You are on page 1of 14

992 Electrophoresis 2006, 27, 992–1005

1*
Abolghasem Jouyban Review
Ernst Kenndler2

1
Faculty of Pharmacy, Theoretical and empirical approaches to
Tabriz University of Medical
Sciences, express the mobility of small ions in capillary
Tabriz, Iran
2
Department for Analytical electrophoresis
Chemistry,
University of Vienna,
Vienna, Austria A discussion is given about the concepts of the ion mobility, the analyte property which
governs migration and thus separation selectivity in CE. It deals with small organic and
inorganic ions, not with charged polymers or large particles like colloids. The discus-
Received September 17, 2005 sion is directed to two main concepts. (i) The first is based on physico-chemistry of ion
Revised November 4, 2005
conductance in solution, and distinguishes three types of mobility. The absolute mo-
Accepted November 7, 2005
bility is the limiting mobility at zero ionic strength; it depends on the solvent and the
temperature. It is obtained by extrapolation of the actual mobilities, those of the fully
charged particles at finite ion concentration. The observed reduction of the absolute
mobility with ionic concentration is related to an ion cloud, and is formulated by the
established theories of ion conductance. It explains the actual mobility as function of
(beside other factors) the ionic strength, the viscosity and relative permittivity of the
solvent, the temperature, the relaxation time of solvent polarisation and the distance of
closest approach between ion and counterion. The effective mobility, finally, is the
mobility when association and dissociation equilibria play a role. Most important are
acid–base reactions, but complexation, ion pairing and homo- and heteroconjugation
were considered as well. (ii) The second approach treats mobility data with different
mathematical methods, and formulates their dependence on variables like solvent
composition with appropriate algorithms. These empirical methods mainly include
least squares and neural network-based methods. The least square methods ranges
from the simplest model, which uses only the molecular weight of the analyte, to more
complicated model requiring three-dimensional structural descriptors of the solutes.
Neural networks have been applied to model the mobility using different input variables
and various architectures. Work comparing the accuracy of least squares and neural
network methods was discussed; the results showed that the neural network method
leads to a more accurate mobility calculation. However, the least squares methods
could give some information to the factors affecting the mobility of the analytes. The
resulting methods allow the prediction of mobilities under different experimental con-
ditions with certain accuracy. It has been shown that using such models, it is possible
to predict mobility of analytes after training the models by a minimum number of data
to speed up the method development stage.

Keywords: Capillary zone electrophoresis / Ionic strength dependence / Mobility /


Organic solvent / Review DOI 10.1002/elps.200500696

Correspondence: Professor Ernst Kenndler, Department for Analyti-


1 Introduction: The mobility of ions
cal Chemistry, Faculty of Chemistry, University of Vienna, Währinger-
strasse 38, A-1090 Vienna, Austria It is a prerequisite for separation in CZE that the analyte
E-mail: ernst.kenndler@univie.ac.at
zones migrate with different velocity, in other words, that
Fax: 143-1-4277-9523
certain separation selectivity is achieved. Whether or not
Abbreviations: ANN, artificial neural network; CODESSA, compre- sufficient resolution of the sample zones is finally reached
hensive descriptors for structural and statistical analysis; DHO,
Debye-Hückel-Onsager; EtOH, ethanol; FA, formamide; MeCN, ace- * Second corresponding author: Dr. Abolghasem Jouyban,
tonitrile; MeOH, methanol; MLR, multiple linear regression; PC, pro- Faculty of Pharmacy, Tabriz University of Medical Sciences,
pylene carbonate; 1-PrOH, 1-propanol; RBFNN, radial basis func- Tabriz 51664, Iran
tion neural network E-mail: ajouyban@hotmail.com

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 993

depends then on the extent of peak broadening that takes electrophoretic mobility in CE with monosolvent and
place during migration (and on the extra-column peak multisolvent BGEs are discussed. We limit the review on
effects). Peak dispersion, expressed by the plate height small inorganic and organic ions, which means that we
or the plate number, and determining the separation effi- will not consider polymers or colloids. However, before
ciency, is not the topic of this review. This subject was discussing such methods we must distinguish different
discussed in previous reviews [1–5]. The present work is mobilities which describe ion migration, depending on the
directed to the mobility of the analyte, which govern the chemical environment of the ion, and on its own ionization
migration velocity in the electrophoretic system. The properties. These are the absolute, the actual and the
migration velocity, ui, of an ion, i, is directly proportional to effective mobilities. The absolute mobility is the limiting
the applied electric field strength, E (Eq. 1a); the factor of mobility at zero ionic strength, whereas the actual mobility
proportionality, mi, is named the mobility (Eq. 1b) is that of a fully charged ion at finite concentration. The
effective mobility takes into account that the ions are not
(a) ui = miE (b) mi = ui / E (1) fully ionized. It has to be mentioned that if there is an EOF
The mobility has the dimension m2V21s21. It is a single ion present in the electrophoretic system, the overall, or total,
property, which depends on external variables like the or apparent mobility is relevant for separation. However,
temperature and the solvent, and on ion-specific proper- this is simply the sum of the individual ionic mobility and
ties like the particle size and shape, and its charge; it is that of the EOF. The mobility of the EOF will not be dis-
dependent on the ionic strength of the solution as well cussed in the present review. Note that we consider the
(see e.g. [6]). In classical chemistry of ions in solution the mobilities as signed quantities; those of cations have
mobility is a parameter derived from the single ion con- positive and of anions negative signs.
ductance, li, by

m1 = li / ziF (2)
2 Models for ion mobility
where F is the Faraday constant, and zi is the valency of
the ion. li/zi is the single ion equivalent conductance; it is 2.1 Absolute mobility
derived from the electrolyte conductance and the trans-
ference numbers of the ions. With the increased distribu- The absolute electrophoretic mobility, m0,i, is the average
tion of CZE mobilities are measured from the migration velocity of a charged species, i, per unit electric field
time, tm,i, the migration distance, Ld and the field strength strength at zero ionic strength. It is obvious that it cannot
according to mi = Ld / tm,iE. It should, however, not to be be measured directly; it can be obtained by extrapolation
overlooked that mobilities or proportional quantities have from single ion conductances or actual mobilities at dif-
been measured also with other CE methods like ITP, ferent ionic strengths. In the most simplified model the
decades before CZE was commercially available. In the medium is considered as a continuum with a certain
textbook published by Everaerts, Verheggen and Beckers macroscopic dynamic viscosity, Z, and the ion as a rigid
[7] in 1976, the concept of mobility in CE was discussed in sphere with the hydrodynamic or Stokes radius, rSt,i, on
detail, and mobilities in different solvents were tabulated. which the electric force is acting (see e.g. [6, 12]). The ion
Kiso, Hirokawa et al. [8, 9], e.g. presented mobility-related is retained by hydrodynamic friction expressed by
values in the 1980s for nearly 300 anions in water, and of Stokes’ law, whereby the friction coefficient, fhydr,i is
60 anions in MeOH (methanol); in mixed solvents with
fhydr,i = 6pZrSt,i (3)
MeOH and DMSO as constituents mobilities were derived
from isotachophoretic step heights [10, 11]. for ‘sticking’ conditions, that is when the solvent wets the
sphere. Due to the balance between the electric force,
It is obvious from Eq. (1) that the analyte migration behav-
zieE, and the frictional force, 6pZrSt,iui, the resulting
iour in CZE can be forecasted given that its mobility is
absolute mobility (at infinite dilution or zero ionic strength,
known. Therefore, it would be an enormous advantage if
when no counterions are present) is then
methods exist by which the mobilities could be predicted.
In this case any structured design to develop a CE method zi e
m0;i ¼ (4)
for analysing a certain analyte and/or set of analytes would 6pZr St;i
speed up the process at the method development stage.
where e is the elementary charge. The hydrodynamic
This contribution gives an overview about the theoretical radius differs in many cases from the crystal radius; this is
models of mobilities in connection with the theory of ion because in solution the ion exists in a solvated form, and
conductance in solutions on one hand. On the other hand, the solvation shell contributes to its size. Equation (4)
the published mathematical models representing the would allow forecasting the mobility in different solvents

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


994 A. Jouyban and E. Kenndler Electrophoresis 2006, 27, 992–1005

with different dynamic viscosity. It could also express the In another additional modification it is tried to consider the
change of mobility with temperature, because viscosity is deviation of the ions from spherical shapes, and the cor-
a temperature-dependent variable. rection factor, f/f0 [20] is introduced. f is the frictional force
on an ion of hydrodynamic radius rSt, and (f/f0) is the ‘fric-
It is clear that this approach is an oversimplification due to tional ratio’ to correct for the asymmetry of the ion. The
the following reasons: (i) the solvent is considered as modified Eq. (4) reads
continuum, but the solvent molecules have about the
same size as the ions. (ii) The ions are taken as spheres, zi e
m0;i ¼ (6)
but most ions are not spherical. (iii) The hydrodynamic 5pZr St;i ðf=f 0 Þ
radius is considered as constant, but different solvents It should be mentioned that for Z normally the macro-
can solvate the ions to a different extent, thus leading to scopic dynamic solvent viscosity is taken. However, in
varying hydrodynamic radii; moreover, even for the same solutions with polymeric additives it is rather the micro-
solvation number the radii can differ due to the different scopic viscosity that plays the relevant role. The former
size of the solvent molecules. (iv) Only hydrodynamic parameter is that determined in the usual way, e.g., with
friction is considered to retain the moving ion, but dielec- viscometers. On the molecular scale, on the other hand,
tric friction might play a role as well. Equation (4) leads to the microscopic viscosity can have values even for solu-
the so-called Walden’s rule, stating that the product of tion with high macroscopic viscosity, which do not differ
absolute mobility and solvent viscosity is constant and, from those for the pure, polymer-free solvent. This was
e.g., independent of solvent and temperature shown by CE in [21] not only for mobilities but also for
diffusion coefficients. In extreme cases the macroscopic
m0,iZ = const. (5) viscosity of an aqueous solution of a long-chain PEG (with
average molecular mass 26106) was nearly ten times
Due to the oversimplifications, the rule is often not well
larger than a polymer-free aqueous buffer solution, but
obeyed; especially water as solvent gives normally very
the mobility (and the diffusion coefficient) of the reference
different products for a certain ion when related to organic
ion remained approximately constant and independent of
solvents. The reason is that water is an excellent solvator
the viscosity.
for anions and cations as well, in contrast to many organic
solvents. Indeed, the more pronounced the solvation An extended model not only considers the hydrodynamic
ability of a solvent, the less pronounced Walden’s rule is friction but takes into account the so-called dielectric
followed when comparison is carried out with other sol- friction as well. This contribution occurs because in each
vents. Constancy of the product in water is observed, on infinitesimal volume element of the solution polarisation of
the other hand, for a particular ion at varying temperature, the solvent is established, the results of the orientation of
T; note that viscosity is a nearly linear function of T over a the solvent molecules under the influence of the electric
not too wide temperature range. However, for organic field of the ion. This polarization is characterised by the
solvents, particularly those with low solvation ability, the dielectric relaxation time, t; this is the time after which the
product deviates from constancy in several cases sur- solvent molecules are randomly oriented again after the
prisingly weakly, especially for large ions with low charge passage of the ion. An electrostatic drag force acts on the
density. To give an example: the absolute mobilities of an moving ion within the relaxation time, leading to retarda-
individual ion differ by 700% in so different solvents like tion additional to the hydrodynamic friction. According to
acetonitrile (MeCN) and propylene carbonate (PC), but the theory of Hubbard-Onsager [22–24] (modifying the
the products stray by only several ten percent [13, 14]. approach introduced by Fuoss [25], Boyd [26, 27] and
Modifying the exponent of the viscosity from 1 to 0.9 led Zwanzig [28, 29], see. e.g. [30]) the coefficient of dielectric
to a variation of the average Walden products of only friction for sticking condition, when the solvent wets the
610% for 17 aromatic acids in mixed aqueous–organic ion, is
solvents [15]. The mixed solvents consisted of aqueous
phosphate buffer with MeOH, ethanol (EtOH), 1-propanol tz2 e2 e  e1 
f diel;i ¼ 0:0607 i 3 (7)
(1-PrOH) or MeCN as organic constituents at concentra- ri e2
tions up to 60% [15–18]. The actual mobilities at 20 mM e is the absolute static permittivity of the medium e = er.e0,
ionic strength were considered (see below). er is the relative permittivity, e0 is the permittivity of
vacuum (8.854610212 F/m), e? is the permittivity at high
An approach that takes into account that the solvent (still frequency. The resulting absolute mobility is then
considered as continuum) does not wet the spherical ions
(slipping conditions) expresses the friction factor by zi e
m0;i ¼ (8)
fhydr,i =4pZrSt,i instead of fhydr,i =6pZrSt,i, and thus leads to the tz2i e2
6pZr i þ 0:0607
modification of Eq. (4) by replacing factor 6 by factor 4 [19]. r 3i e

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 995

It can be seen that it depends two-fold on the solvent: via where M is the molecular mass, and H is the mean of
the dynamic viscosity and the relaxation time. Lucy and waters of hydration of the ion (the sum of the waters
co-workers [31–38] have demonstrated in a series of hydrating the functional groups of the analyte calculated
striking papers dealing with MeOH and MeCN that the using McGowan’s additive method). S is a shape factor,
dielectric friction indeed plays a role especially for higher which is a function of the number of carbon atoms in the
charged ions. For those with low charge (e.g. zi = 1) the longest chain, the number of side chains on the amines (0
effect is, on the other hand, negligible. for primary amines, 1, 2 and 3 for secondary, tertiary and
quaternary amines, respectively) and the average number
Roy and Lucy [37] rearranged Eq. (8) and used of carbon atoms in the side chains. The percentage error
1 6pr St;i tzi e of Eq. (10) for representing the mobility of monoamines
¼ þ 0:0607 3 (9)
Zm0;i zi e r i Ze was 3.7%.
for correlating the anions mobility with charges of 21 to
C0
24 in water–MeOH mixed-solvent buffers and reported m0;i ¼ (11)
R2 values varying from 0.079 to 0.999. Plots of 1/Zm0,i V C1 þ C2 pK
versus t/Ze given in Fig. 1 indeed show that the slope of
was also proposed by Lucy and co-workers [32] to cal-
the fitted lines are dependent on the charge of the ion. It is
culate the absolute mobility of analytes using size and
increasing with increasing charge number of the analyte.
pKa (or pKb) of the analytes. V is the van der Waals mo-
This justifies the superiority of dielectric friction for higher
lecular volume, and C0–C2 are the model constants. The
charged ions in comparison with ions with unit charge.
numerical values of C0 varies between 6.861023 and
Following earlier contributions [39–41], Kiso and Hiro- 7.861023 for acids and bases, C1 is nearly constant for
kawa [42] showed that the absolute mobility can be acid and bases (0.62) and C2 is ,0.66. V0.62 could be
expressed as function of the ion mass. They found that considered as effective surface area as it is confirmed in
the absolute mobilities were inversely proportional to the Offord’s model (which expresses the actual mobility as
square roots of formula weights for a number of n-alkyl- function of the molecule mass, M, see below). The model
ammonium, mono- and dicarboxylate ions analytes. of Lucy and co-workers [32] combines the hydrodynamic
Another empirical attempt was used [31] to compute the and dielectric frictions on the electrophoretic mobility
mobility of analytes taking their size and shape into and has been derived based on Born’s model. Its accu-
account, and used the following equation to test the racy has been checked using acidic and basic analytes
concept on the mobility of monoamines and produced errors less than 5% [32]. The model
should be trained for acidic and basic analytes sepa-
6:3  103
m0;i ¼ (10) rately. In a multilayer neural network using an extended
0:620
ðM þ 0:22H  0:175SÞ delta-bar-delta algorithm proposed by Li and co-workers
[43] employing molecular weight, molecular volume, a
code for the charge of the ion (21 for anions and 11 for
cations) and the pKa of the analytes as input variables
using the absolute mobility of 56 mono-charged analytes
collected from the literature [32] more accurate results
have been reported [43]. The authors divided 56 mobility
data into three subsets, namely training, validation and
test sets. The artificial neural network (ANN) model has
been trained by 40 mobility data of the training set and
the trained model was validated using ten validation data
sets. Then, six data points of the test set have been
predicted using the trained and validated model. The
percentage error for training, validation and test sets
were 2.2, 3.7 and 2.8%, respectively, where the overall
percentage error was 2.5% which is significantly less
than 5% of Eq. (11). This equation was, by the way,
Figure 1. Correlation between change in ion mobility and
derived from theory and its constants have physical
the solvent properties t, e and Z for H2O/MeCN media,
according to the rearranged Hubbard-Onsager equation meaning, whereas the ANN method uses weights with-
(Eq. 9). Solutes: benzenesulfonate, 2,6-naphthalenedi- out any physical meaning. It should be mentioned that
sulfonate, 1,3,6-naphthalenetrisulfonate. From [37] with many equations are unit dependent. In the present
permission. review we do not give the appropriate units for the

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


996 A. Jouyban and E. Kenndler Electrophoresis 2006, 27, 992–1005

quantities the equations consist of. These details are jzc za j l0;c þ l0;a
q¼ (13)
certainly indispensable for book chapters or textbooks. jzc j þ jza j jza jl0;c þ jzc jl0;a
Here the readers are referred to the original papers.
or, written for mobilities instead of single ion limiting con-
ductances, l0
In a modelling work on absolute mobility data, Xue and
co-workers [44] used a radial basis function neural net- jzc za j m0;c þ m0;a
q¼ (14)
work (RBFNN) and seven descriptors of analytes studied, jzc j þ jza j jza jm0;c þ jzc jm0;a
which were calculated from the chemical structures of the
q has the value of 1/2 for symmetrical electrolytes
analytes using comprehensive descriptors for structural
(jzc j ¼ jza j). The actual mobility for monovalent ions can
and statistical analysis (CODESSA) software. Their ana-
be derived from Eq. (12) by transformation of con-
lyte set include 115 carboxylic and sulphonic acids and
ductance into mobility and reads
the RBFNN method gave an overall average relative " #
deviation of 3.8% for all data points [44]. 8:204  105 4:275 pffi
mact;i ¼ m0;i  m0;i þ I (15)
ðeTÞ3=2 ZðeTÞ1=2
A simplified form was used in [45] to match mobilities as a
2.2 Actual mobility function of the ionic strength according to
pffi pffi
The actual mobility, the mobility of a fully charged ion in a mact;i ¼ m0;i  Am0;i I  B I (16)
solution of finite ionic strength, is expressed mainly by
two approaches. One approach is based on the theory of A and B are functions of relative permittivity and dynamic
conductance in solutions; in the most established theory viscosity, and depend thus on the solvent (the data for A
– developed by Onsager, based on the work of Debye and and B in many solvents are tabulatedpffiin [2]). As a con-
Hückel – it is tried to relate the deviation of the actual from sequence, the slope of the mact versus I curve is strongly
the absolute mobility to the presence of an ion cloud or solvent-dependent as well, in other words, the effect of
atmosphere formed by the counterions around the ion. the ionic strength on the mobility varies with the solvent.
The other approaches use mathematical algorithm with This can be seen from the plot in Fig. 2 (for hypothetical
empirical variables to express the mobility as function of ions with absolute mobility of 4061029 m2V21s21 in all
e.g. the solvent composition. Both approaches are dis- solvents) constructed from the data for er and Z for differ-
cussed as follows. ent solvents. Whereas the reduction in formamide (FA) is
even less pronounced than in water, it is much stronger in
the common organic solvents MeOH and MeCN. This
was indeed found [46] and its consequence for separation
2.2.1 From theory of ion conductance in
efficiency in CE clearly demonstrated in [45].
solution
Deviations from Eq. (12) are found at concentrations
It was observed since the beginning of research on ion higher than 1023 M. Attempts were made to extend the
conductance in solution that the relationship between validity of Eq. (12) by adding other concentration-
conductance and concentration of a strong electrolyte
decreases nonlinear with increasing concentration. The
theory to explain the influence of the ion concentration on
the chemical activity of an electrolyte was introduced by
Debye and Hückel and extended to ion conductance by
Onsager (see e.g. [12]). The according Debye-Hückel-
Onsager (DHO) limiting law of the conductance, L, for a
binary electrolyte as a function of the ionic strength, I, is
( )
2:801  106 jzc za jqL0 41:25ðjzc j þ jza jÞ pffi
L ¼ L0  pffiffiffi þ I
ðeTÞ3=2 ð1 þ qÞ ZðeTÞ1=2
(12)

Suffix 0 indicates zero ion strength, I, suffix c stands for


cations, a for anions. The ionic strength for a certain ion pffi
Figure 2. Mobility vs. I in FA, PC, water, DMF, MeOH
concentration, c, is I ¼ 1=2ðz2c cc þ z2a ca Þ or generally and MeCN according to the DHO limiting law (Eq. 15) at
when X ion species, i, are present in solution 257C. Mobilities are in 1029 m2V21s21, ionic strength in
I ¼ 1=2 z2i ci . q in Eq. (12) is mol/L.

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 997

dependent terms. One semiempirical equation for the


conductance was introduced by Hsia and Fuoss [47];
adopted to mobilities it would read

mact;i ¼ m0;i  Sc1=2 þ Eclnc þ A0 c þ B0 c3=2 (17)

Here the factors S and E describe the effect of the ion at-
mosphere on the mobility; for low ionic strengths, S
approaches the Onsager limiting slope. A’ and B’ depend
on the same parameters as S and E, and on the distance
of closest approach between ion and counterion. This
means that this and further approaches to widen the
validity of the DHO limiting law to higher concentrations
take into account that the ions are not point charges, but
have finite radii. In this case an ionic diameter, a, is intro-
duced, which is – in a more strict sense – the distance of Figure 3. Actual mobilities, mact,i, of tetramethyl-
closest approach of ion and counterion. This extended ammonium (TMA), tetraethylammonium (TEA) and tetra-
theory for electric conductance was developed among butylammonium measured as a function of the ionic
others by Falkenhagen, Pitts and their co-workers strength in nitromethane as solvent. Temperature, 257C.
[48–51]. The according equation, again transformed for Full lines are fitted curves according to Eq. (18). From [53]
the ionic mobility of a monovalent ion in a symmetrical with permission.
electrolyte, is

mact;i ¼ m0;i  Equation (21) was also proposed based on classical so-
" # pffi
8:204  105 4:275 I lution theory of electrolytes [33]
 m0;i þ pffi (18) !
ðeTÞ3=2 Zðer TÞ1=2 1 þ 50:29aðer TÞ1=2 I pffi
I
mact;i ¼ m0;i  Az pffi (21)
Also in this equation the mobility is solvent-dependent, as 1 þ 2:4 I
it contains the specific parameters er and Z. Indeed not where A is a constant. The accuracy of Eq. (21) was tes-
only for common solvents like MeOH or MeCN, but also ted on mobility data of carboxylates, phenols and sul-
for unusual solvents in CE like PC, FA, DMF, N,N-di- phonates. It has been shown that pthe
ffi Onsager slope (A in
methylacetamide or nitromethane [14, 52–54] this equation most simplified mact;i ¼ m0;i  A I) deviates from con-
is well obeyed, as can be seen from Fig. 3. Equation (18) stancy for higher zi values as demonstrated in Fig. 4 [33].
can be simplified as follows, with B, B1 and B2 being de- Note that all the models could be used to determine the
pendent on the solvent properties just mentioned. In this absolute mobility of analytes by extrapolation.
form it was used to calculate in turn the limiting mobility of
the analytes in MeOH from the actual mobilities [55] The following equation, similar to Eq. (12), was used by
pffi pffi Roy and Lucy [35] in accordance with basic physico-
mact;i ¼ m0;i  ½B1 m0;i þ B2   ð I=ð1 þ aB IÞÞ (19) chemistry
These limiting mobilities served as input data to recalcu- 82:5
late the actual mobilities by the same equation for mact;i ¼ m0;i  zi þ
Zðer TÞ1=2 F
depicting the electropherogram by computer simulation  pffi (22)
with the program PeakMaster (http:///www.natur.cuni.cz/ 1:40  106 2q I
þ jzþ z j pffiffiffi m0;i pffi
,gas) adopted for methanolic solutions [55]. ðer TÞ1=2 1þ q 1 þ Ba I

All these equations stand for mono-charged ions. For The authors investigated the effects of ionic strength on the
symmetrical electrolytes with higher valency than unity electrophoretic mobility of anions with 21 to 24 charges at
the modified equation different concentrations of MeOH in binary mixed-solvent
  pffi buffers containing 0.05 M NaOH. They assumed that the
q I
mact;i ¼ m0;i  A1 m0;i pffiffiffi þ zi A2 pffi (20) ion size parameter, a, remains unchanged in the mixed-
1þ q 1 þ 1:5 I
solvent buffer. In Eq. (22), q is the electrolyte parameter
was used [56]; A1 and A2 are constants, and q is the same equal to 0.5 for a 1:1 electrolyte. In parameter B
factor as defined in Eqs. (13), (14). For unsymmetrical sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
electrolytes it is necessary to calculate the limiting mobil- 8pNA e2
B¼ (23)
ity from mact by iteration. 1000er k B T

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


998 A. Jouyban and E. Kenndler Electrophoresis 2006, 27, 992–1005

Figure 4. Effect of the solute


charge on the Onsager pffi slope
(A in mact;i ¼ m0;i  A I) for
carboxylates (d) and sulfonates
(u) with charge number be-
tween 1 and 6. From [33] with
permission.

er is the only varying term for the water-MeOH mixture (NA 2.2.2 From empirical algorithms
and kB are the Avogadro and the Boltzmann constants,
respectively). It was shown that the ions with higher A purely empirical equation for actual mobilities as a
charges are strongly influenced by dielectric frictions in function of ionic strength and analyte charge number, zi,
these mixed-solvent buffers. was derived by curve fitting of data measured by CE for a
set of 21 organic anions (sulfonates) with charge numbers
All these equations do not take into account chemical between 1 and 6 [62]. The model was adopted from a
equilibria in which permanent ions are involved (we do not previous model successfully implemented in HPCE
consider at this stage protolysis of weak acids and bases). simulation program for CE [63]. The concentration range
These equilibria can lead to phenomena that hardly occur was up to 100 mM. The resulting equation
in aqueous solutions, but are quite frequent in organic sol- pffiffiffiffiffi
vents, especially in those with low dielectric constant: ion mact;i ¼ m0;i expð0:77 zi IÞ (25)
pairing and homo- and heteroconjugation. The modified
enabled to express the actual mobility for each ionic
Fuoss-Onsager equation for the actual mobility extends
strength with less than 5% deviation.
the Hsia-Fuoss equation by a term for ion pair formation,
the last term of the following equation (see e.g. [19]) An empirical equation was used to calculate the actual
pffiffiffiffiffiffiffiffiffi mobility of peptides in paper electrophoresis [64]
mact;i ¼ m0;i  S ðacÞ þ EðacÞlnðacÞ þ A0ðacÞþ
(24) Kzi
þ B0 ðacÞ3=2  K A mi g2 ðacÞ mact;i ¼ (26)
M2=3
where c is the initial total ion concentration, KA is the ion- where K is a constant, M is the molecular mass and M2/3
pair formation constant, a is the degree of dissociation of could be considered as a function of the effective surface
the ion-pair and g6 is the mean activity coefficient area [64]. It was successfully used for modelling the mo-
according to the Debye-Hückel limiting law. For solvents bility of small ions of vinca alkaloids in aqueous and
with high dielectric constant and for low electrolyte con- aqueous–organic buffers [65] and also the mobility of
centrations the Fuoss-Onsager equation approaches the sulphonamides [66]. The only required data are the mo-
Onsager limiting law. Barthel et al. [57–60] used this lecular weights of the analytes as size indicator. However,
equation to derive ion-pair formation constants for 1–1 there the dependence on ionic strength is missing.
electrolytes in several organic solvents (MeCN, MeOH,
EtOH, 1-PrOH, mixtures of PC and MeCN). In a very In order to overcome the limitations from the assumption
recent work, Dai et al. [61] used the deviations from of spherical geometry of the analytes, and also to intro-
Eq. (22) to derive ion-pair formation constants by CE in duce a discriminative tool between different isomers of
mixed solvents consisting of MeCN and water. the analytes, McKillop and co-workers used three-dimen-

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 999

sional structural features of the analytes, which has been SA is the surface area. The superiority of ANN on MLR
ignored above. It is a fact that a number of isomeric models has been shown [69]. As an example, the reported
compounds having similar shape and identical charges MLR model for the cationic form of sulfonamides is
could be separated in CE and the models representing
the mobility should cover these observations as well. mact;i ¼ 1:091 þ 0:001DHf þ 1:335pK  0:002SA (30)
McKillop et al. assumed that three contributing factors The intercept represents the effects of other affecting
could be considered as: (i) the overall size of the ion, (ii) variables which are not considered in the model, factor
the effective radius of the ion and (iii) the resistance of the 0.001 reflects the effect of heat of formation, factor 1.335
ion due to its natural tumbling motion. The authors pre- means that the higher the pKa values the higher is the
sented these factors: rvdw, the van der Waals radius; rrot, mobility observed and factor - 0.002 means that the wider
the rotational radius and ri, the radius of inertia to tumbling the surface area of the ions, the slower is the mobility of
distance between centre of mass and charge. The fol- the ions. However, this is not the case for the ANN mod-
lowing equation for the actual mobility was proposed els, and one cannot provide such information from ANN
 
3:22 0:570 0:0114 curve fitting parameters; they just provide better correla-
mact;i ¼ þ   0:604  103 (27)
r vdw r rot ri tions between input and output variables as shown by the
authors [69].
based on the data of a series of alkylpyridines [67]. As
evident from the coefficients of the independent vari- In another MLR model, the logarithm of the electropho-
ables, the size factor (rvdw) is the most important factor on retic mobility as dependent variable and computational
the mobility, followed by rrot and ri which are responsible descriptors calculated using HyperChem as independ-
for the observed selectivity of the positional isomers. A ent variables have been employed as
good predictability for the mobility of the analytes has
been shown and the model was successful in representing lnmact;i ¼ K 0 þ K 1 PQ þ K 2 V 2=3 þ K 3 TE þ K 4 DHf þ K 5 MR (31)
the mobility of alkylpyridines using the independent vari-
PQ is the net charge of the most positive (for cations) or
ables computed using a molecular modelling package
the most negative (for anions) atom on the analyte, V2/3 is
called SYBYL [67]. The extension to other pyridine deriva-
the surface area of the molecule with volume V, TE is the
tives gives a slight modification of the constants in Eq. (27),
total energy, a computational term calculated using
and still a high correlation between the measured and the
HyperChem software, MR is the molar refractivity and K0
predicted behaviour is demonstrated. In a work by Jalali-
to K5 are the model constants. The accuracy of Eq. (31)
Heravi and Garkani-Nejad [68], employing a data set of
was tested using five sets of various analytes where the
pyridines, the mobilities were modelled using an ANN
overall percentage errors for correlative and predictive
method with architecture 3–6-1 with the reciprocal of van
analyses were 1.4 and 5.6%, respectively [70]. The main
der Waals radius of the molecules, the principal moment
disadvantage of the mathematical equations representing
of inertia of the molecule around the x-axis and the dipole
the mobility of the analytes using computational descrip-
moment of the molecules as input variables. The ANN
tors, so-called quantitative structure–property relation-
model was trained by a back-propagation method.
ships (QSPRs), is that they calculate the mobility solely
Standard errors of training and prediction sets were 1.0
from chemical structures of the analytes. The main lim-
and 1.2%, respectively [68]. It has to be mentioned that
itation of QSPRs is that they should be trained for each
these models do not consider the ion strength depend-
analytical condition and there is no possibility to predict
ence of the actual mobility. In an earlier work of Jalali-
the mobility under various conditions.
Heravi and Garkani-Nejad’s group, an ANN model with
the architecture of 3–4-2 was proposed to correlate the From cosolvency models representing the solute solubil-
actual mobility of the analytes. The accuracy of the ANN ity in binary solvent mixtures an approach adopted to
method has been compared with that of multiple linear calculate the electrophoretic mobility of analytes in
regression (MLR) models. Their proposed MLR models mixed-solvent buffers gave
were
lnmact;i ¼ f 1 lnm1 þ f 2 lnm2 þ K 0 f 1 f 2 þ K 1 f 1 f 2 ðf 1  f 2 Þ (32)
mact;i ¼ J0 þ J1 DHf þ J2 pK þ J3 SA (28)
where f1 and f2 represent the fractions of the solvent 1
for cationic sulfonamides and and 2 in the mixed-solvent buffer, m1 the mobility of the
analyte in the mono-solvent buffer 1 and m2 the mobility of
mact;i ¼ J4 þ J5 DHf þ J6 PPCH þ J3 SA (29)
the analyte in the monosolvent buffer 2; K0–K1 are the
for anionic sulfonamides. J0–J6 are the model constants, model constants [71]. The accuracy of Eq. (32) has been
DHf is the heat of formation, PPCH is the most positive assessed using various acidic and basic analytes in
partial charge (the most positive atom of the analyte), and aqueous and nonaqueous binary solvents and compared

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


1000 A. Jouyban and E. Kenndler Electrophoresis 2006, 27, 992–1005

with other algorithms [72]. The model provides better two other numerical methods, namely RBFNN and sup-
correlations and more accurate predictions among similar port vector machine (SVM), and showed that the SVM
algorithms and is capable of predicting the electropho- provides more accurate results in comparison with MLR
retic mobility of structurally related drugs in a given binary and RBFNN methods [77].
buffer [73]. The model could also be extended to
Ab initio predictions (prediction without using any experi-

lnm ¼ f 1 lnm1 þ f 2 lnm2 þ f 3 lnm3 þ f 1 f 2 M0 þ M1 ðf 1  f 2 Þþ mental data) of the mobility of analytes using computa-
  0 0
tional methods are not available at the present time. As
þ M2 ðf 1  f 2 Þ2 þ f 1 f 3 M0 þ M1 ðf 1  f 3 Þþ discussed above, the present models possess a number
0   00 00 00  of model constants and require experimental data points
þ M2 ðf 1  f 3 Þ2 þ f 2 f 3 M0 þ M1 ðf 2  f 3 Þ þ M2 ðf 2  f 3 Þ2 þ
 000  as training set. However, it is possible to train the available
000 000
þ f 1 f 2 f 3 M 0 þ M 1 ð f 1  f 2  f 3 Þ þ M2 ð f 1  f 2  f 3 Þ 2 mathematical models using a minimum number of
(33) experimental data points and predict mobilities under the
same analytical conditions. In an earlier work [73] experi-
for correlating the mobility in ternary solvent buffers and mental mobility data of atenolol, alprenolol, labetalol and
provided accurate results using mobility of beta-blockers metoprolol in aqueous buffer under a given analytical
in water–MeOH–EtOH mixtures [74]. In Eq. (33), f3 is the condition, their molecular weight and pKb values were
fraction of the third solvent in the mixed-solvent buffer, m3 employed to train the model of Fu, Li and Lucy [32]; and
the mobility of the analyte in the mono-solvent buffer 3; the trained model was
M0–M2, M0 0-M0 2, M00 0–M00 2 and M000 0–M000 2 are the model
constants. It has been shown that by collecting experi- 394:600
mact;i ¼ (35)
mental mobility data of an analyte in the mono-solvent MW 0:639  2:935pK b
buffers and also three data points from each binary and To show the capability of this model, mobilities of three
ternary solvent buffers (total required data points are 15), beta-blocker drugs, i.e. propranolol, timolol and acebu-
and by computing the model constants of Eq. (33), one is talol, have been predicted with the overall percentage
able to predict the mobility in all solvent compositions of error of 2.8% [73]. Since a large number of chemically or
binary and ternary solvent buffers. In practice, employing biologically related compounds are synthesised and/or
such predictive models, the process of method develop- extracted from natural sources to assess their biological
ment could be facilitated [74]. Recent investigations activities and finding an analytical method has its own
showed that the ANN methods produced more accurate priority in such studies, one can estimate the importance
predictions for mobility of analytes in mixed buffers in of such trained models in pharmaceutical and chemical
comparison with Eqs. (32), (33) [75, 76]. industries [73]. It is obvious that any predictive method
Liu et al. [77] calculated more than 130 descriptors of 26 has some prediction errors. So, the question is what error
aromatic acids using CODESSA software and employed is acceptable in practice. It could be agreed that this error
five of the descriptors to model the mobility data of anions range is less than or equal to the RSD of repeated
in mixed-solvent buffers. Their proposed MLR model is experiments; this is say 1% for mobility data [31]. Under
this precondition the accuracy of prediction is too low. If
mact;i ¼ H1 TEIP  H2 MIA þ H3 Qmin  we accept the threefold of the SD as acceptable, then
prediction is within this tolerable range.
H4 HDSA1 =TMSA  H5 VOL þ H6 (34)

where TEIP is the topographic electronic index (all pairs),


MIA is the moment of inertia A, Qmin is the minimum partial 2.3 Effective mobility
charge, HDSA1 is the hydrogen acceptor-dependent
hydrogen-donor surface area, TMSA is the total molecu- The effective mobility meff of the compound consisting of n
lar surface area, VOL is the molecular volume and H1–H6 ionic forms is
the model constants. Equation (34) should be trained for P
n
each solvent mixture of the buffer. To provide a general ck mact;k zk
model for correlating mobility data of aromatic acids in meff ¼ k¼1 (36)
c
water–MeOH and water–EtOH binary mixtures, the
authors added dielectric constants of solvent mixtures where ck, mact,k and zk are concentration, actual ionic mo-
and the energy of the highest occupied molecular orbital bility and charge number of the kth ionic species,
(EHOMO) of MeOH and EtOH to Eq. (34) and reported respectively, and c is the analytical concentration of the
R = 0.835 for 182 mobility data of the analytes studied. To compound. Again the mobilities are signed quantities in
provide more accurate results, the authors also employed this context.

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 1001

It is clear that all equilibria in which the analytes are (http:///www.natur.cuni.cz/,gas). Expressions for the
involved can influence the effective mobility; this holds, mobility dependence of zwitterions on the pH in mixed
e.g., for the ion-pair formation or conjugation as men- solvents were used based on the mass action law for
tioned above. However, the most abundant type of reac- quinoline derivatives by Barron et al. [83] in order to pre-
tion that is relevant in the context of electrophoresis of dict favourable separation conditions.
weak acids and bases is certainly protolysis. In this case
the application of the Henderson-Hasselbalch equation In principle the effective mobility is more adequate than
transformed to express the degree of ionization, a, for the actual mobility also when other than acid–base equi-
example of a monobasic acid HA to the anion A2 libria are present. In the first view, complexation reac-
according to HA = H1 1 A2 gives tions, ion pairing or homo- and heteroconjugation reac-
1 tions are mentioned. Homo- and heteroconjugation is a
a¼ (37)
1 þ 10pK a pH phenomenon which can occur in solvents with low solva-
tion ability for the ion. Let us consider an acid HA, which
and consequently the expression for the effective mobility
dissociates into H1 and A2. In water, with its excellent
as a function of the pH of the buffer
solvation ability towards both cations and anions, the ions
mact;A are stabilized by hydration. If the solvent has a low solva-
meff;A ¼ (38)
1 þ 10pK a pH tion ability, e.g. MeCN, the anion is not stabilized by the
It is important that this equation is followed well not only in solvent molecule (stabilisation of H1 is reflected by the
aqueous solutions but also in organic solvents like MeOH solvent’s basicity). The anion might be stabilized instead
and MeCN [46, 78–80], even in such unusual solvents like by association with its parent molecule, the neutral parti-
nitromethane [53], FA [54] or DMF [81] (see for example cle HA, under formation of the ion AHA2. This reaction is
Fig. 5). called homoconjugation. It is clear that the resulting ion has
a mobility different than A2. If reaction takes place with a
If not only one group is ionised by protolysis, the appro- molecule HB, e.g. a buffer constituent present in the so-
priate equations are obtained accordingly. So is the lution, heteroconjugation can take place under formation
effective mobility for an acidic amino acids like aspartic or of AHB2. Also in this case the mobility of A2 is influenced,
glutamic acid as a function of the pH and it is a matter of the position of the conjugation equi-
mact;Aþ 10pK a;1  pH  mact;A 10pH  pK a;2 librium to what extent the mobility varies [84, 85]. In prin-
meff;A ¼ (39) ciple we could formally treat these equilibria like the acid–
10pK a;1  pH þ 10pH  pK a;2 þ 1
base equilibria, introduce a kind of the degree of associ-
with Ka,1 and Ka,2 being the dissociation constants of the ation (related to the conjugation constants and the con-
equilibria A1 = A7 1 H1 and A7 = A2 1 H1, respectively. centrations of the reactants) and discuss the resulting
For a more detailed formal treatment of the mobility see mobility as effective mobility rather than as actual mobili-
e.g. [82]; this formalism was used for the simulation pro- ty. It is the same formalism as we apply it for complex
gram PeakMaster, which is freely available on the Internet formation reactions, e.g. with CDs or other ligands. In

Figure 5. Measured effective mobilities, meff,i, vs. the pH of the BGE in organic solvents. Solid lines are
fitted according to Eq. (38). Left: nitromethane as solvent with BGEs of different pH. Right: FA as
solvent; mobility of 3,5-dimethylbenzoate vs. pH of the BGE. Temperature, 25.07C. pH was calculated
by means of the Henderson-Hasselbalch equation, and corrected for the activity of the buffer anion.
From [53, 82] with permission.

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


1002 A. Jouyban and E. Kenndler Electrophoresis 2006, 27, 992–1005

case of biological or pharmaceutical agents we name the complexation of two ions with opposite charge Eq. (43)
method affinity electrophoresis. In all cases the same reduces to
principle lies in the reaction, in the simplest case with 1:1 1
stoichiometry, according to meff;A ¼ mA (44)
1 þ K c ½X

A 1 X = AX (40) because the mobility of the complex is zero. In Fig. 6, an


example is given in which the deviation of the depend-
In the case of protolysis, X is the proton; in other applica- ence of mobility on ionic strength according to Eq. (22)
tions X can be CD, a protein in case of drug–receptor was related to ion pairing in buffered MeCN–water media.
interactions etc. We consider the kinetics of complex for- From this deviations the authors derived ion-pair forma-
mation and dissociation being fast enough. In practice the tion constants [61]. It is worth mentioning that Eq. (43)
ligand X is added to the BGE at a varying concentration (in also holds for acid–base equilibria. In this case the effec-
case of the acid–base reaction it is adjusted by the pH of tive mobility of the anion A2 formed by dissociation of the
the buffer); the analyte is injected as usual in CZE. The acid HA (its mobility is zero) reads after transformation of
effective mobility of the analyte, meff,A, is composed from Eq. (43)
its mobility, mA-, as free species A2, weighted by the frac-
mA mA
tion of free analyte, f, present in the solution, and the meff;A ¼ ¼ (45)
1 þ K c ½X 1 þ 10pK a pH
mobility of the complex, mAB, also weighted by the com-
plex fraction, which is (1 2 f). Fraction f is given by whereas [X] is [H1] and Ka, the ionisation constant, is the
reciprocal of the formation constant Kc. It can thus be
½A
f¼ (41) seen that Eq. (44) is identical to Eq. (38).
½A þ ½AX
which upon application of the mass action law gives The expressions for the effective mobility were derived so
far by simple thermodynamics. An equation not based on
1 the mass action law but on an empirical approach using a
f¼ (42)
1 þ K c ½X MLR model describing the effective mobility as function
where Kc is the complexation constant. The effective of pH and of the concentration of organic modifiers [86] is
mobility as function of the ligand concentration is
lnmact;i ¼ A0 þ A1 lnf 1 þ A2 lnpH þ A3 ðlnf 1 Þ2 þ
accordingly
1 K c ½X m  þ mAB K c ½X þ A4 ðlnpHÞ2 þ A5 lnf 1 lnpH (46)
meff;A ¼ mA þ mAB ¼ A (43)
1 þ K c ½X 1 þ K c ½X 1 þ K c ½X
where Ai is the model constants and, f1 is the fraction of
This equation expresses the effective mobility for reac- organic solvent in the buffer. The model has been tested
tions we summarize under the terms complexation, ion on 12 benzoate derivatives at various f1 and pH values of
pairing, conjugation and affinity. It is obvious that for the MeCN and phosphate buffer mixture where the correla-
noncharged complex, e.g. in case of 1:1 ion pairing or tion coefficient was 0.999 [86].

Figure 6. Deviation of the mobility as for-


mulated by Eq. (20). Conditions: 35/65 MeCN/
buffer, 5.0 mM hydrochloric acid at pH 2.3, and
0–55 mM NaCl. Plot legends: (s) nortriptyline;
(u) amitriptyline; (n) perphenazine; dotted
lines, theoretical predictions based on Eq. (20).
It was assumed that perphenazine is doubly
charged. From [62] with permission.

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 1003

3 Conclusion [11] Kenndler, E., Jenner, P., J. Chromatogr. 1987, 390, 185–197.
[12] Erdey-Gruz, T., Transport Phenomena in Aqueous Solutions,
Advanced theories exist which formulate the mobilities as Akademiai Kiado, Budapest 1974.
function of many experimental parameters and variables. [13] Marcus, Y., Ion Properties, Marcel Dekker, New York, 1997.
Fundamental physicochemical theories based on ion [14] Muzikár, J., van de Goor, T., Gaš, B., Kenndler, E., Electro-
phoresis 2002, 23, 375–382.
conductance in solution with increasing complexity have
[15] Sarmini, K., Kenndler, E., J. Chromatogr. A 1999, 833, 245–
been established. They start with the most simple concept 259.
of the electrical movement of a spherical ion in a con- [16] Sarmini, K., Kenndler, E., J. Chromatogr. A 1998, 806, 325–
tinuum with application of Stokes’ law, and try to explain 335.
the deviations from this model by more sophisticated [17] Sarmini, K., Kenndler, E., J. Chromatogr. A 1998, 811, 201–
209.
concepts like introducing parameters for nonspherical
[18] Sarmini, K., Kenndler, E., J. Chromatogr. A 1998, 818, 209–
size, the presence of an ion atmosphere, the considera- 215.
tion of the ions and counterions approaching each other [19] Fernández-Prini, R., in: Covington, A. K., Dickinson, T.
to a certain distance and the polarization of the solvent in (Eds.), Physical Chemistry of Organic Solvent Systems, Ple-
the vicinity of the ion by its electric field (so-called dielec- num Press, London 1973, pp. 525–614.
tric friction). Association and dissociation reactions have [20] Edward, J. T., Res. Correspondence 1956, 9, 41–42.
the effect that only a part of the analyte species is present [21] Shimizu, T., Kenndler, E., Electrophoresis 1999, 20, 3364–
3372.
in ionic forms. The application of basic thermodynamics
[22] Hubbard, J., Onsager, L., J. Chem. Phys. 1977, 67, 4850–
and the mass action law allows expressing the resulting 4857.
effective mobility as a function of the ligand concentration [23] Hubbard, J. B., J. Chem. Phys. 1978, 68, 1649–1664.
for equilibria in acid–base, complexation, ion pairing, [24] Evans, D. F., Tominaga, T., Hubbard, J. B., Woynes, P. G., J.
homo- and heteroconjugation or affinity reactions. In Phys. Chem. 1979, 83, 2669–2677.
many cases, however, the systems are too complex, and [25] Fuoss, R. M., Proceedings of the National Academy of Sci-
ences 1959, 45, 807–813.
then the mobility changes cannot be related to physico-
[26] Boyd, R. H., J. Chem. Phys. 1961, 35, 1281–1283.
chemical quantities. In these cases mathematical algo-
[27] Boyd, R. H., J. Chem. Phys. 1963, 39, 2376.
rithm can be favourably applied to predict mobilities as a
[28] Zwanzig, R., J. Chem. Phys. 1970, 52, 3625–3628.
function of some experimental variables. The presented
[29] Zwanzig, R., J. Chem. Phys. 1963, 38, 1603–1605.
algorithms could be categorised in MLR and ANN meth- [30] Spiro, M., in: Covington, A. K., Dickinson, T. (Eds.), Physical
ods. Usually ANN methods provided better correlation in Chemistry of Organic Solvent Systems, Plenum Press, Lon-
comparison with MLR. On the other hand, MLR models don 1973, pp. 635–680.
are able to provide slightly more significant information on [31] Fu, S., Lucy, C. A., Anal. Chem. 1998, 70, 173–181.
the factors affecting the mobility of the ion. These MLR [32] Fu, S., Li, D., Lucy, C. A., Analyst 1998, 123, 1487–1492.
and ANN models describing the mobility with respect to [33] Li, D., Fu, S., Lucy, C. A., Anal. Chem. 1999, 71, 687–699.
ionic strength, solvent composition, pH of the buffer and [34] Li, D., Lucy, C. A., Anal. Chem. 2001, 73, 1324–1329.
[35] Roy, K. I., Lucy, C. A., Anal. Chem. 2001, 73, 3854–3861.
structural features of the analytes were presented.
[36] Roy, K. I., Lucy, C. A., J. Chromatogr. A 2002, 964, 213–225.
[37] Roy, K. I., Lucy, C. A., Electrophoresis 2002, 23, 383–392.
[38] Roy, K. I., Lucy, C. A., Electrophoresis 2003, 24, 370–379.
4 References [39] Jokl, V., J. Chromatogr. 1964, 13, 451–458.
[40] Preetz, W., Blausius, E., Z. Anorg. Allg. Chem. 1965, 335, 16.
[1] Porras, S. P., Kenndler, E., Electrophoresis 2005, 26, 3203–
[41] Kiso, Y., Kobayashi, M., Kitaoka, Y., Kawamoto, K., Takada,
3220.
J., J. Chromatogr. 1968, 36, 215–228.
[2] Porras, S. P., Riekkola, M.-L., Kenndler, E., Electrophoresis [42] Kiso, Y., Hirokawa, T., Chem. Lett. 1979, 8, 891–894.
2003, 24, 1485–1498.
[43] Li, Q., Dong, L., Jia, R., Chen, X. et al., Comput. Chem.
[3] Gaš, B., Kenndler, E., Electrophoresis 2004, 25, 3901–3912. 2002, 26, 245–251.
[4] Gaš, B., Kenndler, E., Electrophoresis 2002, 23, 3817–3826. [44] Xue, C., Liu, H., Yao, X., Liua, M. et al., J. Chromatogr. A
[5] Gaš, B., Kenndler, E., Electrophoresis 2000, 21, 3888–3897. 2004, 1048, 233–243.
[6] Bockris, J. O. M., Reddy, A. K. N., Modern Electrochemistry [45] Muzikár, J., van de Goor, T., Kenndler, E., Anal. Chem. 2002,
1: Ionics, 2nd Edn., Plenum Press, New York 1998. 74, 434–439.
[7] Everaerts, F. M., Beckers, J. L., Verheggen, T. P. E. M., Iso- [46] Porras, S. P., Riekkola, M.-L., Kenndler, E., J. Chromatogr. A
tachophoresis, Theory, Instrumentation and Applications, 2001, 924, 31–42.
Elsevier, Amsterdam 1976. [47] Hsia, K.-L., Fuoss, R. M., J. Am. Chem. Soc. 1968, 90,
[8] Hirokawa, T., Nishino, N., Aoki, N., Kiso, Y. et al., J. Chro- 3055–3060.
matogr. 1983, 271, D1–D106. [48] Falkenhagen, H., Leist, M., Kelbg, G., Ann. Physik 1952, 11,
[9] Hirokawa, T., Tsuyoshi, T., Kiso, Y., J. Chromatogr. 1987, 51–59.
408, 27–41. [49] Falkenhagen, H., Elektrolyte, 2nd Edn, S. Hirzel, Leipzig
[10] Kenndler, E., Jenner, P., J. Chromatogr. 1987, 390, 169–184. 1953.

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


1004 A. Jouyban and E. Kenndler Electrophoresis 2006, 27, 992–1005

[50] Pitts, E., Proc. Royal Soc. London 1953, 217A, 43–70. [69] Jalali-Heravi, M., Garkani-Nejad, Z., J. Chromatogr. A 2001,
[51] Pitts, E., Tabor, B. E., Daly, J., Proc. Royal Soc. London 927, 211–218.
1970, 693–707. [70] Jouyban, A., Yousefi, B. H., Comput. Biol. Chem. 2003, 27,
[52] Muzikár, J., van de Goor, T., Gaš, B., Kenndler, E., Anal. 297–303.
Chem. 2002, 74, 428–433. [71] Jouyban-Gh., A., Khaledi, M. G., Clark, B. J., J. Chromatogr.
[53] Subirats, X., Porras, S. P., Roses, M., Kenndler, E., J. Chro- A 2000, 868, 277–284.
matogr. A 2005, 1079, 246–253. [72] Jouyban, A., Chan, H.-K., Clark, B. J., Kenndler, E., Electro-
[54] Porras, S. P., Kenndler, E., Electrophoresis 2004, 25, 2946– phoresis 2003, 24, 1596–1602.
2958.
[73] Jouyban, A., Khoubnasabjafari, M., Chan, H. K., Altria, K. D.,
[55] Vceláková, K., Zusková, I., Porras, S. P., Gaš, B., Kenndler, Clark, B. J., Chromatographia 2003, 57, 191–195.
E., Electrophoresis 2005, 26, 463–472.
[74] Jouyban, A., Grosse, S. C., Chan, H. K., Coleman, M. W.,
[56] Vceláková, K., Zusková, I., Kenndler, E., Gaš, B., Electro- Clark, B. J., J. Chromatogr. A 2003, 994, 191–198.
phoresis 2004, 25, 309–317.
[75] Jouyban, A., Majidi, M. R., Asadpour-Zeynali, K., Farmaco
[57] Barthel, J., Neueder, R., Feuerlein, F., Strasser, F., Iberl, L., J.
2005, 60, 255–259.
Solut. Chem. 1983, 12, 449–471.
[58] Barthel, J., Neueder, R., Rawytsch, P., Roch, H., J. Electro- [76] Jouyban, A., Majidi, M. R., Altria, K. D., Clark, B. J., Asad-
anal. Chem. 1999, 471, 78–87. pour-Zeynali, K., Pharmazie 2005, 60, 656–660.
[59] Barthel, J., Krell, M., Iberl, L., Feuerlein, F., J. Electroanal. [77] Liu, H. X., Zhang, R. S., Yao, X. J., Liu, M. C et al., Anal.
Chem. 1986, 214, 485–505. Chim. Acta 2004, 525, 31–41.
[60] Barthel, J., Wachter, R., Schmeer, G., Hilbinger, H., J. Solut. [78] Porras, S. P., Jyske, P., Riekkola, M.-L., Kenndler, E., J.
Chem. 1986, 15, 531–550. Microcolumn Sep. 2001, 13, 149–155.
[61] Dai, J., Mendonsa, S. D., Bowser, M. T., Lucy, C. A., Carr, P. [79] Porras, S. P., Riekkola, M.-L., Kenndler, E., Chromato-
W., J. Chromatogr. A 2005, 1069, 225–234. graphia 2001, 53, 290–294.
[62] Friedl, W., Reijenga, J. C., Kenndler, E., J. Chromatogr. A [80] Porras, S. P., Riekkola, M.-L., Kenndler, E., J. Chromatogr. A
1995, 709, 163–170. 2001, 905, 259–268.
[63] Reijenga, J. C., Kenndler, E., J. Chromatogr. A 1994, 659, [81] Porras, S. P., Kenndler, E., Electrophoresis 2005, 26, 3279–
403–415. 3291.
[64] Offord, R. E., Nature (London) 1966, 211, 591–593.
[82] Štědrý, M., Jaroš, M., Hruška, V., Gaš, B., Electrophoresis
[65] Mazak, K., Szakacs, Z., Nemes, A., Noszal, B., Electropho- 2004, 25, 3071–3079.
resis 2000, 21, 2417–2423.
[83] Barron, D., Irles, A., Barbosa, J., J. Chromatogr. A 2000,
[66] Lin, C. E., Chang, C. C., Lin, W. C., J. Chromatogr. A 1997, 871, 367–380.
768, 105–112.
[84] Okada, T., Chem. Commun. 1996, 1779–1780.
[67] McKillop, A. G., Smith, R. M., Rowe, R. C., Wren, S. A. C.,
Anal. Chem. 1999, 71, 497–503. [85] Okada, T., J. Chromatogr. A 1997, 771, 275–284.
[68] Jalali-Heravi, M., Garkani-Nejad, Z., J. Chromatogr. A 2002, [86] Guillaume, Y. C., Peyrin, E., Ravel, A., Guinchard, C., J. Liq.
971, 207–215. Chrom. 2000, 23, 2789–2806.

5 Addendum c Concentration
E Electric field strength
List of symbols F Faraday constant
H Mean of waters of hydration
a Ion size parameter, distance of closest approach
DHf Heat of formation
E Elementary charge, charge of an electron
I Ionic strength
f/f0 Shape correction factor, frictional ratio
KA Ion-pair formation constant
f1 Volume fraction of solvent 1
MR Molecular refractivity
f2 Volume fraction of solvent 2
M Molecular mass
f3 Volume fraction of solvent 3
NA Avagadro’s number
fdiel,i Dielectric friction coefficient PPCH Most positive partial charge
fhydr,i Hydrodynamic friction coefficient PQ Net charge of the most positive/negative atom of
kB Boltzmann constant the analyte
q Electrolyte parameter (1/2 for 1:1 electrolyte) S Shape factor, Onsager slope in different equations
rcryst,i Crystal radius SA Surface area
ri Inertial radius TE Total energy
rrot Effective radius V Molar volume
rSt,i Stokes (hydrodynamic) radius of the ion a Degree of dissociation
rvdw van der Waals radius e Absolute static permittivity
zi Charge on the analyte, valency of the ion e0 Permittivity of vaccum

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com


Electrophoresis 2006, 27, 992–1005 CE and CEC 1005

er Relative permittivity mI Mobility of an ion


g6 Mean activity coefficient m0,i Absolute electrophoretic mobility of an ion
Z Dynamic viscosity mact,i Actual electrophoretic mobility of an ion
l Single ion conductance meff,i Effective electrophoretic mobility of an ion
m1–m3 Electrophoretic mobility of analyte in mono- t Dielectric relaxation time
solvent buffer 1–3 L Conductance

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.electrophoresis-journal.com

You might also like