 5% methanol
containing acetate buffer (pH&4.0) and UV detec-
tion at 280 nm. The pulsed amperometric detection
of sulfur-containing AAs, at the low pmol level, was
carried out with an Au working- and an Ag/AgCl
reference electrode, subsequent to their separation
on both cation exchange and on RP columns, ap-
plying as mobile phase 0.1 mol L\1 HClO4/0.15 mol
L\1 NaClO4/5% ACN.
LC of Derivatized AAs
Derivatization studies have concerned the optimiza-
tion of parameters, such as the yield and stability of
derivatives, to separate and quantitate all AAs with
a simple and fast elution procedure.
Post-column Derivatization (Table 1, Figure 3)
Post-column derivatization was the Rrst develop-
ment of IEC in the area of RP/high perfor-
mance liquid chromatography (HPLC), in its pioneer
period. It took time to develop pre-column derivat-
ization concepts which resulted in considerable
advantages.
Drawbacks of the post-column techniques
(Table 1) are long elution times and the need for
costly devices, such as a delivery system for the de-
rivatizing reagent (one or more extra pumps); (a)
mixing chamber for the column efSuent and the re-
agent(s); special thermostable reactors (packed bed,
air-segmented and/or coil reactors) ensuring the ne-
cessary delay for quantitative reactions accompanied
with as small band broadening as possible. Last but
not least, the mobile phase was probably incompat-
ible with the derivatizing reagent. The preferred mo-
bile phase might be inappropriate for the optimum
conditions of derivatization reaction. The early and
current stages of post-column methods can be
illustrated by elution followed by post-column
reaction with ninhydrin (NHYD; Figure 3A), with
o-phthaldialdehyde/
-mercaptoethanol (OPA/MCE;
Figure 3B), or with 1,2-naphthoqunione-4-sulfonyl
chloride (NQS; Figure 3C). All three types of deriva-
tives have been separated in most cases on ion ex-
change resin columns from the early 1970s. Recent
post-column methods are without exception, slow
separations (Table 1). However, the efRciency of the
recently published methods of NHYD derivatives us-
ing short columns is superior, determining 59 com-
pounds in 150 min and 40 compounds in 120 min
respectively.
Pre-column Derivatization (Tables 2 and 3;
Figures 4^6)
Pre-column derivatization offers numerous advant-
ages. It requires less equipment and allows the evalu-
ation of the derivatives in an easier way from the
point of view of their selectivity, sensitivity, various
means of detection, derivatization yield, stability
and storability. All of these phenomena can be con-
trolled and improved by use of modern instrumental
techniques and computerization, both individually
and simultaneously. Potential disadvantages in pre-
column derivatization as procedures can be com-
pletely avoided: contamination from the reagents
(due to their insufRcient purity) and loss of analyte
 III / AMINO ACIDS / Liquid Chromatography 2001

Figure 2 LC of underivatized AAs. (A) Separation of a test mixture using n-propanol gradient. Column: 250;3 mm, RP-8;
temperature"253C. Peaks: 1, CySO3H; 2, Asp; 3, Ser; 4, Glu; 5, Thr; 6, Gly#Pro; 7, Ala; 8, Cys; 9, NH# 4 ; 10, Tyr; 11, Val; 12, Met; 13,
Ile; 14, Phe; 15, Leu; 16, His, 17, Lys; 18, Trp; 19, Arg. (Reproduced with permission from Kraak JC et al. (1977) Journal of
Chromatography 142: 671.) (B) Chromatogram of standard AAs using a column-switching technique. First column, Inertsil C3; second
column, Inertsil ODS-2. (Reproduced with permission from Hanai T and Hirukawa M (1988) Journal of Liquid Chromatography 11:
1741.) (C) Chromatogram of AAs obtained by pulsed amperometric detection. Peaks : 1, Arg; 2, Lys; 3, Gln; 4, Asn; 5, Thr; 6, Ala; 7,
Gly; 8, Ser; 9, Val; 10, Pro; 11, Ile; 12, Leu; 13, Met; 14, system peak; 15, His; 16, Phe; 17, Glu; 18, Asp; 19, Cys; 20, Tyr. (Reproduced
with permission from Frankenberger WT Jr and Martens DA (1992) Journal of Liquid Chromatography 15; 423.)
2002 III / AMINO ACIDS / Liquid Chromatography

Table 1 Advances in the LC of post-column derivatized AAs, obtained with o-phthaldialdehyde/
-mercaptoethanol (OPA/MCE),
with ninhydrin (NHYD) or with 1,2-naphthoquinone-4-sulfonate (NQS)

Author Column size Type Eluents (elution Detector Reagent Analyte RSD Matrix No. of
and temperature 3C) UV (nm) (3C) (nmol L\1) % AAs/
date cm ; mm, m FEx/Em elution
time (min)

Moore, 150 0.9 40 Amberlite Citrate buffers, UV NHYD 100I3000 * AAs in 20/
1958 IR-120, IE 0.2 mol L\1: pH 3.25 440, 570 (I) hydrolysates 24I48 h
for first day (303C),
and pH 4.25 for the
second day (553C)
Grunau, 15 3 5 Pickering Pickering Eluents UV NHYD 20 * Plasma 59/&150
1992 ‘fast run’ A (Li280), B (Li750), 570 (1303C) AAs
C (RG003) (423C)
Iwase, 6 4.6 3 2622, Five eluents: UV NHYD 50 (3 Plasma 40/120
1995 Hitachi, IE PF-1IPF-4, PF-RG, 440,570 (1303C) AAs
cont. Li salts, ethanol,
benzyl alcohol,
thiodiglycol, Brij-35
buffer with pH 2.8,
3.7, 3.6, 4.1, -;
(gradient programme:
28I403C)
Roth, 25 6 Aminex 6, Citrate buffers: F OPA/MC 10 * Model study 14/170
1973 IE pH 3.20, 4.25 and * E (553C)
6.40 for 40, 60 and
70 min (343C for
100 min, then raised
to 553C)
Elrifi, 60 9 * IE Pierce Pico-Buffer FH OPA/MCE 0.63I45.0 * AAs in 23/128
1986 system, Li citrate No data (403C) foods
buffers; pH of A,B,C,D
and E"2.9,
3.1, 3.5, 3.4 and 2.3;
temperature gradient:
0I44 min (343C),
44I128 min (633C)
Haginaka, 30 4.6 5 ODS-5 A: 15 mmol L\1 Na F OPA/MCE 0.25I2.5 (4.5 AAs in 18/&120
1988 #guard octane sulfonate/ 340/450 (603C) hydrolysates
3 4.6 5 21 mmol L\1
H3PO4/9 mmol L\1
NaH2PO4/CH3OH
(20/20/20/1, v/v),
pH 2.8;
B: as A, except
(1/1/1/6, v/v), pH 4.2
(603C)
MCller, 1993 15 3 5 Pickering, A: 0.24 mol L\1 F OPA/MCE 0.6 (11 Physio- 39/180
#guard IE Li citrate, pH 2.27, 340/448 (43C) logical AAs
2 3 5 B: 0.64 mol L\1
Li citrate pH"7.50
(503C)
Saurina, 15 4.6 5 Spherisorb A: 20 mmol L\1 UV NQS 32 (5 AAs in food 18/105
1994 ODS 2 H3PO4#20 mmol L\1 305 (653C) #feeda
NaH2PO4#
15 mmol L\1 SDS;
B: 25 mmol L\1
H3PO4#25 mmol L\1
NaH2PO4#18.5
mmol L\1 SDS/PrOH
(4:1, v/v); (503C)

No data available; IE, ion exchange resin; ahydrolysates.

 III / AMINO ACIDS / Liquid Chromatography 2003

from incomplete interaction, undesirable side reac-
tions and sample handling losses.
Although numerous pre-column derivatization
techniques have been introduced in the last 30 years,
none complies with the criteria of an ideal procedure:
providing rapid and quantitative interaction in aque-
ous media, permitting mild conditions, ensuring in-
teraction with both primary and secondary AAs and
resulting in single and stable derivatives in the case of
all AAs.
OPA Derivatives (Table 2 and Figure 4)
The pioneering work of Roth (1971) on the very fast
reaction of AAs in aqueous solutions with ophthalal-
dehyde mercaptoethanol (OPA/MCE), detectable by
both UV and Suorescence, without the need to re-
move excess reagent, represented a great advance.
Because of the different stability of the isoindoles
obtained from the reaction of AAs with OPA/MCE,
pre-column derivatization with 3-mercaptopropionic
acid (MPA) and several N-alkyl-L/D-cysteines was
proposed. The OPA/MPA and OPA/N-acetyl-L-cys-
teine (NAC) reagents provide more stable isoindoles
compared to those formed with OPA/MCE, and the
optical resolution of enantiomeric amino acids with
OPA/NAC, as well as with other N-alkyl-L/D-cysteine
reagents, has opened a new area in enantiomer separ-
ation of AAs. Due to robotic autosamplers which
provide excellent reproducibility for even moder-
ately quantitative interactions, most AA analyses are

Figure 3 LC of post-column derivatized AAs (for details see Table 2). (A) Chromatographic profile of 59 AAs and related compounds.
Peaks: 3, o-phospho-DL-serine; 6, taurine; 9, o-phosphoethanolamine; 10, N -(1-D-mannityl)-L-glutamine (mannopine); 11, urea;
12,
-ciano-L-alanine; D, L-aspartic acid; 17, o-acetyl-L-serine; T, L-threonine; S, L-serine; N, L-asparagine; E, L-glutamic acid;
Q, L-glutamine; 27, L-homoserine; 29, sarcosine; 34, DL--aminoadipic acid; 36, S-methyl-L-cysteine; P, L-proline; G, glycine;
A, L-alanine; 42, L-citrulline; 45, L--aminobutyric acid; V, L-valine; 47, L-cystine; 50, -methyl-DL-methionine; M, L-methionine;
54, L-cystathionine; I, L-isoleucine; L, L-leucine; 62, L-norleucine; Y, L-tyrosine; F, L"phenylalanine; 67,
-alanine; 68; DL-
aminoisobutyric acid; 69, DL-homocystine; 70, -aminolevulinic acid; 71, 5-hydroxy-L-tryptophan; 72, -aminobutyric acid; 73, DL-
kynurenine; W, L-tryptophan; 76, ethanolamine; 77, -hydroxylysines (DL- and DL-allo); 78, ammonia; 79, -amino-n-caproic acid; 80,
creatinine; 81, L-ornithine; K, L-lysine; H, L-histidine; 85, 3-methyl-L-histidine; 87, 1-methyl-L-histidine; 89, L-carnosine; 90, L-anserine;
91, L-canavanine; 92, S-methyl-DL-methionine; 93, L--amino-
-guanidinopropionic acid; 94, L-leucinamide; 95, N G1-dimethyl-L-
arginine; R, L-arginine; 99, L-homoarginine. (Reproduced with permission from Grunau JA and Swiader JM (1992) Journal of
Chromatography 594: 165.) (B) Separation of OPA/MCE derivatives by gradient IEC chromatography. (Reproduced with permission
from MCller SE (1993) Journal of Chromatography 613: 223.) (C) Determination of AAs by ion pair liquid chromatography with
post-column derivatization using 1,2-naphtoquinone-4-sulfonate (NQS). Peaks: 1, Asp; 2, Ser; 3, Glu; 4, Gly; 5, Thr; 6, Ala; 7, Pro; 8,
Tyr; 9, Met; 10, Ile; 11, Phe; 12, Leu; 13, Nle; 14, Trp; 15, His; 16, Orn; 17, Lys; 18, Arg. Line"elution gradient profile. (Reproduced
with permission from Saurina J and HernaH ndez-Cassou (1994) Journal of Chromatography 676: 311.)
2004 III / AMINO ACIDS / Liquid Chromatography

Figure 3 Continued
 III / AMINO ACIDS / Liquid Chromatography 2005

Table 2 Advances in the LC of pre-column derivatized AAs, obtained with OPA/MCE, OPA/3-ethanethiol (OPA/ET), OPA/mercap-
topropionic acid (OPA/MPA), OPA/N-acetyl-L-cysteine (OPA/NAC), with OPA/isobutyryl-L /D-cysteine (OPA/NIBC) or with
OPA/MPA/fluorenylmethylchloroformate (OPA/MPA/FMOC)

Author Column size Type Eluents (elution Detector Reagent Analyte RSD Matrix No. of
and temperature 3C) UV (nm) (3C) (pmol L\1) % AAs/elution
date cm ; mm, m FEx/Em time (min)

Jones, 75 4.6 3 Ultrasphere A: THF/CH3 OH/NaAc F OPA/ 0.1I80 (1.5 AAs in 48/50
1983 #guard ODS (pH 7.2)"(5:95:900, v/v) 305I395 MCE hydrolysates
45 2.1 40 B: CH3 OH; (!) 420I650 (!)

Fekkes, 12.5 4.6 5 Spherisorb A: (pH 6.72I6.77 and F OPA/ 50 (2 Plasma 40/49
1995 ODS-2 B: (pH 5.95I6.00): 337/452 MCE AAs
250 mmol L\1 Na2HPO4/ (33C)
250 mmol L\1 propionic
acid/ACN/THF/H2O"
(20:20:7:2:51, v/v) C:
ACN/CH3OH/DMSO/
H2O"(28:24:5:43, v/v);
(25I353C)

Hill, 30 3.9 * -Bondapak A: 12.5 mmol L\L F OPA/ 5 * AAs in 20/40

1979 C-18 Na2HPO4 (pH 7.2) 229/470 ET (!) human
B: A eluent/ACN in serum
gradient; (!)

Eslami, 50 4.5 3 ODS IBM Buffer:&2 mol L\1 F OPA/ 40I100 * Model 22/14
1987 Na2 HPO4 (pH 7) 330/480 ET (!) study
A: ACN/H2O/buffer"
(50:425:25, v/v)
B: ACN/H2O/buffer"
(275:200:25, v/v);
(223C)

Godel, 25 4 4 Supersphere A: 12.5 mmol L\1 F OPA/ 1I10 (4.2 AAs in 28/40
1984 CH-8 Na2HPO4 (pH 7.2) 330/445 MPA (!) biological
B: 12.5 mmol L\1 fluids
Na2HPO4 (pH 7.2)/
ACN-(1:1, v/v); (!)

van Eijk, 15 4.6 2I3 Spherisorb A: 12.5 mmol L\1 F OPA/ 35 (3 Plasma 30/28
1993 #guard ODS-2 Na2HPO4 (pH 7.0)# 335/440 MPA (!) AAs
1 4 7 mL THF/1 l eluent
B: 12.5 mmol L\1
Na2HPO4 (pH 7.0)/
ACN/THF"(57:43:7,
v/v); (353C)

Teerlink, 10 4.6 3 Microsphere A: 4.5 mmol L\1 F OPA/ 100 (3.2 Plasma 25/17
1994 #guard ODS K2HPO4 (pH 6.9)# 230/389 MPA (!) AAs
1 2 2 mL THF/1L
B: 4.5 mmol L\1 K2HPO4
(pH 6.9)/CH3OH/ACN"
(50:35:15, v/v); (!)

Schuster, 20 2.1 5 Hypersil Protein hydrolysates, UV OPA/ UV: 2I5 (2.5 AAs in 19/20
1989 20 4.6 5 ODS A: 30 mmol L\1 NaAc 338/266 MPA/ F: protein 38/60
cont. 0.5% THF (pH 7.2); F FMOC 0.02}0.05 hydrolysates
ACN/0.1 mol L\1 NaAc" 230/455 (43C) Plasma AAs
(4:1, v/v); (423C) Plasma 266/310
AAs, A: 60 mmol L\1
NaAc cont. 0.6% THF
(pH 8.0);
B: ACN/0.1 mol L\1 NaAc/
CH3OH"(14:4:1, v/v);
(433C)
2006 III / AMINO ACIDS / Liquid Chromatography

Table 2 Continued

Author Column size Type Eluents (elution Detector Reagent Analyte RSD Matrix No. of
and temperature 3C) UV (nm) (3C) (pmol L\1) % AAs/elution
date cm ; mm, m FEx/Em time (min)

BartoH k, 10 4 3 Hypersil A: 18 mmol L\1 NaAc F OPA/ 50 (1.1 Plant AAs 21/8
1994 ODS (pH 7.2)# 0.02%(v/v) 340/450 MPA/
TEA#0.3% THF (v/v) 264/313 FMOC
B: ACN/CH3OH/NaAc (43C)
0.1 mol L\1 (pH 7.2)"
(2:2:1, v/v); (403C)

Indications as in Table 1.

performed with OPA derivatives. The essential short-
age of an OPA/SH-group reagent (reactive toward the
primary AAs only) was eliminated by Shuster’s prin-
ciple } the automatic two-step pre-column derivatiz-
ation method applying the OPA/MPA/Suorenyl-
methyl chloroformate (FMOC) reagent, which also
ensures derivatization of the secondary AAs. A high
speed elution of OPA/MPA/FMOC derivatives was
shown recently (Table 2, Figure 4: 19 compounds/
8 min). Evaluating the improvements between the
corresponding early and recent procedures, in the
newer methods shorter, thermostated columns of
smaller particle size with autosamplers are now used,
giving greater sensitivity and reproducibility.
Phenylthiocarbamyl (PTC), FMOC, 1-N,N -
Dimethylaminonapthalene-5-sulfonyl (DANS)
and Dimethylaminoazobenzenesulfonyl (DABS)
Derivatives (Table 3, Figures 1, 5 and 6)
Judging by the number of publications in the last
decade, the interest in the PTC and FMOC deriva-
tives has proved to be lasting, while the application of
DANS and DABS derivatives is decreasing. However,
in the direct enantiomer separation of AAs, the use of
DANS derivatives is preferred.
The reaction of AAs with phenylisothiocyanate
(Table 3, Figures 1 and 5), in water-free media at
ambient temperature is quantitative and fast (10 min),
resulting in the highly stable single PTC derivatives
(except for cyst(e)ines in hydrolysates which elute in
one to four peaks). The excess reagent is removed by
vacuum, and the PTC derivatives can be stored in the
freezer for an unlimited time, and for a day after
dissolution in buffer at 43C. UV detection at 254 nm
allows their quantitation in the low pmol range. The
short PicoTag and the short TSK gel columns can
separate 17 AAs within 12 min and 4.5 min, re-
spectively.
The Rrst LC separation of the strongly Suor-
escent DANS AAs has been used earlier in protein
chemistry and in thin-layer chromatography. The de-
creased popularity of this technique in LC can be

Figure 4 High speed RP-HPLC analysis of the OPA/MPA/9-fluorenylmethyl chloroformate derivatives. (Reproduced with permission
from BartoH k T et al. (1994) Journal of Liquid Chromatography 17: 4391.)
 III / AMINO ACIDS / Liquid Chromatography 2007

Table 3 Advances in the LC of pre-column derivatized AAs, obtained with phenylisothiocyanate (PITC), 5-dimethylaminonaphtalene-
1-sulfony1-CI (DANS), 4-dimethylaminoazobenzene-4-sulfonyl-CI (DABS) or with 9-fluorenylmethyl chloroformate (FMOC)

Author Column size Type Eluents (elution Detector Reagent Analyte RSD Matrix No. of
and temperature 3C) UV (nm), (3C) (pmol L\1) % AAs/elution
date cm;mm, m FEx/Em time (min)

Koop, 25 4.6 5 Ultrasphere A: 70 mmol L\1 NaH2 PO4 UV PITC 6000 * AAs in 18/130
1982 ODS (adjusted to pH 6.45 with 254 (!) protein
TEA) hydrolysates
B: ACN; (273C)
Tosohaas, 10 4.6 2 TSKgel A: 50 mmol L\1 NaAc UV PITC 250 * Model 17/4.5
1995 Super-ODs (pH 6.0)/ACN"(97:3, v/v) 254 (!) study
B: 50 mmol L\1 NaAc
(pH 6.0)/ACN"(40:60,
v/v); (403C)
Shang, 15 3.9 5 PicoTag A: NaAc (pH 6.4) UV PITC 5 (1.9 AAs in kelp 17/12
1996 ODS B: ACN; A and B 254 (!)
performed
in gradient (383C)
Bayer, 50 3 10 LiChrosorb, Eluent 10 mmol\1 F DANS 0.1 * Model 17/40
1976 RP 8 Na2HPO4/CH3OH" 340/510 (amb) study
(50:20, v/v) to which
1.5 mL CH3OH/min is
added (453C)
Martins, 15 3.9 4 Nova Pak A: 30 mmol L\1 phosphate F DANS 60 * AAs in 17/35
1996 C 18 buffer (pH 7.4)# 5 mL 338/445 (403C) polypetides
CH3OH#6.5 mL THF
adjusted to 100 mL with
distilled water
B: CH3OH/H2O"(70/30,
v/v); (253C)
Chang, * * 5 * A: 25 mmol L\1 NaAc UV DABS 5 * AAs in 17/40
1983 (pH 6.5) containing 4% 436 (703C) protein
dimethylformamide hydrolysates
B: ACN (403C)
Yang, 15 4.6 5 Hypersil A: 25 mmol L\1 NaAc UV DABS 50 * AAs in 17/40
1993 ODS (pH 6.35) containing 4% 436, (703C) polypetides
dimethylformamide 580
B: ACN (403C)
Einarsson, 50 4.6 3 Spherisorb Eluent: 20 mmol L\1 F FMOC (!) (6.6 AAs in 17/10 and
1983 500 2.26 5 ODS-2 NaAC buffer 265/315 (!) protein 33/100
(pH 4.08I4.31)/ACN hydrolysates,
gradient; (!) in urine

Qu, 15 4.6 5 Hypersil A: 30 mmol L\1 phosphate F FMOC 125 (1.0 AAs in 15/35
1996 ODS buffer (pH 6.5) in 15% 270/316 (!) protein
CH3OH (v/v) hydrolysates,
B: 15% CH3OH (v/v) biological
C: 90% ACN (v/v); samples
(383C)
Bank, 15 4.6 5 Micropak A: 20 mmol L\1 citric acid/ F FMOC 50 (3.6 AAs in 21/35
1996 ODS-80TM NaAc buffer (pH 2.85); 254/630 (!) protein
B: 20 mmol L\1 NaAc hydrolysates
(pH 4.5)/CH3OH"(80:
20, v/v); A, and B both,
cont. 0.01% (w/v) NaN3#
5 mmol L\1 (CH3)4 NCI;
C: ACN; (403C)

Indications as in Table 1.
2008 III / AMINO ACIDS / Liquid Chromatography

Figure 5 Separation of 27 phenylthiocarbamyl AAs. Column 150#(20 guard);4 mm, C18 Hypersil 5
m, temperature, 503C, eluent
A: 0.05 mol L\1. NaAc pH 7.2; B: A eluent/acetonitrile/methanol"46/44/10 (pH"7.2), flow rate: 2.1 mL min\1. Peaks:
1, aspartic, 2, glutamic acids; 3, hydroxyproline; 4, serine; 5, glycine; 6, asparagine; 7,
-alanine; 8, glutamine; 9, homoserine;
10, -aminobutyric acid (GABA); 11, histidine; 12, threonine; 13, alanine; 14, 1-amino-1-cyclopropane carboxylic acid (ACPCA);
15, arginine; 16, proline; 17, homoarginine; 18, tyrosine; 19, valine; 20, methionine; 21, cyst(e)ine; 22, isoleucine; 23, n-leucine;
24, phenylalanine; 25, tryptophan; 26, ornithine; 27, lysine. Hsystem peaks. (Reproduced with permission from Vasanits A and
MolnaH r-Perl (1998) Journal of Choromatography 832:109.)

explained by its two main disadvantages: long reac-
tion times, or elevated temperatures for derivatiz-
ation, and generation of Suorescent side products
(DANS hydroxide, DANS amide) and interference
from excess reagent. The disturbing effect of these
compounds cannot be completely eliminated and
they elute between the AA derivatives. No signiRcant
improvement has been obtained and cannot be
expected.
DABS AAs were Rrst separated applying pre-
column labelling. Derivatization was performed in
Na2CO3/NaHCO3 buffer, at pH&8.9 with DABS
chloride dissolved in acetone under continuous
stirring at 803C for 10 min. DABS AAs can be
stored for 4 weeks in solution at 253C, without any
changes. In spite of the unique stability of DABS AAs
in aqueous media, and the improvement in their
chromatographic conditions, the use of DABS AAs
is dwindling.
FMOC was introduced in 1983, as a Suorescent
labelling agent, reacting rapidly with both primary
and secondary AAs, under mild conditions (borate
buffer, pH 7.7}8.0) to give stable derivatives. The
excess reagent is extracted by pentane. Recent de-
rivatization studies have shown that, depending on
the time (2 and 40 min) and pH (8.0 and 11.4),
considerable differences can be found. At pH &8,
acidic AAs manifest low responses, and slow reaction

Figure 6 (for details see Table 3) HPLC of AAs derivatized with 9-fluorenylmethyl chloroformate (FMOC). Peaks labelled with
one-letter abbreviations for protein AAs, as well as: Hyp, hydroxyproline; R1, FMOC-hydroxylamine; R2, FMOC-hydroxyde; R3,
reagent peak present in blank derivatization. (Reproduced with permission from Qu K et al. (1996) Journal of Chromatography 723:
219.)
 III / AMINO ACIDS / Liquid Chromatography 2009

Table 4 Advances in the chiral separation of amino acids by LC: applying chiral mobile-phase additives (CMA), chiral stationary-
phase columns (CSP) and chiral derivatization reagents (CDR), such as OPA/ NACa and OPA/NIBCb

Author Column size Type Eluents (elution Detector Reagent Analyte RSD Matrix No. of
and temperature 3C): UV (nm) (3C) (pmol L\1) % AAs/elution
date cm;mm , m chiral recognition method FEx/Em time (min)

Takeuchi, 15 0.35 5 Develosil A: 40 mmol L\1 AmmAc F DANS (!) (!) Model 1 pair/30
1992 ODS-5 # 27 mmol L\1 -CD/ 315/539 (!) study
ACN" (3:1, v/v)
B. AmmAc/ACN"(72:28,
v/v); (253C) CMA

Marchelli, 10 8 5 Radialpak Isocratic: ACN/ F DANS (!) (!) Model 3 pairs/130

1996 15 4 5 C18 30 mmol L\1 NaAc 330/560 (!) study
(pH 7.0), containing
0.5 mmol L\1 N2-S-2-
hydroxypropyl-S-phenyl-
alaninamide#5 mmol L\1
Cu(II) Ac"(2:8, v/v);
(21.53C) CMA

Galli, 15 4 5 LiChrosorb Isocratic: ACN/ UV DANS (!) (!) Model 3 pair/30

1994 modified Si 100c 10 mmol L\1 NaAc 254 DABS study 4 pair/30
(pH 7.52), containing (!)
25 mmol L\1 Cu(III)
Ac"(7:3, v/v); (603C)
CSP

Iida 15 6 5 Home A: 100 mmol L\1 AmmAc UV PTC 1000 (!) Protein 18 pairs#
1997 maded (pH 6.5), 254 (!) sequencing 1 single/
B: 100 mmol L\1 AmmAc 150
(pH 65)/CH3OH"
50:50 (v/v); A and B
both contain 1 mmol L\1
butanesulfonate; (20I303C)
CSP

Nimura, 20 6 5 Develosil A: 50 mmol L\1 NaAc F OPA/ 5000 (2.3 D- and 14 pairs/70
1986 ODS-5 B: ACN (253C) CDRc 360/405 NAC L-AAs
(!) in protein
hydrolysates

BruK ckner, 25# 4 5 Hypersil A: 23 mmol L\1 Na F OPA/ 1I1000 (2 D- and 17 pairs#
1995 guard 2.1 ODS acetate (pH 5.95) 230/445 IBLC L-AAs in 5 single/70
2 B: ACN/CH3OH" (IBDC) food
(60:5, v/v); (253C) CDRd (!) hydrolysates

Indications as in Tables, as well as: CD cyclodextrin; aN-acetyryl-L-cysteine; bIBL (D) C, isobutyryl-L(D)-cysteine; c [(S )- and (R )-phenylalanine-
amide were covalently bonded to LiChrosorb Si100 silica gel; home maded silica support treated with PITC#
-CD; DANS, dansyl; DABS,
dabsyl; PTC, phenylthiocarbamyl.

is experienced; histidine and tyrosine give their mono-
and disubstituted derivatives in varying ratios. With
longer reaction times, the amount of disubstituted
histidine decreases and that of tyrosine increases,
together with interfering hydrolysis products of the
reagent. At pH&11.4 faster reaction and less inter-
fering hydrolysis products are found. After 40 min
reaction time, the monosubtituted histidine and the
disubstituted tyrosine are formed in quantitative
yield. Also 30% less hydrolysis product is obtained,
favouring the resolution of the neighbouring alanine.
The separation of the FMOC derivatives and
the presence of interfering substances are shown in
Figure 6.
Chiral Separations (Table 4 and
Figure 7)
The knowledge of the distribution of AA enatiomers
in different matrices, and/or the enantiomeric purity
of AAs, is of primary importance in the quality con-
trol of peptide syntheses for pharmaceuticals/
medicines, as well as in various plant products and in
high AA-containing foods, including baby formulas.
2010 III / AMINO ACIDS / Liquid Chromatography

For separation and quantitation of enantiomers, phenylalanineamide as CMA is very time-consuming

HPLC is the method of choice. The application of the
three main approaches for enantiomer separation is
shown in Table 4 and Figure 7, including direct,
chiral mobile-phase additive (CMA) and chiral sta-
tionary phase (CSP) and indirect methods (chiral de-
rivatization reagent CDR).
Direct Methods
Applying either -cyclodextrin (-CD) or Cu(II)
salts together with N2-S-(N2-R-)2-hydroxypropyl-S-
(Figure 7A): the separation of three AA pairs re-
quires more than 2 h. Thus, CMAs can be regarded
as an inferior approach in the chiral recognition of
AA enantiomers, due to the need for a continuous
supply of the often expensive CMA and to the dis-
advantageous chromatographic conditions. Bonded
N 2-S-(N 2-R-)2-hydroxypropyl-S-phenylalanine-
amide, CSP, allows the comparison of the CMA
and CSP protocols for the same enantiomer separ-
ations. The CSP method resolved four pairs of

Figure 7 LC separation of AA enantiomers. (A) Enantiomeric separation of a mixture of three dansyl AAs. (Reproduced with
permission from Marcelli R et al. (1996) Chirality 8: 452.) (B) Separation of 37 phenylthiocarbamyl AAs. (Reproduced with permission
from Iida T et al. (1997) Analytical Chemistry 69: 4463.) (C) Aminogram of fir honey derivatized with (a) OPA/IBLC and (b) OPA/IBDC.
(Reproduced with permission from BruK ckner H et al. (1995) Journal of Chromatography 697: 229.)
 III / AMINO ACIDS / Liquid Chromatography 2011

Figure 7 Continued

dansylated AAs within 30 min, attesting to the Indirect Methods

superiority of CSP over CMA. Recently, the elut-
ion of the PTC AAs on a new CSP (Figure 7B) permit- Spectacular results have been achieved with the separ-
ted the partial separation of 18 AA pairs within ation of AAs derivatized by CDRs (Figure 7C).
150 min. Performing the separation with both OPA/N-L(D)-
2012 III / AMINO ACIDS / Thin-Layer (Planar) Chromatography
acetyl-cysteinyl and with OPA/N-L(D)-isobutyrylcys-
teinyl AA derivatives gave excellent resolution of en-
antiomers. Consequently, the CDR technique is the
primary importance in a number of practical applica-
tions of the separation of enantiomeric AAs. The
interaction of AAs with the enantiomerically pure
reagents takes place at ambient temperature, without
racemization, resulting in the formation of stable
diastereomer derivatives.
Online LC-MS
In the case of AAs, thermospray ionization has been
displaced by the milder techniques of electrospray
(ES) and atmospheric pressure chemical ionization
(APCI), converting analyte molecules without frag-
mentation into ions. The analyte should contain the
AAs in a stable form: either in the free condition or in
the form of stable derivatives, such as phenyl-
thiohydantoins (PTH) or PTCs. SigniRcantly reduced
Sow rates are essential (100}300 nL min\1) for stable
ES and APCI operation. In automated Edman micro-
sequencing, the ES-MS of PTH derivatives. The pro-
tonated molecules were measured with a linear re-
sponse in the 50}1000 fmol level.
Future Trends
Efforts are needed to extend the life time, plate
number and reproducibility of columns, and to
standardize testing methods. The extended use of
thermostated columns is desirable in order to obtain
reproducibility in absolute and relative retention
times. LC-MS will be more widely used in laborator-
ies as the cost of these instruments falls to the level of
GC-MS, and/or an all-purpose interface becomes
available.
Thin-Layer (Planar) Chromatography
R. Bhushan, University of Roorkee,
Roorkee, India
J. Martens, Universitat Oldenburg, Oldenburg,
Germany
Copyright ^ 2000 Academic Press
Introduction
Thin-layer chromatography (TLC) is a simple and
inexpensive technique permitting a number of sam-
ples to be handled simultaneously, thus yielding
a higher precision than sequential analysis. The inert
character of the thin-layer material makes it ideally
See also: II/Chromatography: Liquid: Derivatization;
Mechanisms: Reversed Phase.
Further Reading
Blau K and Halket J (eds) (1993) Handboook of Deriva-
tives for Chromatography. Chichester: John Wiley.
BruK ckner H, Langer M, LuK pke M, Westhauser T and Godel
H (1995) Liquid chromatographic determination of
amino acid enatiomers by derivatization with o-phthal-
dialdehyde and chiral thiols. Journal of Chromatogra-
phy 697: 229.
Deyl Z, Hyanek J and Horakova M (1986) ProRling of
amino acids in body Suids and tissues by means of liquid
chromatography. Journal of Chromatography 379: 177.
Grunau JA and Swiader JM (1992) Chromatography of 99
amino acids and other ninhydrin reactive compounds in
the Pickering lithium gradient system. Journal of
Chromatography 594: 165.
McClung G and Frankenberger WT Jr (1988) Comparison
of reversed-phase high performance liquid chromato-
graphic methods for precolumn-derivatized amino
acids. Journal of Liquid Chromatography 11: 613.
MolnaH r-Perl I (1998) Amino acids. In: Deyl Z, Tagliaro
F and Teserova E (eds) Advanced Chromatographic and
Electromigration Methods in BioSciences. Amsterdam:
Elsevier.
Snyder LR, Kirkland JJ and Glajch JL (1997) Practical
HPLC Method Development. New York: Wiley Inter-
science.
Spackman DH, Stein WH and Moore S (1958) Automatic
recording apparatus for use in the chromatography of
amino acids. Analytical Chemistry 30: 1190.
Zhou J, Hefta S and Lee TD (1997) High sensitivity analy-
sis of phenylthiohydantoin amino acid derivatives
by electrospray mass spectrometry. Journal of the
American Chemical Society of Mass Spectrometry 8:
1165.
suitable for use with strong corrosive reagents and
one can perform many kinds of chemical reactions on
the plate, both from the points of view of detecting
and locating the spot and of achieving improved
separation. Certain groups of interest can be chemic-
ally bonded to the reactive groups of support mater-
ial, e.g. silanization for reversed-phase studies. Im-
pregnation of the adsorbent with a variety of reagents
adds an additional feature for inSuencing the adsorp-
tion characteristics without covalently affecting the
inert character of the adsorbent. TLC is also success-
ful in providing direct resolution of enantiomers of
a variety of compounds by the proper manipulat-
ion of the support material. The analysis of amino