Professional Documents
Culture Documents
To cite this article: M.Y. CHISTI & M. MOO-YOUNG (1987) AIRLIFT REACTORS:
CHARACTERISTICS, APPLICATIONS AND DESIGN CONSIDERATIONS, Chemical Engineering
Communications, 60:1-6, 195-242, DOI: 10.1080/00986448708912017
Download by: [Cornell University Library] Date: 06 October 2016, At: 23:27
Chem. Eng. Comm. 1987, vol60, pp. 195-242
Photocopying permitted by license only
0 1987 Gordon and Breach Science Publishers S.A.
Printed in the United States of America
Bioreactors of the airlift type are a promising design for aerobic fermentations. The basic knowledge
required for understanding and predicting the performance of these reactors is only now beginning to
emerge. In this review we present our observations and those of other investigators in an attempt to
build up a coherent picture of airlift devices. All the major aspects-mixing and hydrodynamics, mass
and heat transfer-in these reactors are considered. Comparisons between bubble columns and airlift
systems are made where analogies, similarities and/or differences between them provide insight into
airlift systems. Throughout, the areas of particular concern and those in need of further research in
this field are mentioned. Extensive work on all forms of airlift reactors, particularly in non-Newtonian
media-homogeneous and suspensions-remains to be done. Current knowledge does not permit
airlift reactor design with a high degree of confidence. However, the technical feasibility of all types of
fermentations-dant cell, tissue culture, bacterial, fungal,
- and those utilizing yeasts-in airlift vessels
has been demonkrated.
KEYWORDS Airlift reactor Hydrodynamics Mass transfer Heat transfer
1. INTRODUCTION
Bioreactors are the core of any bioprocess. Gas-liquid or gas-slurry reactors are
used in aerobic fermentations. The most common aerobic reactors used in
commercial productive fermentations are of the stirred tank variety' and the
design of these dates from 1940's' when they were used for the first modern
industrial fermentation-that of the antibiotic penicillin. Pneumatic reactors, in
which all agitation is due to bubbling gas, are a relatively recent invention. An
exception is the simple bubble column reactor which, despite its long history of
application in fermentation, is not widely used. Airlift reactors are also pneumatic.
Other completely new designs of fermentation vessels and modifications of some
of the earlier types have been proposed. These include bubble columns with
many forms of internal^,^^ pulsed bubble column^,^ airlifts with stirrers,' tubular
loop which may be suitable for small volume fermentations, and
aeration using a downward directed two-fluid nozzle.'0." Spouted beds1' also find
Airlift reactors consist of a liquid pool divided into two distinct zones only one of
which is usually sparged by a gas. The different gas holdup in the gassed and
ungassed zones results in different bulk densities of the fluid in these regions
which causes circulation of the fluid in the reactor by a gas-lift action. The part of
the reactor containing the gas-liquid upflow is the riser and the region containing
the downflowing fluid is known as the downcomer (Figure 1).
Theoretically, airlift reactors may be employed for any gas-liquid or gas-slurry
contacting process. Practical application depends on the ability to achieve the
required rates of momentum, heat and mass transfer at acceptable capital and
operating costs. The technical and economic feasibility of using airlift devices has
been conclusively established for a number of processes and these reactors find
increasing use in aerobic fermentations, treatment of wastewater and other
similar operations. The simplicity of their design and constr~ction,'~better
defined flow patterns in them and comparatively low power inputs for requisite
transport rates, make them very attractive. Low shear fields, good mixing and
extended aseptic operation made possible by elimination of stirrer shafts, seals
and bearings are important advantages of airlifts in fermentation applications.
Continuous production of beer, vinegar, citric acid, and biomass from yeasts,
bacteria and fungi has been carried out in airlift vessels at different working
capacities.13 The USSR and Eastern Europe have extensively employed airlift
vessels for SCP yeast c~ltivation'~ and in England 1CI has operated at 1500 m3
(working volume) airlift fermenter for the PRUTEEN process.15 Ho et a1.I6 have
mentioned several currently used industrial scale applications of airlift fermen-
ters. Malfait et a[.'' claimed a more than 18% (by weight) enhancement in yield
of a filamentous mould Monascus purpureus in an external loop airlift (0.055 m3,
-
DOWNCOMER ( DOWNFLOW I
GAS SPARGED RISERS
I U P FLOW )
2.1 Classification
A large variety of configurations of airlift reactors have been investigated and
occasionally confused terminology is encountered in the literature. We distinguish
two basic classes of airlifts: (i) the internal loop airlifts where what would
otherwise be a simple bubble column has been split into a riser and a downcomer
by an internal baffle; and (ii) the external or outer loop airlift reactors where the
riser and the downcomer are two quite separate tubes connected by horizontal
sections near the top and bottom. Internal and external loop reactors may be
further subdivided depending on the peculiarities. ~ n t e r n a lloop airlifts, for
example, may be of the split-cylinder type (Figure 2a) or they may have a
concentric draught-tube configuration (Figure 2b). In the latter either the
draught-tube or the annulus may be gas sparged. The draught-tube and the
vertical baffle (in split-cylinder mode) may themselves be divided into vertical
sections to increase communication between the riser and the downcomer (Figure
2c). Multiple concentric draught-tubes have also been studied.= External loop
reactors (Figure 2d) have lesser variety, but several designs of horizontal
connections between riser and downcomer, particularly the top connection, may
be appropriate. Further modifications can be introduced into the headspace
region of the reactor where gas-liquid separation takes place, and, depending on
the separator efficiency the reactor performance may be significantly i n f l ~ e n c e d . ~ ~
Internal and external loop airlift reactors usually have circular cross-sections, but
rectangular and square cross-sections which have practical applications in industry
are also a definite option and have indeed been Gas sparger types,
both in the static and dynamic34 sparger classes, and their location in the riser
and/or downcomer may be altered to give different performances for different
purposes. Multiple gas injection points in the reactor in addition to the primary
sparger may also be beneficial. In addition, the downcomer and/or riser may
contain internals such as sieve plates" and baffles projecting from the walls."
In summary, a myriad of possible variations on basic airlift design exist and
may be advantageously utilized for different applications. Influences of some of
the design modifications on performance are examined in later parts of this
review.
RISER -
FIGURE 2 Airlift reactor types: (a) split cylinder internal loop; (b) concentric draught-tube internal
loop; ( c ) concentric draught-tube (vertically split) internal loop; (d) external loop.
as gas flow is increased, several different flow regimes may be observed. At low
gas inputs the gas bubbles rise almost straight up the reactor with little interaction
between them. This is the bubble flow regime known also as unhindered bubble
flow and homogeneous bubble flow (Figure 3a). With further increases in the gas
flow rate the bubble density in the fluid gradually rises and there is greater
interaction between them-bubble collision frequencies increase-and a coal-
esced bubble flow regime characterized by greater turbulence ensues as shown in
Figure 3b. This is a transitional regime which leads eventually to fully developed
churn turbulent flow where, in addition to many small bubbles, larger bubbles
occur frequently. Because of the very high turbulence fields the large bubbles
often have little definition to their shape which fluctuates quite randomly. Figure
3c depicts churn turbulent flow. Spherical caps or bullet nosed bubbles which rise
rapidly may form at higher gas flows. Frequency and size of spherical caps
M.Y. CHISTI AND M. MOO-YOUNG
C) CHURN-TURBULENT d l SLUGGING
increases with gas flow rate and in reactors with small diameters, particularly in
viscous fluids, these bubbles may attain dimensions approaching those of the tube
through which they rise and this is the fully developed slug flow shown in Figure
3d. Annular film flow and spray flow can be obtained at very high gas velocities,
but these are of little interest for aerobic fermentation purposes. The range of gas
flows over which a particular flow regime occurs and whether it exists at all in a
given application is dependent on the properties of the fluid being handled and on
the physical configuration of the reactor. Earlier transition to slug flow occurs, for
example, in tubes of small diameter than in vessels of larger size; and spherical
cap bubbles form more readily in highly viscous fluids and mycelial media than in
water-like systems. Identification of the flow regime existing in a reactor is
simplified by the use of flow regime maps two examples of which are shown in
Figure 4 for air-water. Figure 4a is more suitable for bubble columns with no net
liquid flow, while 4b may be applied to the riser or downcomer of an airlift when
the gas and liquid flows through these sections are known, and to bubble columns
with large fluid throughputs. Other similar maps, more or less suited for
particular purposes, are available in texts and handbooks on multiphase flow.
The hydrodynamic behaviour of bubble columns and airlift reactors is very
different. The main distinction between cocurrent or counter-current bubble
columns and airlift reactors is that in the latter the rate of liquid circulation
depends on, and is determined by, the gas flow rate, whereas in the former types
the liquid flow is independent of gas flow. Because of the long residence times
that are typically necessary in bioreactors, large liquid throughputs are not
possible in bubble columns without significant recycle rates. In airlift reactors, on
the other hand, quite high linear liquid velocities may be generated without the
AIRLIFT BIOREACTORS
HOMOGENEOUS
I BUBBLY I
FIGURE 4a Approximate dependency o f flow regime on gas velocity and reactor diameter for
bubble columns with no liquid flow (water and dilute aqueous solutions). Based on Shah el 01."
/
/
/
ANNULAR WISPY-
;" 10' I ANNULAR
E I
CHURN 7------
---I
CI,
N YI
,
, I
I
\ BUBBLY
\
\
/
/
BUBBLY -SLUG
'
\
FIGURE 4b Flow pattern map for vertical gas-liquid flow for low viscosity Newtonian liquids.
Adapted from Collier."
C ,f-BUBBLE COLUMNS
FIGURE 5 Operating ranges of gas and liquid velocities in bubble column and airlift reactors. Based
on Weiland and O n k e r ~ . ) ~
where P,, and P,, are the pressures at the bottom and at the top (headspace
presure) of the fluid in the vessel. Q, in Eq. (1) is the molar flow rate of the gas,
Uois the gas velocity in the orifice of the sparger which relates to the gas velocity
just above the orifice through the efficiency q (generally 0.06),'"' and M is the
molar mass of the gas. The kinetic energy term in Eq. (1) (second term on the
right hand side) would generally be found to be negligible. Discrepancies in the
literature have arisen due to underestimation of power input when the kinetic
energy term in Eq. (1) was ignored without justification. Interesting debate^^'.^'
on this subject exist. It should be noted that even when the molar gas flow into a
reactor is constant the power input to the fluid declines with increasing reactor
AIRLIFT BIOREACTORS 203
headspace pressure and the reactor becomes markedly less turbulent. Thus when
pneumatic reactors are operated in a pressurized state-to improve the gas-liquid
mass transfer driving force, for example-the turbulence intensity and parameters
dependent on it (e.g. mixing, overall mass transfer coefficient, and gas holdup)
are sacrificed to various degrees. Pressurized operation, however, is less common
for biotechnological processes because of the added costs and problems as-
sociated with carbon dioxide toxicity.
where T is the reactor temperature and A its cross-sectional area. From Eqs. (1)
and (2) it can be shown that for a reactor of uniform cross-sectional area the
power input per unit liquid volume is
which applies to airlifts when Us, is based on the total riser and downcomer
cross-sectional area. If either the superficial mass velocity of gas or its velocity
under reactor inlet (or outlet) conditions of temperature and pressure are used in
M.Y. CHISTI AND M. MOO-YOUNG
1 r G =4.69x I O - ~ ( IP ~V ~ ) ~ ( . b ~ ~ ~
7 A
Headspace
o 145
101 k P o
Pressure
187 kPo
FIGURE 6 Gas holdup in a draught-tube internal loop airlift reactor in air-water at diflerent
reactor headspace pressures. The apparent reduction in gas holdup with increasing pressure seen in
(a) disappears ( b ) when true superficial gas velocity (Usg) is used for correlation instead of the
superficial gas velocity under inlet conditions as in (a).
correlating gas holdup or gas-liquid mass transfer, for example, then apparent
effects of reactor headspace pressure or of unaerated liquid height on these
parameters may seem to exist. An example of this is shown in Figure 6 which
presents our gas holdup data obtained in air-water in a large concentric
draught-tube airlift vessel (d, = 0.762 m, dci= 0.355 m, total height of draught-
tube = 2.06 m (split vertically into 3 sections: from top to bottom 0.434,0.914 and
0.559 m, respectively, separated by gaps of 0.077 m), draught-tube clearance from
bottom of reactor = 0.202 m, unaerated liquid height = 2.32 m; annulus sparged
by perforated pipe (total holes = 128, hole diameter = 0.002 m) ring (3 concentric
rings) sparger) at reactor headspace pressures (absolute) of 101, 145 and 187 kPa.
In Figure 6a the total holdup obtained by the level displacement technique is
shown as a function of superficial gas velocity at reactor inlet conditions and an
apparent pressure effect on holdup is seen. The same data plotted against actual
superficial gas velocity (Eq. (2)) in Figure 6b all falls on the same line. The
superficial air velocities in Figure 6 are based on the entire cross-section of the
airlift reactor.
3. GAS-LIQUID DISPERSIONS
gas
POROUS PLATE SINGLE NOZZLE PERFORATED PLATE
OR PERFORATED PIPE
DYNAMIC SPARGERS
I
7 shows some of the spargers that may be used. The choice of sparger for a
particular application depends on several factors. Perforated plates and pipes are
cheap to install and operate. Porous plates are more expensive and have higher
operating costs due to greater pressure drops through them. In addition they are
prone to blockages and can be a source of contamination. We have, however,
successfully used porous plate aerators (100 p m pore size) in small (=2 L) airlift
devices during cultivation of mammalian cells for monoclonal antibody produc-
tion. Cell growth on aerator surface or into the pores posed no problems.
Dynamic gas spargers are not commonly used in fermentation practice. They are
more complex to design and build and require some external liquid circulation
mechanism, usually a pump. High shear rates in the dynamic spargers and in
associated pumping machinery rule them out for shear sensitive microorganisms
and tissue cultures. Dynamic sparger design is briefly treated by Z l ~ k a r n i k . ~ '
Apart from those considerations the required gas bubble size in dispersion can
also be influenced to some degree by the type of gas sparger used.
206 M.Y. CHlSTI AND M. MOO-YOUNG
For porous plate spargers operating in air-water the effect of various degrees
of sparger blockage on gas holdup may be computed" from:
where f is the fraction of total free sparger area relative to the cross-sectional area
of riser, and Ub is a characteristic bubble rise velocity. The mean bubble rise
velocity may be estimated using
+
Ub = 0.284 2.7U, (5)
(ms-') (ms-') (ms-')
Equation (5) is applicable to bubbles rising in swarms in air-water dispersions in
bubble columns. This equation was obtained by us using gas holdup data from
several sources including our own4' and it described all the data within f 2 0 %
over a gas velocity range of 0.005 to 0.4 ms-I. In the risers of airlift reactors
bubbles rise faster than in bubble columns due to liquid circulation and the liquid
velocity should be added to the value obtained from Eq. (5) to get the correct rise
velocity. More accurate predictions of mean bubble rise velocity are possible
using the equation
which applies to bubble flow regime (Us, < 0.05 ms-') and
for coalesced bubble flow (U, > 0.05 ms-I). Equations (6) and (7) are for bubble
columns. The use of these equations in Eq. (4) requires an iterative procedure.
Another correlation which may be suitable for the estimation of mean bubble rise
velocity in the riser of an airlift reactor is
+
Ub = 0.24 + 1.35(Usg UL,)0.93
(ms-I) (ms-I) (ms-I) (8)
Equation (8) was obtained by Hills46 for air-water in a 0.15 m diameter bubble
column for UL, > 0.3 ms-'.
which is said to apply when the scale of the energy containing primary eddies is
much greater (>200-fold) than the Kolmogoroff scale of energy dissipating
terminal eddies. In Eq. (9) V I is not the liquid volume in the reactor, but it is the
volume contained in the riser. Another expression for bubble size is
which was reported by Azbel."' Dussap and Gros4' proposed the following
expression for gas-liquid interfacial area in aqueous sodium sulphite in a
concentric cylinder airlift (volume = 0.015 m3, d, = 0.11 m, dci = 0.0756 m, total
height = 1.8 m):
EG = 0 . 2 5 5 e E p;l9
(ms-') (Pas)
Substitution of the relation
and Eq. (15) into (14) yields the following equation for Sauter mean bubble
diameter:
dB = 0.08Gsl3 ;::p
(17)
(m) (ms-') (Pas)
The apparent viscosity in Eq. (17) was determineds0 from
where K and n are the consistency and flow behaviour indices, respectively, of
power law fluids. When Eq. (18) is substituted into (17) and the result is
expressed in terms of the power input using Eq. (3) we get equations of the form
where (Y and /? are 0.096 and -0.19, respectively for the solution with
K = 7.683 Pasn and n = 0.440. Similarly for the fluid of K = 0.095 Pasn and
n =0.697, a and /? are 0.007 and -0.04, respectively. A much weaker
dependence of bubble size on power input is indicated in pseudoplastic
homogeneous fluids than in salt solutions.
The nature of gas-liquid dispersion of water, salt water, and CMC-in-water
solutions has been studieds1 in a split-cylinder (Figure 2a) airlift vessel (d, =
0.15 m, A,/A, = 0.5, height = 1.36 m, perforated plate sparger with 38 holes of
0.0016 m diameter, baffle clearance from the bottom = 0.055 m). Results showed
that for air-water system in the riser the bubble size distribution was dependent
on position along the vertical a x k 5 ' The bubble size distribution in air-water was
observeds1 to shift gradually toward the small bubble region with increasing
distance from the sparger; furthermore, the greater was the distance from the
sparger the more uniform was the bubble size distribution. This behaviour was
explaineds1 by the continuous breakage of primary bubbles as they left the
sparger and were exposed to the turbulent flow field. The data of Glasgow et ~ 1 . ~ '
indicated that at a fixed distance up the riser the bubble size distribution was
influenced by superficial gas velocity (or the power input). Increased power inputs
lead to smaller bubbles; however, coalescence also became important at higher
bubble densities and tended to offset bubble breakup resulting from the increase
in dynamic pressure forces. The dispersion behaviour in CMC solutions (0.5 and
0.8%, K and n not given) was very different from that in air-water dispersions.
AIRLIFT BIOREACTORS
In the viscous systems the survivability of large bubbles in the upflow was
e n h a n ~ e d . ~Which
' was explained as being due to increased coalescence and
decrease in available disruptive turbulent energy at a given gas flow rate. Our
observations have shown that the distribution of bubble size in the downcomers
of airlift reactors is narrower than in the risers. In the downcomer, too, there is
an axial gradation of bubble size. This is discussed elsewhere in this work.
Notice that nearly all the empirical equations for gas holdup dependent
parameters such as interfacial area and bubble diameter presented in this section
followed the general form: parameter = constant . Usg? Similar dependence of
gas holdup on superficial gas velocity (or specific power input) has been well
known, although it was only recently that some theoretical insight into this
dependence became availab~e."~
At present there is insufficient knowledge for reliable estimation of gas-liquid
interfacial areas and bubble sizes in homogeneous power law fluids and even less
so in broths containing solids which simulate mycelial suspensions.
GAS GAS
(a (b)
t
GAS
t
GAS
(c) ( d)
FIGURE 8 The influence of sparger location on gas distribution in airlift reactors: poor distribution
o f gas in internal (a) and external (b) loops. Proper sparger positioning (c and d) for improved gas
distribution. Hatched areas indicate filled-in zones to improve liquid flow and to prevent biomass
settling.
above the base of the downcomer the hydrostatic head on it is lower than if it
were positioned conventionally and hence the power demands are correspond-
ingly low. In the operation of this type of reactor initially all gas is injected at the
base of the riser until a high rate of fluid circulation has been established, then
most of the air flow is gradually transferred to the downcomer sparger. The
hydrodynamic metastability of the flow pattern produced in this manner can be a
serious problem5hnd flow reversal (i.e., downcomer becoming the riser) may
occur leading to operational difficulties. In relatively small sizes of airlift reactors
usually examined in academic institutions stable fluid circulation during down-
comer gas injection is extremely difficult to obtain and maintain and consequently
this mode of operation has not been investigated. Kubota er aLS2have proposed
an algorithm, based on the balance of hydrostatic and frictional pressure drops,
for predicting the gas injection rate necessary to maintain liquid circulation in
these reactors. An additional drawback of this form of operation is that even for a
low oxygen demand gas throughput may not be reduced sufficiently because it is
dictated by the need to maintain a stable liquid circulation. Consequently, high
mass transfer economies would occur only when a large oxygen demand exists.s2
Another novel strategy, not examined so far, would be to position the gas sparger
some distance up the riser. This would reduce power consumption and the ease of
startup and stable operation would remain. The region of the riser below the
AIRLIFT BIOREACTORS 211
sparger could be aerated either by recirculating gas from the downcomer or by
supplementary spargers located either in the downcomer or further down in the
riser.
3.3.1 Substrate injection points The problem of location of the substrate feed
points in an airlift vessel is similar to that of gas injection. For rapidly utilized
substrates, the concentration of which must be kept low for reasons such as
substrate toxicity or substrate inhibition, the microorganisms in a tall airlift may
be starved of the substrate only a short distance downstream of the point of
substrate i n j e ~ t i o n . ~Thus,
' multiple substrate feed points may be necessary
axially up a reactor if product yield reduction due to substrate starvation is to be
avoided. The recent work of Fields and SlaterS5which involved cultivation of the
bacterium Methylophilus methylotrophus on methanol further highlighted this
point. The substrate balance for a differential volume of the riser may be written
as:
The few existing studies (e.g. 57) have treated only the overall pattern of fluid
circulation mostly in water or salt solutions. Here we report some of our
observations in aqueous solutions (water, 0.15 kmolm-' NaCI) and non-
Newtonian media (CMC solutions K = 0.03-0.11 Pas", n = 0.68-0.84, a = 80-
83mNm-') in an internal loop airlift of rectangular cross-section (Figure 9)
sparged with air via a perforated plate (20 holes of 0.001 m diameter) located at
the bottom of each riser. Influence of sparger location on flow behaviour has
already been discussed and will not be repeated here.
At very low gas flow rates the bubbles rose in the riser only and the rate of
liquid circulation was quite low. As the gas velocity increased the liquid
circulation rate also increased and some gas bubbles were dragged by the liquid
into the downcomer. The bubble size in the downcomer was lower than in the
riser since the liquid flowing into the downcomer was incapable of sweeping
larger gas bubbles with high rise velocities into it. The depth of penetration of a
bubble into the downcomer depended on its size and the tiniest bubbles
penetrated the deepest. Having attained a certain penetration depth the gas
bubbles remained more or less static and in this condition the buoyancy force on
the bubble was exactly balanced by the downward drag force due to the flowing
liquid. As the gas velocity increased further, progressively larger bubbles were
also swept into the downcomer and the depth of bubble penetration in it
increased. Eventually a recirculation of most bubbles, i.e. complete flow through
the downcomer, was obtained. Throughout, the bubble size distribution in the
AIRLIFT BIOREACTORS
downcomer was narrower and the mean bubble size smaller than in the risers.
The extent of liquid turbulence in the downcomer was also significantly lower.
There were two reasons for this: (i) some energy was used up in forcing the
bubbles against the buoyancy-dictated direction of flow, and, (ii) the net loss of
energy from the liquid to the gas which was used up in gas compression. Both
riser and downcomer turbulence increased with gas flow. The residence time of
gas bubbles in the downcomer was longer than their residence time in the riser.
The bubble density was lower in the downcomer than in the riser and because of
this bubbly flow existed in the downcomer even when the flow in the risers had
changed to coalesced bubble or churn-turbulent type.
As the liquid exited the downcomer and turned into the risers the downward
component of its velocity reduced so much so that the bubbles were not ejected
to any significant depth in the region between the end of the downcomer and the
reactor base. Instead most of the bubbles which came out of the bottom of the
downcomer entered the riser just under the edge of the downcomer baffle.
Consequently a small zone of the reactor located mainly between the lower end
of the downcomer and the reactor base remained largely free of bubbles (Figure
10a). When the downcomer clearance from the reactor base was increased the
volume of gas free zone also increased. Based on these observations we
recommend that the total cross-sectional area for fluid flow from the downcomer
to the riser, i.e. the area just below the baffles separating the downcomer and the
riser in internal loop devices or the cross-sectional area of the bottom horizontal
connection in the external loop units, should not exceed 1.65 times the
downcomer cross-section. This is not so critical for external loops, however.
The reactor shown in Figure 9 was designed as a two dimensional device-a
section through a concentric draught-tube vessel. Stagnant zones were observed
in the corners on the base of the reactor particularly at low gas flow rates. Filling
in of the corners eliminated this problem. A plastic prism was found to smoothen
the turn around of the liquid from the downcomer into the riser. This is
recommended in annulus sparged concentric draught-tube internal loops in which
it takes the shape of a cone as was used in the work of Kawase and ~ o o - Y o u n g . ~ *
The cone should be used even in reactors with otherwise elliptical bottoms. It is
interesting to notice that concern is being expresseds9 about the shapes of the
bottoms of stirred tanks also. The geometric parameters of stirred tank bottoms
so successfully employed by ChudacekS9 may be used for the concentric
draught-tube airlift vessels by replacing the propeller diameter with the diameter
of the draught-tube. Figure lob shows the liquid circulation currents which were
visualized by following the movement of gas bubbles and by the injection of dye.
In the risers there was some liquid downflow and circulation near the walls, but
this was never as strong as has been observed in bubble columns with batch liquid
under identical conditions where large circulation cells have been claimedm to
exist. Figure 10c shows the flow cellsm in a bubble column for comparison with
Figure lob as observed in our airlift. Studies with dyes showed the backmixing in
the riser to be far stronger relative to the downcomer where the flow
approximated to plug flow except near the downcomer entrance where there was
strong mixing. Good mixing was also observed in the reactor headspace region
M.Y. CHISTI AND M. MOO-YOUNG
STRONG
- ~
DOWNFLOW
N E A R WALLS - - -,,
LlOUlD CIRCULATION
IN DOWNCOMER ENTRANCE
- Z O N E O F LlOUlO CIRCU-
LATION I SPARGER EFFECT)
FIGURE 10 Flow visualizations in the rectangular airlift: (a) gas free region below the downcomer;
(b) liquid flow pattern in the reactor; (c) chaotic circulation cells in a bubble column.
particularly above the downcomer where the fluid streams from the two risers met
and reversed their flow direction. As will be seen later this was in keeping with
our observation, confirmed also by others," that mixing times in concentric
draught-tube airlifts decline with increasing liquid volume in the headspace
region.
In CMC solutions the general flow was the same as that described for water and
salt solution. There were important differences, however. Under otherwise
identical conditions the turbulence intensity in any zone of the reactor was less
than was observed for water. The initial bubble size from the sparger was slightly
larger in CMC solutions than in water or salt water. Even at very low gas flow
rates the bubble coalescence in the riser lead to the formation of large bubbles
(>0.025 m in diameter) which increased in size still further by coalescence as they
moved up the riser. Many very small bubbles were also present. The bubbles
from the spargers were displaced toward the far walls (Figure 8a) of the risers due
to the fluid flowing from the downcomer. However, the intense circulation zones
observed at the entrances of the riser in water (Figure lob) were now so much
calmer that even very small gas bubbles rose straight up through them instead of
circulating. Bubbles which moved down the downcomer were generally smaller
than in water. Unlike in water or salt water however, bubble coalescence
occurred in the downcomer and the bubble size increased as the bubbles flowed
down until it became so large that the direction of flow of bubble reversed and it
moved up the downcomer against liquid flow. Further entrappment of smaller
bubbles increased their size even more until small spherical caps (0.02-0.03 m
AIRLIFT BIOREACTORS 215
diameter) formed. The latter did not rise very stably in the centre of the down-
comer due to the higher liquid velocity in the centre and consequent higher resis-
tance to upflow of bubbles, and they moved towards the walls of the downcomer
and rose up along the walls. As the gas velocity in the riser increased further large
spherical caps formed in it. In the downcomer, too, the size and frequency of the
spherical caps increased. However, never did these bubbles attain the dimensions
of the riser or the downcomer and fully developed slug flow was not obtained.
With increasing gas flow the depth of the downcomer where the formation of
spherical caps commenced moved progressively downward. The upper entrance
region of the downcomer was quite turbulent and as a result when the spherical
caps moving up the downcomer reached this zone they were broken up into
smaller bubbles once again and carried down with the flow. Periodically one or
two spherical cap bubbles escaped the downcomer via its entrance. The reactor
became increasingly turbulent at higher gas velocities. These observations pointed
to the distinct possibility, particularly in highly viscous media, of oscillation of
liquid circulation due to large spherical cap slugs momentarily lodged in the
downcomer of an airlift reactor. T o avoid this the minimum downcomer diameter
should exceed the value in which fully developed slugging is possible. An
equivalent hydraulic downcomer diameter of more than 0.1 m may be satisfac-
tory. The maximum stable size of spherical caps in the riser was apparently lower
at the higher gas flow rates due to greater turbulence. In viscous systems the gas
free region between the end of the downcomer and the reactor base (Figure 10a)
which was observed in water no longer existed. This was because gas bubbles now
experienced a greater drag due to higher fluid viscosity and penetrated this zone
with liquid eddies. More violent agitation of the gas-liquid dispersion surface was
noticed in viscous systems due to frequent bursting of the spherical cap bubbles
on the surface. In CMC solution it was noticed at all gas flows that near the inside
edges of the downcomer entrance (Figure 11) there was a zone of trapped gas
which grew in size with gas flow and, despite high turbulence, lead to rapidly
fluctuating separation of liquid flow from the downcomer wall.
The foregoing flow visualization studies covered a very broad range of
superficial gas velocities: 0.01-0.40 ms-' based on the cross-sectional area of the
risers.
I I DOWNCOMER
FIGURE 11 R o w separation and gas attachment to the inside edge of the downcomer entrance in
the rectangular internal loop at (a) low, and (b) high gas flow rates.
216 M.Y. CHISTI AND M. MOO-YOUNG
5. REACTOR DESIGN PARAMETERS
The k J d , ratio for water and Solka-Floc (grade KS-1016) cellulose fibre
suspensions is shown in Figure 12 as a function of the superficial air velocity
(based on the entire column cross-section) in the rectangular airlift depicted in
Figure 9 and in an identical bubble column (the removal of the downcomer from
the airlift in Figure 9 and replacement of the spargers by a single perforated plate
with holes of 0.001 m diameter converted it to a bubble column). Over a wide
range of gas velocities spanning bubble flow and coalesced bubble flow regimes
the k L / d , ratio was constant in any given fluid irrespective of the type of
pneumatic contactor used (Figure 12). This revealed a direct relationship between
--',* I0
8 -
Airlift I
FIGURE 12 kJd, ratio in 0.15 kmolm-' NaCl (@), I wt./vol.% Solka-Floc (SF)(O), 2 wt./vol.%
SF(O), and in 3 wt./vol.% SF ( A ) in the rectangular airlift and the bubble column at various
superficial gas velocities.
218 M.Y. CHISTI AND M. MOO-YOUNG
the mass transfer coefficient kL and the bubble diameter. The results showed3'
that for air-water dispersions and for suspensions where the suspending fluid was
water-like in its rheology, kL could be calculated using the equation
FIGURE 13 Variation of true mass coefficient k , with bubble diameter d , in air-water dispersions
according to (a) Eq. (24), and (b) Eq. (25).
AIRLIFT BIOREACTORS
FIGURE 14 Comparison of experimental k , / d , with values obtained using Eq. (24): (@)
0.15kmolm-"aCl. 1 wt./vol.% SF ( A ) , 2 wt./vol.% SF (m), and 3 wt./vol.% SF (V)in the airlift
and bubble column reactors.
5.1.1 Overall mass transfer and hydrodynamic performance The overall mass
transfer coefficients, k,aL or kLa,, are more useful for practical reactor design
purposes. These have been studied by a number of investigators,28~31-33~38~57~58~7w79
and ourselves, among others, in external and internal loop airlifts of various
types. All the main correlations available for mass transfer and for gas holdup in
airlift vessels are summarized in Table I. Notice that apart from the gas holdup
correlation of Miyahara et al." all other equations in Table I may be reduced, for
a given fluid and reactor geometry, to simple power law type functions of gas
velocity. Similar equations were reported by other investigators.= Expressions of
this form are now recognized to have more fundamental hydrodynamic basis.45
In general, for draught-tube internal loop airlift reactors there is no influence of
either the presence of a draught-tube or of its relative area with respect to that of
the annulus on the overall gas holdup. We confirmed this by a reanalysis of the
data presented by bell^'^ for the internal loop vessels examined by Bello er aL7'
Thus, the total gas holdup in the internal loop airlifts was exactly the same as in a
TABLE I
s-I Wm-'
and
Wm-'
2. External and internal Liquid velocity effects: As above.
loops as in 1
Bello er
(29)
a = 0.56 (water)
o = 0.58 (salt solution)
3. Draught-tube internal Water; sodium sulphate. glycerol and iso-
loops (draught-tube (31) butyl alcohol solutions; U,8, = 0.015 to
spar ed) Chakravarty 0.20ms-'; d,=0.10m;dc,=0.074, 0.059,
% .
ec a/. and 0.045 m; L, = 0.40 rn; L, = 0.026 m.
(36)
F
9
a
gd
6. Draught-tube internal Equation 37: water, pseudoplastic fluids; V1
loops; bubble columns. (37) 0.008 S U , (ms-I) 5 0.285; 0.14 5 d,(m) 5
Kawase and Moo- 0.35; 1 z n z O . 2 8 ; 0.001 5 K(Pas")S 1.22.
Youngs8 and Equation 38: Pseudoplastic fluids, water;
0.008< U,8(ms-L) < 0.084; 0.14 < d,(m) <
(38) 0.305; 1 > n > 0.543 and 0.00089 < K
(Pas") < 2.82.
TABLE I (Continued)
h)
8
Sr. Reactor type
No. and reference Equation Parameter range
for the concentric draught-tube internal loops. Equations (47) and (48) applied to
air-water only. For water and for rheologically complex cellulose fibre suspen-
sions Chisti et al.33 confirmed a similar relationship:
This equation was obtained in external loop reactors identical to those used by
Bello et From these results we conclude that even though the absolute value
of overall gas holdup in external loops is fluid property and reactor geometry
dependent, the relationship between the riser and downcomer holdups is
independent of these factors. This is to be expected unless the reactor headspace
configuration is changed.
The lower downcomer gas holdups in the external loop airlifts relative to the
concentric-tube reactor^'^ were due to the geometric peculiarity of the external
loop devices in which the horizontal connection between the riser and the
downcomer allowed most of the gas to separate from the liquid and little gas was
carried to the downcomer. In general then while the gas holdups for bubble
column and concentric draught-tube internal loop airlifts tend to be similar under
AIRLIFT BIOREACTORS
0.15
BUBBLE
COLUMNS
AND
INTERNAL
LOOPS
FIGURE 15 Comparison of overall volumetric mass transfer coefficients in salt solutions in bubble
columns, concentric draught-tube internal loops and external loop reactors. Based on Weiland and
Onken."'
identical conditions, the gas holdups in external loop airlifts are usually lower. As
illustrated in Figure 15 an identical trend has been observed for the overall
volumetric mass transfer coefficient. According to Bello et a1.,72the contribution
of the downcomer volume to oxygen transfer in airlift reactors was negligible and
most of the transfer took place in the riser. This was said to be due to the lower
gas-liquid relative velocities (slip velocities) in the downcomer. Others have
made similar claims.83 Declining gas holdup and k,aL with increasing apparent
viscosities of homogeneous power law fluids have been ~n
' airlift
reactors and in bubble column^.'^^^^ Introduction of an apparent viscosity in
correlations (e.g. references 58, 73) makes them questionable for bioreactor
design purposes. This is because for many biological fluids, particularly suspen-
sions such as the fungal fermentation broths, flow indices K and n depend quite
strongly on the shear rate range used in their determination. Furthermore, the
value of the shear rate in the bioreactor should also be known if the n value
determined under given conditions of shear is to apply to it. Shear fields in
bioreactors are position dependent and the estimation of either local or global
shear rate experienced by a fluid in a reactor is not only extremely complex, but it
is beyond analytical treatment with the currently available knowledge. As Table I
shows, the effect of airlift reactor geometry on gas holdup and kLaL has
frequently been expressed in terms of the riser-to-downcomer cross-sectional area
ratio. Some investigator^^^,^^.^^ have found the outer column diameter, d,, to be a
significant influence on performance in internal loops. For example, the gas
holdup equation (Table I) reported by Kawase and M ~ o - Y o u n gand , ~ ~which was
claimed to apply also to bubble columns, contained d, as a variable. The columns
226 M.Y. CHISTI AND M. MOO-YOUNG
used by these investigatorsss were all larger than 0.14 m in diameter and for them
there is substantial accumulated evidences7-" that holdup should be diameter
independent. For otherwise identical conditions in concentric tube internal loop
airlifts we found, based on the results reported by Koide et a1.,75.76inexplicable
differences in the dependence of kLaL on reactor geometry for annulus and
draught-tube sparged modes of operation. For the latter, kLaLshowed a strong
dependence on the diameter ratio of the inner and outer tubes while the influence
of outer tube diameter was very pronounced: kLaLa dkos9 (draught-tube sparged)
versus kLaLa dzw' (annulus sparged). In a draught-tube internal loop no effect of
draught-tube clearance (L,) from the column base on gas holdup or kLaL was
observed7"or 1 / 8 5 Lh/dCi5 1. Similarly, no effect of draught-tube length, L,,
on holdup or kLaL was found75 for 5 5 LJd, s 15 in a reactor with d, = 0.140 m
and dCi= 0.082 m. Most of these geometric variables have not been studied in
enough detail to justify definitive conclusions.
It seems that while there is little influence of the area or diameter ratios of
downcomer and riser on gas holdup in internal loop airlift reactors in water, salt
solutions and other low viscosity media such as iso-propanol solutionss7 this may
not always be true in viscous systems. For example: up to 60% higher gas holdups
were obtained by W e i l a r ~ din~ ~glycerol solution (63.4 wt%), which was 14-times
more viscous than water, when the ratio of draught-tube to reactor diameter was
increased from 0.59 to 0.74. For further rise in this ratio the influence on holdup
was slight.
The gas holdup values of interest in design are the total holdup in the reactor as
a whole, in the riser and in downcomer, and those are the values most often
quoted in the literature. In reality there are positional variations in holdup.
Radial gas holdup variations in the riser of airlift reactors have been
documentedw~" in a fashion similar to that observed in bubble columns.6s Higher
holdups near the core than closer to the riser walls were found.w Axial variations
in riser and downcomer gas holdups are also known to o c c ~ r . ~ ' ~ ~ ~
It should be noted that most gas holdup expressions for airlift reactors correlate
the holdup with superficial gas velocity based on the fresh gas input to the riser.
The additional contribution of the recirculated gas is ignored. Because the extent
of gas-liquid separation in the reactor headspace varies widely depending on
design, the amount of recirculated gas also varies. This would explain at least
partly2y the large difference between gas holdups observed by various inves-
tigators in otherwise similar reactors.
Although the gas holdup in airlift reactors may sometimes be lower than in
corresponding bubble columns this does not automatically mean that kLaLis also
lower. Enhanced kLaL has been o b s e r ~ e d ~ "In some
'~ ~ ~ ' ~ cases in the presence of
draught-tubes relative to bubble column operation. This apparent discrepancy is
explicable in terms of the different bubble size distributions which occur in bubble
columns and airlift reactors under otherwise identical conditions. Further
comparison of airlifts and bubble columns occurs in the work of Bello et aLY4
Although all fermentations involve gas, liquid and solid phases, nearly all the
bioreactor studies in simulated media have been limited to gas-liquid systems. In
most cases the liquid used was either water-like or a pseudoplastic homogeneous
AIRLIFT BIOREACTORS
fluid. In a few studies which made use of suspended slurries in airlift vessels (e.g.,
95-98) the solids employed were not a simulation of microbial, particularly fungal
mycelial, solids. In suspensions which were carefully formulated to simulate
fungal fluids we large reductions in gas holdup and mass transfer in an
airlift reactor and in a similar bubble column relative to solid-free operation.
Under otherwise identical conditions the airlift produced marginally lower
holdups and mass transfer than the bubble ~ o l u m n . ~ * . ~ ~
The properties of liquid such as surface tension, density, viscosity and ionic
strength are known to affect gas holdup and overall mass transfer coefficient. The
magnitude of various influences depends on the reactor type. Gas holdup, for
example, has been found3' to be more sensitive to liquid properties in a bubble
column than in an external loop (dispersion height =8.5 m, downcomer
diameter = 0.050 m, riser diameter = 0.1 m) reactor. The gas holdup in aqueous
2-propanol in the bubble column (riser of external loop used as bubble column)
was up to 100% higher than in pure water.3s For the same two liquids the gas
content in the airlift differed only by about 20%.3s
5.1.3 Modelling for mass transfer and economic modelling In the experimental
determination of the overall volumetric mass transfer coefficient, kLaL, certain
assumptions must be made about the extent of mixing in the liquid (slurry) and
gas phases in the reactor. Usually, either constant composition or plug flow for
the gas phase is assumed, and liquid is considered to be either plug flow or
backmixed. In most cases, even for quite tall airlift reactors, the assumption of
fully backmixed liquid phase is satisfactory. The contribution of very small gas
bubbles to mass transfer is small, while larger bubbles rise so rapidly that their
228 M.Y. CHISTI AND M. MOO-YOUNG
composition is nearly constant. For example, it may be shown1' that for an air
bubble of 1 mm diameter in a water-like fluid the time taken for 63% of the
possible oxygen transfer to have occurred is approximately 50 s. If a bubble rise
velocity of 0.28 ms-I is assumed then the residence time of the bubble in a 5 m
tall pool of liquid is only about 18 seconds. For larger bubbles the extent of
oxygen transfer during the residence time of the bubble in the reactor would be
even lower. For the liquid phase it has been s h ~ w n "that ~ ~ if~ the
~ criterion
k,aLtc 5 2 (t, = liquid circulation time) is satisfied then the well-mixed liquid
phase assumption is correct. A fully backmixed tank-in-series model for liquid
mixing in airlifts, utilizing a different number of tanks in the riser and
downcomer, was proposed by H o et a1.I6 Computer simulation of the dissolved
oxygen profiles based on the model showed these profiles to be quite flat with
column height for airlifts up to 6 m tall.I6 Even in columns as much as 9 m tall,
more or less uniform oxygen concentration profiles were obtained.I6 Similar
results were reported in another s i m ~ l a t i o n We
. ~ ~have experimentally confirmed
quite flat steady state axial dissolved oxygen profiles in the riser of a 6 m tall split
cylinder airlift (d, = 0.243 m, AdIA, = 0.411, L, = 4.8 m, Lb = 0.102 m) in water
as well as in highly viscous cellulose fibre suspensions even for a low riser
superficial gas velocity of 0.039ms-'. Because stage to stage variation in the
dissolved oxygen level is small, there may be some justification for utilizing the,
much simpler, well mixed liquid phase in kLaLcomputations. Observations such
as those of Lin et and Onken and weilandlo4 have shown, however, that
complete backmixing does not always exist particularly in the presence of
microorganisms when oxygen consumption is involved.
Economic modelling of airlift reactors is mostly lacking. For high mass transfer
performance tall airlifts have been r e c ~ m m e n d e d . ' ~ ,Moresi13
~' modelled the
oxygen mass transfer in a draught-tube airlift based on perfect mixing of liquid
and plug flow in gas, and for oxygen consumption appropriate to a continuous
whey fermentation by Kluveromyces fragilis. Operating costs were concluded to
decline asymptotically with increasing aspect ratio (Lld,) of the airlift. For
fermentation volumes s 100 m3, aspect ratios greater than 30 led to lower
operating costs, while for volumes ranging from 250 to 1000m3, the optimal
aspect ratio decreased to 15.13 As far as power consumption per unit of oxygen
transfer was concerned, the resvonses of the model were influenced mainlv. bv. the
equations used for estimating k,a,. Operation of the airlifts at headspace
pressures higher than atmospheric was found less expedient than atmospheric
]@ER DETECTOR
TRACER
INLET
DETECTOR
FIGURE 16 Typical responses to pulse input of tracer in (a) airlifl loop reactors, and (b) batch
bubble columns.
230 M.Y. CHISTI A N D M. MOO-YOUNG
to the experimental tracer response c ~ r v e ; ~with . ' ~ Peclet
~ number (Pe) as the
fitting parameter. In practice Eq. (50) evaluated for the limits j = 0 f 2 may be
s u f f i ~ i e n t .In~ Eq. (50) C and CE are the instantaneous and final (equilibrium)
tracer concentrations, respectively, c is the dimensionless tracer concentration
(= C/CE), 6 is the dimensionless time and j is the dimensionless distance. 0 and j
are given by
and
where t and t, are the instantaneous time and the mean liquid circulation time,
respectively; and x, and x, are the total distance travelled by an element of fluid in
the direction of flow and the distance covered in one circulation, respectively.
The Peclet number in Eq. (50) is defined as
4 was 1.55 and 0.66 for the external and internal loop reactors, respectively, and
had the respective values of 0.74f 0.04 and 0.78f 0.08.113 Higher liquid
AIRLIFT BIOREACTORS 23 1
circulation in external loop reactors relative to internal loop systems occurred
because of the peculiar geometry of the former which allowed substantial gas
disengagement before the liquid entered the downcomer and this resulted in a
higher driving force for circulation in external loops. In almost the same external
loop vessels as used by Bello et d 1 1 3 another i n ~ e s t i g a t i o nconducted
~~ in
non-Newtonian CMC solutions (pap, of 0.015 to 0.5 Pas) showed decreasing
liquid circulation with fluid viscosity according to:
where cu was 0.052 and 0.0204 for bubble- and slug-flow regimes, respectively.
Notice that the exponents on the gas velocity and on the geometric term in Eq.
(56) were almost the same as reported earlier113for water and salt solution thus
pointing to their independence of fluid properties. Decline in liquid circulation
with increasing fluid viscosity has been observed by other researcher^^^^^^^^^^ also;
it arises probably because higher viscosities enhance internal friction losses. In
solutions containing small amounts of a drag reducing polymer such as xanthan
gum the liquid circulation was more rapid than in water.li5 Even in these fluids
increased viscosity at higher polymer concentration slowed down the circulation
of liquid.'15 For relatively large solids (d, =0.0027 m) dispersed in water in an
external loop device reducing liquid circulation with increase in solids loading has
been reported.97 We have observed similar results for mycelia-like solids in
external and internal loop reactors. For draught-tube gas sparged internal loops
weilandS7 recommended d J d , ratios between 0.8 and 0.9 for efficient mixing
(and oxygen transfer). Largest liquid circulation rates were reported at diameter
ratios of 0.5957 and approximately 0.5.1m Another report116 claimed minimal
mixing times for 0.6 < d,ld, < 1.0 (where dCi= d, corresponded to bubble column
operation) and for minimum mixing time equal riser and downcomer cross-
sectional areas (i.e. dcldci= d 2 ) were recommended116 in internal loops. Com-
paring bubble columns and concentric draught-tube internal loops with either one
or two draught-tubes, Margaritis and Sheppardz8 found that the presence of
draught-tubes lead to systematically higher mixing times relative to bubble
column operation. In addition, the double draught-tube configuration resulted in
higher mixing times than the single draught-tube geometry28 probably because of
the extended liquid circulation path in the former. Other similar observations
have been reported.l17 A problem with many existing correlations for liquid
circulation in airlift reactors is that they are not very successful for liquid velocity
prediction in anything other than the particular reactors used for obtaining them.
For low viscosity Newtonian fluids this difficulty has been overcome with a
recently proposed118 theoretical equation which related the superficial liquid
velocity in the riser (UL,) to various geometric and operating parameters as
follows:
232 M.Y. CHISTI AND M. MOO-YOUNG
KT and K B in Eq. (57) are the dimensionless frictional loss coefficients for fluid
turn around in the top and bottom riser-downcomer connecting zones, respec-
tively. Nearly all the published air-water liquid circulation data70~92~97~106~109"11~119
including that of the authors' could be correlated with Eq. (57) within f30%.
The data used included external and internal loop airlift reactors, the latter in
annulus sparged, draught-tube sparged and split cylinder geometries. Broad
ranges of reactor volume, reactor dispersion height, AJAd ratio and circulation
velocity were covered by Eq. (57) which demonstrated its usefulness as a scale-up
tool. The authors118 pointed out that Eq. (57) may be simplified for particular
reactor types and they also discussed the dependence of frictional loss coefficients
on reactor geometry.
Another empirical liquid circulation correlation specific to draught-tube inter-
nal loop reactors was reported79 recently for a limited amount of air-water data.
The mixing characteristics of the riser, downcomer and the headspace in airlift
vessels can be very different. Our tracer studies have shown that the reactor
headspace region is the best mixed whereas the downcomer has the poorest
mixing performance. Similar results have been found in a large external loop
In internal loops with concentric draught-tubes Weiland" observed that
the mixing time declined with increasing volume of liquid in the reactor
headspace above the draught-tube (Figure 17). Our observations partly confirmed
this trend. In a tall split cylinder airlift (d, = 0.243 m, A d / A ,= 0.411, L, = 4.8 m;
Lb = 0.102 m; air-water) we found improved mixing performance with increasing
height (h,) of liquid above the baffle up to a height of about 0.5 m. Further
increase in hH lead to different results: while the mixing in the zone above the
0 0
0 0.10 0.20 0.30 0.40 0.50
DISTANCE h, ( m )
FIGURE 17 Effect of the clear liquid height in the reactor headspace above the draught-tube on
mixing time (95% mixing) in air-water in an internal loop reactor (draught-tube sparged at
U,, = 0.0105 ms-', dJd, = 0.76). Data from ~ e i l a n d . ~ '
AIRLIFT BIOREACTORS 233
baffle (Figure 2a) remained good, the intermixing of the fluid in this zone with
that in the baffled height of the split cylinder slowed down markedly. There was
no contradiction between our results and those of weilands7 because the results
of the latter (Figure 17) did show a definite flattening of mixing time versus h,
curve near hH of 0.4 to 0.5m. Based on our results we recommend that the clear
liquid height above the top of the baffle (or draught-tube) should not exceed
about 0.5m. This value is independent of the total height of the airlift reactor.
Significant reductions in mixing time may also be obtained by horizontally
splitting (Figure 2c) the central draught-tube of internal loops into two or more
vertical sections.99 We are currently experimenting with a split cylinder airlift with
multiple baffles.
Because the liquid circulation is very sensitive to the gas holdup difference
between the riser and the downcomer of an airlift vessel, incorporation of various
types of gas-liquid separators in the headspace would enhance circulation and
mixing depending on the separator efficiency. This in fact is the reason for the
different circulation performances of internal and external loop airlifts. In the
former there usually is no gas-liquid separator whereas in the latter the top
connection between the riser and downcomer acts as a gas-liquid disengagement
unit. Gas-liquid separator design considerations have recently been d i s c u s ~ e d . ~ ~
For best circulation performance separators should be designed for minimum flow
resistance.
Liquid circulation effects in some actual fermentation systems have been
discussed by Schiigerl et al. lZO
where Q (kWm-3) and Ro, (kgm-3s-') are the rates of heat evolution and
oxygen consumption, respectively. Once the heat transfer rate, which equals the
heat evolution rate plus the heat generation due to agitation at steady state, is
established then the reactor wall area needed to obtain this rate is calculated
from:
where h's are the individual heat transfer coefficients and the subscripts i, f, m
and o refer to heating/cooling fluid film, fouling, metal wall, and the film between
the fermentation broth and heating/cooling surface, respectively. Empirical
correlations and methods of calculation for hi, hv and h, abound in chemical
engineering texts and works on heat transfer; her, the fouling coefficient between
the broth and reactor wall, is essentially constant and typical data given in
handbooks should apply. The calculation of h,, the film coefficient between the
reactor fluid and the heating/cooling surface, is the object of research in airlift
and other bioreactors. For airlift vessels with very low rates of fluid circulation
the heat transfer data obtained in batch bubble columns may apply. This work
has been reviewed by other^.^^.^^.'^^ For air-water system in industrial batch
bubble columns (diameters> 0.45 m) Fair el found the film coefficient h, to
increase with superficial gas velocity in the following way:
which also predicted an increase in the film coefficient with gas velocity raised to a
power of 0.25. In Eq. (62) k and C,, are, respectively, the thermal conductivity
and the specific heat capacity of the liquid in the reactor. Compared to bubble
columns, heat transfer coefficients in airlift reactors are more than 2-fold
higher,li2 due to higher liquid velocities in the latter.
Only two correlations for film heat transfer coefficient in concentric-tube
internal loop airlift reactors are available. That of Chakravarty et a1.Iz3 which is:
in which U, is based on the riser diameter. Equation (63) was obtained using
Newtonian fluids ( p L = 0.78 - 5.27 mPas) in draught-tube sparged airlifts over
approximate U, and A,/A, ranges, respectively of 0.008 to 0.16 ms-', and 0.25 to
1.20. The actual values of h, variedIz3 from 0.6 to 2.4 kWm-zOC-l. The second
equation was proposedIz4 by us for air-water in concentric draught-tube internal
AIRLIFT BIOREACTORS
loops:
, .
Equation (64) covered a gas velocity (based on total reactor cross-section) range
of 0.01 to 0.04 ms-'. whereas A -J A ., values of 0.242 and 0.452 were used in its
determination. Equation (64) predicts significantly higher heat transfer
coefficients than does correlation (63). . . The differences could be due to different
types of gas sparging modes used: annulus sparged as used by us us. draught-tube
sparged as used by Chakravarty et ~ 1 . Additionally,
' ~ ~ the geometric configura-
tions of the bottoms of the reactors used by the two groups of investigators were
quite different and may have lead to different liquid velocities for otherwise
similar conditions. S t u d i e ~ ' ~of~ .heat
' ~ ~transfer in airlift vessels in suspensions of
Aspergillus niger and PeniciNium chrysogenum claimed that the presence of
mycelia enhanced film heat transfer coefficients. The data presented'z showed h,
values as high as 8 kWm-'K-'. A more detailed studyJZ4conducted with
cellulose fibre suspensions which simulated mycelial media showed that the
dependence of heat transfer on solid contents can be very complex. Solids may
enhance or reduce heat transfer depending on the location (riser or downcomer)
of heating/cooling surface in the reactor. Useful recommendations on the
practical problem of positioning of heat exchange devices in airlift vessels were
presented by Chisti and coworker^.'^^ Nishikawa et ~ 1claimed . ~ film~ heat transfer
coefficient to be independent of liquid velocity for UL5 0.015 ms-', however for
higher liquid velocities h, depended on the velocity of liquid as follows:
Heat transfer work done in vertical two-phase flows may be applicable to airlift
reactors provided that the fluid properties, gas holdup and relative velocities of
the two phases are identical for the airlift and vertical two-phase flow device.
In many fermentations, particularly those which have high solids content or
which produce very viscous broths, the presence of heat transfer coils and similar
projections in the reactor is undesirable because of the formation of stagnant
zones, problems with cleaning and sterilization. Consequently, jacket form of
heating/cooling is preferred. As the reactor volume increases, however, the
surface-to-volume ratio declines and a point is soon reached beyond which a
jacket alone is insufficient to handle the heat duty. In airlift reactors this problem
is largely overcome by using double walled draught-tubes which provide
additional heat transfer surface, and by other similar modifications.
6. CONCLUDING REMARKS
While the practical feasibility of airlift reactors for a variety of fermentations has
already been established, systematic studies of these devices have only just
begun. Even with very limited knowledge a coherent picture of some aspects of
236 M.Y. CHISTI AND M. MOO-YOUNG
the airlift reactors can be made as has been done ,in the foregoing sections of this
review. Specific design recommendations appear throughout the review. For low
viscosity systems it is now possible to predict liquid circulation velocity for known
gas holdup using well known principles of fluid mechanics. This ability is lacking
for viscous Newtonian and rheologically complex media. All existing heat and
mass transfer and gas holdup correlations for airlift reactors are essentially
empirical although possible underlaying hydrodynamic basis for them has been
discussed. It would be most useful to show theoretically the effects of various
geometric and operational variables on reactor performance parameters. Mechan-
ically simple reactor designs-split cylinder airlift for e x a m p l e n e e d to be
studied. Gas-liquid separator designs as well as multipoint riser/downcomer gas
injection should be examined.
The problem of quantification of local and global shear rates in airlift vessels is
of critical importance not only because of the shear sensitivity of biological
materials, but also because of the shear dependant flow behaviour of non-
Newtonian systems. This problem remains to be addressed.
ACKNOWLEDGEMENT
This work was supported by a grant from the Natural Sciences and Engineering
Research Council of Canada.
NOMENCLATURE
Greek Letters
LY A constant
B A constant
Y Shear rate T-'
EG Overall fractional gas holdup (-)
E G ~ Downcomer fractional gas holdup (-)
EG~ Riser fractional gas holdup (-)
rl Efficiency (-)
0 Dimensionless time (= tlt,) (-)
pmPp Apparent viscosity of nowNewtonian liquid ML-'T-'
Gas viscosity ML-'T-'
PL Liquid viscosity ML-IT-'
AIRLIFT BIOREACTORS
PW Viscosity of water ML-IT-'
PC Density of gas ML-3
PL Density of liquid ML-3
u Surface tension MT-2
4) A constant
111 A constant
Abbreviations
CMC Carboxymethyl cellulose
SCP Single cell protein
SF Solka-Floc
VVM Volume of gas per volume of liquid per minute
REFERENCES
1. Kossen, N.W.F., Paper presented at the 3rd EFB Conference, Munich, 10-14 September, 1984.
2. Bailey, J.E., Chem. Eng. Sci., 35, 1854 (1980).
3. Maclean, G.T., Erickson, L.E., Hsu, K.H., and Fan, L.T., Biotech. Bioeng., 19, 493 (1977).
4. Viesturs, U.E., Sturmanis, I.A., Krikis, V.V., Prokopenko, D., and Erickson, L.E., Biotech.
Bioeng., 22, 799 (1980).
5. Viesturs, U.E., Sturmanis, LA., Krikis, V.V., Prokopenko, V.D., and Erickson, L.E., Biotech.
Bioeng., 23, 1171 (1981).
6. Baird. M.H.I.. and Garstane. J.H.. Chem. Enn. Sci.. 27. 823 (1972) \ -1
7. ~ e i t e l G.,
, and Onken, ~ . , % e r .Chem. ~ n ~ . , - 250
4 , (1981).
8. Russell, T.W.F., Dunn, I.J., and Blanch, H.W., Biotech. Bioeng., 16, 1261 (1974).
9. Ziegler, H., Meister, D., Dunn, I.J., Blanch, H.W., and Russell, T.W.F., Biotech. Bioeng., 19,
507 (1977).
10. Stein, W.A., and Schafer, H., Ger. Chem. Eng., 6, 91 (1983).
11. Stein, W.A., and Schafer, H., Ger. Chem. Eng., 7 , 115 (1984).
12. Zanker, A,, Chemical Engineering, 21 November (1977), pp. 207.
13. Moresi, M., Biotech. Bioeng., 23, 2537 (1981).
14. Blakebrough, N., Shepherd, P.G., and Nimmons, Biotech. Bioeng., 9, 77 (1967).
15. Westlake, R., Chem. Ing. Tech., 58, 934 (1986).
16. Ho, C.S., Erickson, L.E., and Fan, L.T., Biotech. Bioeng., 19, 1503 (1977).
17. Malfait, J.L., Wilcox, D.J., Mercer, D.G., and Barker, L.D., Biotech. Bioeng., 23, 863 (1981).
18. Erickson, L.E., Patel, S.A., Glasgow, L.A., and Lee, C.H., Process Biochem., 18(3), 16 (1983).
19. Koenig, B., Seewald, C., and Schiigerl, K., in Advances in Biotechnology, Vol. 1 (Moo-Young,
M., Robinson, C.W., and Vezina, C., eds.), Pergamon Press (Toronto), (1981). p. 573.
20. Seipenbusch, R., Birckenstaedt, J.W., Blenke, H., and Schindler, F., Abstracu: Fifrh
International Fermentation Symposium, Berlin (1976), p. 65.
21. Hatch, R.T., in Single Cell Protein 11, Tannenbaum, S.R., and Wang, D.I.C., (Editors), The
MIT Press (Cambridge), (1975), pp. 46.
22. Huang, S.Y., Yeh, M.C., and Liou, K.T., Abstracu: Fifrh lnrernational Fermentation
Symposium, Berlin (1976). pp. 68.
23. Smart, N.J., and Fowler, M.W., Journal of Experimental Botany, 35 (153), 531 (1984).
24. Gallo, T., and Sandford, D.S., Paper . presented
. - o.f A I C h E . A ~ r i l .1-5.
at 86rh National Meen'nn
1979, Houston, Texas.
25.
- ~
Hines. D.A.. Bailev. M.. Ousbv. J.C.. and Roesler., F.C.. ~ Paoer
, oresented
. , at the IChemE-
~
413 (1983j.
-
76. Koide. K.. Kurematsu.. K... Iwamoto.. S... Iwata.. Y... and Horibe.. K... J. Chem. Enn. Jon..
' 16f5).
77. Andrt, G., Ph.D. Thesis, University of Waterloo, Ontario (1982).
78. El-Gabbani, D.H., MASc Thesis, University of Waterloo, Ontario (1977).
79. Kawase. Y.. and Moo-Youne. M.. J. Chem. Tech. Biotechnol.. 36. 527 (1986).
80. ~hakravarty,M., Begum, S, Singh, H.D., Baruah, J.N., and l y e n g a r , ' ~ . s ,Biotech. Bioeng.
Symp. No. 4, 363 (1973).
81. Miyahara, T., Hamaguchi, M., Sukeda, Y., and Takahashi, T., Canad. J. Chem. Eng., 64, 718
(1986).
82. ~ishikawa,M., Kato, H., and Hashimoto, K., Ind. Eng. Chem. Process Des. Develop. 16, 133
\11977).
--..,-
83. McManamey, W.J., and John Wase, D.A., Biotech. Bioeng., 28, 1446 (1986).
84. Deckwer, W.-D., Nguyen-Tien, K., Schumpe, A,, and Serpemen, Y., Biotech. Bioeng., 24, 461
(1987)
\----I'
85. Henzler, H.J., Chem.-1ng.-Tech., 53, 634 (1980).
86. Nakanoh, M., and Yoshida, F., Ind. Eng. Chem. Process Des. Develop., 19, 190 (1980).
87. Akita, K., and Yoshida, F., Ind. Eng. Chem. Process Des. Develop., U,76 (1973).
88. Kataoka, H., Takeuchi, H., Nakao, K., Yagi, H., Tadaki, T., Otake, T., Miyauchi, T.,
Washimi, K., Watanabe, K., Yoshida, F., I . Chem. Engr. Jpn., 12(2), 105 (1979).
89. Fair, J.R., Lambright, A.J., and Andersen, J.W.. Ind. Eng. Chem. Process Des. Develop., 1, 33
(1967)
\-'--,'
90. Lippert, I., Adler, I., Meyer, H.D.. Lubbert. A., and Schiigerl. K.. Biotech. Bioeng., U,437
(1983).
91. Menzel, T., Kantorek, H.J., Franz, K., Buchholz, R., and Onken, U., Chem.-lng. Tech., 57(2),
S139 (1985).
92. Merchuk, J.C., and Stein, Y., A I C h E I . , 27, 377 (1981).
93. Moo-Young, M., and Kawase, Y., Canad. I . Chem. Eng., 65, 113 (1987).
94. Bello, R.A., Robinson, C.W., and Moo-Young, M., in Advances in Biotechnology, Vol. 1,
(Moo-Young, M., Robinson, C.W., and Vezina, C., eds.). Pergamon Press (Toronto), (1981).
p. 547.
95. Fan, L.-S., Fujie, K., and Long, T.-R., AlChE Symp. Ser. 80 (241), 102 (1984).
96. Fan, L.4. Hwang, S.-J., and Matsuura, A,, Chem. Eng. Sci. 39, 1677 (1984).
97. Verlaan, P., Tramper, J., van't Riet, K., and Luyben, K.Ch.A.M., Paper presented at the
International Conference on Bioreactor Fluid Dvnamics. Cambridee (Eneland). 15-17 Aoril.
- ~ u