Pneumatically agitated bioreactors in

industrial and environmental bioprocessing:

Hydrodynamics, hydraulics, and transport phenomena
Yusuf Chisti
Department of Chemical Engineering, University of Almeria, E-04071 Almeria, Spain

M a j o r a s p e c t s o f d e s i g n a n d o p e r a t i o n o f p n e u m a t i c a l l y a g i t a t e d b i o r e a c t o r s are r e v i e w e d . T h e fo-
c u s is o n c o n s i d e r a t i o n s t h a t are r e l e v a n t to industrial practice. Airlift b i o r e a e t o r s a r e e m p h a s i z e d .
T h e t r e a t m e n t c o v e r s h y d r a u l i c s , h y d r o d y n a m i c s , g a s - l i q u i d a n d s o l i d - l i q u i d m a s s transfer, h e a t
transfer, m i x i n g , a n d s u s p e n s i o n . N e w t o n i a n a n d n o n - N e w t o n i a n s y s t e m s are discussed. A p p l i c a -
t i o n s in m i c r o b i a l f e r m e n t a t i o n s , a n i m a l a n d p l a n t cell culture, b i o t r a n s f o r m a t i o n s w i t h i m m o b i -
lized e n z y m e s , a n d t r e a t m e n t o f w a s t e w a t e r are outlined. C o m p a r i s o n s w i t h m o r e c o n v e n t i o n a l
b i o r e a c t o r t e c h n o l o g i e s are m a d e . D e s i g n features for sterile p r o c e s s i n g in airlift s y s t e m s are d e -
tailed. T h e e v i d e n c e for s u p e r i o r p e r f o r m a n c e o f airlift b i o r e a c t o r s is o v e r w h e l m i n g . E x c e l l e n t
p r o d u c t i v i t i e s h a v e b e e n d e m o n s t r a t e d w i t h yeasts, bacteria, a n d f i l a m e n t o u s fungi. P r o c e s s e s t h a t
p r o d u c e h i g h l y v i s c o u s broths, i n c l u d i n g several b i o p o l y m e r p r o d u c i n g f e r m e n t a t i o n s , h a v e b e e n
p r o v e n in airlift devices. Similarly, m a n y h y b r i d o m a c u l t u r e s a n d p l a n t cell s u s p e n s i o n s h a v e g i v e n
g o o d results. A s a g e n e r a l rule, v o l u m e t r i c p r o d u c t i v i t y o f airlift b i o r e a c t o r s e q u a l s o r b e t t e r s t h a t
o f c o n v e n t i o n a l stirred tanks. Typically, this level o f p e r f o r m a n c e is a c h i e v e d at s u b s t a n t i a l l y l o w e r
p o w e r i n p u t t h a n in stirred vessels. F u r t h e r m o r e , the p r o b a b i l i t y o f m e c h a n i c a l failure a n d likeli-
h o o d o f loss o f sterility are l o w e r w i t h airlift bioreactors. In w a s t e w a t e r t r e a t m e n t , too, airlift d e -
v i c e s h a v e far o u t p e r f o r m e d c o n v e n t i o n a l s y s t e m s . Airlift b i o r e a c t o r s a c c e p t h i g h e r B O D l o a d i n g s ,
p r o d u c e less sludge, a n d the d e g r a d a t i o n rate is faster; p e r f o r m a n c e i m p r o v e s w i t h i n c r e a s i n g scale
o f o p e r a t i o n . T h i s r e v i e w article i n c l u d e s 328 references.

CONTENTS
I INTRODUCTION .............................................................. ~..... 34
I. I Pneumatically agitated reactors........................................ 34
2 PNEUMATIC AGITATION ..................................................... 35
2.1 Power input ...................................................................... 35
2.2 Gas velocity ..................................................................... 35
2.3 Pressurized operation ....................................................... 36
3 GAS HOLDUP ......................................................................... 36
3.1 Bubble columns ................................................................ 36
3.2 Airlift reactors .................................................................. 38
3.3 Local events ..................................................................... 45
LIQUID CIRCULATION AND HYDRAULICS ...................... 46
4.1 Induced liquid circulation rate ......................................... 46
4.2 Local phenomena ............................................................. 53
4.3 Numerical simulations of flow ......................................... 53
4.4 The gas-lift pump .......................................................... ~.. 55
GAS-LIQUID SEPARATION ................................................... 55
5.1 Separator design .................................................... ........... 56
SUSPENSION OF SOLIDS ...................................................... 57
6.1 Dead zones and flow instabilities..................................... 59
MIXING .......... . ........................................................................ 60
7.1 Liquid and slurry phases .................................................. 60
7.2 Gas phase mixing ............................................................. 65
7.3 Mixing of solids ............................................................... 65
TRANSPORT PHENOMENA ................................................. 66
8.1 Gas-liquid mass transfer .................................................. 66
8.2 Solid-liquid mass transfer ................................................ 71
8.3 Heat transfer..................................................................... 73
9 CONSIDERATIONS FOR BIOFILM REACTORS .................. 75
9.1 The biofilm ....................................................................... 75
l0 HYDRODYNAMIC SHEAR FORCES ................... i................ 77
I 1 UNCONVENTIONAL AIRLIFT REACTORS ......................... 78
I 1.1 A new mode of operation of internal-loop reactors .......... 78
11.2 Reactors with undulating risers ......... . .............................. 79
11.3 Reactors with perforated draft tube ...2.............................. 80
I 1.4 Cascaded airlift reactor..................................................... 81
11.5 Inverse fluidized airlift reactor ......................................... 83
! 1.6 Airlift-packed bed hybrid reactors .................................... 83
11.7 Liquid-impelled reactors................................................... 87
11.8 Reactors with static mixers ............................................... 88
12 APPLICATIONS ....................................................................... 90
12.1 Microbial products ........................................................... 90
12.2 Biodegradation of environmental pollutants..................... 93
12.3 Animal cell culture ........................................................... 96
12.4 Plant cell culture ............................................................... 99
12.5 Algal culture ................................................................... 100
12.6 Chemical processes ........................................................ 101
13 COMPARISON OF PROCESS PERFORMANCE ................. 101
14 OTHER CONSIDERATIONS ................................................. 103
14.1 Sterile processing ........................................................... 103
14.2 Materials......................................................................... 103
14.3 Surface finish .................................................................. 103
15 CONCLUDING REMARKS ................................................... 104
16 NOMENCLATURE ................................................................ 104
REFERENCES ............................................................................. 106

part of FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY,edited by R Pohorecki

ASME Reprint No AMR233 $70
Appl Mech Rev vol 51, no 1, January 1998 33 © 1998 American Society of Mechanical Engineers

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 34 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
I SOLID-LIQUID~
MASSTRANSFER
l i LIGUIDVELOCI
I t
t GAS-LIQUID
MASSTRANSFER
Fig 1. hlterrelationships among the hydrodynamic parameters and .
other bioreactor performance indices.
1. INTRODUCTION
This review concentrates on hydrodynamics, related
transport phenomena, and design of pneumatically agitated
bioreactors. Especial emphasis is placed, on airlift
bioreactors as being the most important among the
pneumatically agitated reactors. Design and operational
features that are relevant to industrial practice are stressed.
The treatment covers hydraulics, hydrodynamics, gas-liquid
and solid-liquid mass transfer, heat transfer, mixing and
suspension. Newtonian and non-newtonian systems are
considered. Applications in microbial fermentations,
animal and plant cell culture, biotransformations with
immobilized enzymes and treatment of wastewater are
discussed. Comparisons with more conventional bioreactor
technologies are made. Design features for sterile
processing are detailed.
Hydrodynamics, transport phenomena, microbial
survival and production kinetics in bioreactors are
interrelated. These interrelationships are quite complex as
depicted in Figure 1. A variable may be directly or
indirectly affected by several others in sometimes not so
obvious ways. As an example, the induced liquid
circulation in an airlift reactor depends on the gas holdup
difference between the riser and the downcomer zones and
on the geometry of the reactor. However, the gas holdup in
individual zones, and the holdup difference, are in turn
controlled by the rate of liquid circulation and the properties
of the fluid. Liquid flow velocity and gas holdup combine
to affect mixing, gas-liquid mass transfer and heat transfer.
Suspension of solids and solid-liquid masstransfer are
affected. Turbulence in the fluid generally determines the
bubble size and hence the interracial shear rate. Shear
stresses associated with turbulence impact upon microbial
morphology, floc size, and physical survival of fragile cells.
The ability to heat and cool a bioreactor, supply oxygen and
remove carbon dioxide depends on hydrodynamics.
Ultimately the kinetics of the bioreaction are affected and so
is the productivity of the reactor and the feasibility of the
process (Chisti and Moo-Young, 1996). An understanding
of the hydrodynamic factors and how they influence the
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 5'1, no 1, January 1998
SUSPENSIONOF
SOLIDS
I TYi.
BUBBLERISE
VELOCITY
GASHOLDUP [~-"
i-
productivity or performance is essential to bioreactor
design, scale-up and operation. Insight into hydrodynamics
and related phenomena is especially relevant to large scale
bioprocesses for bulk products, wastewater treatment and
environmental remediation. Performance and
competitiveness in producing high-value-low-volume
bioproducts may also be improved by attention to
hydrodynamic phenomena.
1.1 Pneumatically agitated reactors
Airlift bioreactors and bubble columns are the two main
types of pneumatically agitated reactors. Because of many
operatiohal advantages, airlift devices are far more
significant than bubble columns in bioprocessing and
bioremediation. Airlift reactors are pneumatically agitated
gas-liquid and gas-liquid-solid contacting devices that are
used in the chemical process industry, bioprocessing and
waste treatment (Chisti, 1989; Onken and Weiland, 1983).
Particularly attractive features of airlift reactors are
simplicity of construction, good solid-liquid and gas-liquid
mass transfer characteristics at low power consumption
(Chisti, 1989; Mao et al., 1992), a well-defined flow
pattern, ease of long-term sterile operation, and a
hydrodynamic environment suitable for fragile biocatalysts
that are susceptible to physical damage by fluid turbulence
or mechanical agitation. Airlift devices outperform bubble
columns in suspending solids (Chisti and Moo-Young,
1994b). Evidence for superior heat transfer capability of
airlift reactors relative to bubble columns has also been
published (Ouyoung et al., 1989). Some of these features
have helped establish niche applications for airlift
technology. For example, in high-intensity-low-volume
activated-sludge type treatment of wastewater, high rates of
oxygen transfer, mechanical simplicity and low operational
costs are particularly valuable. The small volume and land
area demands of airlift devices compared to conventional
treatment processes are decided advantages in many
locations (Moo-Young and Chisti, 1994). In certain other
fermentations--production of monoclonal antibodies by
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 35

the energy input into pneumatically agitated reactors comes

from the compressed gas. The gas imparts energy to the
liquid as it expands isothermally from the hydrostatic
-- LIQUID 2T7~7 '::-~ :. pressure at the bottom of the reactor to the pressure in the
CIRCULATION
'.Lo* F,;J
. - : : ; .. -- DRAF-[
head zone. A second source of power is the kinetic energy
RISER --
of the gas jet at the point of entry to the reactor. In most
......... L: -oo bioreactors, the sparger designs and the gas flow rates are
--DOWNCOMER
such that the kinetic energy of the gas jet is negligible
relative to the energy derived from isothermal expansion
GAS
GAS
(Chisti, 1989). When the kinetic energy is ignored, the
(o) (b) (c) specific power input, or the power input per unit volume of
the liquid (or slurry) is given by

Pc
Fig 2. Airlift reactors: (a) split-cylinder internal-loop; (b) draft- = PLgUc, (1) '
tube internal-loop; and (c) external-loop. VL
for bubble columns, and
suspension culture of hybridomas (Birch et aL, 1987) and
plant cell culture (Doran, 1993), for instance---capabilities Pc PLgUc~
(2)
for sterile operation and a less damaging hydrodynamic VL 1+ A_j_a
environment are important features. New applications of A,.
airlift technology that are now being developed include
anchorage-dependent culture of animal cells for for airlift bioreactors, in equations (1) and (2) the
biopharmaceuticals (Dai et al., 1994; Ganzeveld et al., superficial gas velocity (Uc or Uc~) is actual height-
1995). IrresPective of the area of application, the averaged value, rather than a point value at any axial
predominant types of airlift reactors are the internal-loop location (Chisti, 1989). The specific power input in
devices shown in Figure 2. The external-loop configuration pneumatically agitated bioreactors does not normally exceed
is less commonly used. 3 kWqn3; a value of 2 kWqn 3 being a more typical
As shown in Figure 2, all airlift reactors consist of a maximum. In animal cell culture applications, power
gas-sparged riser and a downcomer which is usually not inputs of less than 100 W.m 3 are the norm.
sparged. The riser and the downcomer are interconnected
n e a r the top and the bottom of the reactor. Difference in
2.2 Gas velocity
gas holdup between the gas-sparged riser and the
downcomer causes a difference in bulk density of the fluid The actual height-averaged superficial gas velocity is
in these zones, resulting in circulation of fluid--up-flow in calculated with the equation (Chisti, 1989)
the riser and downflow in the downcomer (Chisti and Moo-
Young, 1987a). QmRT in(l + PLghL~,
Literature on airlift bioreactors has been reviewed by Uc - hL A PLg ~ ) (3)
Blenke (1979), Onken and Weiland (1983), Merchuk
(1986), Chisti and Moo-Young (1987a), Chisti (1989) and
Joshi et al. (1990). Much more information is available on which applies to bubble columns. In equation (3) Qm is the
bubble columns. Some major sources are Schtigerl et al. molar flow rate of the gas, hL is the static height of the gas-
(1977), Shah (1979), Shah et al. (1982), Deckwer and free liquid, R is the gas constant, T is the absolute
Schumpe (1987; 1993), and Deckwer (1992). In view of the temperature, A is the cross-sectional area of the column,
extensive literature, this work concentrates mostly on and Ph is the pressure in the head zone. Equation (3)
assumes ideal gas behavior which is an acceptable
developments since an earlier comprehensive treaties
(Chisti, 1989). approximation in most cases.
Equation (3) may be used also for an airlift reactor to
calculate a hypothetical superficial gas velocity (Uc) that
2. PNEUMATIC AGITATION can be converted to the superficial velocity (Ucr) in the
riser:
2.1 Power input
Uc A
The power input into a reactor controls most of its U G r "~ . . . . . . i
performance parameters and affects the economics of A~ (4)
operation. In comparing the performance of bioreactors,
specific power input is often used as the basis. Hence, where A is the sum of the cross-sectional areas of the riser
correct calculation of power input is essential. Typically, all and the downcomer.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 36 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
2.3 Pressurized operation
As seen in equation (3), correct calculation of the prevailing
superficial gas velocity requires that the pressure (PD in the
head zone of the reactor is taken into account. Two
identical reactors operated at different headspace pressures
but the same molar gas flow rate would have different
specific power inputs. Sometimes the 'superficial gas
velocity' reported in the literature does not correct for
pressure as in equation (3). In such cases apparent effects
of the headspace pressure on mixing time, gas holdup and
other variables that are dependent on power input may be
observed (Chisti and Moo-Young, 1987a). Correct
calculation of the superficial gas velocity and power input
eliminates these apparent effects. An example of the
apparent pressure effect on oxygen transfer coefficients in
draft-tube sparged airlift reactors was reported by Wu et al.
(1992). Variations in headspace pressure over 1-2.36
atmospheres affected the mass transfer coefficients but the
effect disappeared when the gas velocity was corrected for
pressure as recommended by Chisti (1989, pp. 287-290).
3. GAS HOLDUP
3.1 Bubble columns
The volume fraction of gas in dispersion, or gas holdup
(co), is an important characteristic of bubble columns and
other gas-liquid reactors. By definition,
VG
¢o - Vo+VL'
where VG and VL are the volumes of the gas and the liquid
(or slurry) phases respectively. The reactor must b e
designed to accommodate the maximum expected gas
holdup. The holdup affects many other performance
characteristics in a reactor. Gas-liquid interfacial area for
mass transfer is controlled by a combination of gas holdup
and the mean bubble diameter. In bubble columns,
6 ¢o
aL -- d B ( 1 - - eO ) '
where aL is the gas-liquid interfacial area per unit volume
of liquid, and dB is the diameter of a sphere having the same
surface-to-volume ratio as the gas bubble. In practice a
distribution of bubble sizes is encountered and the mean
diameter is calculated using
N
dB i=1
N ,
En,
i=1
where ni is the frequency of occurrence of bubbles with a
diameter dsi. Bubble size and frequency distributions may
be measured by electrical resistivity probes and other
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
similar sensors (Choi and Lee, 1990a). The efficiency of
utilization of the gas depends on gas holdup because the
contact time between the two phases is holdup dependent
(Chisti, 1989, pp. 176-180),
hL 6o
to - U o ( 1 - co)' (8)
where to is the mean residence time of the gas in liquid.
For columns larger than about 0.1 m diameter, gas holdup
is independent of column diameter and the overall holdup is
independent of the height of liquid (hL). Hence, for a given
superficial gas velocity, the gas phase residence time
increases with column height. Aspect ratios of four or
higher are not uncommon in bubble column bioreactors. In
columns operated batchwise with respect to the liquid, the
gas holdup can be analytically shown to depend on the
mean rise velocity (Ub) of the bubbles,
Uo
Ub" (9)
Hence, at identical gas f l o w rates, sparger-fluid
combinations that generate larger bubbles, or allow them to
exist, may give rise to lower gas holdups than systems that
produce smaller bubbles. In bubble columns gas holdup
shows the following general dependence on gas velocity
eo = a Ubo. (10)
(5)
The parameters a and b depend on the flow regime and the
properties of the fluid. As shown in Figure 3, for air-water
a and b have been found to be 2.47 and 0.97, and 0.49 and
0.46, respectively, for bubble (Uo < 0.05 m.s-1) and chum-
flow (Uo > 0.05 m.s"1) regimes in broad ranges of bubble
column geometries (Chisti, 1989, pp. 161-164; Chisti and
Moo-Young, 1988b). Addition of inorganic salts to water
enhances gas holdup by a few percent up to an ionic
strength corresponding to about 0.15 M sodium chloride
(Chisti and Moo-Young, 1991; Kelkar et al., 1983). This
(6) effect is due to coalescence inhibition that results from
electrical repulsion between like ions at the surfaces of
bubbles (Chisti and Moo-Young, 1991). For any given
ionic strength, the type of ion does not influence gas holdup
(Hikita et al., 1980).
Numerous gas holdup correlations are available for
bubble columns (Deckwer, 1992; Chisti, 1989). An
example is the correlation (Akita and Yoshida, 1973)
cG 4-
c(gd~PL~[fgd3op2z~"~(
~- " -~ -
Uo .1, (I1)
(7)
(1-¢o) k, L J ~, ,UL ) ~.g ~ o j
where c is either 0.20 (non-electrolytes) or 0.25
(electrolytes). Equation (11) covers column diameters (do)
of 0.152 to 0.60 m and gas velocities from 0.004 to 0.33
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 37

I O0
z~ AKITA AND YOSHIDA (1973)
• DECKWER et ol. (1974)
[] FAIR et oi. (1962) (bosed on $G=3.2 UG)
x HIKITA et oi. (1980)

• CHISTI (1989)
• ~ • ®x

I
~J ,m~,~, L =o:o.49ul.,6
% 10
x
(3
to

i '~G=2"47 U0"97
G - ~ Y •
~,-zx
AIR-WATER

I t .I J I tlll I = I I III I t I t , , ,,
I 2 4 6 8 I0 20 40 60 I00 200 400 I000
UGx 10 (m.s -1)

Fig 3. Gas holdup versus superficial gas velocity in bubble colmmls. The data shown cover column diameters of 0.10-1.067 m, and
heights of 1.37-5.875 m (Chisti, 1989, p. 164).

m.s ~. It applies to newtonian media such as water, glycerol fluid. The exponent b has been observed to obey the
and methanol. empirical equation (Chisti and Moo-Young, 1988c)
In homogeneous n0n-newtonian systems the following
equation may apply b = 0.564 n -0"354 , (13)

06 -0.19 as illustrated in Figure 4. Equation (13) is only a rough

6~ = 0 . 2 0 7 U ~ , u . (12)
guide which disregards any effects of the flow regime;
nevertheless, the equation is based on data on a variety of
Equation (12) was developed by Godbole et al. (1984) for
aqueous carboxymethyl cellulose solutions (n = 0.440- 2.0
" ~ +30~.
0.697; K = 7.683--0.095 Pa.s") in churn-turbulent flow ~ b=0.56 n-0"35
regime when the superficial gas velocity did not exceed 0.25
m.s -~. A higher coefficient of 0.255 has been reported for
equation (12) in presence of 0.8 M sodium sulfate. 1.0
The commonly employed shear rate expression 7 = 0.8
5000.UG, due to Nishikawa and coworkers (1977), was used
by Godbole et aL (1984) to calculate the i.t~p in equation ~- 0.6

(12). Although this approach is widely used, it is o Chisli et ol. (1987) Solka-gloc ~-. a
fundamentally unsound and shear rates in bubble columns 0.4 . • Copuder and Kololn| (1984), Ca(OH)2/CaCO3 ~ ,
C. cellulolyticum A ] []
and other pneumatically agitated reactors are difficult to + C. cellulolyticum B I Moo-Youn 9 et al. (1987)
define as clearly demonstrated by Chisti and Moo-Young v N. sitophilo
(1989). a Hoque et al. (1986)
x Kowose and Moo-Young (1986)
As for newtonian media, the variation of gas holdup 0.2
• Godbole et ol. (1984) t CMc/Woter
with gas velocity in non-newtonian fluids is generally [] 8uchholz et al. (1978)
consistent with equation (10). The exponent b seems to = T= I f t I t I f I r I
depend only on the flow regime and the n-value of the fluid, 0.06 0.1 0.2 0.4 0.6 0.8 1.0

while the coefficient a is sensitive to the properties of the n(-)

fluid (K, n, PL and CYL)as well as the flow regime (Chisti Fig 4. The exponent b (equation 10) versus the flow behavior
and Moo-Young, 1988c). Little experimental data exists on index n for non-newtonian pseudoplastic media in pneumatically
the relationship between a and the characteristics of the agitated reactors (Chisti and Moo-Young, 1988c).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 38 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
fluids including fermentation broths of the fung!
Chaetomium cellulolyticum and Neurospora sitophila.
Additional data on gas holdup in slurries that simulate
mycelial fermentation fluids are available elsewhere (Chisti
and Moo-Young, 1988c; Chisti, 1989, pp. 149-201).
3.2 Airlift reactors
Using newtonian media (gL = 1--266 mPa.s) and a
pseudoplastic fluid (K = 0.0256 Pa.s °902) in an external-
loop airlift reactor (0.138 m3; A / A d = 1; aspect ratio ~ 58;
Uor = 0.002.-.0.12 m.sl), Akita et al. (1994) established that
the gas holdup correlation proposed by Akita and Yoshida
(1973) for bubble columns (see equation 11) applied to the
riser of the airlift reactor when the superficial gas velocity
(Uo) was replaced with [Uo, - (Uz:e~,/(1 - ~,))] as the
correlating parameter; thus,
COr C (gdrPL~ [gdrPL~
2 ~ 3 2
1- gOr /
where c is 0.20 for non-electrolytes and 0.25 for electrolytes
(Akita and Yoshida, 1973). The riser diameter & in
equation (14) has no impact on the value of gas holdup. For
water-like fluids, no gas was carried into the downcomer,
because of an effective gas-liquid separator; however, in
more viscous media, complete disengagement of the tiny
gas bubbles could not be attained (Akita et al., 1994).
For pseudoplastic aqueous solutions of carboxymethyl
cellulose (K = 0.286-11.5 Pa.s"; n = 0.441-0.617) in a
draft-tube sparged internal-loop airlift reactor, Li et aL
(1995) reported the following equations for gas holdup:
6o, = 0.441U~r841/tl2p0'135 ,
and
T T0.935 . -0.107
~Od=0.297~,Gr ,uov • (16)
The apparent viscosity in equations (15) and (16) was
calculated by assuming the shear rate to equal 5000.Uo,. for
Uo, >- 0.04 m-s "1, or 1000.(Uc,) °s for Uc, < 0.04 m.s 1. The
apparent viscosity ranged over 0.020---0.85 Pa-s. At 0.055
m 3 the reactor was quite small; other geometric details
were: A / A d = 0.618 and aspect ratio - 26.
For newtonian and non-newtonian media in bubble
columns and airlift bioreactors, Kawase et al. (1995)
proposed the following equation for gas holdup in tile riser:
~Or (U
\ Or /n
/ ~'~'
/
l--SOr
2 o.+,)
, . - -(~v,/, v )"~2(.~,)g',~',
" ( A-A--r~'("÷"
l+
/ PL
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
Development of equation (17) assumed isotropic turbulence
with the requisite energy dissipation being calculated as
recommended by Chisti and Moo-Young (1989). As major
limitations, equation (17) does not account for effects o f
liquid circulation on gas holdup, nor does it consider such
factors as interfacial tension and solids concentration which
affect holdup. Moreover, the 'theoretical' foundation of the
equation is questionable because the assumption of isotropic
turbulence in bubble columns and airlift reactors is not valid
even at high energy inputs in water-like media (Liibbert
and Larson, 1990; Liibbert et al., 1990; Okada et aL, 1993).
The supposed mechanistic foundation of equation (17) is
especially doubtful when applied to fluids that are more
viscous than water. Although counter claims have been
made (Kawase and Moo-Young, 1990), they do not wholly
support the existence of isotropic turbulence.
In an external-loop airlift reactor (A/Ad = 1.0; aspect
ratio = 28) that achieved complete gas-liquid separation
~12
(that is, no gas in the downcomer) Po~rac and Petrovic
(1991) conducted an exhaustive study of gas holdup. Tap
(14) water and aqueous solutions of alcohols (0.5 or 5% vol
methanol; 0.5% vol n-butanol) were used. The superficial
air velocity in the riser varied over 0.01-0.16 m.s -a with
corresponding liquid velocities (UL~) in the range 0.05-0.65
m.s ~. In general, the presence of alcohols enhanced gas
holdup relative to water. This effect was due to coalescence
inhibitory properties of alcohols. Gas holdup in the riser
followed tile equation (Pogarac and Petrovic, 1991)
OC'Gr = 0 . 6 / 3 ; 0 6 2 pO.O69/./;107 r T0-936
LlOr
0.053 0.185 _ X0.474 ' (18)
,LIL O'L (Uor + ULr)
which agreed with 144 data points to within ±7%. Equation
(18) does not included any geometry related effects because
the geometry did not vary. Po~arac and Petrovic (1991)
(15) noted that the correlation proposed by Bello (1981),
:U ho56(
COt = 0.16 |-Y--q-r-I /1 (19)
\UL,J \
gave consistently higher holdup values with an error that
exceeded 70%1 Similarly, the 'universal' equation of
Kawase and Moo-Young (1987), which for newtonian fluids
is
( U2 ~1/3
eOr = 1 07/--~-~-I (20)
" k.gdoJ '
showed yet poorer agreement with the data (Po~arac and
Petrovic, 1991), overpredicting gas holdup by almost 100%.
u.÷2) • (17) Measurements in aqueous s a l t solutions (sodium
chloride; potassium chloride; calcium chloride; sodium
sulfate; and magnesium sulfate in the 0.01-0.2 kmol.m 3
concentration range) and sucrose solutions (0.5-8% wt/vol
sucrose) in an external-loop airlift reactor showed that the
 Appl Mech Rovvol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 39

type of salt or its concentration had very little effect on the

measured holdup (Snape et aL, 1992), in keeping with VENT
similar results in bubble columns (Hikita et aL, 1980;
Kelkar et aL, 1983). The holdup was not particularly
sensitive to the concentration of sugar either; however, the
liquid velocity did increase noticeably as the concentration
V
increased from 0 to 1% wt/vol (Snape et aL, 1992). Further
r
increase in sugar concentration reduced the rate of
circulation apparently because of a viscosity effect. Several
perforated spargers were employed for aeration (5-220
holes; 0.5-3.0 mm hole diameter; 0.2-0.5% free area). For
--L,,,,, /
the most part, the type of sparger did not influence the gas
holdup or the liquid circulation rate (Snape et aL, 1992).
When any sparger influences were observed, they were
confined to the low aeration rate range (Uar = 0.03-0.09 RISER
I DOWNCOMER
m.s'~). Such influences disappeared as the aeration rate
increased (Snape et aL, 1992) and as the turbulence in the
fluid became the controlling influence on bubble size. The
external-loop reactor used was 0.065 m3; the other
geometric details were: A / A d = 9; and aspect ratio based on
riser = 12.5 (Snape et aL, 1992).
In water, aqueous salt solution (10 kg.m 3 sodium
chloride) and an algal culture medium, Fraser and Hill
(1993) observed almost identical gas holdup values at
AIR =
identical conditions in the three media. An external-loop
airlift reactor was used (A/Ad = 2.3; aspect ratio based on
J
riser = 18). The perforated plate sparger used could be
rotated at various speeds. Compared to stationary
operation, the rotating sparger (ca 430 rpm) produced Fig 5. The external-loop airlift configuration used by Ghirardini
somewhat higher gas holdup (Fraser and Hill, 1993), but et aL (1992).
not sufficiently high to justify the additional complexity of
the rotating device. Liquid circulation rates in all media
showed excellent agreement with equation (51) developed
configurations other than the one Shown in Figure 5.
by Chisti et aL (1988).
Surprisingly, the specific- interfacial area" (aD was
For the external-loop airlift configuration shown in
unaffected by the fill factor (equation 23) even though the
Figure 5, Ghirardini et aL (1992) obtained the following
fill volume influenced the overall gas holdup (equation 22).
empirical equations
The common use of the drift-flux model of Zuber and
@~ TTT M I~0.27 40.41 Findlay in correlating the riser gas holdup in air-water
U L r ~ ~ . o t tJ Gr 1" tar (21) system in external-loop airlift reactors has been noted
(Glennon et al., 1993). The Zuber and Findlay model can
- - ~ t ~ R rr0.78 E,0.20,40.42
~'G -- u.J J lJ Or 1" tar (22) be written as

and UGr
~Gr ~- ~ ( U o r -j- U L r ) "~- f l '
0 75
a t = 798.37U~r d /
019
. (23) (24)

where <x, the distribution parameter, has typically ranged

Equations (21) and (22) are for air-water while equation
(23) is for aqueous sulfite solutions. These equations were
developed for A,/Ad = 1.0. The riser diameter (dr) varied
over 0.025-0.1 m; the aspect ratio range used was 20-80,
and the ratio of the separator-to--downcomer diameters
varied over 3--6. The equations applied over a superficial
air velocity range of 0.2-1.2 m.s ~. The 'filling factor,' F,
was defined as the ratio of the non-aerated liquid volume in
the separator to the unaerated volume in the loop. The F-
values varied over 0. I-0.32.
The usefulness of equations (21)-(23) is limited
outside the ranges used in establishing theln and for
over 1.03-112; and ]3, the terminal rise velocity of a bubble,
has varied over 0.25-0.36 m.s 4 (Glennon et aL, 1993). An
a-value of close to unity indicates fairly flat radial velocity
and gas holdup profiles in the riser (Chisti, 1989, p. 56);
more parabolic profiles are obtained for higher values of ~.
The parameters ~ and 13depend on the flow regime and the
properties of the fluid.
Using measurements in two large external-loop airlift
reactors (80 L, A / A d = I, aspect ratio = 19.5; 700 L, A / A d =
1 or 1.8, aspect ratio (based on riser) = 35.9) and additional
data from other sources, Kemblowski et al. (1993)

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 40 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
established tile following empirical correlation for the riser
gas holdup in external-loop reactors
FRO."( q0,,
,¢o~ = 0.203 M ~ O 1 2 \ U°~
UL, ' (25)
where the Morton number (Mo) and the Froude number (Fr)
are, respectively,
4(n-l) + 4n
Mo
g(PL--PG)
~L~
K4(SULr~
\-'~',)
(3n 1~
\---~--)
and
(U Lr+ UGr)2
Fr =
gdr
Equation (25) agreed with tile data to within 20%; it
covered the ranges: eo, = 0.002-0.21; Uc, = 0.001-0.50
m-s-a; Uzr = 0.0%-1.3 m.s'~; Fr = 0.005-14.1; Mo = 2.47 x
10~-0.39; and As/At = 0.11-1. Although not specifically
developed or tested for internal-loop airlift reactors,
equation (25) could well apply to them: the equation
incorporates the liquid circulation velocity and, hence,
indirectly, the effects of geometry.
Performance of a draft-tube sparged internal-loop
airlift reactor (5 L; details unknown) that generated a fine-
bubble dispersion was studied by Yuan et al. (1994).
Experiments were done in water, aqueous ethanol (19'0) and
baker's yeast broth. Aeration was through a nucleopore
membrane (3-5 ~m pore size) sandwiched between a nylon
wire screen and a perforated plate (Yuan et al., 1994).
Extremely small bubbles, 100-200 i.tm diameter, were
produced in the alcohol solution; at 380 ~m, the bubble size
was somewhat larger in water. Because of large specific
interfacial areas, the values of the gas-liquid volumetric
mass transfer coefficient (kzaL) were high. Thus, during
yeast fermentations the mean value of ALaz was about 0.46
sz (aeration rate of 1 vvm). In comparison, the average
kzaL during fermentation in a stirred tank (600 rpm
agitation; 1 vvm aeration rate) was only 0.13 s~. When the
membrane distributor in the airlift reactor was replaced with
a 'jet' distributor, the kzaz declined to 0.1 s-~ even though
the reactor was aerated at 2.5 vvm compared to tile 1 w m
with the membrane sparger. Despite these impressive
results, membrane aerators with fine pore sizes are not
suited to industrial fermentations because of potential
difficulties with clogging.
The data presented showed that the overall gas holdup
values were similar in water and the aqueous ethanol
solution, but much lower in the yeast broth (Yuan et al.,
1994). The maximum gas holdup in aqueous ethanol was
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
about 0.6; it was less than 0.2 in the yeast broth. Liquid
circulation rates were about 4-fold higher in air-water than
in the broth; the values in the ethanol solution were similar
to those in the yeast broth. An interpretation of these
results is that in the non-coalescing media (yeast broth;
ethanol solution), many tiny bubbles must have recirculated
into the downcomer, hence the driving force for liquid
circulation was lower.
In draft-tube sparged concentric tube airlift reactors
(A/Aa = 0.2-1.3; aspect ratios = 10) operated with tap
water, aqueous n-butanol (0.5% wt) and aqueous glycerol
, (26)
(46% wt), Petrovic et al. (1991) did not see much effect of
the A/Ad ratio in water-like media. In the more viscous
glycerol solution, there seemed to be an optimum A/Aa ratio
of about 1/3 for maximum holdup (Petrovic et aL, 1991).
The overall holdup agreed closely with the following
correlation of Koide et aL (1983b)
(27)
/ "x 0 . 9 9 6 / 3 x 0.294
s~ _ 0.124f._Uq_~_ ] f_~_]
4 ,, j gPL J
(28)
/-- / x 0.114
×Iaj~o) (1 - 0.276(1 - e-° 0386Ma))-~ ,
where Ma is the Marrucci's bubble coalescence parameter
(equation 29) that characterizes the frothing ability of
liquids (for non-frothing pure liquids Ma = 0). Koide et al.
(1983b) developed equation (28) for the following ranges:
1.93 x 1 0 "4 _< UG~LL/(Y
L _<
2.85 x 10"2; 1.31 x 106 _<9:rL3/gl.tZ4
_< 6.04 x 10~0; 0.471 _< d/do -< 0.743; and 0 _< Ma _< 67.3.
The equation was developed in draft-tube sparged
concentric tube airlift reactors with aspect ratios of 5-15.
The bubble coalescence parameter Ma is given as
d e ct ( d o ' L ~ 2 (247r c r L / N d s ) 2/3
Ma- -- --
a L CRT \ - ~ - J
din cl J x2 v #
(29)
where Nis the Hamaker constant; c~ is the concentration o f
component 1 in liquid; ~ is the total number of moles of
ions per mole of electrolyte (~ is unity for non-electrolytes);
f is the activity coefficient of component 1 in liquid; xl and
x2 are mole fractions of components 1 and 2 in liquid; vl
and v2 are molar volumes of components 1 and 2 in liquid;
R is the gas constant; and T is the absolute temperature of
the liquid (Koide et al., 1992).
For several newtonian media (9L = 997-1179 kg.m-3;
ktL = (0.894-16.81) x 10.3 Pa-s; crL = (53-72) x 10-3 N.m -1)
in draft-tube sparged concentric tube airlift devices (A/Aa =
0.3-1.2; aspect ratios = 5-16), Koide et aL (1988)
recommended the following equations for gas holdup:
 Appl Mech Rev vol 51, no 1, January ! 998 Chisti: Pneumatically mixed bioreactors 41

60
fl:Uop['/&aL.
[ : - 3'°'283 : a,.1"-° 222
4 = 01601--1(-~ -) k,~--o)
(1-eo) \ O"L ,
CUGr +U<r'~(d2ogpL? -°'88
x 0.415+427
t. .vg-;:. 9t
. . . . . .
9 .
x \~--j

:.o.
+ 1.13/----~-I L~-J J
, [1- ,.610- exp(- 0.00,6,Ma))l-',
t, 4 gd, J (33)

for the riser; and where 8 is the diameter of sparger nozzle. Equation (33)
applied to the following ranges: 1.21 x l 0 "4 ~ UGglJ6L~

CGd 1 50
279 66r
( Ar )2,8( t2,,
(1 - eOd) 5 A,. + Ad
Ld 0.352
× ~,77~'J ( 1 + 0.433
for the downcomer. Equation (30) correlated 140 data
within 12% of the measurements, while equation (31)
applied to 109 data with an average error of less than 24%.
T h e Ld in equation (31) was the length of the draft-tube.
Equations (30) and (31) are apparently more refined forms
of similar equations presented in an earlier paper by Koide
et aL (1984a).
In an air-water system in a large (about 186 L; A,/Aa
= 0.77) draft-tube sparged internal-loop airlift device
equipped with an elaborate gas-liquid separator, Cai and
Nieuwstad (1991) noted that the gas holdup in the riser
correlated well with the equation
go, = 2.47 U~ 97
developed by Chisti (1989, p: 164) for bubble columns
operating in the bubble flow regime. This agreement was
probably because the liquid circulation velocities in the
airlift reactor were relatively low (lnaximum UL~ was less
than 0.4 re.s4).
In a large draft-tube sparged concentric tube airlift
reactor (A/Ad = 1.86; aspect ratio = 11; 4 m 3) with an
expanded head zone for gas liquid separation, Fr/Jhlich et
al. (1991a) noted that the gas holdup in tile riser was
independent of axial position, but the holdup declined down
the downcomer. Tap water, aqueous ethanol solution (1%
vol/vol ethanol) and tap water containing an antifoam agent
were studied.
Ad
....
1.57 X 10"2; 1.18 x 106 _<pLoL31ggz4 N 5.93 x 1010; 0.471 £
d/do ~ 0.743; 7.14 x 10.3 g 81do ~ 2.86 x 10"2; and 0 _<Ma _<
93. The average error in estimating the overall gas holdup
A~ + Ad
with equation (33) was 9% for 175 measurements (Koide et
-1 aL, 1983a).
Clearly, many correlations have been developed for
.
predicting gas holdup. The ones presented here supplement
numerous others compiled in a previous work (Chisti,
(31)
1989). Despite this information, reliable prediction of gas
holdup remains difficult. Most correlations are quite
specific to particular fluid-reactor combinations and, even
then, estimates should be used with caution. Two different
fluids with identical densities, viscosities and inteffacial
tensions, may give very different results in the same reactor.
F o r all but the simplest cases, reliable estimation of gas
holdup for design and scale-up must rely on procedures
such as that outlined in Section 3.2.3.
3. 2.1 Relationship between riser and downcomer gas
holdups
Internal-loop reactors. In airlift reactors without gas-
liquid separators, the gas holdup in the downcomer depends
mainly on the holdup in the riser and tile velocity of the
(32) circulating liquid. Important aspects of performance of
airlift reactors depend not only on the overall gas holdup,
but also on the distribution of holdup between the riser and
the downcomer; hence, a knowledge of this distribution is
essential.
As noted earlier, the holdup distribution controls the
rate of induced liquid circulation. In addition, the gas-
liquid interfacial area for mass transfer and the residence
time of gas in the liquid and the residence time of liquid in
tile riser and the downcomer are affected. Specifically, in
internal-loop airlift devices, the ratio of the gas-liquid
interfacial areas in the riser (ALr) and the downcomer (ALl)
zones is (Chisti et aL, 1995)
AL,. Ar dsd Co,
. . . . , (34)
ALd Ad dBr GOd

In studies with annulus sparged concentric-tube airlift
vessels (aspect ratio = 6-15; AJA, = 0.52-1.23), Koide et
al. (1983a) obtained the following equation for overall gas
holdup:
where Ar and Ad are tile cross-sectional areas of the riser
and the downcomer, respectively. Generally, the mean
bubble diameter in the downcomer (dna) is slightly lower
than that in the riser (ds,.) because smaller bubble are more
easily and preferentially dragged into the downcomer

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 42 SPECIALISSUE: FLUID MECHANICSPROBLEMSIN BIOTECHNOLOGY
(Chisti, 1989, pp. 132-149). Similarly, the time spent by
the recirculating liquid phase in the downcomer and the
riser depend on gas holdup (Chisti et al., 1995),
tLr Ad (1--8Gr)
(35)
ttd Ar (1--6or)"
Because of its significance, the relationship between
the gas holdups in the riser and the downcomer has received
some attention in the published literature. Using air-water,
Bello (1981) established the empirical relationship
e O d = 0.89 eGr , (36)
which applied to annulus sparged concentric draft-tube
airlift reactors (0.033 m 3 approximate dispersion volume;
Ad/A~ = 0.13, 0.35 or 0.56; 0.152 m outer tube diameter in
all cases; 1.8 m gas-liquid dispersion height; 1.55 m height
of draft-tubes measured from the base of the reactors). In a
much larger split-cylinder airlift device (0.23 m 3 working
volume; 5.0 m liquid height; Ad/Ar = 0.411; 0.243 m
column diameter; 0.229 m baffle width; 4.8 m baffle
height), Chisti (1989, p. 218) found a similar correlation:
Equation (37) was determined for air-water and air-salt
e O d = 0.997 6or. (37)
solution (0.15 M sodium chloride). Other investigations
have revealed similar linear relationships between the gas
holdup in the riser and that in the downcomer. Thus, there
is significant empirical evidence spanning different types of
airlift devices, ranges of reactor volumes and varieties of
fluids, for linear dependence between the gas holdups in the
riser and the downcomer. An exception was reported by
Miyahara et al. (1986) who worked with draft-tube sparged
concentric tube reactors of modest size (0.021 m 3 dispersion
volume; 0.148 m outer column diameter; A/Ad = 0.128-
0.808; 1.2 m height of dispersion; 1.0 m height of draft-
= =
tubes; PL 952-1168 kg.m3; ~L 1.0-14.9 mPa.s; ~L =
34.1--72.0 mN.m'~). Miyahara et al. (1986) observed the
following relationships between gas holdups
( Ar ~4"2
COd = 4.51 × 1 0 6 M o ° J 1 5 / : - ~ I 4.2 (38)
• \ Ad) SCr,
when
get < 0.0133 , (39)
d gas-liquid dispersion (t79) and the gas holdup (~r) in the
and
e O d = 0.05 MO -°22 '~Gr) , (40)
when the gas holdup in the riser was greater than that
obtained with the inequality (39). In these equations Mo is
the Morton number defined as
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
Mo = g/.t~/ 3. (41)
/ PLO'L
Equations (38) and (40) were established in reactors with
cross-sectional areas of risers smaller than the areas of
downcomers even though the opposite is the norm for
practical designs of airlift reactors.
Equations (36)-(40) were empirically established.
These equations are inconsistent with the observation that
downcomer remains free of gas (that is ~d = 0) until some
gas holdup has been builtup in the riser. An alternative,
more useful relationship between the riser and the
downcomer gas holdups is the equation
ULr = " 2 g h D ( e O r - eGd)
10.5
(42)
Equation (42) was developed by Chisti et al. (1988) for the
prediction of liquid circulation velocity in internal-loop
airlift reactors without gas-liquid separators. The equation
was shown to apply to low-viscosity, water-like, fluids over
a broad range of scales. Since initial publication, the
equation has been repeatedly validated by other independent
investigators (Cai and Nieuwstad, 1991; Choi and Lee,
1993; Fraser and Hill, 1993; Garcia-Calvo and Let6n, 1996;
Petrovic et al., 1991; Po~arac and Petrovic, 1991). In
equation (42) K8 is the frictional loss coefficient for the
bottom zone of the reactor. For internal-loop reactors, the
parameter K8 is related to the configuration of the bottom
zone in accordance with equation (54) (Section 4.1.1).
A silnple rearrangement of equation (42) leads to
2 K B( a r / Ad) 2
ULr
- 1. (43)
2gho(6~r--6Od)(1--COd) 2
Equation (43) is an implicit relationship between the gas
holdups in the riser and the downcomer. When the
superficial liquid velocity (UL,.) in the riser, the beight of
riser are known, the equation may be used for an iterative
solution for tile downcomer gas holdup (,~d) in a reactor of
given geometry (given At, Ad and AD. The iterative
procedure starts with an assumed value for the downcomer
gas holdup (e~d). The left-hand-side of equation (43)is
computed and compared to the right-hand-side. An
agreement between the two sides indicates a satisfactory
value of ~d, the desired variable.
 Appl Mech Rev vo151, no 1, January 1998 Chisti: Pneumatically. mixed bioreactors 43

Unlike such previous empirical correlations as
equation (36) and equation (37), equation (43) shows the
effects of the geometry of airlift reactors, that is the effects
of KB, hD, Ar and Ad, on the relationship between the gas
holdups in the riser and the downcomer. Geometry of the
head zone has been clearly shown to affect the gas holdup
distribution (Choi et aL, 1995b); mixing time and the liquid
circulation velocity are also strongly affected (Choi et at,
1995a, b). Because equation (43) accounts for reactor
geometry-associated effects, it has the potential for general
applicability to internal-loop reactors irrespective of scale.
Equation (43) has been confirmed for prediction of the
downcomer gas holdup in five internal-loop airlift reactors
when the holdup in the riser and the velocity of liquid
circulation were known (Chisti et al., 1995). Remarkably
good agreement between the measured and the calculated
gas holdups was observed as illustrated in Figure 6 for two
split-cylinder devices (Chisti et al., 1995).
Using split--cylinder airlift reactors, Ganzeveld et al.
(1995) showed that the relationship between the riser and
downcomer gas holdups was unaffected by the loading of
animal cell microcarriers (ps = 1030-1050 kg.m-3; 150-300
p.m carrier diameter; 0-30 kg.m 3 solids loading; A / A d = 1;
aspect ratio = 7.6 and 14.5; Uar = (0--6.7) X 10"3).
The gas holdup distribution in airlift devices operated
such that the initial level of fluid is below the upper edge of
the draft-tube or baffle is discussed in Section 11.1.
External-loop reactors. The head region of external-loop
reactors usually achieves a high degree of separation of gas
and liquid, her/ce, the downcomer gas holdup is much lower
than in the riser. With well designed separators and water-
like media, the downcomer gas holdup is negligible. In
other cases, linear relationships between the riser and the
downcomer holdups usually exist as noted for air-water
systems (Bello, 1981; Chisti, 1989, pp. 184-186) as well as
more complex fluids (Chisti et al., 1987; Chisti and Moo-
Young, 1987a). Such relationships are sensitive to the
design of gas-liquid separator. Thus, for example, the
expression
where IVs and WL were respectively the total weight of solid
and that of the liquid in the slurry. Equation (45) was
obtained with suspensions of calcium alginate beads (pp =
1030 kg.m3; dp = 1-3.6 mm; loading = 0-30% vol).
Overall gas holdup in suspensions of calcium alginate
beads in draft-tube sparged internal-loop airlift reactors
was reported also by Koide et al. (1992). The properties of
the systems were: 0-20% vol solids; 1.88-3:98 mm bead
diameter; i.tL = (0.894-12.5) x 10"3 Pa.s; A,/Ad = 0.3-1.2;
and aspect ratios = 6-16. The beads were suspended in
water, aqueous glycerol, and aqueous salt solutions (0.10 or
0.27 kmol.m"3 barium chloride; 0.4 kmol.m3 sodium
sulfate). The gas holdup declined with increasing
concentration of solids, but was not affected by the size of
the particles (Koide et aL, 1992). The holdup values in
water and salt solutions were similar. The holdup was
correlated with the following empirical equation (Koide et
aL, 1992)
8G
0,3t oj Mo0,,
1 - £G)4 ( 1 - 4.2 ~69)(1 _ 0 . 3 6 9 ( 1 - e -°°46M' ) ) '
(46)
where ~bs is the volume fraction of solids in the gas-free
system, and the superficial gas velocity (UG) is based on the
diameter do of the outer column. In equation (46), the
bubble coalescence parameter (Ma) is defined as in equation
(29) while the definition given in equation (41) is used for
the Morton number (Mo). Equation (46) estimated gas
holdup with an average error of less than 10% for 260 data
in the experimental ranges 2.48 x 10.4 _< UGktL/OL _<3.24 X
10"2; 1.69 X 10"11 __<MO _<6.67 x 107; 0.471 _<d/do ~ 0.743;
0 _<Ma _<53.9; and 0 _<as -< 0.2 (Koide et al., 1992).
0.16 /
+15%/ /
• REACTOR1 // /
O REACTOR2 // ~ •//"

6Ca = 0.79 Se, -- 0.057, (44) {El'd /~////

03 " /// ///
developed by Bello (1981) for air-water in external-loop
airlift reactors did not apply to the measurements reported
:by Kawase et al. (1995). This discrepancy was certainly
because of the very different configurations of the head
Z
w 0.08

0.04 Z / / ///
S
/
/
//
/
/ /

zones for otherwise similar reactors that were used by

Kawase et aL (1995) and Bello (1981).
0 I I I
3.2. 2 Systems with suspended solids 0 0.04 0.08 0.12 0.16
EOd(- ) CALCULATED
With a limited amount of data acquired in gas-liquid-solid
slurries in a draft-tube sparged internal-loop configuration, Fig 6. Measured downcomer gas holdup versus predicted values
Lu et al. (1995) obtained the following empirical equation in two split--cylinder reactors. The calculated holdups were
obtained with a best-fit KB-value of 8.0. The solid line
)I WS'~-0"46 represents exact agreement. Air-water system; AJAa = 2.44 (for
SGd = 0 . 8 S ~ r ( l + 2 0 d p 1+ WL-----J ' (45) both reactors); aspect ratios - 6.6 (reactor 1) and 13.2 (reactor 2);
Chisti et al. (1995).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 44 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

For suspensions of denser solids (glass or bronze (lower viscosity). The overall holdup data were correlated
spheres) in draft-tube sparged internal-loop vessels (A/Ad with the empirical equation (Smith and Skidmore, 1990)
= 0.3-0.8; aspect ratio = 6-15), Koide et al. (1985)
developed the following equation:
6o=(3.128-7.232~s ~.zzO~'sJU6, e q_ & ,.~,-.,,- 2 2 ~ r r0 98 -284/T
, (48)
f" r T • ~ 0.934 [" ~ 0 " 3 ~ 0.292
where ~bs is the volume fraction of solids in the unaerated
(1__ 60) 4 --0.127 --0-L -------7-g,U
L
slurry and T is the absolute temperature.

(a,) _,
Gas holdup data in a 0.35 m 3 airlift reactor of the type
shown in Figure 7 has been published (Heijnen et al.,
1990). Over a superficial air velocity of 0.02-0.06 m.s-~
(based on outer tube), the gas holdup was somewhat higher
0.091[" 3 " ~ 0 . 0 4 3 / !-. N0.067-] -1
than expected for bubble columns. The holdup was not
Cs] IP a£-
x 1+0.17___I I significantly or systematically influenced by presence of
44 tg ) J' sand particles in the 50-250 kg.m 3 concentration range.
Correlations for holdup of gas and solid phases in
where Cs was tl~e mean solids concentration in gas-free three-phase upflow by independent variation of slurry and
slurry and Ut was the terminal settling velocity of a single gas flow rates were obtained by Douek et al. (1995). The
particle in stagnant liquid (Koide et al., 1985). The average procedure used had been detailed earlier (Chisti, 1989, pp.
error in estimating the overall gas holdup with equation 269-271). For an air-water-glass beads (dp = 500--600 p.m;
(47) was 12% for 383 data (Koide et aL, 1985). The Ps = 2550 kg.m 3) system the superficial gas (UGr) and
equation applied to the following ranges: 3.59 x 10 -4 _< liquid (UL,) velocities were independently varied over the
UoktZJCL -< 3.70 x 10-2; 1.69 x 10 -ll _< g~tL4/pL(YL3 _< 2.55 X respective ranges of 0.04-0.12 m.sa and 0.2-0.6 m.s ~ in a
10-6; 0.471 _<d/do -< 0.743; 0 _<Ma _<95.9; 0 _<Cs/Ps -< 8.00 vertical pipe (diameter = 0.08 m). The gas and the solids
x 102; and 1.17 x 10.2 _< Ut/I..JG_<0.844 (Koide et al., 1985). holdups were correlated with the equations (Douek et al.,
1995)
Equation (47) applied also to solids-free systems when
the Cs-value was set to zero (Koide et al., 1985). The
physical properties ranges covered were: PL = 997-1182
UGr
kg'm'3; ktL = (0.894-17.0) x 10-3 Pa.s; ~L = (51.7--73.0) X eGr = ].| l(UGr + ULr q- Usr) + 0 . 2 6 ' (49)
10.3 N.m-~; and Ma = 0-95.9 (Koide et al., 1985).
In suspensions of basalt particles in air-water in a EXHAUST
draft-tube sparged internal-loo p airlift reactor, Cai and DEGASSER FOAM CONTROL
Nieuwstad (1992) observed that the riser gas holdup showed \
excellent agreement with Chisti's (1989, p. 164) equation II\
(32) which applied irrespective of the solids loading
presumably because the loading did not exceed 4%.
Similarly, the overall volumetric gas-liquid mass transfer
coefficient was insensitive to the solids concentration (Cai
and Nieuwstad, 1992). Details of the systems tested follow: SETTLER - -
A / A d = 0.77; total volmne = 0.186 m3; Ps = 2760 kg.m "3, dp ANNULUS - -
= 0.245 ram; the input solids loading = 27-134 kg.m-3 (Cai
and Nieuwstad, 1992).
The overall gas holdup in a draft-tube sparged 'i
concentric draft-tube airlift reactor (A/Ad = 1.3; aspect ratio
5; Uo, = (0.154-1.39) x 10"~ m.s ~) for potential use in
microbial desulfurization of coal was reported by Smith and
Skidmore (1990). Measurements were made in a simulated
I
basal salt medium (composition in distilled water, kg.m~:
(NH4)2804, 1.3; KH2PO4, 0.28; MgSO4.2H20, 0.25;
CaCI2.2H20, 0.07; acidified to pH 2 by sulfuric acid)
containing pulverized coal (74 x 10.6 m particle diameter;
density = 1415 kg.m -3) at 0-40% wt (equivalent to 0-0.315 t
AIR
volume fraction of solids in gas-free medium) at 30, 50 and
72°C. ' At any fixed gas flow rate, increases in fluid
temperature led to small increases in gas holdup (Smith and Fig 7. Microcarrier airlift bioreactor used in wastewater
Skidmore, 1990) possibly because of increased turbulence treatment with inunobilized biofihn (Heijnen et al., 1992).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
and
Usr
Gs~ = 0.93(UG~ + ULrq- U S r ) -- 0.12' (50)
where Usr is the superficial velocity of solids, and 6s~ is the
solids holdup in the three-phase system. The velocity and
holdup of solids were determined by collecting the overflow
from the pipe into a bucket, quantifying the flow rate and
the volumes of the individual phases. Based on equation
(50), a fairly flat radial distribution of solids in the tube was
inferred because the drift-flux parameter was close to unity.
The constant, 0.12 m.s~, in equation (50) amounted to the
terminal settling velocity of solids in the three-phase
system. As expected, this velocity was significantly higher
than 0.08 m.s ~ that was estimated for settling in gas-free
liquid. Correlations such as equations (49) and (50) are
expected to apply to risers of airlift devices (Chisti, 1989,
pp. 269-271).
3.2. 3 A general approach f o r predicting gas holdup
As noted above and in prior reviews (Chisti and Moo-
Young, 1987a; Chisti, 1989, pp. 58--65), many correlations
are available for gas holdup in airlift reactors. However, as
a rule, these correlations are system specific, being of little
use in design or scaleup where the usual requirement is for
estimation of expected performance in larger or
geometrically different reactors or fluids. An alternative,
scale-independent, method for prediction of gas holdup is
now available (Chisti, 1989, pp. 266-272). This procedure
is based on fundamental principles and it applies to the
external-loop as well as the internal-loop geometric types
of airlift reactors. The procedure has been demonstrated for
newtonian and non-newtonian media (Chisti, 1989).
Unlike most other correlations, Chisti's method employs
simultaneous determination of gas holdup and the induced
rate of liquid circulation. This is logical because in airlift
reactors gas holdup and the velocity of liquid are
interdependent: the holdup difference between the riser and
the downcomer drives the circulatory flow which in turn
affects the holdup (Section 4). The method works as
follows: the parameters a and b in the Zuber and Findlay
type expression (equation 24) are experimentally established
• with the fluid of interest. The parameters are determined
over the range of Uor and independently varied UL~that are
expected to occur in the riser of the airlift device. The
experimental details and the model rig for the
measurements have been described (Chisti, 1989). Once the
riser gas holdup is known (equation 24) using an assumed
value for UL, and a selected U~r, the downcomer gas holdup
is calculated. The latter is usually a fraction of the riser gas
holdup (see Section 3.2.1) if the reactor has no gas-liquid
separator. If, on the other hand, an effective gas-liquid
separator is used (see Section 5), the downcomer holdup is
negligible. The holdup values and the assumed value of UL~
are now used in equation (42) to compute a value for Uu.
(see Section 4). If the assumed and the calculated
UL~-values agree, the corresponding gas holdups are the
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 45
desired values. If no agreement is found, the iterative
procedure is repeated with a new assumption for Uzr.
Figure 8 compares the riser gas holdup calculated using
Chisti's procedure with measurements in a large annulus
sparged concentric tube airlift reactor (~1 m~; A / A d = 3.61;
aspect ratio = 3; air-water system).
3.3 Local events
In gas-liquid systems local properties such as bubbling
frequency, bubble rise velocity, bubble diameter and gas
holdup have sometimes been measured by electrical
resistivity and optical probes. Several such studies are cited
by Choi and Lee (1990a) who show that the design of the
sensing probe has a strong influence on the measurements.
Because different designs can give very different results
(Choi and Lee, 1990a), probe-based measurements should
be viewed with caution unless the data have been verified
with an independent, well--established method. Choi and
Lee (1990b) used an optical probe to measure radial
distribution of gas holdup in air-water system in the
annular downcomer of a concentric tube airlift reactor (~70
L; A / A d ~ 0.7; aspect ratio ~ 6.5; U~r = 0.05-0.20 m.s~).
The downcomer gas holdup increased with increasing gas
flow rate in the riser, but the radial holdup profiles were
fairly flat over the dimensionless radial distance (r - rl)/(ro -
ri) = 0.2-0.8 for the entire range of gas flow rates (Choi and
Lee, 1990b). Here, r is the radial position in the annulus, ri
is the outer radius of the draft-tube, and re is the radius of
the outer tube. For all gas flow rates, the cross-Zsecfional
averaged local gas holdup declined down the downcomer
(Choi and Lee, 1990b). The optical probe preferentially
measured smaller bubbles (Choi and Lee, 1990b); thus, the
reported axial decline in holdup was probably less
pronounced than in the real situation. This was likely
because smaller bubbles are known to penetrate deeper into
the downcomer so that the proportion of smaller bubbles
•increases axially downward (Chisti, 1989; Chisti and Moo-
Young, 1987a); consequently, higher than actual holdups
are likely to have been •measured in the lower zones. Data
20
• o
10
%
x
I
• EXPERIMENTAL
o CALCULATED
I I K i t
2 3 10
uc~(m.s-1)xlO2
Fig 8. Measured and calculated riser gas holdups versus
superficial gas velocity in air-water in a ca 1 m3 airlift reactor.
The line is based on the measurements (Chisti, 1989, p. 267).
 46 SPECIAL
ISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl MechRev vol 51, no 1, January1998

presented by Choi and Lee (1990b) suggest that up to a 219; Chisti et al., 1988; Lu et al., 1995), hence equation
superficial gas velocity of 0.05 m.s4 (usual upper limit in (51) simplifies to
many applications) only about 8% of the gas entrained into
the downcomer recirculated into the riser. The recirculated
fraction rose to --.49% of the entrained gas at the superficial
gas velocity of 0.20 m.s4 (Choi and Lee, 1990b). In the
upper portions of the annulus (downcomer) only about 18% ' ULr ---/ (52)
of the fresh gas input to the column was entrained at the
lower gas velocity (Uar = 0.05 m.s'l); this fraction rose to
-29% at the highest aeration rate of 0.20 m.s4 . Even at this
exceedingly high flow rate, the recirculated gas from the For external-loop types of vessels, where the geometry of
downcomer contributed less than 15% to the superficial gas the top and the bottom zone is often quite similar, Kr and
velocity in the riser. As a general rule, therefore, the KB are roughly equal, and equation (51) reduces to
contribution of recirculated gas to the gas flow in the riser
may be neglected for all practical purposes as has also been
0.5
emphasized in the past (Chisti, 1989).
Measurements of 'bubble rise velocity' presented by 2gho(=~r-=~a)
Choi and Lee (1990a) for the air-water system in the riser Uzr = • (53)

of a draft-tube sparged concentric tute airlift device (A/Ad

= 0.6; aspect ratio = 6.5) were actually the sum of the true
bubble rise velocity and the liquid velocity. The latter was
KB (l- =Gr)2 ~.AdJ (]--~Gd)
not documented. Thus, the apparent bubble rise velocities For internal-loop vessels such as the concentric
varied over 0.7-1.4 m.s4 (Choi and Lee, 1990a) whereas cylinder or the split-cylinder designs, the parameter KB has
the expected range for true rise velocity in air-water is 0.2- been empirically correlated with the equation (Chisti et aL,
0.3 m.s-1. 1988; Chisti, 1989, pp. 203-219)

4. LIQUID CIRCULATION AND HYDRAULICS KB=I1.40(A-~10.79. (54)

4.1 Induced liquid circulation rate
4.1.1 Water-like newtonian fluids
A better known equation for predicting the induced liquid
circulation velocity in airlift reactors is
In equation (54) Ab is the area available for flow under the
draft-tube or the baffle as illustrated in Figure 9. The Aa/Ab
range covered by equation (54) is 0.2-1.8 (Chisti et al.,
1988; Chisti, 1989, pp. 203-219). The K8 (or Kr) values for

]0.5 external-loop devices are typically about 5 so long as the

geometry is within the ranges specified by Chisti (1989, pp.
203-219). Note that equations (51)-(53) disregard the wall
2ghD(£Gr-£~a) friction associated losses which have been proven to be
(51)
ULr= "(1 x , " negligible relative to form frictional losses in water-like
flow in typical airlift reactors. Ganzeveld et al. (1995) have
:7DRY K=t.zJ i-=o,)' shown that equation (52) can satisfactorily predict the liquid
circulation velocity in animal cell microcarfier suspensions.
Equation (51) was developed for low viscosity water-like
media by a purely mechanistic approach (Chisti et al., 1988;
Chisti, 1989, pp. 203-219); it was shown to apply to
external-loop and internal-loop airlift configurations over a
broad scale. Since initial publication, it has been repeatedly
validated (Cai and Nieuwstad, 1991; Choi and Lee, 1993;
Fraser and Hill, 1993; Garcia-Calvo and Let6n, 1996;
Ganzeveld et aL, 1995; Gavrilescu and Tudose, 1995;
Petrovic et al., i991; Po~arac and Petrovic, 1991). An
algorithm for simultaneous prediction of the riser and the
downcomer gas holdups and the liquid velocity with
equation (51) has also been published (Chisti, 1989, pp. Droft-tube Split-cylinder
203-219). In equation (51), the parameters Kr and K8 are Configurotion Geometry
the form friction loss coefficients for the top and bottom
zones, respectively. For internaMoop airlift reactors, Kr is Fig 9. The free area Ab (shaded) under the baffle or draft-tube
generally negligible relative to KB (Chisti, 1989, pp. 203- in internal-loop airlift reactors (Chisti, 1989).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
The approach discussed above is well supported by
other independent evidence. Thus, in air-water system in a
large draft-tube sparged internal-loop airlift device (about
186 L; A/Ad = 0.77) equipped with an elaborate gas-liquid
separator, the liquid circulation velocities calculated with
equation (51) generally agreed well with the measured
values (Cai and Nieuwstad, 1991); the single exceptional
case was when the form friction loss coefficient (Kr) for the
constricted flow path at the top was significantly larger than
KB for the bottom. Apparently, the high value of Kr was not
correctly taken into account in calculation of the liquid
circulation rate.
In conformity with Chisti et aL (1988) and Chisti
(1989), Merchuk and Berzin (1995) concluded that wall
friction associated energy losses were negligible compared
to the other sources of energy dissipation in airlift reactors
operated with the air-water system. The conclusion that
"energy dissipation is higher in the bottom of the reactor..."
(Merchuk and Berzin, 1995) also agreed with the earlier
findings (Chisti et aL, 1988; Chisti, 1989) according to
which the form friction loss coefficient (Ke) at the bottom of
internal-loop airlift reactors was the major source of
pressure loss. Merchuk and Berzin (1995) did propose a
complicated model for liquid circulation. Little evidence
supported the model, and it required several questionable
empirical inputs. The view that the energy used in
compressing gas bubbles as they moved down the
downcomer somehow affected the overallenergy balance
(Merchuk and Berzin, 1995) was plainly incorrect because
the same energy was recovered when the gas bubbles
recirculated. The recirculating gas had no impact
whatsoever on the overall energy input.
For internal-loop airlift reactors operated with low
viscosity, water-like media, Wachi et al. (1991) obtained
the same general equation (equation 51) as previously
derived by Chisti et aL (1988); however, the form frictional
loss coefficients, Kr and KB in equation (51), were defined
differently as (Wachi et aL, 1991)
K,--f rq\An/ - 1,
(55)
and
Kq = k~) - 1, (56)
whereAn andAD are the 'effective' cross-sectional areas for
fluid flow in the entry regions of the riser and the
downcomer, respectively (Wachi et aL, 1991). Although
the concept of effective area for flow of streamlines through
sudden contraction or enlargement of flow channel is well
established (Coulson and Richardson, 1977), little
knowledge is available on calculation of such areas,
particularly for the flow geometries that are commonly seen
in internal-loop airlift devices. As a further modification of
equation (51), Wachi et aL (1991) replaced the height of
dispersion (ho) with the length of the draft-tube (La).
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 47
For draft-tube sparged concentric tube airlift reactors
(A/Ad = 0.05-1.48; aspect ratio = 9), Wachi et al. (1991)
experimentally confirmed the pressure drop in the bottom
zone to be greater than in the top region, that is K8 > Kr, as
originally asserted by Chisti et al. (1988) and inferred also
by Lu et aL (1995).
Significantly, Wachi e t al. (1991) showed that at a
fixed aspect ratio, the amount o f gas holdup in the
downcomer is dependent on the downcomer-to--riser cross-
sectional area ratio (Aa/Ar). AsAa/Ar increased from 0.55 to
8.57, the gas holdup in the downcomer declined for any
fixed riser holdup (Wachi et al., 1991). This is sensible,
because as AdA, increases, the reduced flow velocity
through the larger cross-sectional area of the downcomer
becomes less effective in dragging gas bubbles into the
downcomer. Irrespective of the Aa/A, value, the riser and
downcomer gas holdups were found to be linearly related
(Wachi et aL, 1991).
As is evident from the data presented by Wachi et aL
(1991), any fluid properties (for example, viscosity, surface
tension, ionic strength) that affect bubble size have the
potential to alter the gas holdup distribution between the
riser and the downcomer and, hence, affect the induced
liquid circulation rate. Thus, a low surface tension may
reduce the liquid circulation velocity relative to water, while
certain non-newtonian media may actually circulate faster
(Wachi et aL, 1991).
Data on gas holdup and liquid circulation velocity in
several external-loop airlift reactors (A/Ad = 0.11--0.53;
aspect ratio based on riser = 11; air-water) were reported by
Choi and Lee (1993). For any given reactor, the
configurations of the top and the bottom connections
between the riser and the downcomer were identical. The
horizontal length of the connecting pipes (/) varied, so that
the ratio between I and the vertical distance (Lh) between the
top and the bottom connections ranged over 0.091-0.455.
The measured data were used in equation (53) of Chisti et
al. (1988), to calculate the values of the loss coefficient KB
(Choi and Lee, 1993). The calculated values were
empirically correlated with the geometric variables as
follows
KB = 12.70511 + Aa~ z°5(1 + l.~ -u~9 (57)
AJ \ --~h) "
Equation (57), which applied over a narrow range (KB =
10-25), has important weaknesses (Chisti and Moo-Young,
1994c). In particular, why the distance (Lh) between the top
and the bottom connecting zones should affect frictional
losses in these zones is difficult to understand (Chisti and
Moo-Young, 1994c). Further, for any given A/Aa ratio, K8
is expected to increase with the length of the connecting
pipe (I) between the riser and the downcomer; however,
equation (57) predicts the opposite (Chisti and Moo-Young,
1994c). Despite equation (57), the experimental data of
Choi and Lee (1993) did show that for otherwise fixed
conditions, increasing l slowed down the circulation of
liquid. As a result, other associated parameters (gas-liquid
 48 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
volumetric mass transfer coefficient, gas holdup) were
affected.
In another external-loop device (0.165 m3; air-water;
aspect ratio = 16 based on riser; A/Ad = 4) equipped with a
butterfly valve in the bottom pipe between the riser and the
downcomer, the extent of opening of the valve was the main
determinant of the induced liquid circulation rate for any
fixed value of the superficialgas velocity (Verlaan et al.,
1989b). The highest velocity (ULd), when the valve was
fully open, was about 1.4 m.s ~ when the superficial gas
velocity in the riser was about 0.14 m.s ~. " The valve
position affected the bottom friction loss coefficient KB
which ranged from about 5 to infinity with the valve fully
closed in the bubble column mode of operation (Verlaan et
al., 1989b). At any fixed valve opening, the coefficient KB
depended on the Reynolds number, declining somewhat
with increasing Reynolds number (Verlaan et al., 1989b).
For a fixed gas flow rate, the gas holdup in the riser
declined with increasing liquid velocity achieved by
opening the valve; hence, the bubble column mode of
operation gave the highest values of gas holdup (Verlaan et
al., 1989b).
In the air-water system in a draft-tube sparged
internal-loop airlift reactor (A/Ad = 0.77; total volume =
0.186 m3), the measured pressure drop (APB) in the bottom
zone of the reactor showed excellent agreement with the
equation (Cai and Nieuwstad, 1992)
,A P 8 = K B - -
2g
In equation (58), Vz.d is the linear liquid velocity in the
downcomer. Addition of basalt particles (ps = 2760 kg.m 3,
dp = 0.245 mm; the input solids loading = 27-134 kg.m -3)
increased the Kin-value, but equation (58) was obeyed until
a loading of 27 kg.m -~. At higher loadings, KB was no
longer independent of Vzf (Cai and Nieuwstad, 1992).
A large amount of data on liquid circulation rates in
an external-loop airlift reactor that achieved complete gas-
liquid separation was reported by Po~arac and Petrovic
(1991). Tap water and aqueous solutions of alcohols (0.5 or
5% vol methanol; 0.5% vol n-butanol) Were used. The
superficial air velocity in the riser varied over 0.01-0.16
m.s-~ with corresponding liquid velocities (ULr) in the range
0.05-0.65 m.s ~. The relevant reactor geometries were:
A/Ad = 1.0; and aspect ratio = 28. The liquid circulation
data agreed within about 17%, with the equation (51) of
Chisti et al. (1988). On the other hand, predictions of the
correlation proposed by Kawase (1990) were absurdly high
at more than 280% of the data (Po~arac and Petrovic, 1991).
Extensive data in draft-tube sparged concentric tube
airlift reactors also agreed with the predictions of equation
(52); however, for A/Ad ratios less than 0.6, the form
friction coefficient (Kr) for the top zone could not be
ignored (Petrovic et al., 1991). The measurements were in
tap water, aqueous n-butanol (0.5% wt) and aqueous
glycerol (46% wt). The geometric details of the reactors
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
were: A/Ad = 0.2-1.3; and aspect ratios ~ 10 (Petrovic et
aL, 1991).
For liquid circulation velocity in two external-loop
airlift reactors (A/Ad = 2.25 and 4.0; aspect ratio = 14 and
40) operated with air-water, Glennon et al. (1993) reported
the following empirical equations:
1 1"~1"71.-0.409TT 0.420
ULr = l.Vltll, f I..JGr , (59)
for U~, _<0.05 m.s-~; and
U t , = 0.735k/°427 trrO.31s
J Gr J (60)
for U~, > 0.05 m.sk In these equations kf was a friction-
dependent factor given as (Glennon et al., 1993):
k: ZKi ( A~ 2 (61)
2ghD \ Ad; '
where Kis are the individual frictional loss coefficients in
the circulation loop. Equations (59) and (60) are clearly
unsuitable when gas is entrained into the downcomer.
4.1.2 Viscous and non-newtonian media
Broths of mycelial fungi, filamentous bacteria and
microorganisms that produce large amounts of extracellular
polymers, tend to have a viscous non-newtonian rheology.
Such broths can often be successfully cultured in airlift
(58) bioreactors (Cai et aL, 1992b; Chisti and Moo-Young,
1987a; SchOgerl, 1990; Stasinopoulos and Seviour, 1993),
but sometimes, severe limitations on oxygen transfer and
mixing adversely affect production (Bayer et aL, 1989;
Srivastava and Kundu, 1995). For practical purposes, the
behavior of most non-newtonian broths can be described as
pseudoplastic, or shear-thinning. Discussions on whether
Bingham plastic, Casson model, or other flow behavior is
more closely followed by a particular culture fluid are
largely academic (Chisti, 1989; Chisti and Moo-Young,
1991). Pseudoplastic fluids follow the power law model;
thus, the shear stress (t) in the fluid depends on the shear
rate (y) to which the fluid is exposed:
r=Ky", (62)
where K is the consistency index and n is the flow behavior
index of the fluid. By analogy with newtonian fluids, for
which the viscosity is defined as the ratio of shear stress and
shear rate, an apparent viscosity (pap) is defined for tile
power law fluids as follows
,up - r - KZ,,_1. (63)
Y
The apparent viscosity of pseudoplastic fluids decreases
with increasing shear rate; for these fluids, the flow index
 Appl Mech Rev vol 51, no !, January 1998 Chisti: Pneumatically mixed bioreactors 49

(n) is less than unity. For newtonian fluids, n is unity, and

the viscosity is independent of the rate of shear.
For viscous newtonian and non-newtonian media in
airlift reactors, equation (51) does not apply; instead, the ~'L
n2
mechanistic approach that was used in developing equation
(51) can be extended (Chisti and Moo-Young, 1988a; 2(l+n)
Chisti, 1989, pp. 221-229) to yield the following expression

UL~A,(APFr+ APxzd)-pLghDULrAr(GO~-6Od) (68)

~l+n
PL L.A. Kr 2 1 2 × 1+ 3(1 - 6G1
+ 2 (157Gr) + K B (1- eOd) = 0,
l+n( . ' M+nl ~
(64)

where and APFd are the frictional pressure drops in tile

~kPFr
L 1+ ------i~-/ ,
riser and the downcomer, respectively. Equation (64) may
be used to predict the rate of liquid circulation as discussed
below.

In a manner analogous to flow in pipes, the frictional

pressure losses in the riser and tile downcomer are
expressed as Turbulentflow. Turbulent flow exists for Reynolds numbers
higher than those given by equation (66). For this regime
L the shear stress for the flow of gas-free liquid is (Sokolov
AP = rD,
4--;-
a
(65)
and Metkin, 1981; Metkin and Sokolov, 1982)

where XD is the wall shear stress in the gas-liquid system

(note that all fermentation broths without non-biological
suspended solids, may be treated as pseudohomogeneous 0.316n°nlP LU~
liquids), L is the length of the flow channel and d is its (69)
hydraulic diameter. rL = ~K:"---e2/~-'÷2 '

Two possible flow situations--laminar or turbulent

flow--may exist and these are considered separately below, and
following the approach proposed elsewhere (Chisti and
Moo-Young, 1988a; Chisti, 1989, pp. 221-229).
Laminarflow. For non-newtonian pseudoplastic media, the
2xn
flow is laminar when (Sokolov and Metkin, 1982)
fi'D =( In(x-l)+2__
1
rL \ 1 - eGJ
xn 2
3n+ 1 "~(.+1×.(x-0+2)
Re< 2 3 0 0 - - (66)
4n
|+n
2x(l+n) n (70)
3 ( 1 - co)"(~-')
+----~ rE
For this condition, tile shear stress for tile flow of gas-free
liquid is (Metkin and Sokolov, 1981) 1+ Xn

8pLU~ / 2x0+.) / I:. .....

rL - - - , (67)
Re / 1+ - • I ' - '- ' -

and for tile flow of gas-liquid dispersion, this stress is given r L _.i " Y
as (Metkin and Sokolov, 1982)
/

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 50 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Throughout, the Reynolds number and the parameters
and x are given (Sokolov and Metkin, 1981), respectively,
by
8TT2-n :In r~
Re =
I6.+
UL " ~'L (71)
1
(72)
~3:4-7 g~'~(1-~'~) ~ l_oooj,
and
x = 0.70n -°591. (73)
Equation (73) was obtained by Chisti and Moo-Young
(1988a) using the data reported by Sokolov and Metkin
(1981).
Equations (66)-(72) were developed theoretically and
were successfully used by Metkin and Sokolov (1982) for
the calculation of xJxL in the flow of Aspergillus niger
suspensions (5.25-19.8 kg.m "3 dry biomass; n = 0.787-0.92;
K/pL = (0.885-5.14) x 10-4 mLs"'2) in vertical pipes
(diameter = 0.05 and 0.5 m) over a broad range of air
(0.05-0.4 m.s "~) and slurry (0.13-1.2 m.s"~) velocities. The
experimental and the predicted values of the ratio XJXL
were within ±12% (Metkin and Sokolov, 1982).
Equations (64)-(72) may be used to predict the liquid
circulation velocity. The following procedure is necessary
(Chisti and Moo-Young, 1988a; Chisti, 1989, pp. 221-229):
1. Initially, a value of ULr is assumed. The riser gas
holdup must either be known or it must be estimated.
A suitable equation for estimating the holdup in broths
of Aspergillus niger is (Metkin and Sokolov, 1982)
U~r (74)
~Gr "~
U~ + ULr+ 0.9(1+20Csxf~)"
Other similar equations have been reported (Chisti et
al., 1987; Chisti, 1989; Kemblowski et al., 1993).
. The Reynolds numbers Rer and Red are calculated
(equation 71), followed by x (equation 73). Depending
on whether the flow is laminar or turbulent, either
equation (67) or equation (69) is used to calculate x~
and XLr. The parameter 5 is then calculated (equation
72).
. The shear stress XDr (equation 68 or equation 70) is
calculated. Note that for many cases, especially for
the external-loop type of airlift vessels, the
downcomer gas holdup is close to zero and the
calculation of XDd is not necessary. When the
downcomer gas holdup is not zero, equations such as
those reported by Chisti et al. (1987) may be used to
calculate it; XDd is then calculated (equation 68 or 70).
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
4. The pressure drops, AP~'r and APFd , a r e calculated
using equation (65).
5. The dispersion height is obtained as follows:
hL
hD : I CGdAd] (75)
1- 6~, Ar +
Ar+ J
6. The frictional loss coefficients for the top (Kr) and the
bottom zones (Ks) of the reactor are calculated as
noted in Section 4.1.1 (Chisti, 1989, p. 226).
7. Equation (64) is now solved for ULr using the already
calculated values of APFr, APFd, hD, ~r, ~a, KT and
KB.
8. The ULr calculated in step 7 is compared with the
value that was initially assumed (step 1). An
agreement indicates a solution. When no agr~ment is
obtained, steps 1-8 must be repeated with a new
assumption for ULr.
The foregoing procedure has been shown to provide
excellent estimates of the liquid circulation velocity in non-
newtonian fluids (Chisti and Moo-Young, 1988a; Chisti,
1989, pp. 221-229; Kemblowski et al., 1993).
Upon reviewing the available literature on prediction
of the induced liquid circulation rate in airlift reactors,
Kemblowski et al. (1993) favored the approaches developed
by Chisti and coworkers for both newtonian and non-
newtonian media (Chisti, 1989, pp. 203-229; Chisti et al.,
1988; Chisti and Moo-Young, 1988a). In two large
external-loop airlift reactors (80 L, A / A d = 1, aspect ratio =
19.5; 700 L, A/Ad = 1 or 1.8, aspect ratio (based on riser) =
35.9), the liquid circulation rate could be predicted within
• 20% of the actual value using Chisti's methods
(Kemblowski et al., 1993). This further confirmed the
proposed approaches particularly for non-newtonian fluids.
The data of Kemblowski et al. (1993) covered the ranges: Pz
= 998-1286 kg.m3; CL = 42.0-81.6 mN.m'l; n = 0.758-1;
and K = 0.001-0.261 Pa.s".
Among other approaches, methods for predicting gas
holdup and liquid circulation velocity via the drift flux
model have been proposed (Ayazi Shamlou et al., 1994; Lu
et al., 1995). These methods invariably require difficult to
obtain information on the drift flux distribution parameter
and the bubble rise velocity (Ayazi Shamlou et al., 1994; Lu
et al., 1995). Thus, tofit the model to the measured data,
the friction coefficient and the drift flux parameter were
varied arbitrarily by Ayazi Shamlou et aL (1994). The
effect of downcomer gas holdup on circulation of liquid was
apparently not considered. Data in newtonian broth of
Saccharomyces cerevisiae and non-newtonian power law
broth of Saccharopolyspora erythraea were correlated
(Ayazi Shamlou et al., 1994). Surprisingly, none of the
equations used in the model (Ayazi Shamlou et al., 1994)
contained a term for the K-value of non-newtonian fluids.
Apparently, somehow, the values Of the distribution
parameter and the friction coefficient that were used as
 Appl Mech Rev vol 51, no 1, January 1998
fitting parameters accounted for tile K-value. The general
applicability of the proposed procedure remained untested.
In the work reported by Lu et al. (1995) for gas-
liquid-solid slurries, the bottom frictional loss coefficient
KB was calculated using equation (54) of Chisti et aL
(1988). The frictional losses in the head zone were
neglected because, in keepingwith earlier observations
(Chisfi et al., 1988; Chisti, 1989), the clearance between the
liquid level and the top of the draft-tube did not have any
effect on liquid circulation velocity (Lu et al., 1995). The
model required information on the wall frictional losses in
multiphase system, and the viscosity of the slurry. A
limited amount of data from draft-tube sparged internal-
loop configuration did agree with the model (Lu et aL,
1995). In general, the liquid circulation velocity and the
gas holdups in the riser and the downcomer decreased with
increasing loading of solids (calcium alginate beads; p~ =
1030 kg.m3; dp= 1-3.6 mm; loading = 0-30% vol). The
liquid velocity declined with increasing diameter of
particles. The velocity increased with increasing height of
the airlift loop roughly in proportion to (hD)°5, agreeing
with the mechanistic equation (51) derived by Chisti et al.
(1988). This effect of height was further confirmed by
Russell et al. (1994) who obtained the equation
~ ~/- rO.44::tO.03 1-0.47_+0.07
(76)
ULr u.y t_.t G r .l-.Jd •
Equation (76) was obtained in draft-tube sparged concentric
tube airlift reactors (A/Ad = 0.8) with baker's yeast broth.
It applied for the riser superficial gas velocities up to 0.2
m.s -1. The height of draft-tube (Ld) varied over 1.2-3.2 m
(height of draft-tube to vessel diameter ratios of 3.8-10.1).
Additional data in slurries of calcium alginate beads in
draft-tube sparged vessels have been reported by Vicente
and Teixeira (1995).
An analysis of liquid circulation in external-loop types
of reactors was attejnpted by Kawase (1990). No
consideration was given to gas holdup in the downcomer;
therefore, complete gas-liquid separation was implied even
though it is impossible to achieved in many non-newtonian
media. Despite overwhelming evidence to the contrary
(Akita et al., 1994; Chisti, 1989; Chisti and Moo-Young,
1987a; Choi and Lee, 1993), Kawase (1990) assumed that
gas holdup in airlift reactors was governed by equations that
were developed for bubble columns. In addition, isotropic
turbulence was assumed to exist (Kawase, 1990)even
though this assumption has been experimentally discredited
(Ltibbert and Larson, 1990; L0bbert et al., 1990; Okada et
al., 1993). Likewise, the analysis invoked eddy diffusivity
(Kawase, 1990) which is highly uncertain for multiphase
systems. Finally, the important effects of the geometric
configurations of the bottom and the top zones of tile reactor
on liquid circulation were completely disregarded as were
the reactor height associated effects (Kawase, 1990). The
results seemed to indicate a decrease in liquid circulation
velocity with increasing shear-thinning behavior (that is,
decreasing n-value)--an observation that is inconsistent
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 51
with other work (Akita er aL, 1994; Kawase et al., 1994).
In view of these problems, the finding that Kawase's (1990)
correlation overpredicts liquid circulation rates by more
than 280% (Po~rac and Petrovic, 1991) is hardly
surprising.
For two small external-loop airlift reactors (Ad/A~ =
0.458 and 0.204; aspect ratio based on riser = 7.5), Kawase
et ai. (1994) noted that the induced liquid circulation
velocity data in air-water were lower than the predictions of
Chisti's equation (53). The predicted values were
apparently calculated with an unrealistically low assumed
KB-value of 5.0 even though a much larger value would
have been more appropriate for the awkwardly designed,
restrictive pipe sections that connected the riser and the
downcomer. The results in air-water did show a decline in
the liquid circulation velocity with increasing A / A a ratio as
expected from equation (53). A similar behavior was noted
in pseudoplastic solutions of carboxymethyl cellulose
(Kawase et al., 1994).
For liquid circulation time (t,) in external-loop
reactors, Kawase et al. (1994) proposed the equation
A "x-(l'3n+°8)( rr2 .,~1/3
._q_L] ]U_.Ef_[
tc = 5.53n 2/3 l+Ary \gdr.) 2Uor, (77)
L
where L was the length of the circulation path. Equation
(77) is supposedly based on fundamental principles, but its
applicability is quite limited as important factors such as
hydraulic resistance of the flow path, the gas-liquid
separation and the aspect ratio of the reactor have been
completely disregarded in developing it.
Note that equation (5!) which has been well proven
(Chisti, 1989) for prediction of the liquid circulation
velocity may be used to calculate the liquid circulation time
(to). Thus, by definition, in an internal-loop reactor, tc is
' given by
hr ha Ad
tc +
ULr ULr Ar (78/
Typically, tile heights of the riser (hr) and the downcomer
(hd) are roughly equal; hence,
tc = 1+ . (79)
For external-loop type of reactors, equation (78) should also
include the lengths of the top and the bottom connection
pieces.
Using several newtonian media (pL = 997-1179
kg.ln-3; i.tL= (0.894-16.81) x 10"3 Pa.s; cL = (53-72) x 10-3
N.in 1) in draft-tube sparged concentric tube airlift devices
(A,/Ad = 0.3-1.2; aspect ratios = 5-16), Koide et al. (1988)
developed empirical equations for prediction of pressure
drops for fluid turnaround in the tbp (APt). and the bottom
(APs) zones. These equations are:
 52 SPECIAL ISSUE: FLUID MECHANICS PROBLEMS IN BIOTECHNOLOGY Appl Mech R e v vol 51, no 1, January 1998

In these equations Hc is the bottom clearance of the draft-

tube; d~ is its diameter, and tw is its wall thickness, whereas
Ad U Ld do is the diameter of the outer tube.
APe = o.45 tJh,O-
Tile average error of estimations of APB with
(1 n'(di+ tw)h, 1 equations (80) and (83) was claimed to be within 31% for
109 data (Koide et al., 1988). In reactors with H,/di > 0.5,
× Z ) the H/di value of 0.5 was recommended for use in equation

_132 (83). The mean error in estimation of APr with equations

(81) and (84) was within 29% for 119 data (Koide et al.,
+ 2 kl---~oJ \zr(di+ t~)h,(1-~Od) 1988). Typically, (APt + APB) made up 65-85% of the total
pressure drop in the circulation loop (Koide et al., 1988).
+
k d, J' In studies with newtonian media (~L = 1~-266 mPa.s)
(80) and a pseudoplastic fluid (K = 0.0256 Pa.s °9°2) in an
external-loop airlift reactor (0.138 m3; A,/Ad = 1; aspect
and
ratio = 58; Uor = 01002-0.12 m.sl), Akita et al. (1994)
noted that in most cases, the measurements of the induced
ArULr liquid circulation velocity correlated well with the
APt, = 0.45 .n.td i tw, h.[ 1 difference in gas holdup between the riser and the
downcomer raised to a power of 0.5 (Akita et al., 1994), as
expected from the theoretical derivation (equation 51) of
Chisti et al. (1988). However, for highly viscous newtonian
× (] -Z7 liquids (I.tL> 14.9 mPa-s), slight deviations were observed
when the riser gas holdup exceeded about 0.03 (Akita et al.,
1994). For water-like fluids, no gas was carried into the
downcomer, because of an effective gas-liquid separator;
however, in more viscous media, complete disengagement
(80 of the tiny gas bubbles could not be attained. USing the
terminology of Chisti et al. (1988), the form frictional loss
where ey, the gas holdup over the upper end of tile riser, was coefficients for the top (KT) and the bottom (KB) zones of the
defined as reactor were empirically correlated as follows (Akita et al.,
1994)

eI - dZo (82)
KB = (1- 6Od)2 1.3+32 °°Gr L ~ J '
In equations (80) and (81), ht is the height of tile liquid
stream that flows under the lower edge of the draft-tube,
(85)
and h, is the corresponding height for flow over the upper
edge (Figure 10). The parameters ht and h, were calculated
respectively as (Koide et al., 1988)

h, I~/¢1 0"401( ~ t 4 r ~ 0.659r Ad ~1.26

d--~ = 0"434 \ltlr + AdJ \ArT~4d~ '

(83) TUBE
hu h!

and
DRAFT
f N0.0924. 0.818 TUBE

~- = 0.391 ArT~ TOPZONE BOTTOMZONE

(84)
(. A.a_ .~]'36Igf14~-O'OO392(l_SGr),.08
X \ Ar + Ad) "LPL0"3) Fig 10. The heights h. and ht of flow streams at the upper and
lower ends of the draft-tube in concentric tube airlift reactors.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Appl Mech Rev vol 51, no 1, January 1998
and
J/
Kr=( 1 - 6~r) 2 1 . 3 + 2 1 0 \PL)
(86)
Equations (85) and (86) were developed for an external-
loop reactor with a specific geometry. The bottom
configuration of the reactor consisted of two sharp right
angled elbows (Figure 11) which had the same diameter as
the rest of the loop (Akita et al., 1994). Because the A / A d
ratio was unity, there were no contraction or expansion
losses at the bottom; hence, equation (85) does not account
for such factors. Similarly, equation (86) is expected to be
specific to the peculiar geometry used in the top zone with a
gas-liquid separator (Figure 11). Usefulness of equations
(85) and (86) is limited as they do not show how geometric
variables affect Kr and KB.
4.2 Local p h e n o m e n a
Use of non-invasive laser-doppler anemometry for the study
of local flow velocities in air-water in draft-tube sparged
airlift reactors (Ad/Ar of 0.75 and 1.80; aspect ratio = 10)
has been reported (Tan et al., 1995). Low aeration rates,
0.64-2.40 L.min -1 (specific power inputs of 6.8 to 25.5
W.m3), relevant to suspension culture of hybridomas were
used. Little or no gas entrained into the downcomer
because of the low flow rates. Depending on the induced
liquid circulation rate, either parabolic or fiat radial velocity
profiles were observed in the draft-tube (riser). Because of
gas bubbles the profiles were not particularly smooth. In
the downcomer, the profiles were generally flatter and
smoother. Liquid velocity fluctuations as a measure of
turbulence intensity were shown to be lower in the
downcomer than in the riser. However, in both zones, the
magnitude of the fluctuations was fairly uniform over the
cross-section (Tan et al., 1995) indicating a relatively
unvarying level of turbulence in any given zone.
Radial profiles of liquid velocity in risers of airlift
reactors are generally flat at sufficiently high gas flow rates.
Local downflows near the walls of the riser are also known
to exist at yet higher gas flow rates (Chisti, 1989, pp. 132-
150; Chisti and Moo-Young, 1987a; Ltibbert et al., 1990).
In some airlift reactors, the flat velocity profiles have been
demonstrated to become increasingly parabolic with
increasing gas flow rates with corresponding increase in the
extent of downflow zone (negative liquid velocity) in the
periphery of the riser (Liibbert et al., 1990). This happens
because the downcomer of such a reactor is not able to
transport the increased liquid flow resulting from increasing
gas flow beyond a certain value (Liibbert et aL, 1990);
hence, supplementary downflows develop in the riser.
4.3 N u m e r i c a l simulations of flow
Computational hydrodynamic modeling of bubble columns
and draft-tube internal-loop airlift reactors has been
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
/
Chisti: Pneumatically mixed bioreactors 53
p 0.775 -I
I- .!
o o
0.148
LENGTH = 0.51m
RISER DOWNCOMER
(o) (b)
Fig 11. The top (a) and the bottom (b) configurations of the
reactor used in developing equations (85) and (86).
discussed by Svendsen et al. (1992) for the air-water
system. For a draft-tube sparged airlift device, qualitative
agreement was claimed between the model predictions and
the experimental data for radial' gas holdup profiles in the
riser. Similarly, the computed liquid velocity profiles were
qualitatively consistent with the measured data. Gas holdup
in the downcomer could not be satisfactorily predicted;
inadequacies in estimating the drag on bubbles were the
apparent reason for lack of agreement (Svendsen et al.,
1992). For the bubble column, the simulated liquid flow
fields presented by Svendsen et al. (1992) for the entire
vessel, seemed to show a single large circulation cell--a
behavior not consistent with experience (Chisti and Moo-
Young, 1987a) or with other more realistic simulations
(Sokolichin and Eigenberger, 1994).
For uniformly aerated bubble columns, other dynamic
simulations (Sokolichin and Eigenberger, 1994) suggest
that large unstable circulating vortices (or cells) originate at
the sparger and move up the columns (Figure 12a). Time
averaging of such dynamic flows over a long period led to
the general pattern of Figure 12b with up-flow in the center
and downflow near the walls (Sokolichin and Eigenberger,
1994).
Evaluation of flow, pressure and shear stress fields by
computational hydrodynamic methods has been described
for the downcomer of a draft-tube sparged internal-loop
airlift reactor with an expanded head zone for gas-liquid
separation (Zhou et al., 1994). Within the downcomer, the
flow, pressure and shear stress fields were uniform (Zhou et
al., 1994); variations occurred mostly in the expanded head
region where the flow from the riser turned into the
downcomer and accelerated in moving into the narrower
cross-section.
Flow simulations and observations in non-uniformly
aerated bubble columns and an external-loop airlift reactor
have also been reported (Becker et aL, 1994). Simulations
were generally consistent with observations. Of particular
note were the maldistribution of the gas in the riser, and
 54 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

0.75

..,xltllll
.,~lll||l/
..~1111|II
.,~II111/I
,itllItll
,lllIIttl
,rlllltll
,~ttlttlt
,tttlllIl
0.60- 0.60 - ,ttttlltl
.,ttltltll
,tttHIll
,tttlltll
,sttlttll
,ttIIltlI
,rllltIII
,~JlrIltI
,fltllrll
,11I||[|I
,ulllT|II
,olTIT|II
,iiiiiiii
.,111|IIII
.,wll|llrl
0.45" 0.45 - .,1111|It
.,1111||1
.,iI11|11
.,~111|II
.,~I11111
.~I11111

.,attllll

0.30 ¸

0.15

(a) (b)
0.00
( 5 0.00 0.05 0.10 0.15

Fig 12. Simulations of liquid velocity fields in a unifonnly aerated bubble column with an aspect ratio of 5: (a) instantaneous and (b)
long time averaged profiles. Based on Sokolichin and Eigenberger (1994).

existence of a liquid downflow zone in the entrance of the
riser, when the sparger was located below the point at which
the flow from the downcomer entered the riser (Becker et
al., 1994). Such maldistributions were reported also by
Chisfi and Moo-Young (1987a) and Chisti (1989, pp. 132-
148) in several different types of airlift reactors. Chisti and
Moo-Young (1987a) recommended relocating the sparger to
above the point where the downcomer flow joins the riser.
Flow non-homogeneities were corrected by such relocation
(Chisfi and Moo-Y0ung, 1987a; Chisti, 1989, pp. 132-148)
as confirmed also by Becker et al. (1994). However, with
the repositioned sparger, Becker et aL (1994) did note a
reduced rate of liquid circulation. This effect would not
have been significant if the reactor used by Becker et al.
(1994) had an effective gas liquid separator: with the
sparger located at the base of the riser, and consequent flow
maldistribution, bubble coalescence occurred in the riser
and most of the bubbles were relatively larger than when the
sparger was repositioned. This is clearly seen in the
photographs presented by Becker et al. (1994). With few
smaller bubbles, there was less bubble carryover into the
downcomer and the consequent smaller downcomer holdup
translated into a larger driving force for liquid circulation.
With the relocated sparger, the gas holdup in the
downcomer was noticeably higher. Even with good gas-
liquid separation, the placement of sparger within the flow
path in the riser would cause some reduction in liquid
circulation rate relative to positioning at the base o f the

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
riser. However, this effect can be kept negligibly small by
ensuring that the sparger has a large open area for flow
through of liquid. An example is the ladder type perforated
pipe sparger shown in Figure 13 (Chisti, 1989, pp. 94-96).
Generally, a well-designed sparger should block less than
30% of the cross-sectional area of the riser while assuring a
uniform distribution of the gas across the cross-section.
Such a sparger would produce relatively little pressure loss
when the Ad/A, ratio is less than or equal to AyAr, where AF
is the free area for liquid flow at the location of the sparger.
4.4 The gas-lift pump
Transport of liquids and slurries by gas-lift pumps is a well
known application of the airlift principle. While little used
in bioprocessing at present, the airlift pump is capable of
sterile operation and hence has a significant potential.
Unlike airlift bioreactors which are operated exclusively in
bubble flow or churn-turbulent flow regimes, the practical
operation of an airlift pump is restricted to the slug-flow
regime (Zenz, 1993), although there are exceptions.
Operational efficiency, Ef, of the airlift pump is the ratio of
the work required to lift the liquid to the point of discharge
and the work done by isothermal expansion of the gas. The
operational efficiency can be calculated using
G hLig
P.Qln
(87)
where G is the mass flow rate of the liquid, hLi is the height
of lift, Pa is the .atmospheric pressure, P is the pressure at
the point of entry of the gas, g is the gravitational
acceleration, and Q is the volume flow rate of the gas at the
inlet. The maximum operating efficiency is generally 40-
50% and falls off rapidly below or above the point of
optimum slug-flow operation (Zenz, 1993). The point of
maximum delivery of liquid typically occurs at gas flow
•mixing, solid-liquid and gas-liquid mass transfer and heat
t liquid separator always adds to the hydraulic resistance of
AIR
Fig 13. The perforated pipe ladder type gas sparger (Chisti,
1989).
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 55
rates significantly greater than at the point of maximum
efficiency. Aspects of design of gas-lift pumps are
discussed by Zenz (1993) and a more thorough treatment
has been provided by Brown (1967).
As an interesting example, Molina Grima et al. (1994;
1995) used an airlift drive to pump culture fluid through
lengths of horizontal tubes in the solar receiver of a
photobioreactor for culture of microalgae Isochrysis
galbana and Phaedactylum tricornutum for production of
polyunsaturated fatty acids. In both cases, the airlift pump
was 3 m high, and the biomass containing fluid (essentially
water) flowed through 80.8 m length of a 0.026 m diameter
tube. The flow velocity was 0.30 m.s 1 which produced
turbulent f l o w . The corresponding air flow rate was 60
L.min "1. The airlift drive was capable of generating higher
flows. Based on the data presented by Molina Grima et al.
(1995), the relationship between the linear liquid velocity
(Vzo, m.s "l) in the tube and the air flow rate (Q, L.min "l)
was
V ~ = 0.054 Q0.41. (88)
The aeration rate in equation (88) covered the range 20-140
L.min -1"
5. GAS-LIQUID SEPARATION
Many airlift fermenters do not have any especial features for
separating the gas from the liquid before it recirculates into
the downcomer. Examples are configurations (a) and (b) in
Figure 2. Internal-loop types of airlift bioreactors for
animal cell culture usually do not employ gas-liquid
separators because the aeration rates are low and the
resulting liquid circulation velocity is insufficient to drag
gas bubbles into the downcomer. In external-loop devices,
the head region does allow significant gas-liquid separation
even when not specifically designed for that purpose. In
comparison, most split-cylinder and draft-tube types of
airlifts without separators, when aerated at rates typically
required in microbial fermentations with water-like
properties, would have a downcomer gas holdup that
exceeds 80% of the holdup value in the riser.
Because the difference in gas holdup between the riser
and the downcomer provides the driving force for liquid
circulation, the extent of gas-liquid separation in the head
region strongly affects all performance characteristics of
airlift reactors. As noted in Figure 1, the magnitude of
induced liquid circulation velocity affects the gas holdup in
the riser, the intensity of turbulence, hence, shear rates,
transfer, the ability to suspend solid particles, break up
flocs, as well as other performance indices. While a gas-
the liquid circulation loop, with well designed separators
the increase in driving force for circulation is such that the
net effect is a substantial enhancement of liquid circulation
rate. Increases of three- to four-fold over equivalent vessels
without separators are easily achieved. Typically, gas-
 56 SPECIAL
iSSUE:FLUIDMECHANICS
PROBLEMSINBIOTECHNOLOGY ApplMechRevvol 51, no 1,January1998

-Ar ~Ade
¢

GAS GAS L "

:.:.* ___t

RISER RISER-

DOWNCOMER
~'~OWNCOMER
(o) (b) (c)

Fig 14. Gas-liquid separation in external-loop airlift reactors: the various geometric forms of the head region (Chisti and Moo-Young,
1993a).
liquid separators are designed for complete (or near
complete) disengagement of gas, so that the downcomer is
gas-free. In general, effective gas-liquid separation in
airlift bioreactors is feasible only for fluids having a
relatively low viscosity such as water-like fluids. The low
rise velocities of small bubbles in viscous media make
practical separation impossible. Observations by Akita et
al. (1994) being a case in point.
velocities and the cross-sectional areas in these regions
obey tile continuity relationship:
ULr Ar = ULa Ad = U L . xW, (89)
where x and W are the depth and tile width, respectively, of
tile head region (Figure 14b). For gas to completely
disengage, the bubbles from the riser must rise through

5.1Separatordesign height x before the fluid travels the horizontal distance l

(Verlaan et al., 1986), that is
Practical designs of gas-liquid separators for airlift
bioreactors employ one of two principles: Either the x < m
l
distance between the exit of the riser and the entrance of the U b -- U L H ' (90)
downcomer in the head region is increased sufficiently that
the gas bubbles have enough time to rise out of the liquid where Ub is tile bubble rise velocity. But, from equation
before it reenters the downcomer; or, the cross-sectional (89), the superficial liquid velocity in the gas-separating
area of the entrance of the downcomer is enlarged so much head region (UzH) is
so that the downcomer superficial liquid velocity is reduced
to less than the rise velocity of the bubbles. Specific design
procedures are given in the following sections. ULr Zr
ULH
xW (91)
5.1.1 External-loop reactors
Some commonly used separator configurations for external-
Equation (91) taken together with inequality (90) leads to
loop reactors are shown in Figure 14 (Chisti and Moo-
the design relationship
Young, 1993a). Of these, configuration (a) is the least
attractive for several reasons: tile horizontal pipe section
between the riser and the downcomer yields a higher
1 IW
pressure drop than file other configurations shown; the -- _< - - . (92)
separated gas slugs must flow along with the turbulent Ub ULr Ar
liquid, with accompanying redispersion causing relatively
poor gas-liquid separation; and, the limited capability of the Hence, tile rise velocity of tile smallest bubble to be
configuration for being cleaned-in-place for processes that excluded from the downcomer should be such that
require aseptic operation. Because of these severe inequality (92) is satisfied at tile maximum anticipated
limitations, configuration (a) is not recommended in liquid circulation velocity (ULtra). Notice that the ungassed
fermentation applications. Here the focus is on design liquid height (x) in the head region has no effect on gas-
aspects of configurations (b) and (c) in Figure 14. liquid separation performance (Verlaan et al., 1986). As a
Because all the liquid exiting the riser flows through matter of interest, if a circular connecting pipe of diameter
the head region into the downcomer, the superficial liquid Dc (as in Figure 14a) is used, the design criterion works out

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
Cb)
Co) C~)
Fig 15. Gas-liquid separators for internal-loop airlift reactors: (a) split--cylinder; (b) draft-tube sparged internal-loop; (c) annulus-
sparged internal-loop; and (d)-(f) use of baffles for gas-liquid separation.
to (Chisti and Moo-Young, 1993a)
1 ln'Dc
_<
Ub 4ULrmAr'
where.l is the length of the connecting pipe.
Occasionally, because of space restrictions, it may not
•be feasible to provide an adequate channel length. In such
cases, a smaller length may be used to disengage most of the
larger gas bubbles and the cross-sectional area of the
downcomer entrance may be enlarged (Figure 14c) to
reduce the liquid velocity to below the rise velocity of the
smallest bubble that is to be excluded from the downcomer.
A low estimate of bubble rise velocity in the downcomer
entrance region leads to a safer design (Chisti and Moo-
Young, 1993a).
5.1.2 Internal-loop reactors
Some gas-liquid separator designs for draft-tube and split-
cylinder airlift reactors are shown in Figure 15 (Chisti and
Moo-Young, 1993a). Configurations (a)-(c) are based on
the principle of reducing the downward liquid velocity in
the entrance of the downcomer to less than the rise velocity
of the bubbles to be kept out of the downcomer. The
reduction in velocity is achieved by expanding tim cross-
sectional area of the downcomer in the •head zone of the
reactor, taking care to minimize the boundary layer
separation and the formation of stagnant regions. Thus, for
the maximum anticipated liquid circulation velocity in the
riser (UL~), the expanded area of the downcomer is given
by (Chisti and Moo-Young, 1993a)
ArULrm
z4de -- - - J
U zde
where ULde,the downward liquid velocity in the expanded
region, should be less than the bubble rise velocity, Ub.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 57
1 BAF~E ~ _ RISER
Cd)
• E
(e) (0
Separator designs (d)-(f) in Figure 15 make use of
horizontal or vertical baffles to increase the distance over
which gas disengagement takes place. Baffled constructions
such as (d) and (e) have been used in wastewater treatment
(93)
applications (Opie, 1993; Redman, 1987), but are not
suitable for bioreactors that must be run sterile. Baffles
such as (d) and (e) introduce undesired dead spots into the
system, generally lead to higher pressure drop than the
other configurations mentioned, and cause problems with
in-place cleaning of reactors (Chisti, 1992a, b). Baffled
configuration (f~ is satisfactory for hygienic process
applications. Other types of gas-liquid separators for
specialized non-hygienic uses in internal-loop reactors have
been discussed by Choi et al. (1995a, b).
6. SUSPENSION OF SOLIDS
The need to fully suspend various types of solids is
commonly seen in gas-liquid-solid three-phase airlift
devices. Information on the behavior of suspensions and
the gas and liquid flow required to attain the fully
suspended state is essential for design and operation. Many
studies have shown that airlift reactors are better at
suspending solids than bubble columns (Gaspillo and Goto,
1991; Goto et al., 1989; Immich and Onken, 1992). The
gas velocity requirement to achieve suspension increases
with increasing loading of solids (Gaspillo and Goto, 1991;
Goto et al., 1989; Koide et al., 1984b; Mao et al., 1992).
The solids distribution in airlift devices generally tends to
be uniform (Choi et al., 1996; Wenge et al., 1995). Thus,
for air-water-glass bead (dr = 122-308 ~m) systems in a
draft-tube sparged concentric tube internal-loop vessel
(aspect ratio ~ 8; A / A d = 0.56), Immich and Onken (1992)
observed homogeneous distribution of solids. Operation of
(94) the reactor with the draft-tube removed to give a bubble
column led to non-uniform distribution with higher
concentrations closer to the bottom (Immich and Onken,
1992). The three-phase solids holdup did not exceed about
 58 SPECIALISSUE: FLUID MECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech .Rev vol 51, no 1, January 1998

15% in these studies. The minimum gas velocity necessary multicomponent solids (different densities and/or particle
to completely suspend the solids increased with increasing diameters), provided that the sedimenting solids are
diameter and concentration of the beads (Immich and assumed to settle in a slurry of other than the settling solids
Onken, 1992). This behavior was observed in both reactors; (Koide et aL, 1986). In such cases, the FL and PL terms in
however, as shown in Figure 16 which is based on the data equation (95) should be replaced with lad and PsL,
of Immich and Onken (1992), the minimum gas flow rate respectively, where
for suspension in the airlift vessel tended to be substantially
lower than in the bubble column (Immich and Onken,
1992). A method for estimation of the minimum gas flow
was presented (Immich and Onken, 1992).
Using slurries of glass and bronze spheres suspended
in various newtonian media in draft-tube sparged airlift
vessels, Koide et al. (1984b) studied the minimum (or
PSL= ~L( 1+ 1
Cs
31 -1
052
'
(96)

critical) superficial gas velocity (Uo¢, based on cross- and

sectional area of outer tube) required for complete
suspension. The U~¢ was roughly proportional to U/~,
where Ut was the terminal settling velocity of the particles
:,, (97)
(Koide et aL, 1984b). The Uo~ increased with increasing
concentration of solids according to Uoc oc Cs°27. An aspect In equation (97) Ps, is the average density of non-settling
ratio of about 12 produced a minimum in Uo~. The data fraction of solids in the slurry and Cs is the volume fraction
were correlated with the equation (Koide et al., 1984b) of non-settling solids in the gas-free slurry (Koide et al.,
1986). Note that equation (95) disregards the effects of the
Uoc = 4 6 0 I C s 1 ° 2 7 3 ( p s - pLIO'750I UtILlL1-0'634 bottom clearance of the draft-tube on Uoc although such
effects are expected to be significant.
u, .,.7J ,. J CTF) A simplified form of equation (95) of Koide et al.
(1984b) was successfully used by Petrovic et al. (1993) in
• 2 -0.340 0.546 0.454
predicting Ucc-values for complete suspension of particles
in the circulating bed regime in draft-tube sparged
concentric tube airlift vessels. The observations spanned
( ( 4 ~ 0.2~ the ranges: dp = 1--6 mm; concentration of solids = 10--70
X 1+ / kg'm3; ps = 2508-2631 kg.m3; PL = 998-1113 kg.m3; P-L=
(1.01-3.43) x 10.3 Pa.s; CL = 0.0568--0.0728 N.m'~; aspect
ratio ~ 11; andA/Aa= 0.2-1.6.

x + 1.47 x 10 -4

I f A ~0"997(d~gp'/°'z)°'172~ PARTICLE DIAMETER(p.~n',

× 1-1.32/1--"--"[ /, • , ~ . ~ e ~ 8
0.08
\ .do) ) !
• ~ CBUBBLNE
OLU
(95)

/ i
w

0.06
where Ld was the height of draft-tube (varied over 0.70-
2.10 m). Equation (95) agreed with the data to within
L~ m

±13% for 142 measurements (Koide et aL, 1984b)~ The o 0.04

equation was established for the following ranges: 8.55 _< bJ
>
Cs/ps -< 0.i60; 1.12 _< (ps - PL)/PL -< 7.80; 1.31 x 10.4 _< (/1 /
U,~tL/~z -< 9.67 x 10-4; 1.36 x 103 _<do2gpzJcL _< 1.22 x 104; (.9
/ .
0.02
1.68 x 10-n _<g~L4/PL~L3 _< 1.62 X 10"6; 0.18 _< d/do -< 0.5; • j r BUBBLECOLUMN
~E
4.99 x 10"4 ~ Ut/(gLd)°'5 _< 2.10 x 10"2; 4.67 _< L/do <_ 15; Z , , ,
and 0.222 _<A,I(A, +Ad) -<0.552. The specific properties of 5
solids and liquids, and the reactor geometries used were: ps 0
0 2 4 6 8 10 12 14 16 18
= 2500, 4680 and 8770 kg.m3; dp = 79--498 pan; SOLIDS CONCENTRATION (VOL.%)
concentration, Cs = 25---400 kg.m3; PL = 997--1178 kg.m'3;
I-tL = (0.894--15.2) X 10.3 Pa.s; PL = (51.7--72.0) x 10"3
N.m'~;A/Ad = 0.3-1.2; and aspect ratio = 6-15. Fig 16. Minimum superficial gas velocity for complete
Evidence suggests that equation (95) may be used to suspension of glass beads. Based on the data of Immich and
predict Uoc for different particle fractions in suspensions of Onken (1992).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 59

A hysteresis phenomena in which a certain minimum derivation is flawed in that in concentrated slurries the wall
gas velocity is required to bring solids into suspension, but a friction is not negligible as equation (98) assumes. Note
lower gas flow rate is sufficient to keep them in suspension, that the equation (51) was derived for water-like fluids
has been widely noted in three-phase airlift reactors (Chisti, (Chisti et aL, 1988); it had to be modified for application to
1989, p. 52; Douek and Livingston, 1994; Heck and Onken, more viscous fluids (Chisti, 1989, pp. 221-229; Chisfi and
1986; 1987; 1988; Petrovic et aL, 1989; Pogarac and Moo-Young, 1988a). However, for some three-phase
Petrovic, 1988). Below a certain superficial gas velocity, systems, particularly relatively dilute suspensions of solids
the three-phase airlift reactors with non-neutrally buoyant in water, equation (51), developed for gas-liquid systems,
• solids may stall (that is, cease to circulate). Increasing can satisfactorily predict the liquid circulation velocity as
solids loading decreases liquid circulation rate (Douek and demonstrated by Ganzeveld et al. (1995).
Livingston, 1994; Ganzeveld et aL, 1995; Heijnen et aL, Based on equation (98), a gas-liquid-solid airlift
1990) and increases the value of gas velocity at which reactor would stall when Pd = Pr (1 - ~ ) . Note that there
stalling takes place (Douek and Livingston, 1994; Heijnen would be no gas in file downcomer in the stalled state. In
et aL, 1990). This type of flow instability has been reported theory, a gas-liquid reactor could also stall if the gas
in quite large reactors. Thus, in a 0.35 m3 airlift reactor holdups in the riser and the downcomer become equal;
similar to the one shown in Figure 7, Heijnen et aL (1990) however, such a condition would be unstable, some gas
noted that below certain low aeration rates, the circulation would escape from the downcomer and the reactor would
was unstable in solids containing systems and the reactor recover.
stalled. Increasing amount of solids raised the value of gas In developing equation (98), Livingston and Zhang
flow rate at which the instability occurred. At solids (1993) did not experimentally prove whether unequal
concentration of 40 kg.m3 the circulation ceased at a distributions of solids between the riser and the downcomer
superficial gas velocity of 0.006 m-s"~ whereas, at a did in fact exist. For solids up to 2.5 times as dense as the
concentration of 150 kg.m3, stalling was seen at 0.015 m.s~ suspending liquid and 70-150 ~m in diameter,
(Heijnen et aL, 1990). Consideration of tiffs phenomenon is measurements in a large split-cylinder airlift device
essential for selection of acceptable minimum aeration rates revealed no appreciable difference in solids concentrations
in three-phase airlift devices. in tile riser and the downcomer (Wenge et al., 1995).
The foregoing studies notwithstanding, the concept of Similarly, Immieh and Onken (1992) and Choi et al. (1996)
a critical gas velocity for suspension of solids in airlift noted homogeneous solids distribution in airlift vessels.
reactors is not particularly useful because for the same solid However, for large high-density solids, the particles may
in different reactors, U~¢-value may be very different. A distribute unevenly. Hence, in suspensions of relatively
better approach is to specify the critical liquid velocity for heavy basalt particles in air-water in a draft-tube sparged
suspension based on the sedimentation characteristics of the internal-loop reactor (A/Aa = 0.77; total volume -- 0.186
solids and then to calculate the gas flow rate required to m3; Ps = 2760 kg.m3, dp = 0.245 mm; the input solids
achieve the critical liquid velocity. loading = 27-134 kg.mS), Cai and Nieuwstad (1992)
In a recent study, Livingston and Zhang (1993) correlated the distribution with the equation
surmised that in gas-liquid-solid systems in airlift reactors,
when the solids are denser than the liquid, the concentration #sr 6 (ULd + Ut~( Ad~,
of the solids would be different in the riser and the Csa -- 0.9 8~- ~ _ - ~-) \-~-j (99)
downcomer. The authors used the approach developed by
Chisti et aL (1988) to arrive at the following equation for
liquid circulation velocity where the solids volume fractions in the riser (¢Sr) and the
downcomer (¢sa) are for the gas-free system. In equation
e Lr (99), ULr and U~ are the superficial velocities of the liquid
in the riser and the downcomer respectively, and Ut is the
0.5
terminal settling velocity of the solids. Equation (99) was
determined for a single type of solid and a single value of
I 2ghD[Pd(l_Sod)_Pr(l_~o~) ] the Aav/Arratio.
PrKr + K ( Ar')2 1 6.1 Dead zones and flow instabilities
Studies by Cai et al. (1992a) highlight the importance of
(98) eliminating dead zones near tile bottoms of airlift reactors if
sedimentation of solids is to be minimized. Cai et al.
where p, and Pd are the densities of the solid-liquid slurry (1992a) studied suspensions of relatively heavy basalt
in the riser and the downcomer, respectively, while es, and particles (Ps = 2760 kg.m3, dp = 0.245 mm) in air-water in
eSd are the respective solids holdups in those zones. a draft-tube sparged internal-loop airlift reactor (A/Ad =
Equation (98) requires a knowledge of holdups of all phases 0.77; total volume = 0.186 m3); the input solids loading
in the riser and downcomer. The equation has not been ranged over 27-134 kg.ma. The conical trough that
adequately demonstrated by experimental evidence and its constituted the bottom of the reactor did create a significant

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 80 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
dead zone. The clearance of the draft-tube from the bottom
varied over 25-90 ram, but the actual clearance for the flow
of the slurry was lower because the solids settled and
reduced the free area for flow. As the gas injection rate was
decreased from a high initial value (2.5 m3.h~), the induced
superficial liquid circulation velocity in the downcomer
decreased until a critical value, Utac, below which the
circulation could not be sustained because the solids blocked
the zone below the draft-tube. The critical velocity was a
function of the available volume (Vat) below the draft-tube,
the packed volume (Vs,) of the settled solids and the vertical
clearance (HD between the draft-tube and the bottom of the
reactor; thus,
u ,o
H,g = O.
o ('v,,h - 0.036.
Equation (100) which is dimensionless, is expected to be
specific to the types of solids and the reactor geometry
employed by Cai et aL (1992a).
Other work has shown that poorly designed
recirculation paths can lead to flow instabilities especially in
three-phase systems (Douek and Livingston, 1994). In
unstable flow, the circulation velocity oscillates significantly
even though the gas flow rate remains constant. The
oscillations arise when the solids settle in constricted zones,
the cross-sections change, and slugs of liquid suddenly
bring solids into suspension. Other possible factors may
play a role. Flow oscillations in gas-liquid two-phase
reactors have also been reported (Chisti, 1989) but they are
a natural characteristic of internal-loop airlift reactors in
which substantial amounts of gas bubbles are carried into
• the downcomer with the downflow of liquid.
7. MIXING
7.1 Liquid and slurry phases
Mixing in airlift reactors has generally been analyzed using
the axial dispersion model which considers mixing as being
analogous to a diffusion process (Chisti, 1989, pp. 27-29;
80-83); thus, for a tracer added to the reactor,
~3Cr 1 02Cr OCt
c3A------~--Pe .Opz Op '
where Cr is the dimensionless tracer concentration, Xt is the
dimensionless time, p is the dimensionless distance, and Pe
is the Peclet number that characterizes the mixing. The
dimensionless variables are defined thus:
C , . &, = t J " and Pe = VL°L--------L',
where Ct is the tracer concentrationat time t, j is the total
distance traveled, Ct~ois the final equilibrium concentration
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
of tracer, Lt is the distance between the tracer measuring
points, tc is the circulation time, VLo is the linear velocity of
the liquid and EL is the axial dispersion coefficient. In
theory, fully backmixed behavior obtains when EL is infinite
and plug flow occurs when EL is zero (Chisti, 1989, pp. 27-
28). In practice, fully backmixed state is approached for
Peclet numbers less than 0.1, and plug flow behavior is
found for Peclet numbers greater than 20 (Pavlica and
Olson, 1970). If mixing behavior is represented by a series
of nr ideally mixed tanks, the axial dispersion and the tank-
in-series models may be interrelated as follows (Blenke,
1979)
Pe
nr ~ 1 + w (103)
(100)
2
Equation (103) applies for Pe > 8 (Blenke, 1979).
Lu et al. (1994) used equation (101) to analyze the
mixing behavior in a draft-tube sparged concentric tube
airlift reactor (AJAd = 1; aspect ratio of 7.8; Uar = 0.015-
0.15 m.s-I) in air-water with and without suspended solids
(polystyrene pellets or calcium alginate beads, both 2.4 mm
equivalent diameter, up to 30% vol loading).
Measurements of tracer response at tile entrances and exits
of the riser and the downcomer ensured that the dispersion
coefficients could be evaluated separately for the riser, the
downcomer, the head zone, and the bottom region. In
general, for the 5% vol polystyrene pellet system, Peclet
numbers were 20-30 in the riser, 40-70 in the downcomer,
about 10 in the top section, and 10-20 in the bottom zone
(Lu et al., 1994). Thus, the top and the bottom zones were
the best mixed, the riser was less so, and the downcomer
was tile least mixed. Both tile riser and tile downcolner had
t
distinct plug flow characteristics as the Peclet numbers were
greater than 20. This general trend in relative mixing
performance of the various zones is consistent with other
well known observations (Chisti, 1989, p. 81).
Lu et al. (1994) noted clear reduction in the overall
dispersion coefficients when the height of the draft-tube
was increased by about 54%. This behavioL too, is
generally expected because increasing aspect ratio, or
height, should reduce the contribution of tile relatively
better mixed top and bottom zones to the overall mixing in
the reactor (Ganzeveld et aL, 1995).
The dispersion coefficients increased linearly with
(lOl) increasing gas flow rate; however, the coefficients in the
riser increased much more rapidly than in the downcomer
(Lu et aL, 1994). Data presented by the authors suggested
that the mixing behavior may depend on the type of solid.
Thus, at the same loading and particle size, the overall
Peclet numbers for the hydrophobic polystyrene beads were
consistently lower than for the hydrophilic calcium alginate
solids (Lu et aL, 1994). The hydrophobic solids tended to
attach to the rising air bubbles, which apparently aided
mixing. For a given gas flow rate, the overall axial
(102)
dispersion coefficients declined with increasing solids
loading until about 15% vol loading; further increase in the
loading to 30% vol increased the overall dispersion
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 61

coefficients (Lu et al., 1994). This behavior was explained
(Lu et al., 1994) as arising from two counteracting effects:
increased solids loading enhanced bead-to--bead collisions
which contributed to mixing, but the gas holdup declined
with increasing solids concentration, which probably
reduced the transport of liquid in the bubble wakes.
The following correlations were empirically obtained
(Lu et al., 1993) for axial dispersion coefficients:
( L ~ -1.241
Er~r,,n = 38713 I ×
• k. hLJ
I 853+35208Ws-3387×lO6W2s(PsfL-PL)I
1+
(104)
with f-values of-1•893 for polystyrene and -2•917 for the
alginate beads, respectively;
zone for gas-liquid separation, Fr6hlich et al. (199!a) noted
that the overall liquid-phase Peclet numbers declined with
increasing gas flow rate (Uor ~ 0.06--0.17 m.s'l), except in
the antifoam containing solution in which the numbers were
constant at ca 70. The induced liquid circulation rates were
comparable for all fluids. On the other hand, the liquid
phase mixing times in the antifoam containing medium
were generally lower than in the other two fluids•
Circulation rates for the liquid phase were larger than for
the gas phase. Although no explanation was given, the
results confirmed that for different fluids in a given airlift
vessel, at a constant liquid circulation rate, the mixing times
may be different if the bubble size and/or gas holdup are
different• This difference would exist even if the fluids had
' identical density and viscosity. These differences arise
because there are two distinct contributions to mixing:
firstly, liquid circulation velocity and consequent
turbulence; and, secondly, the mean relative velocity
between the gas and the liquid. The latter depends on
bubble size and gas holdup (see Sections 3 and 4.1).

( Ld'~ -0.217 Extensive data compiled by Petrovic et al. (1995)

ELr = 5 9 6 4 2 U ~ f f 5| - - ~ - ] confirm that correlations for mixing time in airlift reactors
" k, hL) must consider the effects of geometry and the fluid
(105) properties; correlations based on the superficial gas velocity

x 1+1000--
Ws(p--p31"
-- / ,
as the sole variable do not apply to reactors other than those
used in obtaining the correlations• Using air-water data
WL PL J from several sources, Petrovic et al. (1995) empirically
obtained the following equations:
with f-values of 0.018 for polystyrene and -0.517 for the
alginate beads, respectively; and f h ~o.6o
__ 509uor
-0.31 r v°l'vo 3v; °", (107)
/ ' L '~ -0.093 k . d r v/
T T0"495 | d]
ELd=408•72t"& ~,-~LJ x
and
F]-~ '~ 0.12

[ 1+ 1~0-~
1'
" '

(106)
tat =
5. 2 t~/-r_0.31 [ ' ~ r [
-'.-'~-,o~ \doJ
vO.19vO.50V_0.26
r d • •

Equations (107) and (108) applied to external-loop and

(108)

with f-values of -2.657 for polystyrene and -4•117 for the
alginate beads, respectively. In equations (104)-(106) Ld is
the length of the draft-tube, and Ws and WL are,
respectively, the weights of the solids and the liquid in the
reactor. These equations applied to the ranges: 0.007 _< Uor
_<0.15 m.s'l; 0 _< Ws <- 10.55 kg; WLof about 13-18 kg; and
0.69 _<Ld/hz ~ 0.92. In the absence of solids, the 9s term in
these equations is replaced with the density of the liquid (Lu
et aL, 1993). The empirical nature of equations (104)-
(106) and their base on limited data in a specific type of
reactor is bound to restrict their usefulness. Mixing time
data in slurries of calcium alginate beads in draft-tube
sparged reactors were reported also by Vicente and Teixeira
(1995)•
In mixing studies in a draft-tube sparged concentric
tube airlift reactor (A/Ad = 1.86; aspect ratio = 11; 4 m3;
fluids: tap water, 1% vol/vol aqueous ethanol solution, tap
water containing an antifoam agent) with an expanded head
draft-tube internal-loop reactors respectively. In these
equations, Vr, Fd and Fs are the volumes of the riser, the
downcomer and the gas-liquid separator, respectively; hr is
the height of the riser (or draft-tube) and do is the diameter
of the outer vessel in the internal-loop units. For the latter,
the entire head region above the draft-tube was considered
as the gas-liquid separator (Petrovic et al., 1995), The
mixing time (tM) was for 95% homogeneity. Equations
(107) and (108) lack a fundamental basis and they have
other drawbacks. Significantly, the effects of the geometric
configuration of the top and the bottom zone are not taken
into account, but these geometries are known to have a
major impact on the rate of induced liquid circulation.
Comparison of the mixing time data obtained by Ganzeveld
et al. (1995) in two split-cylinder vessels with equation
(108) is shown.in Figure 17. The data were obtained in 0.1
M sodium chloride. The two reactors had identical unit
values of the A/Ad ratio, but the aspect ratios were different
at 7.6 and 14.5. The U& varied over (0-6.7) x 103 m.s1.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 62 SPECIALISSUE:FLUID MECHANICS PROBLEMS IN BIOTECHNOLOGY
10~ - RAT.__..~._O
ASPECT
. • 7.6
o 14.5 +~o~.-
W s~
65 ~s s¢
s o s~ J
45 ss~ J~
25 %'/
.,, $4
5
5 25 45 -65
t M (MEASURED) (s)
Fig 17. Comparison of mixing time data obtained in two split-
cylinder airlift reactors (Ganzeveld et al., 1995) with predictions
of equation (108) of Petrovic et al. (1995). Data are for aqueous
salt solution (0.1 M sodium chloride). The reactors had identical
A,/Ad-values at 1.0, but the aspect ratios were different.
The design of the head region in external-loop airlift
reactors has been shown to affect the mixing performance
(Choi et al., 1995a, b). In internal-loop devices without
gas-liquid separators per se, the clearance B (Figure 18)
between the static liquid level and the top of the draft-tube,
or baffle, affects mixing time (Chisti, 1989, pp. 146-147).
Weiland (1984) first demonstrated that increase in the
clearance B improved mixing up to a B-value of 0.5 m.
Similar results were confirmed by Chang et al. (1993), Lu
et al. (1993), Russell et al. (1994), and Sukan and Vardar-
Sukan (1987). Chang et al. (1993) presented mixing data
in terms of the overall axial dispersion coefficients which
were shown to increase as the clearance B increased over
the range 0.1-0.5 m. Increasing B much beyond 0.5 m
worsens the mixing performance as documented for a large
split-cylinder airlift device (Chisti, 1989, p. 146). For B-
values greater than about 0.5 m, the intermixing between
the fluid in the unbaffied upper zone and the main airlift
circulation loop deteriorates because the circulatory flow
can no longer sweep through the entire head zone. Instead,
most of the flow from the riser just flows over the edge of
the baffle, or the draft-tube, into the downcomer (Chisti,
1989, pp. 146-147). Similar behavior was noted in a large
draft-tube sparged reactor (Russell et al., 1994).
In draft-tube sparged concentric tube airlift reactors
(A/Ad = 0.8; height of draft-tube varied over 1.2-3.2 m;
height of draft-tube to vessel diameter ratios of 3.8-10.1;
baker's yeast broth), for any given gas flow rate, the mixing
time was observed to increase with increasing height of the
draft-tube (Russell et al., 1994). This is qualitatively
consistent with other data (Ganzeveld et al., 1995; Kawase
and Moo-Young, 1989; Petrovic et al., 1995), but the exact
magnitude of the effect has not been consistent. Depending
on the type of reactor, the mixing time may increase with
height according to tM oc (heighty where the exponent may
range over 0.12-2.
In mixing time (100% homogeneity) studies in air-
water in draft-tube sparged concentric tube airlift devices,
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev voi 51, no 1, January 1998
-/// ~ o
//
j .
s ~
ss
-~B
85 105
Y AIR
Fig 18. The upper clearance B abOve the baffle or draft tube in
an internal-loop airlift bioreactor.
Sukan and Vardar-Sukan (1987) varied the clearance of the
draft-tube from the base of the reactor so that the ratio of
the bottom clearance to the width of the annulus was 1.8,
1.0 or 0.5. A value of unity was most effective for mixing
(Sukan and Vardar-Sukan, 1987). Note however that a
clearance ratio of one is not the geometry for minimum flow
resistance as the cross-sectional area of the annulus
(downcomer) would always be greater than the area for flow
under the draft-tube. Shorter draft-tubes produced
somewhat lower mixing times than taller tubes. For
otherwise identical conditions, increasing height of liquid
above the draft-tube reduced mixing time until a minimum
value was reached. Further increase in liquid level
extended the mixing time (Sukan and Vardar-Sukan, 1987).
This behavior is general to internal-loop airlift reactors as
noted earlier. The geometric details of the reactors were:
A / A a = 0.37 or 0.78; working aspect ratios z 5--6; draft-
tube heights = 0.25 or 0.27 m (Sukan and Vardar-Sukan,
1987).
Data on overall axial dispersion coefficients in air-
water in a relatively large external-loop airlift reactor
(0.165 m3; aspect ratio = 16 based on riser; A / A a = 4) were
reported by Verlaan et aL (1989a, c). Apparently no gas
was carried into the downcomer because of efficient gas-
liquid separation in the head zonel A butterfly valve in the
bottom pipe between the riser and the downcomer was used
to vary the flow resistance of the circulation loop (Verlaan
et aL, 1989a). The overall axial dispersion coefficient (EL)
increased with increasing superficial gas velocity in the
riser, but the EL-values were not particularly sensitive to the
 AppI.Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 63

extent of opening of the butterfly valve. The mixing time operating volume to 18 L for comparison with a stirred tank
(95% homogeneity) and tile circulation time were correlated reactor of that size. The stirred vessel was agitated at 200
as follows (Verlaan et al., 1989a, c): rpm without aeration. The airlift device was aerated at a
riser superficial gas velocity of 2.9 x 10.2 m.s"l. Under these
tM conditions, the mixing times in the airlift and the stirred
= O.093Pe, (109)
tc device were 10.3 s and 44 seconds, respectively (Cheng et
aL, 1992). The time course of penicillin conversion was a
where Pe is the overall Peclet number defined as UiJ-,/Ez. little faster in the airlift reactor; however, at 150 W.m3 the
Here, L is the length of the circulation loop, EL is the overall energy input in the airlift system was only about a third that
axial dispersion coefficient for the liquid, and Uz is the in the stirred tank.
mean circulation velocity. Mixing times (90% homogeneity) in draft-tube
Peclet numbers in individual zones of the reactors sparged concentric tube airlift reactors operated with air-
were also computed (Verlaan et al., 1989c). In the water-glass beads systems were reported by Petrovic et al.
downcomer and the head zone, the Pc-values were virtually (1990). The relevant details were: aspect ratio ~ 11; A/Ad
independent of the gas flow rate (Uo,) which varied over 0.2-1.6; UG= 0.004-0.06 m.sa; dp= 1, 3 and 6 mm; weight
0.01-0.14 m.s-a. Pc-values of about 40 were noted in the fraction of solids = 4, 6 and 8%; and density of solids =
downcomer and approximately 10 in the head region 2550 kg.m"3. In the vessel with anA/Ad value of 0.36, with
(Verlaan et aL, 1989c). In the riser, the Pc-value did the solids settled, increasing aeration rate such that the
depend somewhat on the superficial gas velocity; the Pc- packed bed of solids remained undisturbed, caused a decline
values varied over 20-35. The overall Peclet number for in mixing time. From the point of incipient fluidization to
the whole circulation loop ranged over 40-60, increasing incipient circulation of solids, the mixing times increased.
with gas flow rate (Verlaan et aL, 1989c). Further increase in gas flow rate reduced the mixing time.
Residence time distribution behavior in a small The position of the maximum in the tM versus Uo plots
external-loop airlift reactor during culture of a highly apparently depended on the fluidization properties of the
flocculating strain of Saccharomyces cerevisiae was beads; the lnaxima shifted toward higher gas flow rates as
reported by Roca et al. (1995). The relevant details of the the particle size increased. While the solids did influence
reactor were: A/Ad = 2.6; aspect ratio = 7.4 based on the the mixing behavior relative to solids-free operation for all
riser; and 2 L overall volume. The residence time reactor-solids combinations, no systematic general relations
distribution was determined using lactose tracer. The were identified among mixing time, solids loading, particle
measured distribution curve fitted a model curve for a size and the A/Ad ratio.
system consisting of two unequal well mixed stirred tanks Mixing behavior in animal cell microcarrier
with fluid exchange between them. For the small reactor suspensions in split-cylinder airlift vessels has been
used, the circulating loop behaved as a well-mixed zone reported by Ganzeveld et al. (1995). The mixing time
whereas the head region behaved as a second, smaller well- variation with aeration rate followed the pattern shown in
mixed region (Roca et al., 1995). In contrast, for most Figure 19 that is typical for such reactors. Increasing
large scale airlift reactors, while the head region and the loading of the carriers reduced the mixing performance.
circulating loop region remain distinct mixing zones, the The properties of the carriers and the reactors were: Ps =
circulating loop does not approximate to a well-mixed tank. 1030-1050 kg.m'a; 150-300 ~m carrier diameter; 0-30
Other reports confirm that the liquid circulation velocity kg.m"3 solids loading; A/Ad = 1; aspect ratio = 7.6 and 14.5;
and hence the mixing time are strongly affected by the and Uor = (0-6.7) × 10-3 m.s"1 (Ganzeveld et aL, 1995).
design of the head region (Choi et al., 1995a, b). Mixing in shallow cylindrical tanks (1.6 m3, 1.26 m
In draft-tube sparged airlift reactors (A/Ad = 0.8; diameter, 1.30 m height; 4.1 m3, 2.0 m diameter, 1.6 m
aspect ratio = 6.7-8.9) Cheng et al. (1992) noted that height) having multiple, small-bore (0.025 m maximum
mixing times declined with increasing gas flow rates and diameter) draft-tubes, sparged with air, has been detailed by
aspect ratios. Addition of polystyrene pellets (2.7 x 3 x 3 TriUeros and Recio (1990). Such tanks may be of use in
mm; Ps = 1020 kg.m"3) in air-water over the solids loading biological wastewater treatment, although the geometry
range of 10-30% (vol/vol) increased the mixing time, but would have to be significantly modified as in the 'biotower'
further addition of solids to 50% (vol/vol) improved mixing. devices (see Section 12.2.1) being marketed by Bayer,
This effect depended on the magnitude of the aeration rate, Krupp Uhde and Lurgi (Fouhy, 1992).
being most pronounced at lower aeration rates (Cheng et
aL, 1992). The plastic pellets were used to simulate 7.1.1 Local effects
immobilized enzyme particles as was done also by Mao et Based on observations of local mixing effects on scales of a
aL (1992). In separate experiments, Cheng et al. (1992) centimeter in the riser of an airlift reactor (air-water with 1
used the airlift reactor with a reduced aspect ratio of about kg.m"3. Sodium chloride), Liibbert and Larson (1990)
2.2 to study the hydrolysis of penicillin with pellets of concluded that the local level mixing mechanism in bubble
immobilized penicillin acylase. The acid produced was columns and airlift reactors were identical; however, at
neutralized with sodium hydroxide (4 M) to control pH equal specific power input, the local level mixing was
between 7.0 and 7.5. The lower aspect ratio reduced the stronger in bubble columns. This implied that in

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 64 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rov vo151, no 1, January 1998

liquid velocity. As a result, the relative velocities of the two

SPLIT-CYLINDER AIRLIFT (Ar/A d = 1.0) phases are different along the tube radius with a minimum
AIR-0.1M NoCI-GLASS BEADS relative velocity in the center (Schmidt et aL, 1992).
(ps= 1030 kg.m -3 ; diameter = 150-210#m)
Published evidence (Schmidt et al., 1992) suggests that the

,,°t
120

IO0
contribution of bubble wake mechanism to local mixing is
relatively small closer to the walls of the riser because of the
low gas holdup in that area. Closer to the center of the

I.-
80 O• ASPECT
RATIO BEADS
CONCENTRATION
riser, there seems to be somewhat lower contribution of
wakes to local mixing because this is the region of least
difference between the velocities of bubbles and liquid.
6o 125 Z 1.2 Non-newtonian media
X
- -------.---.L.__ o.o A mechanistic equation for axial dispersion coefficient in
power-law fluids in risers and downcomers of airlift
z~ 40 t reactors was developed by Li et al. (1993). The equation
20 was
• • 0.9
0
0 1
I
2
I
3
I
4
I
5
I
6
I
7
Re Ar ~(]-.)/O+.),
UGr X 10 3 (m's -1)
Sc= * rr (A,+ Aa)) (110)

Fig 19. The typical variation of mixing time with superficial gas where tile Schmidt number, tile Reynolds number and the
velocity in airlift reactors. The data are for two split-cylinders Froude number were defined as
(A/Ad = 1.0; aspect ratios of 7.6 and 14.5) in air-salt solution
• with and without microcarriers (0.1 M NaCI; ps = 1030 kg.m'3;
150-210 l.tmdiameter). Based on Ganzeveld et aL (1995).
SC=pLEL ~ ' (III)

comparison with bubble columns, the deterministic flow p ..inTT2_ n

velocity was enhanced in airlift reactors at tile expense of Re = L~o c, Gr (112)

random mixing motions (L0bbert and Larson, 1990).
Mixing in radial direction in the riser was observed to be of
diffusional type; however, in the axial direction, diffusional
mixing was supplemented with convective mixing due to
the wakes of the rising bubbles (Liibbert and Larson, 1990).
•Even at a relatively high superficial gas velocity of 0.108
m.sl, mixing deviated from that expected of local isotropic
turbulence. By comparison, axial and radial mixing in
single phase flow in the same flow geometry as the riser, at
identical Reynolds number (= 60,000, based on liquid
velocity) as in the two-phase system, was dominated by
turbulent dispersion (Liibbert and Larson, 1990). Thus,
Liibbert and Larson (1990) concluded that the
Kolmogoroff's isotropic turbulence theory cannot generally
be assumed to apply in multiphase flow of the type that
exists in bubble columns and airlift reactors. Assertion of
earlier work (Chisti et al., 1988; Chisti, 1989, pp. 300-301)
that bubble wakes are the primary contributor to energy
dissipation in the riser were supported (Liibbert and Larson,
1990).
The contribution of bubble wake associated mixing to
mixing in any region of the reactor should depend on the
prevailing gas holdup in that region (Schmidt et aL, 1992).
Within the risers of airlift reactors radial profiles of liquid
velocity and gas holdup exist. Gas holdup is generally
higher closer to the center of the riser, but the bubble rise
velocity is also higher. In the bubble flow regime,
particularly when the liquid velocity is low, the radial
variation in bubble velocity is not as pronounced as that in
K
and
Fr= (113)
dog"
where do is tile diameter of the outer tube. Data in power-
law fluids (K = 0.286-11.5 Pa.s"; n = 0.441-0.617) in a
draft-tube sparged concentric tube airlift vessel (A/Aa =
0.618; aspect ratio z 25) having an expanded head region
for gas-liquid separation produced C-values of 0.0697 and
0.0593, respectively, for the riser and the downcomer (Li et
aL, 1993). Equation (110) applied with an average error of
about 27% over a riser superficial gas velocity range of
0.01-0.1 m.s l. For any given air flow rate, the value of the
axial dispersion coefficient increased with increasing
viscosity of the liquid (Li et al., 1993). Slugging in the
liquid explained this effect which became more pronounced
as the gas flow rate increased.
For estimation of mixing time in bubble columns,
Kawase and Moo-Young (1989) obtained the equation
6.3 3 F h~
IM = 220+,)/3 n ~|13 j4/3 T rl13' (114)
tJo L/G
where tile parameter F depended on the location of the
tracer monitoring point and on the specified degree of
mixing holnogeneity (Kawase and Moo-Young, 1989). For

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
limited data in newtonian media, equation (114) did give
reasonably good predictions, within ±20% of measured data
(Kawase and Moo-Young, 1989; Kawase et a L , 1994);
however, in non-newtonian fluids the equation performed
poorly (Kawase and Moo-Young, 1989).
7.2 Gas phase mixing
Studies of gas phase residence time distribution in a large
external-loop airlift reactor (A/Ad = 5.13; volume = 5 m3;
nominal height = 24 m) in water and baker's yeast culture
medium revealed equivalent behavior in the two fluids
(ROller et aL, 1994). Plug flow of gas was confirmed for
the riser and the downcomer. Compared to the riser, most
gas bubbles in the downcomer were smaller in keeping with
other observations (Chisti, 1989, pp. 132-150; Llibbert et
aL, 1990). The small bubbles in the downcomer contained
little oxygen (Riiffer et al., 1994); however, because of the
short liquid residence time in the downcomer, the
productivity of the yeast fermentation was not significantly
affected by cycling through the oxygen limited zone.
Studies clearly show that from the time of injection at
the base of the riser, a part of the gas phase passes up the
reactor without recirculation (Liibbert et al., 1990). When
no gas-liquid separators are used, or separators are
ineffective, a portion of the gas recirculates through the
downcomer. The recirculated gas contributes to
backmixing in the gas phase. In exceptional cases more
than 20% of the gas may recirculate CLiibbertet aL, 1990).
In air-water, results suggest that most of the recirculating
gas bubbles do so only once; a very small portion may
recirculate a second time (Liibbert et al., 1990). Although
no data are available in other media, if visual observations
are any guide, the mean age of recirculating bubbles is
expected to depend on the mediuln. For example, in viscous
media, extremely tiny bubble persist for long periods and
may recirculate many times.
Observations of two different characteristic bubble
sizes have led to the use of a 'two-bubble-class'
hydrodynamic model for interpretation of gas-phase mixing
in air-water system in bubble columns (Shetty et al., 1992).
The model assumes a bimodal distribution of bubble sizes
with large bubbles flowing in a plug-flow manner and small
bubbles partially backmixed (Shetty et al., 1992).
In a draft-tube sparged concentric tube airlift reactor
(A/Ad = 1.86; aspect ratio = 11; 4 m3) with an expanded
head zone for gas liquid separation, Fr6hlich et al. (1991a)
noted that typically, the mean gas-phase residence time in
the downcomer was longer than in the riser which is
consistent with visual observations (Chisti, 1989). Gas-
phase Peclet numbers in the riser were 4- to 25-fold larger
than in the downcomer (Fr6hlich et al., 1991a). Peclet
numbers in tim downcomer were not particularly sensitive
to the gas flow rate (Uor = 0.06-0.17 m.sa), but those in the
riser declined with increasing aeration rate. Over most of
the range of gas flows, the riser gas-phase Peclet numbers
in tap water with antifoam were significantly lower than in
other media (tap water; aqueous ethanol solution, 1%
vol/vol ethanol).
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 65
7.3 Mixing of solids
Axial distribution of biomass and other solids in a small
draft-tube sparged, concentric tube airlift vessel (5 L; A / A a
= 0.96; aspect ratio = 4.95) has been examined (Assa and
Bar, 1991). Over a riser superficial air velocity range
(0.19-2.81) x 10-2 m.s-~, the solids in a baker's yeast
suspension (46 kg.m-3 dry weigh0 were uniformly
distributed in the vessel (Assa and Bar, 1991). This
observation is quite general for yeasts, non-filamentous
bacteria, unclumped animal cells and hybridomas. For
plant cells which are usually larger (60-100 ~m diameter)
than microbial and animal cells, and tend to clump, non-
homogeneous distribution of biomass may occur at certain
combinations of biomass concentrations and aeration rates.
Thus, Assa and Bar (1991) noted non-uniform distribution
of clumped plant cells (595-707 gm clump size) in
suspension culture of Phaseolus vulgaris when the biomass
concentration was 12.5 kg.m"3 (dry weight) and aeration
rates were lower, than 0.95 v v m (UGr= 0.018 m@). The
local biomass concentration increased down the
downcomer, but the concentration in the riser was not
particularly sensitive to axial position. With reduced
amount of solids in suspension (for example, 8.6 kg.m"3 dry
biomass), the axial distribution was more homogeneous.
Measurements in suspensions of calcium carbonate
(particle diameter _<37 gin; 55 kg.m"3 solids) indicated
incomplete suspension of solids below certain aeration rates
(Assa and Bar, 1991). Axial solids concentration profiles
were observed in the riser and the downcomer even at
aeration rates as high as 1.45 vvm (U~r = 0.028 m.sl);
however, the difference in concentrations between the top
and the bottom did not exceed a few percent of the mean
value (Assa and Bar, 1991). While non-homogeneous
distribution of solids has been hypothesized (Livingston and
Zhang, 1993), measurements have often revealed uniform
dispersion (Choi et al., 1996; Immich and Onken, 1992;
Wenge et al., 1995). Nevertheless, as noted by Assa and
Bar (1991), axial variations in concentration of suspended
solids can occur during certain fermentations in
pneumatically agitated reactors. This behavior is most
likely for large particles, low aeration rates, systems having
a relatively large difference between the densities of the
solid and the liquid, and low viscosity liquids.
While axial variations in concentration may have no
process implications, local point sampling in these systems
would not provide an estimate of the amount of solids in
suspension. This problem can be overcome by measuring
the mean solids concentration between two widely spaced
axial locations in the riser and the downcomer using the
method developed by Wenge et al. (1995). This method
does not require withdrawal of samples and, hence,
overcomes the errors associated with multiphase sampling
even in suspensions that have a homogeneous distribution
of solids (Wenge et aL, 1995).
Residence time of solids measured by using magnetic
tracer particles in aqueous slurries of non-magnetic solids in
a draft-tube sparged concentric tube airlift reactor (A/Ad =
0.6; aspect ratio = 8.6; U~ = 0.01-0.07 m.s~) was reported
 66 SPECIAL ISSUE: FLUID MECHANICS PROBLEMSIN BIOTECHNOLOGY
by Kikuchi et al. (1995). Polystyrene (Ps = 1020 kg.m3; dp
= 0.90 mm) or glass (Ps = 2500 kg.m3; d p = 1.88 mm)
beads were used up to a solids volume fraction of 0.3. The
residence times in various zones (riser, downcomer, head
region, bottom) were unaffected by the average solids
concentration for the lighter beads. For polystyrene beads,
the concentrations (calculated from residence time data) in
the riser and the downcomer were virtually unaffected by
the gas flow rate. For the heavier beads, too, the gas
velocity had little effect on distribution of solids, but
concentration in the riser was somewhat higher than in the
downcomer (Kikuchi et aL, 1995).
A procedure for estimation of transport of particles in
bubble wakes in airlift reactors has been presented by Snape
and Thomas (1992). The significance of the procedure is
limited because in most fermentations the cells tend to be
fairly homogeneously distributed in the broth. Furthermore,
the procedure requires additional experimental
substantiation.
8. TRANSPORT PHENOMENA
8.1 Gas-liquid mass transfer
The rate of transfer of a sparingly soluble gas such as
oxygen from the gas phase to a liquid is expressed as
d C L _ kLaL(C*-- CL), (115)
dt
where CL is the dissolved gas concentration at time t, C* is
the liquid phase saturation concentration of the transferring
species in equilibrium with the gas, and kLaL is the overall
volumetric gas-liquid mass transfer coefficient. The
coefficient kLaL and the concentration driving force (C* -
CD for mass transfer must be known if the rate of supply of
oxygen or other soluble gas is to be quantified. The mass
transfer driving force, which is easily determined, depends
on temperature and pressure. The coefficient kzaz depends
on the properties of the fluid, the hydrodynamic regime and
the configuration of the bioreactor. Prediction ofkzaL, or its
equivalent kLaD, is an important part of bioreactor design.
In the overall volumetric mass transfer coefficient, the ~
term is the true mass transfer coefficient and az is the gas-
liquid interfacial area per unit liquid volume. The specific
interfacial area may also be expressed in terms of the
volume of dispersion. The latter is represented as aD.
Because kLaL (or kLaD) is relatively easily measured directly,
it is the parameter that is commonly reported instead of the
separate variables kL and aL (or aD). Moreover, the
available separate measurements do not usually apply to the
systems of interest. The two representations of the overall
volumetric mass transfer coefficient are related, thus
kLaD
kLaL = 1--S~. (116)
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
. Appl Mech Rev vol 51, no 1, January 1998
A more complete treatment of gas-liquid mass transfer has
been presented elsewhere (Chisti, 1989), and only a few
more recent major studies are reviewed below. A general
approach to predicting kLaz is discussed in Section 8.1.4.
8.1.1 Newtonian media
For annulus sparged concentric-tube airlift vessels, Koide et
al. (1983a) developed the equation:
g "~0.500f 3'~0.136
kL a D G L [ ill L ~I [PLo'L|
= 225/-- -I /--z-i
DLgPL " ~.PLDL] ~, gktz )
(117)
( , .y0 0 o,
Equation (117) applied over the experimental ranges: 3.71
x 102 _< gdpLDL -< 6.00 X 104; 1.18 X 106 _< pzc%3/gP.L4
5.93 × 10t0; 0.471 -< d/do -<0.743; 7.14 x 10"3 _<5~do _<2.86
x 102; 0.0302 _<~ _<0.305; aspect ratio = 6-15; and Aa/Ar
= 0.52-1.23. The average error in estimating the kLaD with
equation (117) was 12% for 175 measurements (Koide et
al., 1983a),
A different equation applied to draft-tube sparged
internal-loop vessels (Koide et aL, 1985):
d2 / ,,0.500,"~__: 10.715
(118)
[gdoPz.] ~,~34
Equations (118) spanned the ranges: PL = 997=1182 kg.m'3;
~L = (0.894--17.0) x 10.3 Pa.s; aL = (51.7-73.0) x 10.3
N.m'l; DL = (0.145-2.42) x 10.9 m2.st; A/Ad = 0.3-0.8; and
aspect ratio = 6-15.
During culture of baker's yeast in a draft-tube sparged
concentric tube internal-loop airlift reactor (A/Ad = 0.8;
A~/Ad = 0.9; aspect ratio = 10; 0.25 m3), Russell et al.
(1995) noted that the steady state dissolved oxygen
concentration in the riser and the downcomer increased
with increasing aeration rate. The dissolved oxygen levels
were always several fold higher in the downcomer than in
the riser (Russell et al., 1995). In the riser, the dissolved
oxygen levels tended to be _<30% of air saturation. The
oxygen concentration did not vary radially either in the riser
or in the downcomer. Similarly, no significant variations
were observed axially in the downcomer (Russell et al.,
1995). Data were presented for the volumetric oxygen
transfer coefficient (Russell et al., 1995), but some of tile
values were unrealistically high. Contrary to expectations,
lneasurements in different parts of the reactor gave identical
kLaL values at low aeration rates (poorer mixing), but widely
divergent values at high superficial gas velocities. These
puzzling results suggest an unreliable measurement method
and other possible errors.
 Appl Mech Rev vol 51, no 1, January 1998
Using the annulus-sparged mode of the same vessel as
was employedby Russell et al. (1995), Pollard et al. (1994)
found that the axial dissolved oxygen profiles in the riser
showed a constant zero level of oxygen over a broad range
of superficial gas velocities (0.036-0.136 m.s~) during
baker's yeast fermentations. Surprisingly, the dissolved
oxygen concentration increased down the downcomer,
peaking close the bottom of the draft-tube, and declining
sharply to zero as the fluid moved into the riser. These
anomalous results seemed to indicate better oxygen transfer
in the downcomer relative to the riser (Pollard et al., 1994).
In view of the observed profile, the cells experienced
extreme dissolved oxygen heterogeneity in the reactor;
however, there seemed to be no effect on the culture
performance. Installation and operation of a marine
propeller (diameter = 0.16 m; 1000 rpm) near the bottom o f
the draft-tube, combined with a high level of aeration
(0.358 m3.min~), enhanced liquid circulation and oxygen
transfer so much so that the dissolved oxygen level did not
drop below 35% of saturation anywhere in the reactor.
Based on data obtained in several concentric draft-
tube airlift reactors that were sparged either in the draft-
tube or the annulus (de/do = 0.5-0.7; aspect ratios = 6-11;
UG = 0.019-0.12 m.s]), Koide et al. (1984c) suggested that
in air-water the contribution of the downcomer zone to gas-
liquid mass transfer was "relatively large." Furthermore, in
liquids such as aqueous salt solutions in which the bubbles
are smaller, the riser and the downcomer were claimed to
contribute to the total mass transfer in proportion to the
fraction of tile total volume that those zones occupied
(Koide et al., 1984c). These observations in relatively small
.reactors are not entirely consistent with other findings
(Bello, 1981; Chisti, 1989; Fraser et al., 1994).
Using measurements in a small external-loop airlift
vessel operated such that no air entrained into the
downcomer, Fraser et al. (1994) concluded that the
apparent measured gas-liquid volumetric mass transfer
coefficient for the reactor was lower than the actual value in
the riser where mass transfer took place. This conclusion is
obvious from the analytical expression (Chisti, 1989, pp.
258-259.)
~4r(kL~,lL)r q- Ad(~LaL)d
k L ~7/L)measur~ --
A~ + Ad (119)
which relates the measured kzaz to the separate,
hypothetical, contributions of the riser and the downcomer.
Because Fraser et al. (1994) had no gas in the downcomer,
(kLaz)d was zero; hence, the actual volumetric mass transfer
coefficient in the riser was
( Ar+ A~)
(k LaL) r = (k LaL)measured A~
020)
With an A j A r ratio of about 0.43 (Fraser et al., 1994), the
(kLaz)~ was always about 43% greater than the value
measured for the reactor as a whole. Equation (120) leads
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 67
to tile general conclusion that for any airlift device in which
gas-liquid mass transfer is confined to the riser, the actual
volumetric mass transfer coefficient in the riser would be
larger than the apparent measured value; the difference
being dependent on the A/Ad ratio. Equations (119) and
(120) apply to reactors in which the riser and the
downcomer have approximatelY equal heights. In cases in
which a significant portion of the liquid resides in the head
zone, these equations should be modified to account for that
volume.
Data on gas-liquid mass transfer in air-water in a
relatively large external-loop airlift reactor (0.165 m3;
aspect ratio = 16 based on riser; A/Aa = 4) were reported by
Verlaan et aL (1989a). Apparently no gas was carried into
the downcomer because of efficient gas-liquid separation in
the head zone. The flow resistance of the circulation loop
was varied using a butterfly valve located in the bottom pipe
between the riser and the downcomer (Verlaan et aL,
1989a). The volumetric oxygen transfer coefficient was
correlated as follows (Verlaan et al., 1989a):
/,. -, 0.83
kLaLL IUOr)
- 45t -y Pe-°°,
where the Pe-value was for the entire circulation loop (see
Section 7.1).
A draft-tube sparged airlift device for oxygenation and
suspension of small volumes of live aerobic cultures
undergoing measurements in an NMR spectrometer was
described by Kramer and Bailey (1991). The observed
oxygen transfer capability of tile device was sufficient to
support moderate to high cell density cultures of E. coil
(Kramer and Bailey, 1991).
In comparative studies of bubble columns and small
internal-loop airlift devices without separators, similar
values of the gas-liquid volumetric mass transfer
coefficients have generally been reported. This behavior
was recently reconfirmed by Goto et al. (1989).
8.1.2 Non-newtoman liquids
In aqueous solutions of carboxymethyl cellulose (K = 0.286-
11.5 Pa.s"; n = 0.441--0.617) in a draft-tube sparged
internal-loop airlift reactor (0.055 m3; A/Aa = 0.618; aspect
ratio - 26), Li et al. (1995) correlated the gas-liquid mass
transfer coefficient as follows:
1.-2 r ro5:4 -0.255
k L a z = 3.43 X ,,, ~'~r flop " (122)
The apparent viscosity in equation (122) was calculated by
assuming the shear rate to equal 5000.U~r for U~r 2 0.04
m.s"1, or 1000.(U~) °5 for U~r < 0.04 m.s -1. The apparent
viscosity ranged over 0.020--0.85 Pa.s. The exponent on the
viscosity term was substantially greater than -0.66 or -0.89
reported by others (see Chisti, 1989, pp. 74-79). Such
discrepancies have in part been associated with the
questionable practice of defining shear rate as a quite
 68 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vo151, no 1, January 1998

arbitrary linear function of the gas velocity (Chisti and
Moo-Young, 1989; Chisti, 1989, pp. 74-79).
Suh et al. (1992) studied production of the
extracellular polysaccharide xanthan during batch
fermentations of the obligate aerobic bacterium
Xanthomonas campestris. Production was compared in a
bubble column (0.05 m3; aspect ratio z 19) and a draft-tube
sParged concentric tube airlift bioreactor (1.2 m3; A/Ad = 1;
aspect ratio = 12). The gas-liquid volumetric oxygen
transfer coefficient (kLaL) declined with increasing
concentration of xanthan in both reactors as shown in
Figure 20. The kzaL values in the bubble column were about
40% higher than in the airlift reactor; but the latter was
aerated at a somewhat lower specific Ix~wer input of 0.9
kW.m3 compared to 1 kW:m 3 in the bubble column (Suh et
al., 1992). In both reactors, the oxygen transfer data
correlated with the equation (Suh et al., 1991)
~L " 2 ( 10"5( 2 1 0.20
aLdo Pop gdoPL
DL - 0018 • )
~ X 0.62 / x 0.5 l
gd3op~| f U~ )
× -T - - -
compared to tile basic annulus-sparged airlift configuration
(A/Ad = 0.8; aspect ratio of about 10; 0.25 m3). For similar
highly viscous broths of the microfungus N. sitophila in
annulus-sparged draft-tube reactors (1.3 m3; A/Aa = 3.7;
aspect ratio of about 3), Moo-Young et aL (1993) had
earlier established that the airlift configuration
supplemented with a low-shear axial flow impeller in the
draft-tube was better than the pure airlift device. Moreover,
such airlift-impeller hybrid reactors were shown to be better
than the more conventional, Rushton turbine stirred
fermenters (Moo-Young et aL, 1993; Moo-Young and
Chisti, 1988).
10-I
o BUBBLE COLUMN
AIRLIFT
i
to
" ~ 10-2
(1 + 0.12Wi) -~, o
_J

(123) D

where Wi was the Weissenberg number (ratio of the first

norlnal stress and the shear stress in viscoelastic fluids).
The apparent viscosity in equation (123) was calculated by
1 0 -3 I I I I
assuming the shear rate to equal 2800.Uo as recommended 0 5 10 15 20
by Schumpe and Deckwer (1987). The arbitrariness of this XANTHAN (kg.m -3)
approach has already been pointed out.
Effect of the aeration rate on production of xanthan in
the bubble column is shown in Figure 21 reproduced from Fig 20. Variation of the volumetric gas-liquid oxygen transfer
Suh et aL (1991). The figure clearly reveals the critical coefficient with concentration ofxanthan gum in a bubble column
importance of oxygen supply for the highly aerobic xanthan and an airlift fermenter (Suh et al., 1992). The specific power
fermentation. In the airlift reactor, the productivity was inputs were 1.0 and 0.9 kW.m3, respectively, for the bubble
significantly lower than in bubble column. This was traced colunm and the airlift vessel.
to cycling of the oxygen depleted broth through the
unaerated downcomer (Suh et al., 1991). The airlift
fermenter used had not been designed for a highly viscous
and aerobic fermentation. As a result, the unaerated 4O
downcomer constituted about half the volume of the vessel.

,¢¢¢:
Ideally, the A/Ad ratio should have been much larger than u~(m.s-~) ^ ~.,j:>..~.~_.o~
30
unity. In viscous aerobic fermentations the performance of E 0.361 0.165 o..,,.~ - 51
airlift reactors can be significantly enhanced by installation
of an axial flow impeller in the downcomer to improve the z 20
circulation of the fluid. This approach has been 112
I--
Z
demonstrated with fermentations of the mycelial
X 10
micro~ngus Neurospora sitophila by Moo-Young et al.
(1994). It has also been found useful with less viscous yeast
broths (Russell et al., 1995). 0 I I I I I
In highly viscous, non-newtonian, broths of 0 40 " 80 120 160 200 240
TIME ( h o u r s )
Saccharopolyspora erythraea, the use of a marine propeller
• (diameter = 0.16 m; 1000 rpm) located near the bottom of
the draft-tube has been reported (Pollard et al., 1994) to Fig 21. Effect of the superficial gas velocity on production of
enhance the yield of the antibiotic erythromycin by ca 45% xanthan byX. campestris in a bubble colunm (Sub et al., 1992).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
8.1.3 Slurries
For suspensions of glass or bronze spheres in draft-tube
sparged internal-loop vessels, Koide et al. (1985) reported
the equation:
/ -,0.500: d2 x 0.715
kLaDdZo = 2 6 6 [ 'ut I [ ga;pt']
DL " \PL DLJ \ aL )
( --3 _2"~0"251 #__ x -0.429
IgaopL. I Jar. I 6~34
) 5-oJ
I r~ N0.069f • 3"~0"023/ . . N0046~ -1
0099[Csl [&all /u'/
× 1+ . ~7sJ ~-~-) k,Uo)
(124)
The average error in estimating ~ZaD by equation (124) was
within 17% for 383 measurelnents (Koide et al., 1985).
The equation applied over the following experimental
ranges: 3.71 x 102 _< gL/OLDL _N 9.92 x 104; 1.36 x 103 _<
gdo2plJcJL _< 1.22 x 104; 1.29 x 108 _<gdoSpm2/gg2 _< 1.26 x
10n; 0.471 _< d/do -< 0.743; 3.99 x 10.2 _< ~ _< 2.73 x 10"1;
1.69 x 10"11 _< gP.L4/pLUL 3 _< 2.55 X 10"6; 0 _< Cs/Ps -< 8.00 x
10"2; 1.17 x 10.2 _< GB.J~ -< 0.844; A,Z4d = 013-0.8; and
aspect ratio = 6-15 (Koide et al., 1985).
A draft-tube sparged concentric draft-tube airlift
reactor (A,/Ad = 1.3; aspect ratio ~ 5; U~,. = (0.154-1.39) x
10~ m.s ~) for potential application to microbial
desulfurization of coal by the bacterium Sulfolobus
acidocaldarius was studied by Slnith and Skidlnore (1990).
The mass transfer coefficient was measured in a simulated
basal salt medium (composition in distilled water, kg.m-3:
(NH4)2SO4, 1.3; KH2PO4, 0.28; MgSO4-2H:O, 0.25;
CaClv2H20, 0.07; acidified to pH 2 by sulfuric acid)
containing pulverized coal (74 x 10.6 m particle diameter;
density = 1415 kg.m 3) at 0--40% wt (equivalent to 0-0.315
volmne fraction of solids in gas-free medium).
Measurements were made at 30, 50 and 72°C (Smith and
Skidmore, 1990). The kLag-values were strongly enhanced
by increasing temperature (Smith and Skidmore, 1990),
almost certainly because the molecular diffusivity of oxygen
must have increased. Increasing concentration of solids
over 0-5% wt (equivalent to qbs of 0-0.035) increased the
kcac-values slightly, but further increase in solids loading
strongly lowered the mass transfer coefficient (Slnith and
Skidmore, 1990). This phenomena was explained as being
due to penetration of fine solid particles in the stagnant
liquid film around gas-bubbles, reducing the effective
diffusion length (Smith and Skidmore; 1990). However, as
the concentration of solids increased beyond 5% wt, the
solids caused a blocking effect which reduced the effective
mass transfer area (Smith and Skidmore, 1990). In the
region of enhanced mass transfer (0 _<qbs < 0.035), the kLac
was correlated with (Smith and Skidlnore, 1990)
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 69
kt aL = (5.7 - 144.8 ¢ s + 5048.3 #Zs)
(125)
X I v~4t-./GrT
T1-55e -2990/T ,
whereas in the 'blocking' region (0.035 _< dps ~ 0.315), the
following equation was obtained (Smith and Skidmore,
1990)
+ 2
ktat=(2.6-16.8~s 29.4~s) x ,iJu,-,6u~r
r rl 45 -4130/T
e .
(126)
Equations (125) and (126) agreed with the data to within
/ about ±20%.
) "
Performance data for a 0.35 m 3 airlift reactor of the
kind shown in Figure 7 were provided by Heijnen et al.
(1990). Over a superficial air velocity range of 0.02-0.06
m.s-1 (based on outer tube), and a sand concentration o f 250
kg.m s (~8% vol/vol), the volumetric oxygen mass transfer
coefficient correlated linearly with the superficial gas
velocity (Heijnen et al., 1990); thus
kLa = 0.639~U~. (127)
Equation (127) was determined at 35°C.
Measurements in suspensions of relatively low density
calcium alginate beads in draft-tube sparged internal-loop
airlift reactors were reported by Koide et al. (1992). The
relevant systems details were: 0-20% vol solids; 1.88-3.98
mm bead diameter; A/Aa = 0.3-1.2; and aspect ratios of 6-
16. The beads were suspended in water, aqueous glycerol
and aqueous salt solutions (0.10 or 0.27 kmol.ms barium
chloride; 0.4 kmol.m3 sodium sulfate). The viscosity of the
liquids ranged over (0.894-12.5) x 10.3 Pa.s and the
molecular diffusivity of oxygen varied over (0.194-2.42) ×
10-9 m2.s1 . The mass transfer coefficient declined with
increasing concentration of solids, but was not affected by
the size of the particles (Koide et aL, 1992). The values in
water and salt solutions were similar. The data were
correlated with the equation (Koide et al., 1992)
kLaDd2o 4.04 I ,u z 10"5I g d o2P L 1 067
DL - 1 + 2 ~ - - - - - 5 k ~ - ~ ) \ Z -)
(128)
f __3__2"~0"26/__ \-0.047
Iga;pL| [a, I ebs4,
where qbs is the volume fraction of solids in the gas-free
system and the superficial gas velocity (Up) is based on the
diameter do of the outer column. Equation (128) covered
the ranges: 3.71 x 102 _< gL[DLpL _<5.52 x 104; 2.66 x 103 _<
pLdo2g/OL _< 1.22 X 104; 2.35 x 108 _< pL2doSg/l.tL2 _< 3.29 ×
1011; 0.471 _<d/do -<0.743; 1.69 x 10"11 _<Mo _<6.67 x 10"7;
3.79 x 102 _< ~ _< 2.24 x 10"1; and 0 _< qbs -< 0.2 (Koide et
 70 SPECIALISSUE:F L U I D MECHANICS PROBLEMS IN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

aL, 1992). Equation (128) estimated the mass transfer
coefficient within an average error of 14% for 260 data
(Koide et al., 1992).
Mass transfer measurements in suspensions of agar-
filled soft polyurethane foam particles (dp = 3 mm; Pt, =
1030 kg.m 3) in uninoculated penicillin culture medium (I.tz
= 1.9 x 10~ Pa.s) in draft-tube sparged concentric tube
airlift reactors havebeen reported (Karamanev et aL, 1992).
Increasing volume fraction of solids over the 0-0. I range,
enhanced the volumetric mass transfer coefficient by 15-
20% over solids-free operation (Karamanev et aL, 1992).
Further increase in solids loading to 40% (vol/vol) caused a
decline in kzaL values. The mass transfer coefficient values
in airlift reactors were up to 3-fold higher than in
comparable fluidized beds. Surprisingly, for a range of
aeration rates, the maximum values of kzaL were observed at
an Ad/A~ value of 3 (Karamanev et aL, 1992). Gas holdup
declined with increasing solids loading up to 20% (vol/vol)
solids, but was little affected by further increases loadings
up to 40% (vol/vol).
8.1.4A general approach to predicting the mass transfer
coefficient
Purely analytical reasoning has been used to show that the
volumetric mass transfer coefficient (kLaL), gas holdup (~),
the Sauter mean bubble diameter (dB) and the true mass
transfer coefficient (kD are related as follows (Chisti, 1989,
pp. 251-265)
.k.: =
1. Measure the gas holdup and the kLaL in a small model
reactor with the fluid of interest over the range of gas
flow rates that are of interest.
2. Use the data from step 1 to calculate the constant kc/dn
ratio (equation 129).
3. Measure or calculate the riser gas holdup ( ~ ) in the
full scale reactor. (See Section 3.2.3 for simultaneous
determination of gas holdup and liquid circulation
velocity.)
4. Calculate the kLaz-value for the large reactor:
kL ¢o, A~
kLaL (131)
de (1- 6o~) A~+ Aa"
In equation (131) the gas holdup and the geometric
parameters A~ and Aa are for the large reactor; the
kzJdn-value is from step 2.
For bubble columns, the riser gas holdup in equation
(131) should be replaced with the overall gas holdup.
Furthermore, Aa should be set to zero. The rationale for
equation (131) has been discussed by Chisti (1989, pp. 251-
265).
Figure 23 compares the measured kzaz-values with
prediction of the foregoing procedure for eleven reactors
including bubble columns and external- and internal-loop
airlift devices. The comparison spans water-like media as
well as pseudoplastic slurries of cellulose fiber. In all cases
the predictions agree remarkably well with the
measurements (Figure 23).
• Note that the 'Solka Floc' cellulose fiber slurries in

(129) aqueous salt solutions used in Figures 22 and 23 were

d8 6¢o meant to simulate the fermentation broths of mycelial fungi
and filamentous bacteria (Chisti, 1989; Chisti and Moo-
Calculations of the kbldB ratio from the measured kLaL and
gas holdup have shown that in bubble columns and airlift 10 I
devices this ratio is constant irrespective of the aeration rate i
AIRLIFT
. . . . . . . . . . . . .__g- - ~--a . . . . . . + 4.9 %
(Figure 22). The value OfkL/da depends on the properties of 3~ 6 LLO . . . . . . . - o - - r - - ' - - - ~ . . . . . m - ~ - - -4.9%
%
the fluid. Thus, in slurries of mycelia-like cellulose fiber 4'
. . . . . . . . . . . . . .--i~--o- O-- +8.8%
. . . . . . . . . . . . .0. . . . . .0. . . . w -8.8%
particles, kzJdn is affected by the concentration of solids
(Figure 22). For aqueous salt solution (0.15 M sodium 2
chloride) with or without suspended cellulose fiber, the
kzJdB ratio has been expressed as (Chisti, 1989, pp. 251- BUBBLE COLUMN
265) 8
. . . . . . . ._o_ . . . . . . n_.2.__.__,~_-. + 7.2 %
6 -'0- . . . . . . . . . -0 . . . . . ~ ---- . . . . . 7.2 %
1
--.CI- . . . . . . . . . . . . . . . . . . -a-O---El--~÷ 14-9 %
4
kc S g P2L"3-"ff LD L']I-: e -°a3'c} 030)
dn =5"63 X 10- L ,UL J m
. . . . L ____ __ -- ....

D
.g 14.9 %

~----7----., ~. . . . ~-- + 9.1%

-~ . . . . . . . . . 9J%
where Cs is the concentration of solids (% wt/vol).
Equation (130) applies to airlifts as well as bubble columns. I I I , , rl , , t , i
Chisti's (1989, pp. 251-265) observations of constant t~z/dn 4 6 8 10 20 40 60
ratio in pneumatically agitated reactors have also been UG xlO z ( r n . s -~)

confirmed by Bakker et aL (1995).

The observed constancy of kL/dn has been used to
devise a method for prediction of ~ZaL in large scale
bioreactors from data easily measured in a small scale
model unit (Chisti, 1989, pp. 251-265). This proven
technique consists of the following steps for airlift reactors:
Fig 22. The mass transfer coefficient-to--bubble diameter ratio
versus superficial gas velocity in two pneumatically agitated
reactors. Fluids: ( • ) 0.15 M sodium chloride solution; slurries of
'Solka Floe' (SF) cellulose fiber in 0.15 M sodium chloride
(concentration of SF, wt/vol: ( o ) 1%, ([]) 2% and (/x) 3%.
From Chisti and Moo-Young(1987a; 1988d).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 71

"
EXTERNAL LOOPS

Ruid
Ad/A r hLOrh~'
( - ) ! (m) ./
/
/

_ . /
/
/

(
lOL BUBBLE COLUMN ."/"

~t~ + 0.11 ./ u O/ 8- h (m~ / ,.T.~/

~'~" ~ W a t e r 0.25 1.8 / O/
X 0 44 / 7 x - 5.875 .~'~x/ .-"
" • 2%SFin ! 0125 ,I / , 6 v - 3.925 ~'~""
I " • 3.15MNaCI 0.44 1.75 /# O// • //
z~ - 1.50 /" ~/" '-_i/=/ -
. ' Y
ooE!_ ,, o / +
,d" /40,. " " v 4 i • .,LI

:6 ' / / " +."~

0.04" / ~ "/ + 1 5 g ~ i,,~ ~ ;" " [

j I ~' x ,~ INTERNAL LOOPS
o </f. . . ;
/- a - ~ s
~,-, 7 " ? i F~- t-)' iA~' - 15g " AIR-WATER
~,,~ , . x o O.ISMN:Clo' 1.5o
002 - .,,I~,/~ ,.ll/ v P,/.EinNaCI 0 5.875 I I I I I I I I I
,1~,~ / o OA3 2 3 4 5 6 7 8 910
I.~'I~ "I • Water 0.35 ' 1.8
,liE,;'" I , I ,1~ , . o.s6 , ( % / v L)xlo -= w=-~.
O005 0,02 0.04 0.06 0'08

kLOL (experimental) (s-l)

Fig 24. Mass transfer coefficient versus specific power input in
bubble columns. No effect of the liquid height (hL) is seen. The
Fig 23. Predicted versus measured overall volumetric gas-liquid solid line follows the equation kcaL = 2.39 x 10.4 (Pc/VL)°86
mass transfer coefficient in bubble colmrms, external-loop and (Chisti, 1989, p. 239; Chisti and Moo-Yotmg, 1987).
internal-loop airlift devices in 'Solka Floe' (SF) cellulose fiber
slurries and solids-free media (Chisti, 1989). The solid line
represents exact prediction. 8.2 Solid-liquid mass transfer
Mass transfer between suspended solid particles and the
Young, 1988d). Such broths are usually pseudoplastic.
suspending fluid may become an important consideration in
Table 1 compares the K- and n-values of cellulose fiber
three-phase bioreactors where the suspended solids (or
slurries with broths of the microfungus Aspergillus niger
liquid droplets) may be a sparingly soluble substrate,
(Chisti, 1989; Chisti and Moo-Young, 1988d). The
particles supporting animal cells, microbial films or
properties of the simulated fluids are quite close to those of
immobilized enzymes. Transport of a soluble nutrient,
the ffungal broths (Table 1). including dissolved oxygen, from tile liquid to the particle
In bubble columns operated with water and similar may be a rate limiting step, or removal of an inhibitory
fluids, the height of liquid does not affect the overall mass metabolite. Solid-liquid mass transfer limitations have
transfer coefficient as shown in Figure 24 (Chisti and Moo- indeed been seen in diverse bioprocesses (Mojovic etaL,
Young, 1987; Chisti, 1989). In aMift reactors, the overall 1994; Park et al., 1994; Schorr-Galindo et al., 1995;
volumetric gas-liquid mass transfer coefficient is sensitive Wagner and Hempel, 1988).
to the overall height because the rate of liquid circulation The rate of mass transfer to or from the suspended
and hence the gas holdup are strongly affected by the height particle depends on the solid-liquid mass transfer
(Chisti, 1989). coefficient (Ks), the total solid-liquid interfacial area (as)
and the concentration driving force; thus,

Table 1. The rheological properties of celMose fiber slurries and VL-

--dCL = K,a,(C'-CD, (132)
Aspergillus niger broths (Chisti, 1989).

Solids Contents n K where CL is the concentration of the transferring component

(kg.m3) (-) (Pa.s") in the liquid at time t, and C* is the saturation concentration
of the transferring component. Equation (132) is written for
2 0.55 0.17 transport of a component from the solid to the suspending
5 (20)* 0.33 (0.32) 1.62 (1.46) liquid. As seen from equation (132), a knowledge of the
transfer coefficient Ks is essential to quantifying the rate of
10 (30) 0.2 (0.24) 9.06 (6.13) mass transfer.
The behavior of the solid-liquid mass transfer
The data in parentheses are for the 'Solka Floe' fiber slurries
in aqueous sodium chloride (0.15 M); the rest are for A. niger coefficient as a fimction of the hydrodynamics has been
broths ofReuss et al. (1982). investigated mostly in bubble columns (Deckwer, 1992);
few studies exist in airlift reactors. Irrespective of the

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 73

8. 2.2 Draft-tube internal-loop reactors In another study, Gaspillo and Goto (1991)
Mass transfer measurements in draft-tube sparged internal- characterized the effects of Kenics-type twisted ribbon static
loop airlift reactors have been reported by several mixers (see Figure 36) on solid-liquid mass transfer in
investigators (Gaspillo and Goto, 1991; Goto et aL, 1989; airlift vessels. Mass transfer was enhanced by about 20%
Kushalkar and Pangarkar, 1 9 9 4 ) . Generally, the relative to operation without the mixers. The a- and c-
observations have been similar to those in external-loop values in equation (136) were 0.133 and 1.68, respectively,
devices (Mao et aL, 1992). Thus, in comparison with when mixers were used in draft-tube sparged internal-loop
bubble columns, concentric draft-tube internal-loop airlift reactors. Similar data were reported in an external-loop
reactors have a better mass transfer performance (Goto et device (Goto and Gaspillo, 1992).
aL, 1989) and the concentration of solids up to 7.7% vol
does not affect the Krvalue (Goto et aL, 1989). The mass 8.3 Heat transfer
transfer coefficient has been correlated with the equation • Heat transfer considerations are essential in thermal
(Goto et al., 1989) sterilization of bioreactors, subsequent cooling and
temperature control during fermentation. Metabolic heat
3 4 a
output of most industrial fermentations is 3-5 kW.m3 and
C( E p L d p ) S c 0.33,
(136) heat transfer does not usually become limiting. Higher heat
loads and more difficult heat transfer problems arise in high
biomass density fermentations and those utilizing highly
where a and c were 0.173 and 1.01, respectively. Equation reduced carbon sources such as hydrocarbons and methanol.
(136) was determined using Amberlyst 15 ion exchange Typically, temperature is controlled by heating or cooling
resin (dp= 0.55-0.92 mm) suspended in dilute aqueous through external jackets and internal coils. Less frequently,
sodium hydroxide. No effect of the Ar/A d ratio (varied over additional double walled baffles, draft-tubes or heat
0.1-1.4) was noticed on Ks. Figure 27 compares equation exchangers located inside the fermentation vessel are
(136) with equation (133) that was developed by Mao et al. needed to provide sufficient heat transfer surface area. The
•(1992) for an external-loop reactor. heat exchange system must remove not only the metabolic
Using a different system, 0.5% wt benzoic acid heat of fermentation but also that generated by pneumatic
granules (alp = 0.55-3 mm) in air-water in draft-tube agitation. All pneumatic energy input may be assumed to
sparged internal-loop airlift reactors (A/Ad = 0.17-1.29; dissipate as heat. Usually the pneumatic agitation/aeration
UG = 0.08-0.35 m.s'l), Kushalkar and Pangarkar (1994) heat load is no more than 25°,6 of the total load. For
obtained the equation temperature control, the rate QH of heat removal must
match the heat generation rate. The reactor wall area An
f 3 4 "x 0165
needed to obtain the required heat transfer is calculated
sco4, (137) using
\ ,UL Y

In agreement with Goto et al. (1989), the solid-liquid mass

transfer coefficient was unaffected by the A/Ad ratio. A H - - QH
-
UnAT' (138)

1992~
1 O0
where AT is the mean temperature difference driving force
for heat transfer and UH is the overall transfer coefficient.
MAO ET AL,~ The latter depends on the resistance to heat flow due to the
metal wall, the fouling films on either sides of the wall, the
film of heating/cooling fluid and the film of the
10 fermentation broth in contact with the heat exchange
I
surface.
Estimation of heat transfer resistances of walls,
fouling films and films of heating and cooling fluids is
discussed in readily available handbooks and heat transfer
literature. Estimation of heat transfer resistance, or its
J inverse the heat transfer coefficient ho, for the film of liquid
10 100 1000 or broth in the fermenter is detailed here. For air-water in
E*/SpL d p 4/5 industrial batch bubble columns (diameters > 0.45 m) the
fihn coefficient is known to increase with the superficial gas
Fig 27. Comparisonof the solid-liquid mass transfer velocity as follows (Fair et al., 1962)
performance of external-loop (Mao et al., 1992) and draft-tube
sparged internal-loop airlift reactors (Goto et aL, 1989). ho = 8,850U6--, (139)

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 74 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
indicating an independence of column diameter. Equation
(139) may be used for water-like media such as some
bacterial fermentation broths or animal cell culture fluids.
A correlation for a greater range of fluids in bubble columns
was developed by Deckwer (1985); it took the form
1
3 2-~
(140)
,PzCpUo-0"l~-~-I kr J ) '
which also predicts an increase in the film coefficient with
gas velocity raised to 0.25. In equation (140), krand Cp are,
respectively, the thermal conductivity and the specific heat
capacity of the liquid in the reactor. For most fermentation
broths the Cp and kr values are close to those of water.
Literature on heat transfer in airlift reactors is sparse.
Equations (139) and (140) may be used for a low estimate in
airlift vessels when the induced liquid circulation rates are
low. Under other conditions, the coefficient in airlift
reactors can be more than two-fold greater than in bubble
column (Onken and Weiland, 1983). As pointed out in an
earlier review, only two correlations for film heat transfer
coefficient in internal-loop airlift reactors are available
(Chisti and Moo-Young, 1987a). Chakravarty et al. (1971)
proposed the equation
-°.'
h°=8'71OU~r ~A--d)" ~--kr--) " (141)
Equation (141) was obtained in newtonian fluids (P.L =
0.78-5.27 mPa.s) in draft-tube sparged airlifts over
approximate gas velocity and A,IAd ranges, respectively.of
0.008-0.16 m.s-1, and 0.25-1.20. The actual values of ho
varied from 600 to 2400 W.mL°C-1. The second equation
was proposed by Ouyoung et aL (1989) for air-water in
concentric draft-tube internal-loops
ho =13,340(l + Ad]-°7U~275. (142)
ArJ
Equation (142) covered a gas velocity (based on the total
reactor cross-section) range of 0.01 to 0.04 m.s-~, whereas
Aa/Ar values of 0.242 and 0.452 were used in its
determination. Equation (142) predicts significantly higher
heat transfer coefficients than does correlation (141),
possibly because of the higher induced liquid circulation
rates in the reactors used by Ouyoung et al. (19.89).
Studies of heat transfer in airlift vessels in suspensions
of Aspergillus niger and Penicillium chtysogenum claimed
that the presence of mycelia enhanced film heat transfer
coefficients (Blakebrough et aL, 1983a, b). The data
presented showed ho-values as high as 8 kW.m2.K1
(Blakebrough et al., 1983a).. A more thorough study
conducted with cellulose fiber suspensions which simulated
mycelial media showed that the dependence of heat transfer
on solid contents can be very complex (Ouyoung et al.,
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Moth Rov vol 51, no 1, January 1998
............ ] ..... I < ,o
12
11~ 13
I f .........
RETURN ))
WASTE
WATER
COOLN
I GR~TURNWA)TER
Fig 28. Temperaturecontrol on a fermenter.
1989). Solids may enhance or reduce heat transfer
depending on the location (riser or downcomer) of
heating/cooling surface in the reactor (Ouyoung et aL,
1989). Useful recommendations on the practical problem of
positioning of heat exchange devices in airlift vessels were
presented by Ouyoung et aL (1989). Nishikawa et al.
(1977) claimed film heat transfer coefficient to be
independent of liquid velocity for UL --< 0.015 m.s'~;
however, for higher liquid velocities ho depended on the
velocity as follows
ho-~ uLrrl/4 (0.015<UL(m.s-1)_<0.139). (143)
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 in the airlift and the vertical two-phase
flow device.
Certain fermentations, particularly animal cell
cultures, can be extremely sensitive to temperature
fluctuations and require good temperature control without
developing hot spots on the heat transfer surfaces.
Typically, the temperature is controlled by circulation of
thermostated water through the jacket of the fermenter.
Because the heat exchange system must accommodate the
high heat loads encountered during sterilization and
subsequent cooling, tile operational details can be quite
complex as shown in Figure 28. Valves 1, 10 and 5 in
Figure 28 are control valves that modulate the flow of steam
and cooling water; valves 2 and 9 are vacuum breakers that
allow the pipework to drain under gravity; valve 3 is a
pressure relief valve to protect the jacket and the pipework
against pressurization above the safe acceptable limit; and
all other valves are pneumatically operated devices that are
either fully open or fully closed.
 Appl Mech Re~/vol 51, no 1, January 1998
For sterilization the jacket is heated by steam. Valves
4--6, 8 and 10 are closed and the pump 7 is off. Valve 11 is
opened to drain the jacket. After a short period, valve 11 is
closed.and valve 12 is opened. Valve 1 is opened to let
steam into the jacket. The condensate drains through the
steam trap 13. The temperature and pressure of the circuit
are monitored at PI1 and TIC1.
Cooling is carried out by closing valves 12 and 1;
valves 8 and 4 are opened. Cooling water enters the circuit
at valve 4, flows up the jacket and out to the cooling water
return line (Figure 28).
Temperature control during fermentation is
maintained by a closed loop recirculation. Valves 8 and 6
are opened and the pump 7 is turned on. The water is
pumped through the plate heat exchanger (PIIE) where it is
indirectly heated by controlled (control valve 10) flow of
steam. The heated water now passes the temperature
indicator/controller TIC1 and enters the jacket. The water
recirculates via valve 8 and pump 7. Cold water is injected
(control valve 5) in the circuit to maintain the temperature.
Any excess water leaves the loop via the relief valve 3. In
some designs, direct steam injection into the circulating
-loop may be used instead of the heat exchanger. In small
fermenters, the exchanger may be replaced by electric
heating. Notice the flexible connections between the
fermenter and the temperature control pipework (Figure
28). These connections are necessary for fermenters that
rest on load cells; the connections allow the fermenter to
move freely.
9. C O N S I D E R A T I O N S F O R BIOFILM R E A C T O R S
Three-phase airlift reactors are increasingly being used in
wastewater treatment. The suspended solid phase (usually
sand, glass beads or carbon particles) supports a biofihn of
waste-degrading microorganisms. A high concentration of
active biomass is maintained and the carriers are easily
separated from the treated water (Heijnen et al., 1992).
Degradation of phenols and nitrogen--containing aromatics
by microbial biofilms supported on sand and activated
carbon particles in an external-loop airlift reactor has been
reported (Koch et al., 1991). Applications in biological
reduction of sulfate (van Houten et al., 1994), degradation
of naphthalene-2-sulfonate (Wagner and Hempel, 1988),
dichloromethane (Dunn, 1994) and other substances have
been described (Htippe et al., 1990; Ulonska et al., 1995).
In such systems hydrodynalnics and transport
considerations play an important role. These considerations
include suspension or fluidization of the solids, adequacy of
shear forces to control the biofilm thickness without
dislodging it altogether, gas-liquid mass transfer and solid-
liquid mass transfer.
In studies of degradation of naphthalene-2-sulfonate
by P s e u d o m o n a d s immobilized as biofilms on sand particles
suspended in an external-loop airlift bioreactor, Wagner
and Hempel (1988) noted that solid-liquid mass transfer at
the biofilm-liquid interface had to be considered in
explaining the experimental degradation rates. Solid-liquid
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 75
interfacial concentrations of oxygen and naphthalene-2-
sulfonate could not be assumed to be the same as in the bulk
liquid. These observations were for particles about 200 gm
in diameter, covered with ~75 gm thick biofilm, suspended
in the reactor under fairly turbulent conditions (230--440
W.m 3 specific power inputs). At solids loadings of 20-30
kg.m"3, only about 15-20% of the particles were covered
with biofilm, but high solids loadings were necessary to
control the film thickness at 75±10 gm by providing a high
attrition environment (Wagner and Hempel, 1988).
Naphthalene-2-sulfonate degradation rates up to 36
kg.m3.d ~ were achieved (Wagner and Hempel, 1988).
An example where gas-liquid mass transfer was
limiting was encountered by van Houten et al. (1994). In
reduction of sulfate with biofilms on particles suspended in
a draft-tube sparged airlift reactor (aspect ratio = 7; A / A d =
0.6), van Houten et al. (1994) observed that mass transfer of
hydrogen controlled the maximum sulfate conversion rate.
Mixtures of l\vdrogen and carbon dioxide were used for
aeration. Hydrogen was the electron donor for the reduction
of sulfate to sulfide, while carbon dioxide was the carbon
source and the pH control agent. Conversion rates of up to
30 kg SO42-.m-3.d-~ were obtained (van Houten et al., 1994).
9.1 The biofilm
In a typical wastewater treatment operation, the thickness of
the aerobic portion of the biofihn tends to be 75-200 p.m
with oxygen up-take rates of the order of 7-15 g O2.m-Ldq
(van Loosdrecht and Heijnen, 1993). Even when oxygen
limitation is not an issue, diffusion of other substrates limits
the useful thickness to less than 400 ~m in most cases (van
Loosdrecht and Heijnen, 1993). Hence, wastewater
treatment with biofilm airlift reactors depends on
developing thin films with large specific surface areas (van
Loosdrecht and Heijnen, 1993). Carrier surface area in the
reactor is usually 2000-3000 m2.m"3 (Tijhuis et al., 1994c).
Oxidation and nitrification capacities of biofilm airlift
reactors have been rated at 20 kg COD.m3.d ~ and 5
kg.m3.d "1, respectively (van Loosdrecht and Heijnen, 1993;
Tijhuis et al., 1994c). Large installations with 300 m 3
reactors are in operation (Tijhuis et al., 1994c).
The development of biofihn on carriers in a
concentric-tube (riser) airlift device (2 L; aspect ratio = 20;
A / A d = 4) shown schematically in Figure 7 was investigated
by Heijnen et aL (1992). Gas-liquid-solids circulated in
the primary airlift zone. Gas separation took place in the
central part of the head region above the draft-tube, and
gas-free slurry flowed over the weir into the annulus of the
head zone (Figure 7). The solids sedimented into the
downcomer while solids-free treated water was withdrawn
from the outer edge of the head region (Figure 7). The
carrier loading was 125 kg-m3, and the superficial gas
velocity was constant at 0.038 m.st. The hydraulic
retention time of wastewater was one or two hours. Based
on tests with 21 carriers, carrier diameter and surface
roughness were the major factors that influenced biofilm
formation (Heijnen et al., 1992). Smaller particles, 0.1-0.3
 76 SPECIALISSUE: FLUID MECHANICS PROBLEMSIN BIOTECHNOLOGY
mm in diameter, tended to show better biofihn development.
This was said to be due to more frequent collisions among
larger particles causing biofilm detachment (Heijnen et al.,
1992); however, with larger particles, a greater severity of
collisions would have been a better explanation. For equal
mass loadings of solids in two systems, the system with
larger particles should have fewer particles per unit volume
and, therefore, a lower frequency of particle-particle
collisions. Cell disruption experiments in agitated slurries
of glass beads confirm this (Garfido et al., 1994). The
explanation for the impact of surface roughness on biofilm
formation lay in the fact that a rough surface provided many
relatively protected, low shear sites where microbial
colonies could take hold in the process of gradually
spreading over the whole surface. Based on the criteria of
surface roughness, particle size, cost and the requirement of
a high density carrier for ease of separation, basalt (ps =
2900 kg.m "3) was the best carrier (Heijnen et aL, 1992).
Evaluation of support matrices for immobilization of
Butyribacterium methylotrophicum on particles suspended
in a gas lift device for fermentation of coal derived synthesis
gas has also been reported (Chatterjee et aL, 1996).
Irrespective of carder type, biofilm formation is
favored at relatively high dilution rates when any detached
cells are rapidly washed out (Heijnen et aL, 1992; Tijhuis et
al., 1994c). This is illustrated in Figure 29 where, for a
waste COD loading of 5 kg-m3.d 1, the influence of the
hydraulic retention time on the relative amounts of biomass
on the carriers and in the liquid is shown. The data in
Figure 29 are for basalt carriers at a carrier solids loading of
175 kg.m3 and a superficial gas velocity of 0.075 m.s~.
Other similar data have been presented (Tijhuis et aL,
1994a). Tijhuis et aL (1994c) have shown that during start-
up at high dilution rates, the concentration of bare carriers
declines as an increasing number are covered with the
biofilm. The total amount of the biomass in the reactor
increases, but the amount per carrier remains constant until
most of the carriers are covered. With few bare carriers
remaining, the extent of abrasive loss of fihn declines and
the film thickness begins to increase until a stable thickness
is reached (Tijhuis et aL, 1994c). These observations
suggest that varying the reactor operation to suite different
stages of film development may be worthwhile.
In other similar work, detachment of biomass from
non-growing biofilms supported on carders suspended in
four configurations of draft-tube sparged internal-loop
airlift reactors was examined (Gjaltema et aL, 1995). The
reactor geometries were: A / A d = 0.35-1.14; Ad/Ab = 1.04-
2.82; aspect ratios = 8-18; volume = 0.09-2.1 L, and Uc,.
0.03-0.09 m.s-~. The detachment of biomass from biofihn-
covered carriers in presence of bare solids was dominated by
collisions between bare carriers and the covered particles
(Gjaltema et aL, 1995). The concentration of bare solids
and the particle rouglmess strongly influenced the
detachment rate (Gjaltema et al., 1995). For otherwise
identical conditions, biofilm detachment due to bare carriers
increased in the order: glass, sand, basalt. The glass
particles were the most round and smooth; the basalt
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vo151, no 1, January 1998
:~ 1 2 0
o
w 96
t~J
t~
b_
~b,1 7 2
<
o 48
121
N 24
-J
O
~_ 0 1 2 ,3 4 5
RETENTION TIME ( h )
Fig 29. Influence of the hydraulic retention time on the relative
amounts of biomass in the earrier-suplx~rted biofilm and in
suspension (Heijnen et aL, 1992).
carriers were tile roughest. While biofilms develop more
easily on rough surfaces (Heijnen et aL, 1992), once
developed, the fihn is readily damaged by biofihn-free
rough carriers. Gjaltelna et aL (1995) observed that a
"change in flow regime from bubbling to slug flow
considerably increased the detachment rate." This change
apparently occurred at a Ucr-value of ,--0.06 m.s~. The
"slug flow" regime noted by Gjaltema et al. (1995) was
more likely the highly chaotic churn turbulent regime in
which particle-to-particle collisions are more frequent and
severe than in the bubble flow regime. The bottom
clearance of the draft-tube had no effect on detachment rate
(Gjaltema et al., 1995), but only a small range of the Ad/Ab
ratios was investigated. The volume fraction of carders was
3.3-3.8% and other properties were: basalt, dp = 0.32 mm,
Ps = 3010 kg.m3; sand, dp = 0.31 mm, Ps = 2606 kg.m-3;
and glass, dp = 0.28 ram, Ps = 2887 kg.m "3 (Gjaltema et aL,
1995). Another similar study focused on the effects of
particle shape, structure and concentration on biofilm
detachment (Gjaltema et al., 1997).
Strategies for development of biofilms with
nitrifcation activity on carriers suspended in draft-tube
sparged internal-loop airlift reactors were investigated by
Tijhuis et al. (1995a) and van Benthum et al. (1996). The
development of the nitrifying flhns was similar to what had
been reported for heterotrophic biofihns (Tijhuis et al.,
1994a-c; 1995b); however, the nitrifying films were more
robust (Tijhuis eta/., 1995a). This suggests that biofllms
formed by different microorganisms can have markedly
different physical properties and strength of attachment to
substrate; hence, the operating regime of the reactor and, in
extreme cases, the design peculiarities are likely to be
affected.
With a cartier loading of 69 kg.m 3, once the nitrifying
biofilm had dally established, the ammonia nitrogen
conversion was 5 kg nitrogen.m3.d1 at a hydraulic
retention time of 1 h in the reactor (Garrido et al., 1997;
Tijhuis et al., 1995a).
 Appl Mech Rev vol 51, no 1, January 1998
In further studies of the dynamics of biofilm
detachment from carriers suspended in airlift reactors,
Tijhuis et aL (1994b; 1995b) noted that the common
assumptions of uniform growth and detachment of biomass
over the surface of the film under fixed hydrodynamic
conditions were inconsistent with observations. Instead,
non-uniform local detachment of biofilm with formation of
cracks and fissures on its surface and rapid filling of the
spaces formed with new growing biomass was proposed as a
more realistic picture of film dynamics (Tijhuis et al.,
1994b).
10. HYDRODYNAMIC SHEAR FORCES
The magnitude of the fluid mechanical forces, often
expressed as shear stress or shear rate, has important
implications in culture of fragile biocatalysts such as animal
and plant cells. While the effects of hydrodynalnic forces
on cells are relatively well known through studies in defined
flow geometries (Doran, 1993; Molina Grima et al., 1997;
Papoutsakis, 1991), little information exists on
quantification of shear fields in bioreactors. Intense shear
fields can disintegrate even the most robust of
microorganisms (Chisfi and Moo-Young, 1986).
Hydrodynamic shear stress may also affect production of
extracellular microbial polysaccharides (Dreveton et al.,
1996; Moo-Young and Chisti, 1988).
Shear rate expressions for bubble columns have been
summarized by Chisti (1989, pp. 74-79), Chisti and Moo-
Young (1989), and Shi et al. (1990). These expression
generally take the form
y = o~Uo, (144)
where the constant tx has ranged over 1500-5000 m l .
Often, and quite arbitrarily, similar expressions have been
applied to airlift reactors (Chisti and Moo-Young, 1989).
In bubble columns, the shear rate numbers produced by the
various correlations are so vastly different as to be
meaningless (Chisti and Moo-Young, 1989). Logically, the
degree of turbulence in a reactor must depend not only on
the power input (or superficial gas velocity) but also on the
momentum transport properties of the fluid. Indeed, it is
well known that the bubble size in a turbulent field depends
on the viscosity and the density of the fluid as well as on the
specific energy input. It is therefore reasonable to assume
thai correlations that express the shear rate as a function of
Uo (or Uor) alone are incomplete (Chisti and Moo-Young,
1989). Furthermore, correlations such as equation (144)
have generally been based on observations of phenomena at
solid-liquid interfaces (for example, heat transfer from coils
or jackets), and their extension to phenomena at the gas-
liquid interface or the bulk fluid is absurd at best.
An attempt at quantifying the effective shear rate in an
external-loop airlift reactor (A,Z4a = 9.2; aspect ratio = 6)
was made by Shi et aL (1990). The effect of viscosity on
the induced liquid circulation velocity in the downcomer
was determined using newtonian glycerol solutions (laL =
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 77
0.012-0.271 Pa.s) at various gas flow rates. Tile maximum
Reynolds number in the downcomer was about 3200, barely
in the turbulent regime. Pseudoplastic media
(carboxymethyl Cellulose and xanthan gum solutions; K =
0.078-0.333 Pa.s"; n = 0.404-0.937) were now used in the
reactor. Effective viscosity (Nov) of these fluids in the
circulation loop was determined as being equal to the
viscosity of the newtonian glycerol solutions, when the two
systems were at identical aeration rates and liquid
circulation rates. The effective viscosity and the known
values of K and n were used in the power law equation to
calculate the prevailing shear rate:
1
(145)
The calculated shear rates were correlated with the
superficial gas velocity in the riser (Shi et al., 1990); thus,
y = 3.26 - 3.51 x 102 Ucr + 1.48 x 10 4 U~r, (146)
which applied over the range 0.004 m-s-1 < Uor < 0.06
m-s1. Although written in terms of the superficial gas
velocity in the riser, equation (146) may usefully be
expressed in terms of the specific power input in the reactor
as recommended elsewhere (Chisti and Moo-Young, 1989).
Because the procedure employed by Shi et al. (1990)
equated the viscosity associated reduction in tile liquid
circulation velocity in different fluids, it gave in some sense
a shear rate in the vicinity of the interface between the fluid
and the walls of the reactor; shear rate in the bulk flow,
which is tile quantity of interest in most cases, was not
quantified. Furthermore, the method of analysis used (Shi
et al., 1990) applies strictly to a laminar flow regime, quite
unlike the flow situations encountered in most practical
operations.
Compared to bubble columns, equation (146) yields
quite low values for shear rates in airlift reactors as noted by
Shi et aL (1990). In such comparisons, care needs to be
taken to ensure that the devices are being compared at
identical values of specific power inputs (Chisti and Moo-
Young, 1989). Although Shi et al. (1990) did not adhere to
this criterion, the specific geometry of tile reactor they used
was such that the error was small. Unlike what Shi et al.
(1990) concluded, equation (146) is not suitable for
correlating mass transfer from gas bubbles or suspended
solids because it does not give shear rates at gas-liquid or
particle-liquid interfaces. An alternative, mechanistic
approach to quantifying the bulk shear rate in various zones
of airlift bioreactors has been developed by Molina Gfima et
al. (1997). Because the hydrodynamic environment in
various zones of airlift reactors tends to be quite different,
characterization of shear rate by a single global value is not
sensible. The overall shear rates can be deceptively low
even though damaging levels may be experienced in the
high-shear zones (Molina Grima et al., 1997). Hence, the
approach used by Molina Grima et al. (1997) is preferred.
 78 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Spectra of acoustic signals associated with such events
as bubble formation at the sparger, bubble disengagement at
the surface, and liquid flow over the baffle into the
downcomer, have been measured in water, a salt solution
and aqueous glycerol media in a bubble column and split-
cylinder airlift reactor (Glasgow et aL, 1992). Because
several events occur simultaneously in bubbling reactors,
there were substantial ambiguities in assignment of the
noise signals to specific events. The practical significance
of the results was unclear even though the intended aim was
to somehow relate the data to damage of fragile cells that
has been reported by others in such reactors. Yet other
work (Okada et al., 1993) has examined the structure of
turbulence in water and power law solutions (K = 0.0194
Pa.s°973; 0.0596 Pa.s°958) in an external-loop airlift reactor
(A/Ad = 1.0; aspect ratio ~ 20). The reactor achieved
complete gas-liquid separation and there was no gas in tile
downcomer. Measurements of local root mean square
velocity fluctuations as an indicator of turbulence intensity
showed a slight decrease from the center of the riser to the
wall. These measurements were at a constant gas velocity
of 2.05 x 10-2 m.s-1. The magnitudes of the velocity
fluctuations were similar for all media; however, the
velocity fluctuations were much lower in the gas-free
downcomer than in the riser despite similar values of
Reynolds numbers in the two zones.
One dimensional energy spectra were measured in the
center of the riser (Okada et al., 1993). Based on these
spectra, turbulence could not be considered isotropic
particularly in the power law fluids. Other substantial
evidence for the absence of isotropic turbulence in airlift
and bubble column reactors has been documented (Liibbert
and Larson, 1-990; Liibbert et al., 1990).
11. UNCONVENTIONAL AIRLIFT REACTORS
Several unconventional designs of airlift reactors have been
described. A few of these have gained acceptance in
specific applications while others are less practical. Some
of the novel configurations and operational schemes are
discussed here.
11.1 A new. mode of operation of internal.loop reactors
In conventional operation of airlift bioreactors tile static
height of the liquid or the slurry exceeds the height of the
draft-tube or the baffle between the riser and the
downcomer (Figure 2). A different mode of operation with
the initial, static, gas-free liquid or slurry level lower than
the upper edge of the baffle, or draft-tube, has been
discussed by Wenge et al. (1996). This configuration has
certain advantages in being able to achieve greater
separation of the gas from the liquid than is possible in the
conventional operation. Moreover, this mode of operation
can potentially increase the range of operating aspect ratios
and turndown ratios of existing airlift devices, thus partly
overcoming an important limitation of these systems in
comparison with bubble columns and stirred tank
bioreactors.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Roy vol 51, no 1, January 1998
A mechanistic analysis (Wenge et al., 1996) shows
that the minimum gas holdup ( ~ ) necessary to attain
incipient circulation of liquid in these systems is given by
(147)
A,i (1 - Po
6o~ hb 1+ A t \
where hb is tile height of the baffle. Note that in the mode
of operation considered, the initial ungassed level of liquid
(hL) is always less than the height ofbaffe.
Until the gas-liquid dispersion level reaches the upper
edge of the baffle, or draft-tube, the riser behaves purely as
a bubble column operating without flow through of liquid.
For such columns, the general relationship between
superficial gas velocity (Uo,) and the gas holdup is given by
equation (10). Equation (10) and equation (147) can be
combined to obtain an equation for the minimum superficial
gas velocity (U~,~,) required to establish recirculation of the
liquid; thus,
l/b
gor,,=l'ahbTi~Ad-~'-~l)[ (148)
k. k. Ar \ PLJJJ
Equation (148) is applicable to split--cylinder and draft-tube
internal-loop airlift reactors.
For extensive data in air-water, a- and b-values have
previously been reported (Chisti and Moo-Young, 1988b) to
be 2.47 and 0.97, respectively, for the bubble flow regime
(U~r < 0.05 m.s'l). Because in the bubble flow regime the
exponent b is close to unity, in this regime plots of the
minimum superficial gas velocity for incipient liquid
circulation (Uc,,) against the dimensionless parameter hc
(= (hb - hz)/hb) are expected to be linear. Experimental
observations shown in Figure 30 confirm this expectation
(Wenge et aL, 1996), thus validating equation (148).
Furthermore, the experimentally determined slope of the
line in Figure 30 is in remarkably good agreement with the
theoretically predicted value (Wenge et aL, 1996).
Earlier work in conventionally operated split--cylinder
and draft-tube internal-loop airlift reactors without special
gas-liquid separators has established that in these devices
the gas holdup values in the riser and the downcomer are
linearly related in the general form (Bello, 1981; Chisti,
1989, pp. 184-186, 218; Chisti et aL, 1995)
~Gd = k ~Gr , (149)
where tile dimensionless constant k is about 0.8 or 0.9,
depending on the geometric configuration of the reactor
(Section 3.2.1). The behavior in tile mode of operation
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumaticallymixed bioreactors 79

critical riser gas holdup, the downcomer and riser holdups

0.6 could be correlated with the equation (Wenge et al., 1996)

0.5 8or = (0.889 + 2 . 9 7 2 h J e a r - 0.642h~. (150)

Note that for the normal mode of operation (he ~ 0) a value

0.4 of zero should be used for hc in equation (150)irrespective
I
of the actual value. For zero he, equation (150) reduces to
E 0.3 the same form as equation (149) with a k-value of 0.889
E hc which is consistent with previously reported data (Bello,
0.2 1981; Chisfi, 1989, p. 218). As shown in Figure 31, for
normal mode of operation (that is, when h~ _<0), most of the
data does indeed fall just below the dashed parity line.
0.1
The behavior of the operational configuration
I described by Wenge et al. (1996) differed markedly from
0.0 i i t
the norms for internal-loop reactors operated
0.00 0.03 0.06 0.09 0.12 0.15
conventionally. Specifically, (i) circulation of liquid did not
hc(-) develop until the gas flow rate had exceeded some critical
value that increased with increasing dimensionless baffle
Fig 30. The minimum gas velocity for incipient liquid circulation clearance, h4 (ii) the downcomer remained free of gas until
(UG,m) versus the dimensionless height h~. The UG,,,-valueshave a critical gas holdup.had been established in the riser; thus,
been scaled by ×10 to ease presentation (Wenge et aL, 1996). the relationship "~d = k ,~,, that is well known to apply to
the conventional mode of operation, broke down; (iii) while
the gas holdup in the downcomer was strongly affected by
the he-value, the holdup in the riser was only slightly
0.25 affected; and (iv) mixing time, liquid circulation time and
I hL(m)I hi;(-) circulation velocity were sensitive to the he-value (Wenge et
/
o 11.50 I 0.120 ,/
+11-55 I o.os8 al., 1996).
0.20
v l 1-60 I 0.059 In summary, operation with he-values greater than
a 11.65 I 0.029 //
x 11.7o I 0.ooo zero can significantly extend the limits on the minimum
0.15 11.75 I-0.029 usable volume of airlift reactors. However, as shown by
I Wenge et al. (1996), the magnitude of hc affects all the
g major hydrodynamic performance parameters of the reactor.
0.10 Thus, utility of the proposed mode of operation depends on
the process-specific implications of the changes in
operational behavior that accompany changes in the he-
0.05
value.

0.00 I I 11.2 Reactors with undulating risers

0.00 0.05 0.10 0.15 0.20
~cr (-)
Fig 31. Effect of the static liquid height on the relationship
between the riser and the downcomergas holdup in split-cylinder
vessels (Wenge et al., 1996).
investigated by Wenge et al. (1996) was quite different. As
shown in Figure 31, for h, > 0, the downcomer remained
free of gas until a significant gas holdup had been built-up
in the riser and the level of dispersion had reached the
upper edge of the baffle. Some time after the gas-liquid
dispersion had spilled over into the downcolner, further
increase in the riser gas holdup led to a rapid rise in holdup
in the downcomer (Wenge et aL, 1996). The critical value
of the riser gas holdup at which the sudden increase in the
downcomer holdup began depended on the value of h, and
increased with increasing h,-value (Figure 31). Beyond the
0.25
A sequence of papers (Ghosh et al., 1993a, b; 1994a, b;
Ghosh and Bhattacharyya, 1993) have described a small
external-loop airlift reactor in which the riser was made of
a series of convergent and divergent nozzles so that file
riser-to-downcomer cross-sectional area ratio varied
between 0.25 and 1.00. This vessel (aspect ratio : 12, based
on downcomcr; volume = 12 L) was compared with an
identical reactor having a straight tube for the riser (A/Aa =
0.55). Studies showed that the reactor with variable A / A a
ratio produced consistently and significantly higher gas
holdups (Ghosh et aL, 1993b) and gas-liquid mass transfer
coefficients (Ghosh et at., 1993a) than the reactor with
constant Ar/Ad ratio. Variations in the cross-sectional area
of the riser caused variations in the local flow velocities
along the riser, generating turbulence that enhanced
performance. Although no data were presented, the
increased hydraulic resistance of the configuration with
variable A,/Ad ratio must have slowed down the circulation
of liquid compared to the reactor with the .straight riser.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 80 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Despite superior oxygen transfer capability of the
modified airlift, during fermentations of baker's yeast,
under optimal conditions, this reactor produced only a
16.7% higher biomass yield than the conventional airlift
device (Ghosh et al., 1994b). Continuous culture of baker's
yeast in the modified airlift was also described (Ghosh and
Bhattacharyya, 1993).
11.3 Reactors with perforated draft-tube
The hydrodynamic and process performance of concentric
draft-tube airlift reactors with either perforated draft-tubes
or tubes that were made of wire meshing have been
examined by a number of investigators (Bando et aL, 1992a,
b, 1995; Hsiun and Wu, 1995; Jong e t al., 1995; Roddguez-
Mendiola et al., 1991; Tzeng and Young, 1996; Wu and
Jong, 1994; Wu and Wu, 1990, 1991, 1992; Wu et al.,
1992). Use of a porous draft-tube made'of a geotextile
material has been reported in one case (Karamanev et al.,
1996). A summary of these studies follows.
T h e mixing time and specific gas-liquid interfacial
area in bubble columns and concentric draft-tube internal-
l o o p airlift reactors were characterized by Bando et al.
(1992a). Draft-tubes with 3 mm diameter perforations
covering 37% of the wall area were used. Draft-tube
sparged and annulus sparged modes were tested. Mixing
times were determined in the air-water system for 95%
homogeneity of a salt tracer. Significant movement of gas
and liquid in either direction through the perforated wall of
the draft-tubes was observed (Bando et al., 1992a).
Relative to the bubble column mode of operation,
introduction of the draft-tubes decreased mixing times (t~t)
as follows (Bando et al., 1992a):
tM _ 1--4.8 1 - C 1 -
t~
(151)
where C~ was 1.3 and 2.6, respectively for gas dispersion
into the draft-tube and into annular space. The mixing
time (tMn) in bubble columns (no draft-tube) was correlated
with the equation
t~ =0.40do do (152)
Equations (151) and (152) applied for the geometric ranges:
0.164 _<do _<0.300 m; 4.3 _<ho/do _<20; and 0.43 _< d/do _<
0.79 (for equation (151)). In these equations di is the
diameter of the draft-tube and do is that of the outer tube.
The calculated mixing times from equations (151) and
(152) agreed with the experimental data to within ±25%
(Bando et aL, 1992a).
The specific gas-liquid interfacial areas were
measured in air-sodium sulfite solution. For the bubble
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
columns the interracial area depended only on the
superficial gas velocity as follows (Bando et al., 1992a):
az~ = 1,900U~ 74, (153)
for the earlier specified geometric ranges. For reactors with
perforated draft-tubes, the interfacial area was enhanced
relative to bubble columns as follows:
-°'
aL _ 1 = 0.090 - (154)
OLB
when the column diameter was 0.164 m; and
aL
OLB
_ 1 = 0.096 (~O 410"4IY~O
- ¢1-0'6 (155)
when the column diameter was 0.300 m (Bando et al.,
1992a). The L/do range for these equaiions was 0.20--0.80,
where the characteristic length L¢ was the diameter of the
draft-tube when the gas was dispersed in the tube, and it
was equal to (do - di)/2 when the gas was dispersed in the
annulus.
In another study with draft-tube sparged internal-loop
reactors (A/Ad = 0.22; aerated aspect ratio = 14.0), Bando
et al. (1995) used aqueous non-newtonian solutions of
carboxymethyl cellulose (K = 0.11 Pa.s 1s8 and 0.26 Pa.s °81)
as well as the air-water system. The effects'of diameter of
perforations (0-14 mm) and of perforated area fraction (0-
51%) of draft-tube were examined on mixing time and the
gas-liquid volumetric oxygen transfer coefficient. In
general, mixing time was lower with perforated tubes than
with an unperforated one. Variation of mixing time with
perforated area fraction showed a minimum at 37%
perforated area. For otherwise fixed conditions, increasing
diameter of perforations decreased mixing time up to a
perforation diaineter of 5 mm. For the most viscous fluid
(K = 0.26 Pa-s°8~) used, the size of perforations or the
perforated area fraction had little effect on the mass transfer
coefficient. Only with water was any significant
enhancement in the volumetric mass transfer coefficient
seen with perforated draft-tubes. The mixing results were
explained in terms of the relative amounts of axial flow
through the draft-tube and the radial leakage through the
perforations (Bando et al., 1995). With the most viscous
medium used, apparently the axial flow through the draft-
tube predominated. Limited data were presented; all having
been obtained at a fixed superficial gas velocity (based on
outer tube) of 0.06 m-s~.
Liquid phase axial dispersion coefficients in draft-tube
sparged concentric tube airlift reactors (A/Ad = 0.33-1.78;
aspect ratio ~ 9; U~ up to 0.06 ln.s~; porous plate sparger
with 40-50 i.tm pore size) with wire mesh draft-tubes (24
mesh size) were reported by Wu and Jong (1994). The
oscillatory tracer response pattern that is characteristic of
conventional airlift reactors was not seen (Wu and Jong,
1994); instead, the response behavior corresponded to that
 Appl Mech Rev vol 51, no 1, January 1998
expected in bubble columns, indicating substantial
intermixing between the riser and the downcomer zones
through the walls of the draft-tubes. Observations with
dyes confirmed this intermixing. Consequently, the axial
dispersion coefficients were calculated as for bubble
columns (Wu and Jong, 1994). No correlations were
proposed, but for A/Ad values up to 0.61, axial dispersion
coefficients increased asymptotically with increasing
aeration rate, reaching a maximum of about 4.5 x 10-3
m2.s1 at superficial gas velocities of 0.04 m-s1 or higher
(Wu and Jong, 1994). This maximum dispersion coefficient
value was about 2-fold greater than expected in
conventional bubble columns at identical gas flow rates.
For A lAd values larger than 0.61, tile dispersion coefficients
increased with increasing gas flow rate, peaking at a gas
velocity of 1.2 x 10.2 m.sl; tile coefficients declined with
further increase in aeration rate (Wu and Jong, 1994). In a
different study, Hsiun and Wu (1995) noted that draft-tube
sparged internal-loop airlift reactors with wire mesh draft-
tubes (mesh number 24; A/Ad = 0.25-1.09; aspect ratio =
12.7-22.2; air=water) were consistently better mixers (had
lower specific mixing times) than corresponding bubble
columns (Hsiun and Wu, 1995), suggesting that the
directional character of flow due to the draft-tubes aided
mixing.
The gas-liquid oxygen transfer .characteristics of
reactors used in the mixing work of Wu and Jong (1994)
were reported by Wu et aL (1992). Unlike most earlier
work on mass transfer, the headspace pressure was varied
over 1-2.36 atmospheres. No major effect of the A/Ad ratio
was observed on the overall volumetric oxygen transfer
coefficient. This was consistent with the earlier
observations that the mixing behavior in these reactors
resembled that in bubble columns (Wu and Jong, 1994).
The leakage of the gas into the downcomer through the
openings in the 'walls' of the draft-tubes further reduced or
eliminated the effects of A/Ad ratio that are expected in
conventional airlift devices having solid draft-tubes.
Plots of the oxygen transfer coefficient against
superficial gas velocity showed an apparent dependence on
the headspace pressure (Wu et al., 1992), which
disappeared when the superficialgas velocity was corrected
for pressure as recommended by Chisti (1989, pp. 287-
290). An apparent pressure effect on gas holdup had been
reported earlier by Chisti and Moo-Young (1987a) whose
uncorrected data showed declining gas holdup with
increasing pressure (Chisti and Moo-Young, 1987a). The
apparent decline in the overall oxygen transfer coefficient
that was observed by Wu et al. (1992) was likely related to
the previously described apparent influence of pressure on
gas holdup. Note that the interfacial area for mass transfer
is dependent on holdup. In defiance of their own evidence,
Wu et al. (1992) concluded that pressure had a significant
effect on the volumetric oxygen transfer coefficient.
The overall gas holdup and the volumetric gas-liquid
oxygen transfer coefficient in draft-tube sparged concentric
tube airlift vessels (A/Ad = 0.3; aspect ratio = 14; Uc = 0-
0.05 m.sl; porous plate sparger, 40-50 pJn pore size) were
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chist!" Pneumatically mixed bioreactors 81
reported also by Wu and Wu (1990). Tile size of the draft-
tube wire mesh varied in this work (3-24 mesh size). In
comparison with a conventional vessel (solid draft-tube),
tile gas holdup was not particularly affected by the mesh
size. The mass transfer coefficient in a conventional airlift
were significantly lower than in reactors with 'leaky' draft-
tubes. With perforated tubes, the mass transfer coefficients
were comparable to those obtained in the bubble column
mode of operation (Hsiun and Wu, 1995; Wu et al., 1992).
The size of the draft-tube wall openings did not have a
major impact on the mass transfer performance over most of
the operating range.
In microbial cultivation, airlift reactors with wire
mesh tubes consistently outperformed bubble columns (Wu
and Wu, 1991; 1992). Thus, in fed-batch culture of
Saccharomyces cerevisiae in a wire mesh draft-tube
sparged concentric tube reactor (24 mesh size; A/Ad = 0.3;
aspect ratio = 14), an aeration rate of 1.5 w m was sufficient
for a level of productivity that could be obtained in an
equivalent bubble column only at a much higher aeration
rate of 2.5 vvm (Wu and Wu, 1991). Another similar study
described successful fed-batch culture of Bacillus
thuringiensis for tile production of the insecticide
thuringiensin (Jong et al., 1995). The reactor was shorter
(aspect ratio -8), but the other geometric details were the
same as employed by Wu and Wu (1991) in yeast culture.
The wire mesh draft-tubes have been claimed to be
especially useful in fed-batch culture where the level of the
liquid may be below the upper edge of the tube and increase
as the culture is fed.
In another work, comparing glutamic acid production
by Brevibacterium divaricatum in wire tube airlift and •
bubble column reactors, Wu and Wu (1992) noted that in
the airlift mode, increasing aeration rate from 1.5 to 3.5
vvm increased biomass and glutamic acid productivities,
respectively, by 1.6- and 1.8-fold for this highly aerobic
fermentation. For identical aeration conditions
fermentation productivities were significantly lower in the
bubble column mode of operation. For example, at 3.5 vvm
aeration rate in the two reactors, the biomass and glutamic
acid productivities of the airlift were 1.3- and 39-fold that
in bubble column. Superior performance of the airlift was
associated with consistently higher gas-liquid volumetric
mass transfer coefficient in the airlift relative to the bubble
column at all air flow rates. The mass transfer coefficients
were measured by oxygen balancing during fermentations.
Other evidence for poorer oxygen transfer performance of
the bubble column came from measurements of the
fermentation byproduct, lactic acid, in the broth. Up to 4-
fold higher concentrations of lactic acid were detected in the
bubble column compared to the airlift. Insufficient oxygen
is known to alter the metabolic pathway causing higher
production of lactic acid.
11.4 Cascaded airlift reactor
A cascade of three internal-loop airlift reactors combined
into a single unit, as shown in Figure 32a, has been
employed by Bakker et al. (1994) for hydrolysis of sucrose
 82 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
(a)
L tSTRATE
OUT UT
Ii l i"-I---eAFFLES
1 ~ 1 1 1 I"
Fig 32. A cascade of three draft-tube internal-loop reactors combined into a single unit(a; Bakker et al., 1994); and (b) the split-
channel version of the cascade.
to glucose and fructose using immobilized invertase. Tile
sucrose solution entered the downcomer of the central airlift
unit, moved into the riser and overflowed into the
downcomer of the second unit and on to the outermost
reaction unit. The three units operated at successively lower
dispersion levels. The reactor as a whole behaved as a
series of three stirred tanks (Bakker et al., 1994). This
behavior occurred probably because of the relatively small
aspect ratio of the device. Stainless steel sieve plates at the
outlets of each airlift section retained the catalyst beads in
various compartments. A possible advantage of the
proposed reactor over multiple separate units is tile
elimination of interconnecting pipework and pumps
(Bakker et aL, 1994). In addition, a series of different
reactions, aerobic or anaerobic, may be carried out within
one vessel in the proposed configuration (Bakker et al.,
1994). These advantages notwithstanding, the reactor
configuration is quite restrictive. A potential case for
successful application may be in biological wastewater
treatment, especially if the configuration is modified to a
'split-channel' design shown in Figure 32b. Separate,
multiple, airlift vessels, arranged in cascades, are more
flexible and have indeed been used in commercial practice
(Steiner, 1987).
Hydrodynamics, hydraulics and transport behavior of
the airlift cascade reactors, as in Figure 32a, may be
analyzed by treating each compartment as a separate airlift
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
_l
-I
WASTEWATER
TREATED
il"
EFFLUENT
AIR
(b)
device. Thus, Bakker et aL (1993) noted that the behavior
of individual compartments was generally consistent with
well known observations for internal-loop reactors (Chisti,
1989). For example, i n air-water, the riser and the
downcomer gas holdups were linearly related with the latter
being about 90% of the former (Bakker et al., 1993).
Apparently, the equation KB = 11.4 (Ad/Zb) 0"79 (Chisti, 1989,
pp. 206-219; Chisti et aL, 1988) did not apply to the
reactors tested by Bakker et al. (1993), possibly because the
Aa/Ab values for most of the units tested were barely within
tile recommended range of applicability of the equation
(Chisti et al., 1988).
Oxygen transfer measurements in the middle airlift
zone of a cascaded airlift device of the type shown in Figure
32a were detailed by Bakker et al. (1995). The vessel was
0.034 m 3 and A j A , values were 0.31, 0.43 and 0.91; the
static liquid height was 2.26 m. For comparison, a
rectangular channel internal-loop airlift reactor (length=
0.8 m, width = 0.1 m, height = 1.5 m, volume = 0.115 m 3,
static liquid level = 0.85-1.34 m, A j A r = 0.98 or 0.42) was
also evaluated. Data were obtained in aqueous potassium
chloride (0.02 kmol.m3, ~L = 1 x 10"3 Pa.s, aL = 73 x 10-3
N.m~). In the cascaded device, the kLaL was always lower
than in a comparable bubble column, but apparently the
comparison was not based on identical Specific power
inputs: the superficial gas velocity in the airlift was based
on the cross sectional area of the riser whereas that in the
 Appl Mech Rev vol 51, no 1, January 1998
bubble column was based on the total cross sectional area.
Thus, the airlift kLaL was artificially lowered. In the airlift,
the kLaL decreased with increasing Ad/Ar ratio. The same
effect was noted in the rectangular channel configuration
(Bakker et aL, 1995). In the latter, variations in the
working aspect ratio over the range 8.5-13.4 did not affect
the kzaz values. For any given reactor configuration, a
constant value of the kz/dn ratio applied as originally noted
by Chisti (1989, pp. 251-261). For the rectangular channel
airlift vessel, the kz/d8 was 0.071 s4 which was comparable
to a value of 0.06 s4 reported by Chisti (1989, p. 254). The
cascaded airlift system had unusually low kz,/dB ratios at
0.026 ± 0.004 s4 (Bakker et aL, 1995).
11.5 Inverse fluidized airlift reactor
Gamier et al. (1990) described a mode of operation that
combined inverse fluidization with an airlift configuration.
Solid particles that were less dense than the liquid were
carried into the d0wncomer and held there by the downward
flowing liquid against the buoyancy force. This
configuration was said to have the advantage of retaining
particles in the low-shear, gas-free, downcomer zone
(Gamier et a1.,1990) while offering lower pressure drops
than packed beds. Shear sensitive biofilms could potentially
developed on the particles, particularly for wastewater
treatment. As the fihn thickness increased, the panicle
would become denser and eventually would be carried
through to the relatively turbulent riser where some of the
biomass may be sheared off. The particle would then return
to the downcomer. This mechanism could effectively
control the thickness of the biofihn. However, because the
particles would be less densely packed than in packed beds,
and would be confined largely to the downcomer, the useful
surface area available for the biofihn would be small--much
smaller than in the biofilm suspension airlift bioreactors
described by van Loosdrecht and Heijnen (1993). The
hydrodynamic performance of such 'inverse fluidization'
airlift bioreactors was investigated by Garnier et al. (1990)
with low density polystyrene particles (ps = 60-400 kg.m3;
d v ~ 150-200 gm). The maximum acceptable liquid
velocities in the downcomers of such reactors would be
severely restricted to prevent particle washout, unless larger
particles were used.
11.6 Airlift-packed bed hybrid reactors
Airlift reactors containing packed beds of biocatalysts in the
downcomer are potentially useful in immobilized enzyme
and cell culture applications. Such a reactor is shown in
Figure 33. The success of this type of design depends on
the ability-of the airlift drive to move the liquid through the
packing at the required rate. The rate depends on the
oxygen requirements of the packed bed and on the need to
attain particular levels of solid-liquid mass transfer in the
packing. Anchorage dependent animal cells have been
grown on the surface of microbeads in beds irrigated by
airlift driven culture fluid (Chiou et al., 1991; Spier and
Whiteside, 1976; Whiteside and Spier, 1981). Such devices
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 83
0 o
0 0
0
o
RISER - - °°
eO
0
/
O° 0
0
o
0 0
0
.o Oo
O0
A
AIR
(uGr)
PACKED
BED IN
DOWNCOMER
Fig 33. Airlift-packedbed hybrid reactor (Chisti and Moo-
Young, 1993b).
have been applied at 0.1-100 L scales for the production of
foot-and-mouth disease vaccine (Whiteside a n d Spier,
1981) using cells supported on 3 x 10-3 m diameter glass
beads. A superficial liquid velocity o f - 3 x 10-3 m.s"1
through the packing has been reported (Whiteside and
Spier, 1981). Similar reactors have been used to grow
hybridomas for the production of monoclonal antibodies
(Murdin et al., 1987) and murine bone marrow cells
(Highfill et al., 1996). These examples demonstrate that
the airlift-packed bed combination is a practical idea, not
just a curiosity. The riser of the hybrid reactor offers a
better mixed zone than the packed section; the pH,
temperature and the dissolved oxygen levels of the
recirculating fluid can be easily controlled in the riser.
Hence the hybrid design overcomes some of the
shortcomings of packed bed bioreactors. Despite these
enhancements, the use of packed bed airlifts is unlikely to
be feasible for microbial cultivation (except perhaps for
non-growing immobilized cells) because of possible
difficulties with clogging of the bed by biomass;
immobilized animal cells and immobilized enzymes are
expected to be the main users of airlift-packed bed
bioreactor technology. Although fluidized bed bioreactors
may be employed for reactions involving solid biocatalysts,
packed systems can attain higher solid-liquid mass transfer
rates (at lower energy inputs) because of the higher relative
velocities between the liquid and solid phases in these
systems (Mao et al., 1992). In addition, the kinetics of
 84 SPECIALISSUE: FLUID MECHANICS PROBLEMSIN BIOTECHNOLOGY
particular reactions may have a plug flow requirement
which is easily met in packed beds. Hence an evaluation of
the liquid circulation phenomenon in the airlift-packed bed
combination is necessary. Hydrodynamic and mass transfer
models for airlift-packed bed reactors have been reported
(Chisti and Moo-Young, 1993b; 1994a; Kumar and Shuler,
1995; Murakami et al., 1991).
The model presented by Murakami et aL (1991) was
developed for a draft-tube internal-loop airlift reactor in
which the annulus was packed with a bed of fibers for
supporting anchorage-dependent animal cells. A occur and are not considered in equation (156).
complimentary article (Chiou et aL, 1991) detailed culture
of cells in that device. The complicated model required a
knowledge of the two-phase frictional losses, the variable
drift-flux distribution parameter and bubble rise velocities.
Complete disengagement of tile gas was assumed even
though at the high gas flow rates that would be needed in
realistically sized reactors, complete gas disengagement
cannot occur in the type of device employed by Murakami et
al. (1991). In the small reactor (4 L; aspect ratio about 1.3),
the diameter of the draft-tube (riser) was varied over 0.8-
5.6 cm (Murakami et al., 1991). Maximal values of liquid
circulation velocity were observed in an unusual range of
the A/Ad ratio (0.006-0.03). Maldistribution of liquid must
have inevitably occurred in the low-aspect ratio packed bed
with such small values of the A/Ad ratio.
The behavior of insect cells immobilized on support
material held in the downcomer of a hypothetical split-
cylinder airlift bioreactor has also been modeled (Kumar
and Shuler, 1995). Although the model agreed with some
published data (Kumar and Shuler, 1995), it was not
rigorous, and essential aspects of hydrodynamics and gas-
liquid mass transfer were ignored. More robust
hydrodynamic and oxygen transfer models are available
(Chisti and Moo-Young, 1993b, 1994c). Such thorough
models can be dependably used for scale-up as discussed in
the following sections.
11.6.1 Liquid circulation
As in any airlift system, the difference in gas holdup
between the riser and the downcomer causes the circulation
of liquid through the packed bed. To ensure that the gas
bubbles do not build up in the packing to the detriment of
liquid circulation, the reactor should be designed such that
all gas bubbles escape the liquid prior to it recirculating
(Section 5). This gas-liquid separation has been discussed
and comprehensive design guidelines for ensuring complete
disengagement have been published (Chisti and Moo-
Young, 1993a).
Prediction of liquid circulation velocity is based on an
energy balance over the airlift loop: rate of energy input into
the reactor equals the rate of energy dissipation, or
E~ = ER + E v + EB + E r + EF, (156)
where E~ is the energy input due to isothermal gas
expansion in tile riser; ER is the energy dissipation due to
wakes behind bubbles in the riser; Ev is the energy
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
dissipation due to the bed of solids; EBm is tile energy loss
due to fluid turn around at the bottom (top) of the reactor;
and EF is the energy loss due to friction in the riser and the
downcomer. Equation (156) disregards the kinetic energy
input associated with sparging of the gas because this input
is usually negligible for well designed spargers (Chisti,
1989, pp. 287-292). Further, to ensure stable operation,
airlift-packed bed reactors must be designed for complete
gas-liquid separation in the head region; hence, energy
losses due to stagnant gas in the downcomer would not
Because viscous fluids and non-newtonian media
containing solids are not suitable for use in packed bed
airlifts, only newtonian low-viscosity (water-like) flows
need be considered. Under these conditions, the wall
friction associated energy losses can be neglected (Chisti et
al., 1988) in comparison with other terms in equation (156);
thus EF will be ignored. The energy dissipation due to the
wakes behind the bubbles in the riser (En) can be obtained
(Chisti et al., 1988) by an energy balance on the riser as the
control volume. En is given by
(157)
ER = E~ - p L g h D U L ~ Ar~o~.
The energy loss in the top and tile bottom sections of the
airlift is given as (Chisti, 1989, pp. 203-206):
1
Ee + Er = ~ P L ArU3r
z
x --- 2 ~KB
0-, 3 (1-
(158)
where KT and KB are the friction loss coefficients for the top
and the bottom connection sections between the riser and
the downcomer. Because the downcomer is assumed to be
free of gas, ~d = 0, and equation (158) simplifies to
EBq-ET=~pLArU3Lr 12~-r) +KB •
(159)
By analogy with pipe flow, tile energy dissipation rate
in the packed bed (Ep) can be written as
E e = ULd Ad AP, (160)
where AP is the pressure drop through the bed, and UZd is
the superficial liquid velocity in the downcomer. Because
the continuity relationship for incompressible flow governs
the flow between the riser and the downcomer, we have
ULr Ar = ULd Ad, (161)
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 85

which can be substituted into equation (160) to obtain Ep in flow (Coulson and Richardson, 1978). In terms of UL,,the
terms of UL~: equation is

PL U~S(1- q~)Lp
Ep = ULr A~ AP. (162) AP--
(166)
Substitution of equations (157), (159) and (162) into
equation (156) leads to x (5 Re -l + 0.4 Re -°1

Ei = E~ - PL ghn ULrA~ cot where qb is the void fraction of the bed, Lp is the depth of
mr 2
packing, and S (= 6/dp) is the surface area per unit volume
1 3 of the particles. The definition of the Reynolds number (Re)
_~PL A~ULr(.(1 Kr 2 (163)
+ _77 J employed by Coulson and Richardson (1978) in equation
(166) can be adapted for the airlift:
+ ULrArAP,
or ULrPL ( A ~
Re - sC17~-pL . (167)
U~r( KT Substitution of equation (166) into equation (164) gives an
o = --gho r + . 2. .I.(1--
.
expression which can be solved for the liquid velocity, ULr.
(164)
+ -AP
- Rashig rings. Equation (166) gives consistently low results
(Coulson and Richardson, 1978) for hollow packing. Flow
PL through such packing is better described (Coulson and
Richardson, 1978) by the semi-empirical Ergun equation
Equation (164) is the basis of the method for prediction of
written below in terms of ULr
the liquid circulation rate (Chisti and Moo-Young, 1993b).
The parameters KB and Kr (equation 164) may be calculated
as detailed in Section 4.1 for reactors without packed beds.
Equation (164) requires a knowledge of the riser gas holdup
(~r) which depends on the superficial gas velocity (UGr) ( 1 )2 d~ +175(1-q~)pLULr(A~
q~3 dp2
and the liquid flow (ULr), according to the well known
Ellis's equation cited by Hills (1976) and given below in a
rearranged form: ×(A,/A,)Lp
(168)

UGr
°~Gr 0.24+ 1.7(ULr+ UGh)°7" (165)

Equation (165) applies to air-water system when (UL,. + Following the approach of Coulson and Richardson (1978),
UGr) < 1.3 m.s-~. The calculation of pressure drop through the particle diameter (dp) in equation (168) can be replaced
the packing (AP) for use in equation (164) is explained in with 6/S, and upon rearrangement we obtain
the following sections.
11.6.2 Pressuredrop through the packed bed
Two modes of catalyst deployment are potentially
significant in immobilized biocatalysis: (a) beds composed
of spherical catalyst particles such as those used in
anchorage dependent cell culture (Spier and Whiteside,
1976; Whiteside and Spier, 1981) or enzyme colmnns
(Chaplin and Bucke, 1990); and (b) beds of more open
structure as produced by ring packing such as Rashig rings.
These cases are treated separately below.
(169)
Sphericalparticles. Pressure drop through beds of uniform
particles can be satisfactorily correlated by the Carman- Substitution of equation (169) into equation (164) leads
Kozeny equation applicable to laminar as well as turbulent once again to an expression which may be solved for ULr.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 86 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

11.6.3 Calculation of the superficial liquid velocity Oxygen flow into bed - oxygen flow out of bed -
For calculation of liquid circulation velocity in a reactor of oxygen consumption in tile bed = 0, or
specified geometry (that is, At, Ad, hL, S, Lp, and ¢), fluid (170)
(that is, PL and ~z), and riser gas flow (U~,.), the following
Q G - Q c~,~, - q = 0,
procedure is used:
where Q is the volume flow rate through the downcomer
1. A value for ULr (0--1 m.s "~) is assumed.
containing the packed bed, Ci is the oxygen concentration in
2. The riser gas holdup (~r) is calculated using tile
the fluid flowing into tile bed, Ccr, is the cntical oxygen
assumed ULr (step 1), and known Uc,., in equation
concentration for animal cells, and q is the total oxygen
(165). consumption in the bed. For animal cells, as for microbial
3. The pressure drop through the packed section (AP) is cells, oxygen consumption is zero order in oxygen
calculated using either equation (166) or equation concentration for concentrations greater than or equal to the
• (169), depending on the type of packing (spheres or critical oxygen concentration, C~rit (Miller et al., 1987).
Rashig rings). Thus, oxygen consumption is independent of axial location
4. The height of gas-liquid dispersion (hD) is now in the bed. The total oxygen consumption, q, can be
estimated (Chisti, 1989, p. 298): hD = hL/(1 - ~,.). calculated with the equation
5. The geometry dependent parameters Kr and KB are
calculated as explained in Section 4.1.
6. The right hand side of equation (164) is solved using
q = Nt Ro, (171)
the parameters determined in steps 1-5. If the RHS =
where Nt is tile number of cells in tile bed and Ro is the
0, tile initially assumed Umr (step 1) constitutes a
oxygen consumption rate per cell. Assuming,
solution. For RHS ¢ 0, steps 1-6 must be repeated
conservatively, that a monolayer of cells develops on the
with a new assumption for Uzr. The ,~r and hD
entire surface area of the packing, we have
corresponding to a solution are the predicted riser gas
holdup and the gas-liquid dispersion height,
S u r f a c e a r e a o f the p a c k i n g
respectively. N, = , (172)
Illustrative calculations (Chisti and Moo-Young, 1993b, P r o j e c t e d a r e a o f cell
1994a) show that induced circulator 3, flow in airlift-packed
bed bioreactors can be sufficiently high for anchorage- or
dependent animal cell culture and immobilized enzyme
catalysis. The selection of type, size and depth of Ihe 4S AjZ~(1-C~)
packing is based on the required range of liquid flow, the N, = nd~ ' (173)
acceptable riser gas velocity, and the desired surface area
for supporting the catalyst. The liquid flow requirements
where do. is the diameter •of tile cells.
are determined by the axial depletion of oxygen in tile
Substitution of equations (171) and (173) in equation
packed zone and solid-liquid mass transport considerations.
(170), followed by rearrangement leads to
The approaches described can be easily adapted for
use in airlift reactors having static mixers (Chisti et al.,
1990) or other internals (for example, heat exchangers). Q(C,-C~,~t)zd2c
The ability to predict the rate of circulation is a prerequisite Zc= 4 S R o A d ( 1 - q ~ ) " (174)
to predicting the solid-liquid mass transfer and heat
transfer coefficients within the bed. These transport Because the flow rate of liquid, its superficial velocity and
properties have been thoroughly studied in packed beds as the cross-sectional area of tile downcomer are related,
evidenced by major reviews (Lemcoff et al., 1990; Zenz and
Othmer, 1960). Once the airlift driven liquid circulation
velocity through the packing has been quantified, the Q = ULd Aa, (175)
transport rates in the bed can be calculated with existing
correlations (Lemcoff et al., 1990; Mao et al., 1992; Zenz equation (174) is modified to
and Othmer, 1960).

11.6. 4 Depth of packing to prevent oxygen starvation ULd(C,-Cc,~t)~d2c

As the cell free culture fluid, initially saturated with oxygen, Zc= 4SRo(1-¢) (176)
percolates through the packed bed, oxygen is depleted
axially down file bed due to consulnptionby the
Equation (176) expresses the critical depth of packing in
immobilized cells. The depth of packing up to which there
terms of the superficial velocily of liquid in the downcomer
is no oxygen starvation of cells, the critical depth (Z¢), can
(Uzd) and the specific oxygen consumption rate of the cells
be calculated by an oxygen balance on the bed. Thus,
(Ro).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vo151, no 1,.January 1998 ' Chisti: Pneumatically mixed bioreactors 87

For a large scale (1.5 m 3, A/Ad =1, air-water) airlift-

24
packed bed reactor containing a 4 m deep bed.of 3 x 10.3 m
spherical particles, the superficial liquid velocity in the • @
downcomer (U~) was calculated using the procedure
described by Chisti and Moo-Young (1993b). For a
20

16
_ /D CHO CE LLS~
superficial riser gas velocity range of 0-0.12 m @ , the
"•"
N(J
calculated superficial liquid velocity varied over 0--0.014
12
m.s "1. These velocities were used in equation (176) to
calculate the critical depth (Z¢) of packing. Figure 34 shows
8
a plot of critical depth against liquid flow velocity for tllree
different oxygen consumption rates in the range 0.05-0.5
mmol O2/(cell.h). This is the range of specific oxygen
consumption rates for most cultured mammalian cells as
noted by Fleishaker and Sinskey (1981). In addition, the
o#
2 4 6 8 10
,
12 14
,
6
calculated critical depths are shown in Figure 34 for the ULd x 10 3 ( m s -1)
commonly used anchorage dependent CHO cells for which I l I l I I I I
Gfiffiths (1992) reported an specific oxygen consumption 0 2 4 6 8 10 12 14
rate of 1.5 x 10~° mmol O2/(cell.h). For calculation of Z¢ Uor x 10 2 ( m s -1)
using equation (176), the diameter of cells was assumed to
be 20 x 10-6 m which is typical for lymphocytes and many • Fig 34. Critical depth of packing (Zc) versus superficial liquid
other animal cells (Butler, 1987). The inlet oxygen velocity in the downcomer for various specific oxygen
concentration, C~, in the bed was assumed to be the air consumption rates 0mnol O2/(cell:h)): 03 5.0 × 10n; ® 1.5 ×
saturation value for aqueous media at 203.13 mmol.m 3 (6.5 10"l°; ® 2.5 × 10=°; ® 5.0 × 101°. The superficial air velocities
required to generate a given level of liquid flow in the
ppm). Considering the low flow rates and, consequently,
downcomer are also shown (Chisti and Moo-Young, 1994a).
long residence times of the liquid in the air-sparged riser,
the liquid entering the downcomer will indeed be oxygen
saturated as can be shown using the methods described by packing density in the beds, the hybrid reactor can
Chisti (1989). The critical oxygen concentrations for potentially sustain more than twice as many cells as would
animal cells are quite low, and a typical value of 0.5% of air be possible to support in a stirred tank microcarrier culture
saturation determined by Miller et al. (1987) was used in system of equal volume. In addition, the retention of the
this analysis. packing in the reactor can ease cell removal by
As shown in Figure 34, when, for a given liquid flow trypsinization; the packing can be cleaned and sterilized in-
place, and reused. During start-up, slow, percolating flow
(Uzd ~ 0.014 m.s -~) and oxygen consumption rate (Ro), the
tllrough the bed may be used to favor attachment of cells to
calculated critical depth (Z¢) is greater tlmn or equal to the
the carriers as has been shown in small scale experiments.
actual depth of 4 m, the airlift-packed bed combination can
Other significant advantages of airlift-packed bed devices
be successfully used for culturing the cells. Furthermore,
are an absence of contact between cells and air bubbles
even for cells for which the oxygen demand is at the upper
during most of the operation, and elimination of collisions
limit of 0.5 x 10.9 mmol/(cell.h), sufficiently high rates of
among carriers or between carriers and the impeller.
flow of liquid can be generated to maintain a viable cell
population throughout the packing. The aeration rates
required (Figure 34) to maintain circulation of the medium
11.7 Liquid--iml}elled reactors
are not expected to cause foaming in the increasingly A 'liquid-impelled' loop reactor based on the principle of
common protein-free media. When serum containing airlift bioreactor has been described by van Sonsb~k et al.
media are employed, any foam formation may be controlled (1990a, b; 1992). Instead of a gas and a liquid, two
using chemical antifoam additives or mechanical foam immiscible liquids are used. The continuous phase is
breakers as recently described for industrial cell culture contained in a vessel witll a riser-downcomer loop
reactors (Chisti, 1993). High rate ~aeration of bicarbonate configuration. Droplets of a dispersed phase are injected
buffered media can lead to unwanted pH changes; however, into one arm of tile loop. The dispersed phase containing
this problem is easily overcome by CO2-supplementation of ann may behave as a riser when the dispersed phase is
the aeration gas typically at 5% vol level (Chisti, 1993). • lighter than the continuous phase. Alternatively, when the
dispersed phase is heavier, the dispersed phase containing
11.6. 5 Productivity of the airlift-packed bed arm becomes the downcomer. Figure 35 illustrates these
In addition to oxygen supply, another important alternatives. The difference in the bulk density of the fluid
consideration in establishing the merit of the airlift-packed in the two arms drives the circulation of the continuous
bed combination relative to tile conventional stirred tank phase. The kinetic energy input at the point of injection of
system is the final cell concentration attainable in these the dispersed phase can make a significant contribution to
devices. A comparison of these systems has been reported circulation. A model for liquid circulation in such reactors
(Chisti and Moo-Young, 1994a). Because of the high has been presented by van Sonsbeek et al. (1990a, b), but

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 88 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

DISPERSED
PHASE
IIIIIII. . . .
o o oo
~o o o o Oo° o°O o
o oo0oo to o DO ool

RISER
O o o
)oOo
- CONTINUOUS
O~oo0
ooOo~l
o 0°1
- - RISER

PHASE
'~° 0 ° ° °
o° t:oOo
i ° °o;i.!
;°r OoOl
t
CONTINUOUS
o~o C DOWNCOMERS ;° ro~, ol PHASE
3 o o Do o 0 o ol
Do O O o c o ° Ual
o o ot ,o ° o o ~
°ou2 ° I O C ^ u cI

°°2 *~ )o° o-oOel

,o~°oo
Illllll '.*.~ ....................
DISPERSED
PHASE

PUMP

Fig 35. Liquid-impelled loop reactor configurations: (a) dispersed phase lighter than the continuous phase; (b) heavier dispersed phase
(based on van Sonsbeek et al., 1990a).

the model disregards tile contribution of kinetic energy.

The liquid-impelled loop reactors may have some limited
applications especially in liquid-liquid extraction
operations.

11.8 Reactors with static mixers

Static mixers installed in bubble columns and risers of
airlift reactors may usefully be employed in enhancing gas
holdup, gas-liquid mass transfer and solid-liquid mass
transfer (Chisti et al., 1990; Gaspillo and Goto, 1991;
Kawasaki et aL, 1994; Kawase, 1994; Potucek, 1990;
Stejskai and Potucek, 1985; Yamashita, 1993; Zhao et aL,
1994). In bubble columns the mass transfer is enhanced at
the expense of axial mixing of liquid. In airlift reactors, the
mixers slow down the circulation of liquid because of the
increased resistance to flow in the circulation path. Mixing
elements break up large gas bubbles into smaller ones, tile
gas-liquid interfacial area increases and, hence, the mass
transfer coefficient (kzaL) increases relative to mixer-free
operation (Clfisti et aL, 1990).
Potucek (1990) reported gas-liquid interfacial areas in
an aqueous sulfite oxidation system in a draft-tube sparged
concentric tube airlift vessel (A,Z4d = 0.1) with and without
Kenics static mixers (Figure 36) in the draft-tube. Over a
superficial gas velocity range of 0.01-0.06 m-s"1 (based on
outer tube), gas holdup and interracial area were enhanced
somewhat by the mixers. Following equations were
obtained for specific interfacial area (Potucek, 1990):
i
Fig 36. Six elements of twisted ribbon Kenics type static mixers
ao = 4.17 × 1 0 3 U ~ 1 (no mixers), (177) inside a tube. The flow undergoes alternate clockwise and
anticlockwise rotation as it moves up the tube.
and
Equations (177) and (178) may be used to calculate the
aD = 8.33 × 103U~ z ( w i t h mixers). (178) relative effect of Kenics-type mixers in aqueous salt
solutions, but they are unlikely to apply to fermentation

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vo151, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 89

broths. The Kenics mixers are useful for liquid mixing, but
poor at dispersing bubbles. Thus, in work reported by 0.15 kmol.m -3 NaCI
30 /"
Gaspillo and Goto (1991), the twisted ribbon mixers did
V WITH STATIC MIXER
enhance gas-liquid mass transfer performance, but not as , WtTHOUTSTATIC MIXER 7
much as observed by Chisti et al. (1990) with mixers 20
designed for redispersing bubbles.
The effect of static mixers on gas holdup and kLaL is
I
most pronounced in viscous fluids which give rise to large I/)
v

spherical cap bubbles (Chisti et al., 1990); less dramatic,

but still large improvements are observed with water-like
m

O
media. Generally, for a given fluid, the mass transfer
8
coefficient and the gas velocity in the riser of a given airlift x 7
reactor may be correlated in the form 6
_J
.~ 5
kLaL = ctU~r. (179)

For a given fluid, the exponent 13 in equation (179) is not I I I I I I I

affected by the presence of static mixers in the reactor; 3
2 3 4 5 6 7 8910
however, the mixers significantly enhance the value of ot
(Chisti et al., 1990). The improvement in ot depends UGr x 10 2 ( m . s - 1 )
mainly on the consistency index (K) or the 'thickness' of the
fluid: generally, the thicker the fluid, the greater the effect
Fig 37. Effect of static mixers on gas-liquid volumetric mass
of static mixers (Chisti et al., 1990). The approximate
transfer coefficient in aqueous salt solution in an external-loop
relationship among ~, ~XM(the ~-value in presence of static airlift reactor (Chisti et al., 1990).
mixers) and K has been established to be (Chisti et al.,
1990)

20
a M = 4.43K0.12. (180)
6t

Equation (180) was determined for approximately 104-fold

variation in K. Tile flow behavior index n (varied between 10 m

9 m

0.5 and 1.0) had no direct impact on enhancement of kzaz ~'~ 8

or gas holdup by the mixers (Chisti et al., 1990). 7 7
Depending on the fluid, the static mixers were found to ~ 6 m

enhance kLaz by 30-500% relative to mixer-free operation o 5

(Chisti et aL, 1990). The constants in equation (180) are 4
x
expected to depend on the type of static mixer used. Figures .-,,J
37 and 38 illustrate the enhancement of kLaz by static
mixers in an aqueous salt solution and a non-newtonian
o
...1
3
/
0.6~ wt/vol. (9/100 rnL) C M C /
pseudoplastic medium, respectively, in an external-loop _ IN 0.15 kmol.m -3 NoCI /
airlift reactor. /
o WITH STATIC MIXER
In general, fewest possible mixing elements consistent • WITHOUT STATIC M I X E R /
with the mass transfer needs should be used so that loss of
/
liquid circulation is minimized. The optimal interval I I K I IIII
between elements depends on the type of fluid. The more 2 ,33 4 5 6789
coalescing the fluid, the .closer must the mixers be
positioned for maximum performance. There is an upper UGr x 102 ( m . s - 1 )
limit to the viscosity that may be used in airlift reactors with
static mixers (Chisti et al., 1990). In excessively viscous Fig 38. Effect of static mixers on gas-liquid volumetric mass
media, just a few mixing elements can Slow down the liquid transfer coefficient in a pseudoplastic carboxymethyl cellulose
so much so that the reactor stagnates. A suitable motionless (CMC) solution (K = 9.55 Pa.s°5) in an external-loop airlift
mixing element, the Sulzer SMV-12, which, limits pressure reactor (Chisti et al., 1990).
drop because of its open structure, is shown in Figure 39.
The element shown can be used in fluids containing a types of internals such a s perforated plates or baffles
certain amount of suspended solids. The location of such (Kawasaki et al., 1994; Kawase, 1994; Yamashita, 1993;
mixers in an external-loop reactor is illustrated in Figure Zhao et al., 1994) offer higher pressure drops, are
40. Adjacent elements are installed 90 ° out of phase. Other susceptible to blockages, and present difficult cleaning

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 90 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Fig 39. The Sulzer SMV-12 static mixer elements (Chisti et al., 1990).
-V !:V
MIXER
L_ v
ELEMENT
t
AIR
Fig 40. Arrangementof static mixers in an external-loop airlift
reactor.
problems in bioreactors that must conform to hygienic
design standards (Chisti, 1992a, b).
Improved productivity of cephalosporin C by
Cephalosporium acremonium was noted by Zhou et al.
(1993) when static mixers were used in an airlift reactor.
Identical mixers had earlier been shown to substantially
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
enhance the gas-liquid volumetric oxygen transfer
coefficient and gas holdup especially in non-newtonian
pseudoplastic media (Chisti et al., 1990), The impact of
static mixers on hydrodynamics was further characterized
by R~ffer et aL (1995). Mostly local values of variables
such as gas holdup were reported (ROffer et al., 1995) and
the data had little direct relevance for reactor design. The
performance of the reactor was not evaluated in the absence
of the mixers, hence no satisfactory comparison could be
done. The mixers used were similar to the ones employed
by Chisti et al. (1990) and Zl~ou et aL (1993).
The impact of static mixers on solid-liquid mass
transfer is discussed in Section 8.2.
12. APPLICATIONS
12.1 Microbial products
Airlift bioreactors have been widely used in microbial
fermentations and biotransfonnations as noted in earlier
reviews (Chisti, 1989; Chisti and Moo-Young, 1987a;
Onken and Weiland, 1983). Some recent representative
examples follow.
12.1.1 Antibiotics
Production of antibiotic neomycin by the filamentous
bacteriuln Streptomycesfradiae was investigated by Park et
al, (1994). A draft-tube internal-loop airlift reactor was
used. A maximum neomycin concentration of 1.176 kg.m3
 Appl Mech Rev vol 51, no 1, January 1998
was obtained in the viscous broth of freely suspended
biomass. Oxygen enriched air (60% vol/vol oxygen) was
used for aeration at a relatively high flow rate of 1.5 vvm.
Addition of cellulose beads (5 mm diameter; 7.5 kg.m"3)
during culture immobilized part of the biomass (ca 40%
immobilized), reduced viscosity, possibly improved oxygen
transfer, and enhanced the final antibiotic concentration by
36%. However, further increase in the proportion of the
immobilized cells by increasing the bead loading led to
reduced production even though the viscosity of the broth
declined (Park et aL, 1994). Apparently, the morphological
changes accompanying immobilization reduced production
(Park et aL, 1994). Solid-liquid mass transfer effects may
also have been responsible.
Methods for large-scale immobilization of Penicillium
chrysogenum on porous Celite beads (5% wt/vol) suspended
in airlift reactors were detailed by Keshavarz et al. (1990).
The Celite (a diatomaceous earth) slurry in soft water was
first sterilized in the reactor (2 h, 121°C); a suspension of
spores was added (20-50 spores per bead) to the cooled
(26°C) fermenter and pneumatic agitation was continued for
two hours during which 70-80% of the spores were taken
up by the porous Celite beads; the slurry was allowed to
settle and the water containing free spores was removed;
finally, fresh water and concentrated sterile medium were
added to initiate the culture (Keshavarz et aL, 1990).
Other examples include production of the polyketide
antibiotic ambruticin S by myxobacterium Sorangium
cellulosum (Hopf et al., 1990); penicillin V by Penicillium
chrysogenum (SchiJgerl, 1990); cephalosporin by
Cephalosporium acremonium (Bayer et aL, 1989; Schiigerl,
1990; Srivastava and Kundu, 1995; Zhou et al., 1993);
erythromycin by Saccharopolyspora erythraea (Pollard et
aL, 1994); and tetracycline by Streptomyces aureofaciens
(Schtigerl, 1990).
12.1.2 Organic acids
Among organic acids, production of itaconic acid from
xylose with Aspergillus terreus mycelia imlnobilized in
polyurethane foam carriers has been reported (Kautola,
1990). Citric acid has been produced using the yeast
Yarrowia lipolytica entrapped in alginate or K-carrageenan
beads suspended in the reactor (Kautola et aL, 1991).
Compared to freely suspended cells, the volumetric
productivity of the immobilized systems was significantly
higher. In immobilized systems, oxygen transfer seemed to
be a limiting factor as a decrease in bead size increased
productivity (Kautola et al., 1991).
Production of lactic acid by continuous fermentation of
deproteinated whey using a mixed culture of yeasts with
Kluyveromyces Jhagilis being the predominant strain, was
described by Pigache et al. (1992). The process productivity
was known to be limited by oxygen transfer. A 120 m3
concentric tube airlift fermenter aerated in the annulus was
used. The vessel had a low overall aspect ratio at ~4.2; the
working aspect ratio was still lower. The overall volumetric
oxygen transfer coefficient values were 0.12-0.28 s~, but
the corresponding aeration rates were not specified (Pigache
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 91
et aL, i992). Based on simulations of oxygen transfer, the
level of dissolved oxygen at the bottom of the riser was
concluded to be lower than further up. No experimental
evidence was presented. The simulations disregarded the
effects of aspect ratio. The latter is known to have a strong
influence on the rate of induced liquid circulation and,
hence, affect the dissolved oxygen profiles (Chisfi, 1989).
Another study described the production D(-)lactic acid
via anaerobic culture of freely suspended and aggregated
cells of Bacillus laevolacticus (de Boer et aL, 1990). Either
pure nitrogen or a nitrogen and carbon dioxide mixture
were used for aeration of the gas-lift device. Production of
glutamic acid by Brevibacterium divaricatum was reported
by Wu and Wu (1992).
12.1.3 Soy sauce
A novel process for continuous production of soy sauce in
airlift bioreactors was described by Hamada et al. (1991).
Conventionally produced moromi was filtered and passed
through a series of four reactors, containing, respectively,
immobilized glutaminase, immobilized bacterium
Pediococcus halophilus, and immobilized yeasts
Zygosaccharomyces rouxii and Candida versatilis which
ocCur in various stages of the conventional mixed culture
moromi fermentation. The last two reactors in the train
were draft-tube sparged airlift vessels aerated,, respectively,
at 0.005 vvm and 0.08 vvm. The yeasts in these reactors
were immobilized in freely suspended calcium alginate
beads having a mean diameter of 2.5 mm. The process was
demonstrated over periods exceeding 100-days of
continuous operation. Compared to the six month long
conventional koji-moromi process, the multi-reactor train
reduced production time to about two weeks. The quality of
the product was comparable to the conventionally produced
material.
In further studies of ethanolic fermentations of moromi
by calcium alginate bead immobilized Z. rouxii, Hamada et
aL (1989) tested three draft-tube sparged (0.08 vvm)
concentric tube airlift reactors (1 L; aspect ratios = 1.45-
3.57). The ethanol content of the product increased with
increasing residence time in the reactors as well as with
increasing aspect ratio. A notable decrease in ethanol
content occurred when pure nitrogen was used for aeration
or when the proportion of nitrogen in the feed gas was
increased (total flow kept at 0.08 vvm). Stable, long-term
continuous fermentations could be carried out in airlift
reactors while producing a product with appropriate alcohol
content and aroma within 28 h compared with 3-4 months
for conventional processes (Hamada et al., 1989).
12.1.4 Biopolymers
Production of biopolymers in airlift reactors has included
synthesis of pullulan using Aureobasidium pullulans (Cai et
al., 1992b); xanthan from Xanthomonas campestris (Suh et
al., 1992); and extracellular polysaccharides by
Acremonium persicinum (Stasinopoulos and Seviour, 1993),
among several others.
 92 SPECIAL
ISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
12.1.5 Enzymes and proteins
The filamentous fungus Phanerochaete choesosporium and
its lignin peroxidases are potentially useful in treating
hazardous wastes such as polychlorinated aromatics in soils
and industrial effluent; other possible applications of this
fungus or its enzymes occur in the paper pulp and forest
industries. P. chrysosporium generally grows as pellets in
submerged culture and bioreactor scaleup has proven
difficult. In a comprehensive review of P. chrysosporium
lignin peroxidase production, Linko (1992) lists the various
bioreactor cultivation methods. In comparison with other
reactor types, airlift bioreactors stand out as generally
permitting higher production of ligninase activity in free
suspension culture (Bonnarme and Jeffries, 1990) as well as
in immobilized packed-bed-loop modes of operation (J~ger
and Wandrey, 1990).
Production of lignin degrading enzymes lignin
peroxidase and manganese peroxidase from P.
chrysosporium in a stirred tank, a bubble column and a
concentric cylinder, draft-tube sparged airlift reactor (A/Ad
= 0.64) was examined by Bonnarme et al. (1993). The two
pneumatically agitated reactors had identical aspect ratios,
at ca 5.5, and aeration rates (Ua = (0.8-2.8) x 103 m.s~).
The maximum lignin peroxidase activity obtained in the
stirred tank was about 16% of that produced in the bubble
column and only 11.5% of that in the airlift reactor.
Similarly, the manganese peroxidase production in the
bubble column was 1.85-fold higher than in the stirred
tank; in the airlift, the production was 3.15-fold greater
(Bonnarme et aL, 1993). The relative values of the enzyme
yield coefficients on glycerol in the stirred tank, the bubble
column and the airlift reactors were in the ratio 1:4.2:5.5
(lignin peroxidase) and 1:1.5:2.1 (manganese peroxidase).
Clearly, the pneumatically agitated reactors had a distinct
advantage over the stirred tank. Between the bubble
column and the airlift reactor, the latter was significantly
better. Culture of P. chrysosporium immobilized on nylon
wire mesh draft-tube in a concentric tube airlift device has
also been detailed (Laugero et al., 1996).
Production of the extracellular enzyme tx-amylase by
Bacillus subtilis immobilized in carrageenan beads (,.--4 mm
diameter) suspended in an airlift device was investigated by
Guo et al. (1990). The volumetric productivity of the
immobilized systems was up to 5.8-fold greater than that of
freely suspended cell suspension. Intraparticle ox3,gen
transfer is likely to have been a limitation in the large beads
used. The data presented by Kautola et al. (1990) did
indicate such a limitation for immobilized yeast cells in
beads of a similar size as used by Guo et al. (1990). Note
that the oxygen transfer requirements for yeasts are
generally lower than for the faster growing bacteria.
Among other proteins, production of the insecticide
tlmringiensin by Bacillus thuringiensis has been reported
(Tzeng and Young, 1996; Jong et al., 1995). Several
studies on producing recombinant proteins using
Escherichia coli in relatively large draft-tube sparged
internal-loop airlift reactors are available (Brandes et al.,
1993a-c; Kracke-Helm et al., 1991).
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
12.1.6 Biomass
Studies of baker's yeast production in airlift bioreactors are
numerous (Fr6hlich et al., 1991b; Ghosh et al., 1994b;
Ghosh and Bhattacharyya, 1993; Pollard et al., 1994; Roca
et aL, 1995; Russell et aL, 1995; Wu and Wu, 1991). Some
of this work was done on a relatively large scale, for
example the 4 m 3 reactor used by FrOhlich et al. (1991b).
Among other yeasts, production of Trichosporon cutaneum
(Yonsel et al., 1991), Kluyveromyces marxianus (Mota and
Teixeira, 1990) and Candida utilis (Ichii et aL, 1993) has
been reported.
An external-loop airlift bioreactor with a
hydrocyclone cell separator integral to tile outer loop was
used by Mota and Teixeira (1990) to select a flocculating
strain of the non-flocculant yeast Kluyveromyces marxianus
by continuous culture at dilution rate of 0.075 h 1.
Apparently, the flocculating cells (larger aggregate size)
were easier to sediment and retain in the reactor while the
free cells washed out. This approach may apply to selecting
other flocculant microorganisms. Batch culture was not
successful in producing a flocculant strain (Mota and
Teixeira, 1990). Continuous culture of a highly flocculating
strain ofSaccharomyces cerevisiae in a small external-loop
airlift reactor has also been described (Roca et al., 1995).
A high performance internal-loop draft-tube airlift
reactor for commercial production of ribonucleic acid
(RNA) form Candida utilis grown on molasses was reported
by Ichii et al. (1993). Culture at high specific growth rates,
achieved by high dilution rates, was essential to enhance
synthesis of RNA. Cultivation at elevated specific growth
rates required a good oxygen transfer capability and high
heat relnoval capacity. The fermenter was sparged in the
draft-tube which contained two perforated baffle plates (10
mm hole size) located near the bottom, across the cross-
section. The overall aspect ratio of the vessel was about 5.6.
The draft-tube occupied a little less than 60% of the height
of the vessel. The riser-to-downcomer cross-sectional area
ratio was 0 . 8 . Banks of vertical heat transfer tubes were
located in the annular downcomer, extending from the base
of the draft-tube to about 60% of its height. After a batch
phase, continuous culture was carried out at a dilution rate
of 0.43 h l (Ichii et at., 1993). The aeration rate was
maintained at 2.4 vvm, or a superficial air velocity (based
on vessel diameter) of 0.3 m.s "l, which corresponded to a
quite low specific power input of about 40 W-m 3. The
fermenter achieved an oxygen transfer rate of 9.9 kg
O=.m-3.h~ and maintained a steady state yeast concentration
of 22.8 kg dry cell.m3. The productivities of biomass and
RNA were 9.79 kg dry cell.m3.h1 and 1.38 kg RNA.m-3-h "1,
respectively (Ichii et al., 1993). These results were
achieved at 145 m 3 (75 m 3 working volume).
12.1.7 Biotransformations
A wide range of biotransformations have been reported in
airlift reactors. Conversion of glycerol to 1,3-propanediol
by Clostridium bu~ricum in reactors up to 2 m 3 was
reported by Gi~nzel et al. (1991), while Roy et al. (1992)
detailed a steroid biotransformation. The latter was carried
 Appl Mech Rev vol 51, no 1, January 1998
out with a Mycobacterium sp NRRL B 3683 in a small
external-loop airlift device. A three phase gas-liquid-
liquid system was used with the cells in the aqueous phase
and the steroid substrate in the water-immiscible organic
phase. Successful biotransformation required attention tO
operational hydrodynamics to ensure complete suspension
of the substrate yet prevent 'wash out' of the organic
dispersed phase into the head region by entrainment in gas
bubbles.
An intensely aerated draft-tube airlift device (50 mL;
6-7 vvm aeration rate) was used by Nishii et al. (1989) for
the biotransformation of dihydrooxoisophorone to 4-
hydroxy-2,2,6-trimethylcyclohexanone by a thennophillic
Bacillus stearothermophilus. Despite the unusually high
aeration rate and the high temperature (50°C), loss of the
substrate or the product by volatilization seem not to have
been a concern.
Lipase-catalysed interesterification of pahn oil
midfraction was reported by Mojovic et al. (1994). A draft-
tube sparged concentric tube airlift reactor (80 mL •working
volume; A,IAd = 0.84; aspect ratio = 3.55) sparged with
nitrogen was used. Rhizopus arrhizus lipase immobilized
by adsorption on Celite particles (20-50 p.m diameter; 720
kg.m 3 density) was suspended in n-hexane in the reactor.
The concentration of immobilized catalyst was low at 2.1%
solids. Complete suspension of solids was achieved at
superficial gas velocities higher than about 0.016 ln.st. Up
to the point of complete suspension, increase in superficial
gas velocity enhanced the rate of interesterification
implying that solid-liquid mass transfer effects were a
limitation in insufficiently agitated environments.
Compared to shake flasks, the rate of reaction was faster in
the airlift device, possibly because of the better mass
transfer characteristics of the airlift. The productivity of the
airlift reactor was 2.8-fold that of shake flasks at a specifc
energy input of only 4% of that of the flasks.
Hydrolysis of penicillin V to 6-aminopenicillanic acid
• using immobilized penicillin acylase pellets (0.6 mm
diameter, 1.5 mm length, cylindrical pellets) suspended in
airlift and stirred tank reactors was examined by Chang et
al. (1993; 1994). Identical rates of hydrolysis were noted in
the standard stirred vessel and the draft,tube airlift reactor
under identical mixing conditions; however, the power
input in the airlift unit was less than one quarter that of the
stirred tank (Chang et al., 1994). Moreover, using
polystyrene pellets that simulated the enzyme catalyst
particles, Chang et al. (1994) showed that over a solids
contents range of 0-30% vol, the specific power input in the
airlift device was not sensitive to the concentration of solids,
whereas that in the stirred tank increased with increasing
amounts of solids even though the agitation speed was held
constant. The airlift reactor not only had a better mixing
efficiency than the stirred tank, the advantage of the airlift
increased with increasing loading of solids.
In the stirred tank, under the normal operating
conditions (two 6-bladed Rushton turbines agitated at a tip
speed of 0.995 m.s~), the enzyme pellets were susceptible to
mechanical damage, even though the impeller speed was
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 93
not particularly high. After 18-days of continuous
operation, turbidity measurements (an indicator of pellet
disintegration) in the stirred vessel were 3.4 times greater
than in the airlift device (Chang et al., 1994). Thus, the
airlift reactor allowed for better mechanical stability of the
catalyst. These results indicate that airlift reactors may
potentially improve the economics of production of 6-
aminopenicillanic acid over conventional commercial
processes that use stirred tank reactors (Chisti and Moo-
Young, 1991).
Bakker et al. (1994) reported on a cascaded airlift unit
for hydrolysis of sucrose to glucose and fructose using
immobilized invertase.
An external-loop gaslift device sparged with carbon
dioxide recycled from the headspace was used by Schorr-
Galindo et al. (1995) to convert extracts of Jerusalem
artichoke to ethanol and fructose enriched syrups. The
conversion was achieved with the yeast Saccharomyces
diastaticus. Yeast flocs developed in continuous culture in
the gaslift device were larger and denser than in a
mechanically agitated vessel. Solid-liquid mass transfer
seemed to have been a limitation.
12.2 Biodegradation of environmental pollutants
12. 2.1 .A4unicipal and industrial wastewaters
The deep shaft technology. The 'deep shaft' wastewater
treatment process is an advanced activated sludge process
that relies on high hydrostatic pressure i n a deep airlift
column t o significantly enhance oxygen transfer. In
comparison with conventional processes, oxygen transfer
rates that are 10-fold greater may be achieved (Winkler,
1981). The transfer rate at peak load is about 1 kg
oxygen.m-3.h1 (Opie, 1993). Several factors combine to
yield this high level of performance, • including long gas-
liquid contact times and intense turbulence in the
circulating fluid with Reynolds numbers of the order of 105
or higher (Winkler, 1981). The shaft is 30-220 m deep
(Winkler, 1981), 0.5-10 m in diameter, and partitioned
vertically into a riser and a downcomer. Air is injected into
the downcomer, about 20-40 m below the surface (Winkler,
1981), except during start-up when the riser is aerated
(Figure 41). To ensure that air bubbles move down the
downcomer, a superficial liquid velocity of 1-2 m.s-1 must
be generated in the downcomer. Because the gas is not
recirculated, the downcomer region above the sparger is free
of bubbles. Oxygen transfer efficiencies of 3-5.5 kg
oxygen/kWh may. be achieved (Winkler, 1981). The
construction features of a 60 m deep, 5.7 in diameter, deep
shaft split-cylinder device have been detailed by Opie
(1993) and Redman (1987) notes additional attributes of
similar systems.
For deep shaft plants, Winkler (1981) cites a BOD
loading of 0.9 kg BOD/kg sludge solids.d. A retention time
of about 1.5-hours has been mentioned for 92% BOD
removal (Winkler, 1981). Volumetric BOD removal is of
the order of 3.7-6.6 kg BOD.m3.d 1 which is generally
associated with high-rate treatment processes. Sewage is
 94 SPECIALISSUE:FLUIDMECHANICSPROBLEMSINBIOTECHNOLOGY Appl Mech Rov vol 51, no 1, January 1998

0.5

O u
2._.gh_o6o__.z .[
f Lr (181)
Kr 4 ( Ar'~2|
OoC
0 ¢
0
o
~0 0
RISER oo ¢ OWNCOMER
o C
which is a modified form of the equation (51) developed by
~ o
0 o Chisti et al. (1988) for the normal mode of operation of
~o o
0 ¢ airlift reactors. In equation (181), ~ r is the gas holdup in
o o
~OoC the upper part of the riser, above the elevation of the gas
Oo c
) o
sparger that is in the downcomer.
o 0
p o While successful, the deep shaft mode of circulation is
0
o o i~AIR
¢ inherently unstable and flow reversal can easily occur.
o
)Oo~ Compared to the conventional airlift mode, the deep shaft
oo c
) o
method is operationally less flexible; the need to prevent
o o
, o flow reversal constrains the range of practicable aeration
o 0

AIR ~
Fig 41. A deep shaft airlift reactor. Aeration conunences in the
riser; once circulation is established, most or all of the air flow is
gradually transferred to the downcomer sparger.
treated without primary sedimentation (Opie, 1993); only
preliminary de-gritting is needed (Winkler, 1981).
Sedimentation of grit at the bottom of the shaft is prevented
by ensuring that the flow velocity at the bottom exceeds 1
m.s4 (Opie, 1993).
Deep shaft plants produce less sludge than
conventional processes. Thus, 0.85 kg sludge solids were
produced per kg BOD removed in one case while a
comparable conventional process produced 1.42 kg sludge
solids per kg BOD removed (Winkler, 1981). The
characteristics of the sludge seem unaffected by cyclic
pressure fluctuations during circulation in the reactor.
Because of reduced sludge generation, high degradation
rates and elimination of the primary sedimentation step, the
land area requirements of deep shaft plants tend to be about
half that of conventional plants with similar capacity
(Winkler, 1981). The deep shaft technology is widely used
in Japan where roughly 76% of the World's deep shaft
plants are located (Opie, 1993).
Deep shaft plants generally operate with complete
separation of the gas and liquid in the head region. The
region of the downcomer that is above the gas sparger is
free of gas (Figure 41). The downcomer gas holdup below
the sparger compensates for an equivalent holdup in the
part of the riser that is below the plane of the downcomer
sparger. Thus, the driving force for circulation of liquid is
provided only by the riser gas holdup that is above the level
of the sparger. For air-water in such a reactor, the liquid
circulation velocity may be estimated with the equation
rates.
Removal of heavy metals from sewage sludge by
microbial solubilization in a continuous flow concentric
draft-tube device (AJAd = 1.25; aspect ratio = 12) operated
in tile deep shaft mode was detailed by Tyagi et aL (1991).
Aeration began in the annular riser, but once the circulation
had established, all or part of the air stream was transferred
to the downcomer:
In addition to the deep shaft, a 'biotower'
configuration is increasingly being used in wastewater
treatment (Figure 42). The biotower units consist of a
relatively shallow pool of liquid with multiple draft-tubes
(downcomers). The towers handle streams with 2-12
kg.m "3 chemical oxygen demand (Foully, 1992). The
biomass sludge produced rises to the top with the bubbles
and is separated in an integral settling zone. Airlift
bioreactors with immobilized biofilms are also used in
wastewater treatment as noted in Section 9.
Sulfide removal. A process for biological removal of sulfide
from wastewater by a mixed culture of sulfide-oxidizing
Thiobacilli suspended in a concentric draft-tube airlift
rector was described by Janssen et aL (1995). The airlift
reactor was aerated by recirculating disengaged gas from
the headspace, with suitable fresh air feed and b l e e d
streams. The aim was to achieve partial oxidation of sulfide
to sulfur for recovery and not to go to the sulfate stage
(Janssen et aL, 1995). Such processes have been put into
commercial use, particularly for desulfurization of flue gas
from thermal power generation plants. Success depends on
preventing the formation of sulfate through attention to
bioreactor design and operation. Similar draft-tube airlift
reactors are being tested for potential use in microbial
desulfurization of coal by the bacterium Sulfolobus
acidocaldarius (Smith and Skidmore, 1990).
Nitrogen removal. Biological nitrogen removal from
wastewater is a two step process. An initial aerobic step
converts the ammonium nitrogen to nitrate which
undergoes denitrification to nitrogen gas in a second,
anaerobic step. Hano et al. (1992) used a draft-tube

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 95

i i~~i~ii~iii!!i~:i~iii

i ~ ; " ....

2
1
P Z

Fig 42. The 'biotower' wastewater treatment unit developed by Krupp Uhde GmbH.

sparged concenti'ic-tube airlift device (A,Z4d = 1.1; aspect
ratio ~ 14; 11.5 x 10-3 m 3) to ingeniously combine these two
steps into a single reactor. The wastewater entered the
annular, gas-free, anaerobic downcomer near the top and
circulated to the aerated riser. The circulation rate of mixed
liquor between the aerobic and anaerobic zones was the
most important operational parameter (Hano et al., 1992).
A hydraulic retention time of about 7 hours was optimal for
maximum nitrogen removal (ca 90% removal). At this
retention time, the optimal aeration rate in the riser was
about 3.3 x 10.6 m3.s -] (Uar = 8.2 x 10.4 m.s'~). Increasing
aeration rate up to the optimum value increased the total
nitrogen removal as the rate of nitrification enhanced.
Further increase in aeration rate, led to aerobic conditions
in tile downcomer, suppressing denitrification (Hano et al.,
1992). At constant retention time (7 h) and aeration rate
(3.3 x 10-6 m3.s~), changes in the liquid circulation velocity,
achieved by altering the bottom clearance of the draft-tube,
revealed an optimum ULr of ca 1.5 x 10.2 m.s ] (Hano et al.,
1992). Ideally, for such a two stage process, the Ar/Ad ratio
and the liquid circulation velocity should be selected to
obtain an optimal combination of residence time split
between the aerobic and anaerobic zones while ensuring
suspension of tile sludge and an oxygen-free downcomer.
Hano et al. (1992) noted that the gas-liquid volumetric
mass transfer coefficient, gas holdup and the liquid
circulation velocity in the activated sludge suspensions were
similar to those for the air-water system so long as the
concentration of the sludge did not exceed 3 kg.m -3
(MLSS). Gas-liquid mass transfer coefficient and gas
holdup declined appreciably when MLSS exceeded 5 kg.m "3
(Hano et al., 1992).
Nitrate removal from wastewater may be done also
with photosynthetic microalga. Alginate bead entrapped
Chlamydomonas reinhardtii was used by Vilchez and Vega
(1995) in airlift reactors. Immobilized biofilm airlift
reactors have also been used (see Section 9).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 96 SPECIAL ISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Recalcitrant organics. A two stage process for biological
treatment of coal tar refinery wastewater was reported by
HOppe et al. (1990). Two airlift reactors were u s e d I a
draft-tube sparged internal-loop unit for the first stage and
an external-loop device for the second. The first stage
operated as an activated sludge unit whereas the second
stage contained suspended sand particles (20-70 kg-m3
concentration) with an immobilized biofihn that was
especially adapted to degrade the recalcitrant organics that
escaped the first reactor (Hiippe et al., 1990). The two
stages had volumes of 1.5 m 3 and 0.16 m 3, respectively.
The respective chemical oxygen demands (COD) in the
inflows were 4.9 kg.m 3 and 1.0 kg.m 3. For both units the
volumetric COD loadings were in the 5-15 kg.m3.d ~ range
(Hiippe et aL, 1990); however, in comparison with the first
reactor, the residence time was much shorter in the second
vessel. For a fixed carrier loading, the oxygen uptake rate
(OUR) in the second stage increased with the concentration
of immobilized biomass until a biomass concentration of
about 12 kg.m 3 when tbe OUR was about 5 mg oxygen.
Ll.min -~. Further increases in tile biomass concentration
did not augment the oxygen uptake rate~ suggesting that
beyond a certain thickness of tile biofihn only a portion of
the biomass was effective in degrading the waste (H0ppe et
al., 1990). The overall COD removal by the two stages
exceeded 85% (H0ppe et aL, 1990).
Activated sludge in a draft-tube sparged concentric
tube airlift reactor has also been used to degrade o-cresol
(de Jong et al., 1991).
Continuous flow degradation of 3-chloroaniline using
Pseudomonas acidovorans CA28 entrapped in calcium
alginate beads (0.9 mm mean diameter) suspended in a
draft-tube sparged concentric tube airlift reactor (A,JAd =
0.23; aspect ratio ~ 2.6) has been reported (Ferschl et al.,
1991). The beads were kept in suspension despite a
relatively low aeration rate (Usr = 0.01 m.s-~).
Degradation of quinoline by Comamonas acidavorans
in a continuous flow draft-tube sparged airlift reactor (A,/Ad
= 0.27; aspect ratio = 5.3; Uo~ = 6.8 × 10"2 m.s "1) was
reported by Ulonska et al. (1995). The microorganism was
grown as a film immobilized on porous glass beads in
suspension (density of wet beads = 1600 kg.m3; biolnass-
free solids loading of 8% vol/vol). High dilution rates were
used initially to aid biofilm formation. Over a broad range
of dilution rates (D = 0.025-0.35 h~), the quinoline
degradation rate and the carbon dioxide production rate
were identical to those in a stirred tank reactor (UIonska et
al., 1995). Although the specific power input data for the
stirred vessel were not given, at 147 W.m 3 the power
demand of the airlift device was probably significantly
'lower.
Heavy metals. Removal of heavy metals by microbial
solubilization from sewage sludge in a continuous flow deep
shaft type of concentric draft-tube airlift device (A,gAd =
1.25; aspect ratio ~ 12) was reported by Tyagi et al. (1991).
The removal efficiency was equivalent to that in a stirred
tank.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
12. 2. 2 TreaOnent o f gaseous effluent
The potential of airlift devices for treating gaseous
pollutants has been pointed out (Moo-Young and Chisti,
1994); two examples of that use have since appeared (Dunn,
1994; Ritchie and Hill, 1995). An external-loop airlift
reactor (12.3 L; A l A n = 1.52) has been used to cleanse a
phenol-saturated air stream (Ritchie and Hill, 1995). The
pollutant level in the exhaust gas was reduced to below
detection. The reactor contained a culture of Pseudomonas
putida which is commonly used to degrade phenol and other
aromatics in wastewaters (Allsop et al., 1993). Ritchie and
Hill (1995) reported degradation rates of 21.5-194 mg.h"~
when the phenol concentration in the inlet gas varied over
650-850 mg.m 3. Potentially faster degradation rates are
likely with internal-loop reactors which allow longer gas-
liquid contact and better mass transfer than external-loop
devices.
Degradation of dichloromethane in an external-loop
airlift reactor was detailed by Dunn (1994). Aeration
velocities ranged over (1.25-5.0) x l0 2 m.s"~. The pollutant
entered the reactor in the gas phase. Data on degradation
kinetics were presented, but the small reactor size limited
the usefulness of the hydrodynamic and oxygen mass
transfer data. Sand panicles (50-100 kg.m "3 solids
concentration; 0.25-0.4 mm particle size) suspended in
water carried the waste degrading biofihn (0.14 kg biomass
per kg sand). The degradation rate depended on the oxygen
transfer performance and other factors.
12.2.3 Solid wastes
Contaminated soil fines can potentially be treated in airlift
bioreactors. Chisti and Moo-Young (1994b) have detailed
possible uses in relnediation of hydrocarbon contaminated
soils.
12.3 Animal cell culture
12.3.1 Suspension culture
Gas bubbles have often been associated with damage to
animal cells (Molina Grima et al., 1997; Papoutsakis,
1991), but the exact cause of damage remains uncertain.
Processes associated with bubble formation and detachment
at the sparger, bubble rupture at the surface, and bubble rise
have variously been claimed to be the causes of damage.
Bubble associated damage may be reduced by controlling
bubble size, aeration rate and by tile use of protective
additives such as serum, tile surfactant Pluronic F-68, or
viscosity enhancers such as dextran. However, the effects of
additives are not always predictable. Thus, Zhang et al.
(1995) observed no protective effect of Pluronic F-68 on
porcine erythrocytes.
Bubble associated damage notwithstanding, most large
scale bioreactors for animal cells are gas sparged. Direct
sparging seems to be the only easily scalable, simple, robust
and proven means of oxwgen supply for large scale culture.
Gas sparged airlift bioreactors are among the biggest
industrial cell culture reactors that have been described
(Arathoon and Birch, 1986; Birch et al., 1987; Rhodes and
 Appl Mech Rev vol 51, no 1, January 1998
Birch, 1988) and continue to be used. Similarly, Chisti
(1993) has employed sparged stirred tank bioreactors for
industrial culture of several hybridomas.
Fluid turbulence has also been associated with cell
damage (Cherry, 1993). Potentially, damage to cells may
occur if the dimensions of the energy dissipating terminal
microeddies found in isotropically turbulent fluids are
smaller than, or of the same size as the suspended cells or
cell-supporting microcarriers. Although substantial
evidence exists for absence of isotropic turbulence in bubble
columns and airlift bioreactors (L(ibbert and Larson, 1990;
L(ibbert et al., 1990; Okada et al., 1993), if turbulence is
assumed to be isotropic, the length scale e of the
microeddies may be calculated with the equation
g=
I ~L 31 3/4
p -Wl/
L/5 )
(182)
where E is tile energy dissipation rate per unit mass of fluid.
For an air-water system aerated at a relatively high (for
animal cell culture) power input of 100 W.m 3 (E = 0.1
W.kgl), the size of terminal eddies works outto about 56
p.m which is significantly larger than the dimensions of a
typical hybridoma at approximately 20 p.m. Consequently,
turbulence in the fluid is unlikely to damage animal cells.
A similar analysis can be used to show that microeddies do
have the potential to damage cells anchored on the surface
of microcarriers that are typically 150-300 p.m in diameter.
However, this damage is not limited to pneumatically
agitated bioreactors; cells in stirred reactors are equally
susceptible. Methods for evaluating shear rates in various
zones of airlift bioreactors have been developed by Molina
Grima et al. (1997) (see Section 10).
Insect cells are generally thought to be more fragile
among animal cells. Successful culture of insect cells in
airlift bioreactors was demonstrated by Maiorella et al.
(1988). A 21 L draft-tube sparged concentric tube vessel
was used to grow Spodoptera frugiperda Sf9 cells in
suspension. The aeration and consequent liquid circulation
rates were such that gas bubbles were not carried into the
downcomer. The bubble size was controlled at 0.5-1.0 cm
by choice of sparger hole diameter (Maiorella et al., 1988);
the aeration rate ranged over 0.02-0.06 vvm. A mixture of
nitrogen, air and oxygen was used for aeration and the
dissolved oxygen level was controlled at 20% of air
saturation. The final cell concentration exceeded 5 × 1 0 6
cells.mLl which was comparable to timt obtained in spinner
flasks. The medium contained 10% fetal calf serum and the
non-ionic surfactant Pluronic F-68 (0.1% wt/vol). The
surfactant was effective in controlling foam; no additional
antifoam was used (Maiorella et al., 1988).
For the Sf9 insect cell line, Murhammer and Goochee
(1988) observed that a small airlift reactor (0.57 L working
volume; aeration rate = 0.0035-0.005 vvm; porous
membrane sparger, < 0.2 ~m pore size) performed as well
as spinner flasks when the culture medium in the reactor
was supplemented with 0.2% Pluronic F-68. In another
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 97
study with the same cells cultured in two airlift bioreactors
(aspect ratios of 4.5 and 7; other details unknown), the
authors noted that supplementation of the medium with
0.2% (wffvol) Pluronic F-68 provided ~11 protection to cells
in one case, but no protection in the other (Murhammer and
Goochee, 1990). The aeration rates and bubble sizes were
comparable in the two reactors (,--0.003 vvm aeration rate
and dB= 0.3-1.0 _mm), but the sparger designs differed.
The reactor sparged through porous stainless steel spargers
(5 p.m or 20 p.m pore size) damaged cells even in the
presence of the Pluronic F-68 polyol. The vessel sparged
through a membrane gas distributor (< 0.2 p.m pore size)
did not produce cell damage when the medium was
supplemented with the polyol. Murhammer and Goochee
(1990) concluded that events near the gas sparger were
responsible for the damage, but the specific nature of such
events was not established. No information was given about
the bottom clearances of the baffle, or draft-tube, from the
base of the reactors.
For the cell line used by Murhammer and Goochee
(1990), Maiorella et al. (1988) had preferred aeration with
much bigger bubbles in pilot scale airlift reactors. In
industrial cell culture bioreactors for hybridomas, Chisti
(1993) too had favored aeration with larger bubbles because
smaller bubbles 'washed' the cells out of tile culture
medium into a stable layer of foam by a froth floatation type
of mechanism. Larger bubbles, because of faster rise
velocities, did. not remove cells from the broth and produced
less stable, easily controlled foam (Chisti, 1993).
Successful culture of a cell line in a certain
configuration of an airlift device does not guarantee that the
same cells would thrive in a different geometry. Thus, Wu
et al. (1990) observed that Spodoptera frugiperda cells
adapted to suspension culture in shake flasks did not survive
aeration with bubbles that were smaller than 500 p.m in
diameter in an external-loop airlift reactor in serum
containing media even when supplemented with Pluronic F-
68. The aeration rates used were relatively high at 0.02-
0.18 m-s1. This report contrasts with other observations
(Murhammer and Goochee, 1990; Maiorella et al., 1988)
for the same cells in larger internal-loop airlift vessels, thus
highlighting the importance of culture methodology,
bioreactor design and operation.
Because growing cells adapt to their hydrodynamic
environment, evaluation of alternative reactor geometries
for culture is ideally done with non-growing cells. Zhang et
al. (1995) employed porcine erythrocytes as model cells that
were incapable of multiplying in vitro; hence, the cells were
not susceptible to adaptation and the measured lysis rates
reflected actual cell lysis as opposed to the net effect of
growth and death processes (Zhang et al., 1995). Survival
of erythrocytes was investigated in a split-cylinder airlift
vessel (A/Ad = 1; aspect ratio = 8.8) agitated at a low power
input (0.34 W.m "3, UGr = 7.0 X 10 -5 m,s-1). Cells were
suspended in citrated isotonic saline at a concentration of
~3 × 106 cells-mL~ which is a typical high value that may
be attained in suspension culture. Supplementation of the
medium with albumin enhanced survival as shown in
 98 SPECIALISSUE: FLUID MECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vo151, no 1, January 1998

1.00C
7 = 120
Dimension
in mm

7
v
z* 0.100
Z
NOADDITIVES /
t.O
tO

0.010 , I , I , I I '
0 40 BO 120 160
TIME (hours)

Fig 43. Fraction of erythrocytes remaining as a function of

exposure time in the airlift vessel (Zhang et al., 1995).
RING - -
S P A R GE R
/
1.000 L

Fig 45. Microcarrier airlift bioreactor of Chen et al. (1994).

0.100
7
v
zo uF 12. 3.2 Microcarrier culture
ucr=7.~;~.~-' ~ \ Microcarrier culture of anchorage-dependent cells has been
Z
oDeI SHAKEFLASKj
l O0RPM
~
claimed to be "impractical" in airlift and bubble column
bioreactors because of alleged accumulation of the carriers
at the air-medium interface (Papoutsakis, 1991). This view
is disproved by our experience and that of others (Dai et aL,
0.001 t i I i I w
0 40 80 120 160 1994). In addition, a hydrodynamic study of microcarriers
TIME (hours) in split-cylinder airlift reactors has shown that
accumulation at the surface is not a problem (Ganzeveld et
Fig 44. Comparison of cell survival in shake flask and airlift al., 1995).
bioreactor for 0.1% albumin-supplemented medium (Zhang et Anchorage dependent Veto cells were cultured by Dai
al., 1995). et aL (1994) in a 'draft-tube sparged airlift reactor (0.2 L;
A / A a = 0.15; aspect ratio = 3). The Cytodex 1
Figure 43. Survival in the airlift vessel was identical to that microcarriers were used at a relatively low concentration of
in a shake flask (100 rpm agitation; surface aeration only) 2 kg.m-3. The medium was supplemented with bovine
for the first 66-hours (Figure 44). Subsequently, the lysis serum (10% vol). In the tested aeration range of _<0.12
was faster in the airlift device (Figure 44); however, the 66- vvm, the cell growth was unaffected by aeration rate (Dai et
hour window was considered sufficient for most practical al., 1994). ComPlete confluence was achieved, but because
culture requirements (Zhang et al., 1995). The data in of the low concentration of carriers, the final cell yield was
Figures 43 and 44 are expressed as normalized cell counts low at approximately 11 x 10s cells.mL4.
relative to the initial value (No). A novel airlift configuration (Figure 45) that is
Suspension culture of hybridomas in airlift bioreactors potentially usefful for anchorage dependent cells was
has been widely reported. As an example, Kwong et al. described by Chen et al. (1994). By careful attention to
(1989) used suspension culture as well as cells immobilized hydrodynamic design the configuration could ensure that
in alginate beads suspended in airlift reactors. A volumetric the cell-supporting microcarriers never came in contact
antibody productivity of 44 mg.d 4.L 4 was reported for the with the aeration gas. The microcarrier-free culture fluid
immobilized cell reactor (Kwong et aL, 1989). In free was aerated in a gas-sparged riser (zone A in Figure 45)
suspension, a high cell concentration of 2 x 1 0 6 cell.n~ 4 and flowed down a gas-free central downcomer (zone B).
was attained, but the viable cell count dropped off faster The oxygen rich flow from the bottom of the downcomer
than in a stirred tank. Consequently, the maximum suspended the microcarriers contained in region C (Figure
concentration of the antibody in airlift suspension culture 45). The flow velocity in zone C was too low to move the
was about 60% of that in the stirred tank (Kwong et aL, carriers out of the zone; hence, carrier-free fluid
1989). recirculated into the sparged riser. For the reactor

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no !, January 1998 Chisti.: Pneumatically mixed bioreactors 99

configuration shown in Figure 45, the volumetric gas-liquid to that observed in roller bottles, but it was higher than in
oxygen transfer coefficient value was 3.38-g26:2 'h ~ when shaki~flasl~'and stirred jars (Hong et al., 1989). The final
the aeration rate ranged over 0..1-0.75 vvm, indicating cell concentration in the airlift device was ~11 kg.m 3 (dry
sufficient oxygen transfer capability for supporting cell wt), again comparing well with the best case result at ~13
concentrations of the order of 10 6 cells.mL~ (Chen et aL, kg.m -3 in a roller bottle.
1994). The corresponding liquid velocities in zone C varied Production of benzophenanthfidine alkaloids in
over 0.05-6.15 cm.sl - s u f f i c i e n t to suspend the suspension cultures of Eschscholtzia californica in
microcarriers having a density of 1050 kg.m -3 and a multiphase airlift fermenters was reported by Byun and
diameter of 150-620 ~m (Chen et al., 1994). Pedersen (1994). The alkaloid was continuously extracted
An extensive study of animal cell microcarrier into a dispersed silicone oil phase. Compared to the two-
suspensions in two split-cylinder airlift reactors was phase gas-liquid system in the same reactor, the three-
detailed by Ganzeveld et al. (1995). The liquid circulation phase gas-liquid-liquid system had up to 40% greater
velocity and the gas holdup declined with increasing productivity (Byun and Pedersen, 1994). Despite potential
loading of carriers (ps = 1030-1050 kg.m3; 150-300 ~m applications in extractive fermentations and other similar
carrier diameter; 0-30 kg.m -3 solids loading). The mixing areas, engineering characterization of airlift reactors having
time increased with increasing solids loading. The carriers two immiscible liquid phases has not been done in any
roughly followed the fluid eddies and, as long as the flow detail.
was laminar (Re _< 3200), there was little interaction 12. 4. 2 bnmobilized cell culture
between particles or between solids and the walls of the
vessel (Ganzeveld et al., 1995). Thus, in laminar flow the Airlift bioreactors have been used successfully for
cells are unlikely to be damaged 1V bead-to-bead or bead- immobilized culture of plant cells. Culture of Catharanthus
to-wall collisions. The rate of circulation of the roseus cells immobilized by adhesion to the surface of
microcarfier suspension agreed closely with the predictions nonwoven, spiral wound, polyester sheet matrix held
of equation (52) of Chisti et al. (1988) even though the stationary in the downcomer of an airlift reactor has been
equation was developed for solids-free media. Tile equation described by Archambault et al. (1990). A small riser-to--
indirectly took into account the effect of solids through gas downcomer cross-sectional area ratio (A/Ad = 0.03) was
holdup which was affected by the solids concentration used, dictated apparently by the need for a larger
(Ganzeveld et al., 1995). The geometric details of the downcomer to maximize the surface area of the
reactor were: A / A d = 1; aspect ratio = 7.6 and 14.5; UOr = immobilizing matrix. Such a configuration may not sustain
(0-6.7) X 10.3 m.s -] corresponding to power inputs of 0-33 adequate circulation or oxygen supply upon scaleup.
A multiphase external-loop airlift reactor (250 mL;
W.m-3 (Ganzeveld et al., 1995). Data on shear rates in the
A,Z4d = 2.5; aspect ratio = 10) for potential use in culture of
reactors used by Ganzeveld et aL (1995) were reported by
immobilized plant cells was investigated by Sajc et al.
Molina Grima et al. (1997). The latter also reviewed the
(1995b). Calcium alginate beads (pp = 1070 kg.m-3;
effects of hydrodynamic forces on cells.
equivalent diameter = 2.5 mm; solids holdup = 0-0.5) were
suspended in Murashige and Skoog medium. A water
12.4 Plant cell culture
immiscible silicon oil (density = 960 kg.m3; viscosity =
Suspension culture of plant cells in airlift and bubble 0.05 Pa.s; surface tension = 22 x 10.3 N.m ~) or n-
column bioreactors has been proven repeatedly (B0ch et al., hexadecane (density = 773 kg.m-3; viscosity = 0.003 Pa.s;
1991; Doran, 1993; Hegarty et al., 1986; Husemann, 1982; surface tension = 25 x 10.3 N.ln ~) were continuously added
Kato et al., 1976; Moo-Young and Chisti, 1988; Reinhard to the reactor at the base of tile riser and withdrawn as
et al., 1989; Scragg et al., 1989; Smart and Fowler, 1984; separated phases from the top of the downcomer. The
Spieler et al., 1985; Townsley and Webster, 1983). More superficial air velocity in the riser ranged over (2-12) x 10-3
recently, hairy roots (Rodriguez-Mendiola et al., 1991; m.s"~, but beyond a Uor of about 6 x 10-3 m.s ~ the dispersed
Sharp and Doran, 1990; Whitney, 1991) and immobilized phase droplets recirculated through the downcomer.
cells (Archambault et al., 1990; Sajc et al., 1995a) have Because recirculation was not wanted, the gas velocity was
also been cultivated. Some recent examples follow. restricted to 6 x 10.3 m.s 1 (corresponding to a U~ of ,-0.8
m.s "l) in most experiments. The size of dispersed phase
12.4.1 Suspension culture droplets decreased with. increasing gas velocity until at a
During suspension culture of strawberry cells in a small Uor of about 6 × 10 -3 m.s~ a droplet size of ,-.4 mm was
split-cylinder airlift vessel and other devices, Hong et al. attained (Sajc et aL, 1995b). Further increase in gas
(1989) noted that, at 0.44, the cell yield coefficient on velocity did not affect the size of drops.
carbohydrate was higher in the airlift bioreactor than in The gas holdup in air--culture medium-silicon oil
other systems. Cells were inoculated at ~2 kg.m 3 (dry wt) system was almost the same as in oil-free system. Addition
in nutrient supplemented Murashige and Skoog medium. of alginate beads reduced holdup which declined with
Aeration rate ranged over 0.09-0.36 vvm, increasing as the increasing concentration of solids. The mixing
cell concentration increased. The average specific cell performance of the vessel as indicated by Peclet numbers
growth rate of 0.14 d z in the airlift device was comparable was not affected by the solids loading (Sajc et aL, 1995b).

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 100 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
The overall volumetric mass transfer coefficient (kLaz)
values in air-water and air-culture medium (both at a
constant silicon oil flow rate of 4.0 x 104 m.s-~) were about
the same, but declined by 5-10% with increasing
concentration of solids (solids holdup = 0-0.5). The
volumetric liquid-liquid mass transfer coefficient for
transfer of a plant secondary metabolite (1,8-
dihydroxyanthraquinone or danthron) from the aqueous
medium to the silicon oil droplets was 1.25 x 10~ s~ at a
constant oil superficial velocity of 4.0 x 104 m.s~ (Sajc et
al., 1995b). The coefficient was virtually independent of
the superficial gas velocity, the induced liquid circulation
velocity and t h e loading of alginate beads (Sajc et al.,
1995b). Apparently, the transfer coefficient was controlled
by the aqueous fihn, but because the drop size was ahnost
constant and the reactor environment was fairly quiescent,
the density difference between the droplets and the
suspending fluid determined the value of the fihn
coefficient. In this situation the volumetric coefficient is
expected to depend only on the specific interfacial area of
the droplets. Apparently, the interfacial area remained
constant because the flow rate of the dispersed phase as well
as the drop size were fixed.
The reactor was eventually used to culture alginate
bead immobilized cells ofFrangula alnus for the production
of anthraquinones (Sajc et al., 1995a). The beads were
suspended in Murashige and Skoog medium aerated at 0.7
vvm. The product was continuously extracted with the
silicon oil (flow rate = 0.05 vvm) or n-hexadecane.
Compared to shake flasks that contained the oil but were
not continuously extracted, the cells in the airlift vessel had
10-30-fold higher productivity (Sajc et al., 1995a). Based
on estimations of effectiveness factors, neither intraparticle
diffusion nor solid-liquid mass transfer limited the
production rate (Sajc et al., 1995a ).
12. 4.3 Hairy root culture
An important development in plant cell culture technology
has been the ability to transform the cells with the soil
bacterium Agrobacterium rhizogenes to produce cultures
that grow stably as hairy roots. Engineering of bioreactors
for these cultures poses new challenges because the roots
grow as an interconnected network that cannot withstand
mechanical agitation. Bubble columns and airlift reactors
are well suited to this type of culture and have been
successfully employed. As the biomass grows and fills up
the reactor, mixing deteriorates, particularly in bubble
columns. In airlift devices, where the biomass can be
confined to either the riser or the downcomer, better
nutrient supply can be maintained for longer periods.
Whitney (1991) presents an example of hairy root culture of
Nicotiana tabacum in a split-cylinder airlift device aerated
at 0.3 vvm. A 'mist bioreactor' in which the root mass is
held unsubmerged in an aerated chamber and sprayed with
a mist of recirculating nutrient medium (Whitney, 1991) is
a possible alternative to submerged culture in airlift
systems. Comparative evaluations are lacking to establish a
preferred culture methodology. A properly designed
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
submerged culture system is expected to outperform the
mist device in maintaining a uniformly good supply of
nutrients to the entire root mass.
Hairy root culture of Trigonella foenum-graceum and
Nicotiana rustica in two draft-tube sparged concentric tube
airlift reactors was reported by Rodriguez-Mendiola et al.
(1991). The two reactors were geometrically identical (9 L
nominal; aspect ratio = 6.4; A,gAd = 0.8; aerated at 0.1 vvm)
except that the draft-tube in the culture vessel used for N.
rustica was unconventional in being made of nylon mesh.
No attempts were made to confine growth to either the riser
or the downcomer. In the solid draft-tube configuration,
root growth occurred mostly in the annulus where the
centering arms on the draft-tube provided sites for initial
attachment of the rootlet in the inoculum. Flow of the
medium through the root network did not pose a limitation
during eight weeks of culture (Rodriguez-Mendiola et al.,
1991). Because the nylon mesh tube offered many more
attachment points, and allowed unconfined growth in all
directions from these points, growth occurred throughout
the volume of the reactor (Rodriguez-Mendiola et aL,
1991). Better volume utilization and even distribution of
biomass was realized in that system. At harvest, the nylon
mesh was fully embedded in biomass. Tiros, inexpensive,
disposable mesh tubes may be necessary for routine use.
Alternatively, more durable stainless steel mesh structures
may be employed. The latter may be reused after cutting off
the root mass.
Production of atropine by hairy root culture of Atropa
belladonna in shake flasks and a concentric draft-tube
airlift device (2.5 L) has been compared (Sharp and Doran,
1990). The level of atropine in the reactor grown biomass
was -1.5-fold that in the shake flask grown roots; however,
the alkaloid and the biomass yields on sugar were poor in
the reactor. The inferior performance of the airlift reactor
was associated with poor mass transfer of nutrients to the
highly entangled, high density root mass. Mass transfer
limitations were suggested by the observed development of
necrotic tissue within the root mass (Sharp and Doran,
1990). The low aeration rate used (0.08 vvm) probably
contributed to insufficient mass transfer; moreover the
geometry of the reactor had not been optimized to attain
high levels of liquid circulation. Design of large scale
airlift devices for hairy root culture would require a
knowledge of how the root network affects the circulation.
Beds of compressed polymer fibers or metal wire may be
used to simulate the root system in such studies.
12.5 Algal culture
Culture of the marine microalga Dunaliella tertiolecta in
artificial sea water in a bubble column and airlift reactor
was reported by Suzuki et al. (1995). Tile bubble column
had a rectangular cross-section (20 x 50 mm; 350 mm tall)
and could be converted to the airlift configuration by
inserting a vertical baffle (20 × 130 mm) with a 30 mm
clearance from the bottom. Under optimal conditions
(30°C; 15 klux light intensity; 6% mol carbon dioxide in
air), no growth was observed in the airlift reactor whereas
 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors ' 101

hydrodynamic smoothing of the upper and lower parts of

3 the baffle to prevent flow separation. In tile bubble column,
the algal growth was sensitive to aeration rate, increasing
2 with superficial gas velocity until a velocity of about 0.6
m.min~ (Figure 47). Further increase in aeration rate
reduced growth, apparently because of hydrodynamic
I 10 - I stresses in the fluid (Suzuki et al., 1995). Under non-
E growth conditions (no light), the specific death rate in the
7
bubble column was shown to increase with superficial gas
velocity for velocities exceeding 0.6 m.min-~ (Suzuki et aL,
< 4 1995). At a fixed aeration velocity (Uo = 1 m.min~), the
0
GO specific death rate decreased with increasing height of the
culture fluid in the column (Suzuki et al., 1995), probably
because the specific power input and, hence, the turbulence
AIRLIFT BIOREACTOR intensity declined. [For similar findings with animal cell
culture see Molina Grima et al. (1997).] For otherwise fixed
10 -2 I I I conditions, increasing the number of aeration nozzles--that
.0 25 50 75 100 is, decreasing the gas jet velocity---caused a marginal
reduction in the specific death rate. This led to the
TIME (h)
conclusion that the events at the gas sparger did contribute
to cell death (Suzuki et aL, 1995), although the effect was
Fig 46. Growth of D. tertiolecta in a bubble colmml (circular minimal.. Unlike Suzuki et al. (1995), Molina Grima et al.
symbols) and an airlift bioreactor under optimal conditions: 30°C; (1994; 1995) were quite successful in culturing microalga in
15 klux light intensity; and 6% vol carbon dioxide in aeration pilot scale airlift devices (see Section 4.4). Continuous
gas. Superficial gas velocities (based on outer column diameter): culture of the green alga Chlorella pyrenoidosa in a
0.2 m.minJ ( o , A ) and 0.4 m.minl ( e , • ). From Suzuki et aL
(1995). rectangular airlift device with multiple light-emitting
diodes was reported by Matthijs et al. (1996).

12.6 Chemical processes

Pneumatically agitated reactors are suitable for many
4 multiphase contacting chemical processes. Some of the
early applications were noted by Chisti (1989) and others
(Onken and Weiland, 1983; Steiner, 1987). As a recent
'E 2 example, Dedek (1990) studied the catalytic hydrogenation
of sunflower oil in a small draft-tube sparged concentric
m 10 -1 tube airlift reactor (1 L; A,/Aa = 0.8; aspect ratio --6). The
m
< 7 flow regime, whether bubble flow or slug flow, strongly
_o affected the rate of hydrogenation. The reaction rate
4 [] o.1 increased with increasing hydrogen flow rate over 0.5-2
~-/-------•~ • 0.2 vvm, but higher flow rates, up to 4 vvm, reduced the rate of
22I A
0
•
0.5
0.4
0.6
reaction (Dedek, 1990). Apparently, coalescence of gas
bubbles at high flow rates reduced the specific gas-liquid
10- I I I I I interfacial area. The conversion efficiency of the airlift
0 25 . 50 75 100 125 150 device was comparable to that of the mechanically agitated
systems tested so long as the latter were operated within the
TIME (h)
rotational speed limits (< 13 s~) that are encountered in
Fig 47. Growth ofD. tertiolecta in a bubble colunm (30°C; 15 larger industrial scale devices (Dedek, 1990).
klux light intensity; and 6% vol carbon dioxide in aeration gas). The examples noted here highlight only a selection of
From Suzuki et al. (1995). the many possible applications of airlift reactors in
bioprocessing, environmental remediation and the chemical
industry.
good growth occurred in tile bubble column (Figure 46).
Microscopic examination showed significant disruption of
the algal cells in the airlift device (Suzuki et al., 1995). 13. COMPARISON OF PROCESS PERFORMANCE
This was associated with the hydrodynamic stresses Many of tile specific performance characteristics of airlift
generated as the culture flowed over the upper edge of the bioreactors and other types of devices have been compared
baffle into the downcomer (Suzuki et al., 1995). This effect elsewhere in this work as well as in other reports (Chisti,
could have been avoided, or at least minimized, by 1989; Chisti and Moo-Young, 1987a; Choi et al., 1996;

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 102 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Immich and Onken, 1992; Mao et al., 1992; Ouyoung et
aL, 1989). This section deals with comparison of the
overall process performance in some representative
applications.
In a careful comparative study of stirred tanks and
draft-tube internal-loop airlift reactors Hopf et al. (1990)
confirmed identical biomass and antibiotic production in the
two types of systems for cultures of the myxobacterium
Sorangium cellulosum producing the polyketide antibiotic
ambruticin S. Identical performance occurred only so long
as the dissolved oxygen levels in the two systems remained
above ca 40% of saturation. Overall, the airlift devices
were superior because of lower power consumption, reduced
contamination risk and less likelihood of mechanical
failure. Also, the capital expenditure on an airlift device is
expected to be lower than on a stirred vessel of equal
volume.
In another investigation, using cultures of strictly
aerobic yeast Trichosporon cutaneum, identical biomass
production rateswere observed in airlift and stirred tank
reactors when the oxygen transfer rates were identical; once
again, the airlift system attained the requisite levels of
oxygen transfer at lower power inputs (Yonsel et aL, 1991).
Thus, over a range of oxygen transfer rates (0.7-2.7 kg
oxygen transferred m3.h~), the oxygen transfer efficiency of
the airlift reactors was nearly constant at ca 2 kg O:/kWh,
whereas that of the stirred tank declined form a high of
about 1 kg O2/kWh to 0.5 kg O2/kWh (Yonsel et aL, 1991).
Identical performance of stirred tanks and airlift
systems was noted also in anaerobic cultm-e of Clostridium
butyricum during production of 1,3-propanediol from
glycerol in reactors up to 2 m 3 (Giinzel et aL, 1991).
Reporting on extracellular polysaccharide production
by the filamentous fungus Acremonium persicinum,
Stasinopoulos and Seviour (1993) observed that biopolymer
production levels in Rushton-turbine stirred tanks were
approximately 15% of those found in shake flasks or in an
airlift fermenter when the agitator tip speed in the stirred
tanks were 0.46-0.92 m.s 1. For a reduced tip speed of 0.18
m.s ~, the polymer production in stirred tanks was
comparable to that in shake flasks; however, this low
agitation speed led to poor mixing causing sampling
problems and an inability to control the level of dissolved
oxygen.
In evaluating reactors for possible use in highly
viscous non-newtonian fermentations for producing
pullulan using Aureobasidium pullulans, Cai et al. (1992b)
determined that a draft-tube sparged internal-loop airlift
device with the draft-tube split into short sections was the
most suitable reactor. Splitting the draft-tube improved
communication between the riser and downcomer, reducing
the likelihood of oxygen depletion in the unaerated
downcomer. Performance of this 3 m 3 airlift reactor was
compared with an equal volume stirred vessel agitated by
two Rushton turbines at 120-180 rpm. For identical
aeration conditions in the two systems, the substrate
conversion in the airlift reactor was about 45% higher but
the power consumption was only about 64% of that in the
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
stirred tank. In tile airlift device the specific power input
ranged over 1.6-2.4 kW.m 3. The aeration rates varied over
0.6-0.95 vvm. The high power inputs were a reflection of
the highly viscous broth; the power law parameters K and n
for the pseud0plastic broth were 7.5 Pa.s" and 0.4,
respectively, toward the end of the 100-hour batch
fermentation (Cai et al., 1992b).
Upon comparing the performance of stirred tank and
airlift fermenters in production of penicillin V by
Penicillium chrysogenum, cephalosporin by
Cephalosporium acremonium and tetracycline by
Streptomyces aureofaciens, Schiigerl (1990) concluded that
airlift reactors were superior for all three fermentations.
However, in all eases, the processes needed to be adapted to
airlift reactors in terms of inoculation strategies, selection of
conditions, media and strains for the preferred growth
morphology (Schiigerl, 1990). In the stirred fermenters,
cell mass concentration and the volumetric productivity
tended to be higher than in airlift vessels, but the power
requirements were extremely high at 3-5 kW.m "3. Airlift
reactors were better in terms of oxygen transfer efficiency,
and specific productivities with respect to substrate input,
power input and oxygen consumption (Schiigerl, 1990).
The product yield coefficients based on substrate and on
oxygen were also higher in the airlift reactors. Overall the
airlift devices were better, because for these fermentations
the major costs were due to raw materials (for example, 50-
60% of the final cost of product) and power consumption
(for example, 10-20%); thus, high specific productivities
were preferable to high volumetric productivity (Schiigerl,
1990).
In producing cephalosporin C by highly viscous broths
of the mold C. acremonium in an external-loop airlift
reactor (60 L; aspect ratio based on riser = 18.3; A/Ad =
9.1), Bayer et aL (1989) reported a final antibiotic
concentration of 5 kg.m 3 which was about 70% of what had
been obtained in a large stirred fermenter. The poor
performance of the airlift device was probably because the
processes had not been adapted to the reactor. Another
study, using a draft-tube sparged internal-loop airlift
reactor, reported the final yield of cephalosporin C at ~80%
of that in a stirred fermenter (Srivastava and Kundu, 1995).
In production of thuringiensin from Bacillus
thuringiensis, an airlift device with a wire mesh draft-tube
gave twice the yield (2.2 kg.m 3 thuringiensin) of a stirred
reactor (Tzeng and Young, 1996).
For removal of heavy metals from sewage sludge by
microbial solubilization, Tyagi et aL (1991) compared the
deep shaft operation of a concentric draft-tube airlift device
with stirred tank reactors. The removal efficiencies were
equivalent for the two types of reactors. Although Tyagi et
al. (1991) were able to maintain a stable circulation of
liquid, this type of operation is inherently unstable and flow
reversal can easily occur.
In conclusion, suitably designed airlift bioreactors
perform as well as or better than conventional stirred
fermenters. Airlift reactors invariably use less power, are
less susceptible to Contamination and mechanical failure.
 Appl Mech Rev vol 51, no 1, January 1998
14. OTHER CONSIDERATIONS
To ensure satisfactory performance of bioreactors, attention
to transport phenomena, hydrodynamics and hydraulics
must be combined with consideration of other essential
factors. Most food and biopharmaceutieal processing
requires that the equipment comply with the standards of
Good Manufacturing Practices. In some cases, the design of
the reactor may require special primary containment
features consistent with the biosafety-level classification of
the process. The principal design considerations for
contamination free operation are discussed in the following
sections.
14.1 Sterile processing
Bioreactors other than those used in waste treatment are
typically designed for sterile operation. Industrial
bioreactors for sterile processing are normally designed as
pressure vessels capable of sterilization with saturated steam
at 16 psig (Chisti, 1992a). In most instances, the vessel is
designed for a maximum allowable working pressure of 40-
45 psig at a design temperature of 150-180°C. The vessel
is designed to withstand full vacuum (Chisti, 1992a). In
North America, the design conforms to ASME, Section
VIII, Division 1, Boiler and Pressure Vessel Code. Other
codes may be accepted in other jurisdictions. The vessel is
hydrostatically tested to 150% of maximum working
pressure by the builder (Chisti, 1992a). The vessels are
invariably jacketed, and unless there is some special need,
the jacket is designed to the same conditions as the vessel.
Chloride-free fiberglass insulation, fully enclosed in a
protective shroud is applied to the jacket. The jacket and
the reactor are provided with over pressure protection: a
relief valve on the jacket or its associated piping, a bursting
disc on the bioreactor. Usually, AISI Type 316L stainless
steel or graphite burst disc is used. The latter type has the
advantage that it does not develop pinholes or cracks
without complete failure (Chisti, 1992a). Thus, sterility is
less likely to be compromised with graphite discs.
The vessel should be designed to drain fully. It should
have a minimum number of ports, nozzles, connections, and
other attachments consistent with the current and
anticipated future needs of the process (Chisti, 1992a). The
bioreactor should be free of crevices and stagnant areas
where pockets of liquid and solids may accumulate. The
vessel should have few internal. Steam connections should
allow for complete displacement of air pockets. Other
essential design features for sterilization and sterile
operation have been treated by Chisti (1992b).
During cleaning, removable spray balls (or nozzles)
are inserted into the vessel through one of the ports on the
head plate. Permanently installed spray heads are not
recommended because of potential difficulties with
sterilization (Chisti and Moo-Young, 1994d). The spray of
cleaning solutions should reach all part of the vessel; hence,
careful design of static spray balls is essential. Better
coverage of the tank is possible with rotating spray nozzles.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 103
Additional details on design for cleaning-in-place have
been given by Chisti and Moo-Young (1994d).
14.2 Materials
Materials of construction, must withstand the physico-
chemical conditions encountered during cleaning,
sterilization and processing. The materials that come in
contact with the process fluids should be non-reactive, non-
additive, non-absorptive, and durable. For a great majority
of applications, austenitic stainless steels are the preferred
material of construction for bioreactors. The bioreactor
vessel is usually made in Type 316L stainless steel while the
less expensive Type 304 (or 304L) is used for the jacket, the
insulation shroud, and other non-product contacting
surfaces (Chisti, 1992a). The L grades of stainless steel
contain less than 0.03% carbon, which reduces chromium
carbide formation during welding and lowers the potential
for later intragranular corrosion at the welds. Cold-rolled
steel sheets should be used in vessel making for ease of
polishing later.
14.3 Surface finish
The finish on surfaces that come in contact with the process
material and, to some extent, the finish on external surfaces
affect the ability to clean, sanitize and sterilize the
bioreactor and the general processing area (Chisti, 1992a).
The surface finish has implication on stability and reactivity
of the surface, and it may have process implications relating
to microbial or animal cell adhesion to surfaces.
The surface may be polished mechanically or by a
combination of mechanical and electrolytic processes. The
mechanical polish is achieved by abrasive action of a
sandpaper type material on metal. The surface finish may
be specified by grit number, for example, 240 grit polish,
which refers to the quantity of particles per square inch of
the abrasive pad or belt (Chisti, 1992a). The higher the grit
number, the smoother the finish. More quantitative
measures of surface profiles in terms of "arithmetic mean
roughness," (Ra), and "root mean square roughness," (Rq),
are now coming into use~ From a bioprocessing point of
view, i t should be noted that even a 320 grit polished
surface will have an Ra of the order of 0.23-0.30 Ixm.
Clearly, such a surface could provide many potential sites
for attachment to microbes and other microscopic particles.
Hence, for internal surfaces of bioreactors, electropolished
surface is preferable to mechanical polish alone (Chisti,
1992a).
Electropolishing is an electrolytic process which
preferentially removes the sharp microscopic surface
projections arising from mechanical polishing; the result is
a much smoother finish. Electropolishing significantly
reduces the metal surface areas, and hence the product-
metal contact area. The treatment imparts corrosion
resistance to stainless steel by removing the microscopic
regions of high local stress; it creates a passivated steel
surface, rich in protective chromium oxide (Chisti, 1992a).
 104 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

The bioreactor surface specifications should state the 16. NOMENCLATURE

level of mechanical polish needed prior to the electropolish A Cross-sectional area of column or riser and
step (Chisti, 1992a). Scanning electron micrographs o f downcomer (m2)
electropolished surfaces reveal that there is little advantage Ab Cross-sectional area for flow under baffle or draft-
to starting with a much better than 220 grit (Ra about 0.4- tube (m2)
0.5 p.m) polished surface. If mechanical polish alone must Ao Effective cross-sectional area for flow in entry region
be used, it should be at least 240 grit (Chisti, 1992a), and of downcomer (m2)
Ad Cross-sectional area of downcomer (m2)
the direction of polish should be controlled to produce a
Aae Cross-sectional area of expanded entrance of
vertical grain for good drainage (Chisti, 1992a). The downcomer (m2)
surface should receive a nitric acid wash treatment as a AF Free area for liquid flow at the location of sparger
minimum. The orientation of grain does not seem to be of (m2)
consequence if the surface is to be electropolished (Chisti, An Area for heat transfer (m2)
1992a). ALa Total gas-liquid interfacial area in downcomer (m2)
Electropolishing removes surface impurities from Az~ Total gas-liquid interfacial area in riser (ms)
steel. In the region of impurity, the surface may develop An Effective cross-sectional area for flow in entry region
such flaws as cavities and grooves from where the impurity of riser (m2)
has been dissolved away. For superior surface with few Ar Cross-sectional area of riser (m2)
flaws, pure stainless steel should be used in construction. a Constant in equation (10) (mb.Sb); exponent in
equations (133) and (136) (-); specific gas--liquid
Thus, Type 316L from a consumable electrode or
interfacial area (basis unknown, m"l)
electroslag removal process is preferred (Chisti, 1992a). aD Gas-liquid interfacial area per unit volume of
The welds on the internal parts of the bioreactor dispersion (m"1)
should be ground flush with the internal surface, and az Gas-liquid interfacial area per unit liquid volume
polished in the same way as the rest of the vessel. Welds (m"l)
are difficult to notice in high-quality construction (Chisti, az~ Gas-liquid interfacial area per unit liquid volume in
1992a). bubble column (ml)
as Total solid-liquid interfacial area (m2)
B Unbaffied height in the head zone of internal-loop
15. CONCLUDING REMARKS reactors (m)
b Exponent (-)
Pneumatically agitated bioreactors, especially the airlift BOD Biochemicaloxygen demand
designs, have proven effective in numerous kinds of Cl Constant in equation (151) (-)
bioprocessing. Compared to the more conventional stirred Ccrit Critical concentration of oxygen (mmol.m"3)
fermenters, airlift devices typically require less power for a C~ Oxygen concentration at inlet of bed (mmol.m3)
given performance, are more mechanically robust and less CL Instantaneous concentration of transferring
susceptible to contamination. The airlift technology has component in liquid (kg-m"3)
been demonstrated at extremely large scale and continues to CMC Carboxymethyl cellulose
displace the stirred bioreactors. However, pneumatically COD Chemical oxygen demand
agitated bioreactors are less flexible and unforgiving of Cp Specific heat capacity (J.kgq.°C"l)
design flaws. Stirred tanks, though inefficient, expensive Cs Concentration of solids in slurry (kg.m3)
Cr Dimensionless tracer concentration (-)
and failure prone, have the agitation speed and aeration rate
C~ Tracer concentration at time t (kg.m"3)
as two independent major operational degrees of freedom.
C,~ Final equilibrium concentration of tracer (kg.m"3)
Furthermore, if a performance problem is encountered, the C* Saturation concentration of transferring gas or solute
stirrer motor and the impeller type, number, location or in liquid (kg-m"3)
diameter are relatively easily modified in attempts to correct Parameter defined by equation (72)
it. The pneumatically agitated bioreactors have no such c Constant in equations (14), (133) and (136) (-)
modifiable features and, therefore, there is little tolerance cl Concentration of component 1 in liquid (mol.m3)
for design shortcomings. As discussed in this review, the Dc Diameter of the 6onnecting pipe (m)
knowledge base for design and operation of pneumatic DL Diffusivity of gas or solute in liquid (mLs"1)
bioreactors is substantial, yet much remains to be learnt. d Hydraulic diameter of pipe (m)
How the many geometric design variables and operational ds Sauter mean bubble diameter (m)
parameters affect mass and heat transfer, mixing and shear dsd Mean bubble diameter in downcomer (m)
rates, and gas holdup is not fully established. Aspects of ds~ Diameter of ith bubble (m)
dBr Mean bubble diameter in riser (m)
hydraulics are better understood, but uncertainties remain.
dc Diameter of cell (m)
With some exceptions, design and scaleup continue to dj Diameter of draft-tube (m)
require pilot plant studies. Nevertheless, experience do Diameter of column or outer tube (m)
suggests that reliable design is generally feasible based on a dp Diameter of particle (m)
combination of the available procedures and suitable pilot dr Diameter of the riser (m)
plant data. Therefore, the trend toward greater use of the E Energy dissipation rate per unit mass (W.kg"3)
airlift technology is likely to persist.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998 Chisti: Pneumatically mixed bioreactors 105

E~ Energy dissipation due to fluid turnaround at bottom No Initial cell concentration (cells.m"3)
(W) N, Total number of cells in reactor (-)
E~ Energy dissipation due to wall-friction (W) n Flow behavior index (-)
Efficiency of airlift pump (-) n~ Frequency of occurrence of bubble with diameter dBt
Ej Energy input rate' (W) (-)
Overall axial dispersion coefficient of liquid (mZs"1) nr Number of ideally mixed tanks in series (-)
E~ Axial dispersion coefficient of liquid in downcomer OUR Oxygen uptake rate
(mLsl) P Pressure at the point of entry of the gas (Pa)
ELf Axial dispersion coefficient of liquid in riser (m2.s"l) Pa Atmospheric pressure (Pa)
Ee Energy dissipation rate in the packed bed (W) Pe Peclet number (-)
ER Energy dissipation due to bubbles in riser (W) Po Power input due to isothermal expansion of gas (W)
Er Energy dissipation due to fluid turnaround at top (W) PHE Plate heat exchanger
F Ratio of liquid volume in unaerated head region to Ph Pressure in the headspace (Pa)
volume in loop (-) P Dimensionless distance (-)
Fr Froude number (-) AP Pressure drop (Pa)
f Exponents in equations (104-106) (-) APs Pressure drop in the bottom zone of the reactor (Pa)
A Activity coefficient of component 1 in liquid (-) APpa Frictional pressure drop in downcomer (Pa)
G Mass flow rate of the liquid (kg.s4) APpr Frictional pressure drop in riser (Pa)
g Gravitational acceleration (m-s"2) . APr Pressure drop in the top zone of the reactor (Pa)
Vertical clearance between draft-tube and bottom of Q Vohnne flow rate of gas or liquid (m3.sd)
reactor (m) Qn Rate of heat removal (W)
N Hamaker constant (J) Qm Molar flow rate of gas (mobs"l)
hb Height of baffle (including the lower clearance) (m) q Total oxygen consumption in the bed (mmol-h"l)
h~ Dimensionless distance between level of liquid and R Gas constant (LKl.mol "l)
the upper edge of baffle (-) Ro Specific oxygen consumption rate of cells
Height of gas-liquid dispersion (m) 0mnol/(cell.h) or mmol/(cell-s))
ha Height of dovmcomer (m) RNA Ribonucleic add
Height of gas-free liquid (m) Ra Arithmetic mean surface roughness (lain)
hLi Height of lift (m) Re Reynolds number (7)
ht Height of fluid stream at lower opening of draft-tube Red Reynolds number in downcomer (-)
(m) Rer Reynolds number in riser (-)
ho Film heat transfer coefficient inside reactor Rq Root mean square rouglmess (I.tm)
(W'm'L°C "') r Radial position in the annulus (m)
hr Height of riser (m) ri Outer radius of draft-tube (m)
h, Height of fluid stream at upper opening of draft-tube ro Radius of outer coltnnn (m)
(m) S Surface area per refit volume of particle (m~)
J Total distance traveled (m) Sc Schmidt number.(-)
K Consistency index of fluid (Pa-s") Sh Sherwood nmnber (-)
Ks Frictional loss coefficient for the bottom (-) T Absolute temperature (K)
K, Individual friction loss coefficients in circulation path t Instantaneous time (s)
(-) t~ Liquid circulation time (s)
K, Solidqiquid mass transfer coefficient (m.sl) to Residence time of gas in liquid (s)
Kr Frictional loss coefficient for the top (-) ILd Liquid residence time in downcomer (s)
k Constant in equation (149) (-) tL~ Liquid residence time in riser (s)
kf Friction factor defined by equation (61 ) (-) tM Mixing time (s)
kL Gas-liquid mass transfer coefficient (m-s"1) tMB Mixing time in bubble colmnn (s)
( kLaL), o~d Overall volumetric mass transfer coefficient in riser t~ Wall thickness (m)
or downcomer (S"1) AT Mean temperature difference (°C)
kl" Thermal conductivity (W-mL°C q) Ub Mean bubble rise velocity (m.s4)
L Length of circulation loop, pipe or separator (m) Uo Superficial gas velocity based on outer column cross-
L~ Characteristic length (m) sectional area (m.sq)
Ld Length of draft-tube (m) U~ Critical gas velocity for suspension of solids (m.s1)
Lh Vertical distance between upper and lower riser- Uor Superficial gas velocity in the riser (m.s"l)
downcomer connections (m) Uo,,, Minimum superficial gas velocity in riser to initiate
Depth of the packed bed (m) circulation of liquid (m.s4)
L, Distance between the tracer measuring points (m) UH Overall heat transfer coefficient (W.m'2.*Cq)
l Length of the connecting pipe or distance between UL Mean superficial liquid velocity in the reactor or pipe
riser and downcomer in separator (m) (m.s-t)
g Length scale of microeddies (m) Uza Superficial liquid velocity in the downcomer (m.sq)
Ma Marrucci parameter (-) ULd~ Critical superficial downcomer liquid velocity for
MLSS Mixed liquor suspended solids (kg.m"3) suspension of solids (m.s"1)
Mo Morton number (-) ULd, Superficial liquid velocity in the expanded zone of
N Cell concentration at time t (cells.m3) downcomer (m. s'l)

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 .106 SPECIAL ISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Moch Rov vo151, no 1, January 1998

ULn Superficial liquid velocity in the head region (m.s q) pSL Density of slurry (kg.m"3)
ULr Superficial liquid velocity in the riser (m.s "l) aL Interfacial tension (kg.s2)
Uz,,, Maximum anticipated superficial liquid velocity in Shear stress (Pa.s)
riser (m.s q) ~D Wall shear stress in gas-liquid dispersion (Pa.s)
Up Slip velocity (m.s q) ZDa Wall shear stress for flow of dispersion in downcomer
USr Superficial velocity of solids in the riser (m.s "1) (Va.s)
Ut Terminal settling velocity of a single particle in liquid ~Dr Wall shear stress for flow of dispersion in riser (Pa.s)
(m-sq) ZL Wall shear stress for flow of gas-free liquid (Pa.s)
Va Volume of downcomer (m 3) ZLa Wall shear stress for flow of gas-free liquid in
Vat Available volume below drafl-tube (m 3) downcomer (Pa.s)
Vo Volume of gas (m 3) Tl.,r Wall shear stress for flow of gas-free liquid in riser
VL Volume of liquid or slurry (m 3) (Pa-s)
VLd Linear liquid velocity in downcomer (m.s q) ¢b Constant in equation (110) (-)
VLo Linear velocity of the liquid (m.s 1) Void fraction of the packed bed (-)
V, Volume of riser (m 3) Volume fraction of solids in gas-free slurry (-)
V, Volume of gas-liquid separator (m 3) Volume fraction of solids in gas-free slurry in
V= Packed volume of settled solids (m 3) downcomer (-)
vvm Volume of gas flow per unit volume of liquid per Volume fraction of solids in gas-free slurry in riser
minute (rain q) (-)
W Width of gas-liquid separator (m)
Wi Weissenberg number (-)
Wz Total weight of liquid in slurry (kg) REFERENCES
Ws Total weight of solids in slurry (kg) Akita K and Yoshida F (1973), Gas holdup and volumetric mass transfer
coefficient in bubble columns, Ind Eng Chem Process Des Develop 12,
x Depth of liquid in separator (m); parameter defined 76-80.
by equation (73) Akita K, Nakanishi O, and Tsuchiya K (19947, Turn-around energy losses
xz, x2 Mole fractions of components 1 and 2 in liquid (-) in an external-loopairlift reactor, Chem Eng Sci 49, 2521-2533.
Z~ Critical depth of packing (m) Allsop PJ, Chisti Y, M0o-Young,M, and Sullivan, GR (1993), Dynamicsof
phenol degradation by Pseudomonas putida, Biotech Bioeng 41, 572~
580.
Greek symbols Arathoon WR and Birch JR (1986), Large-scalecell culture in biotechnol-
ct Distribution parameter (-); parameter in equation ogy, Science 232, 1390-1395.
(179) (mP.s<P+°); parameter in equation (144) (tnq) Archambault J, VoleskyB, and Kurz WGW (1990), Developmentof biore-
actors for the culture of surface immobilized plant cells, Biotech Bioeng
CtM Value of a in presence of static mixers (mP.s"<p+°) 35, 702-711.
13 Parameter in the drift-flux model (m.sl); exponent in Arters DC and Fan L-S (1986), Solid-liquid mass transfer in a gas-liquid-
equation (179) (-) solid fluidized bed, Chem Eng Sci 41, 107-115.
Assa A and Bar R (1991), Biomass axial distribution in airlift bioreactor
Parameter in equation (114) (-) with yeast and plant cells, Biotech Bioeng 38, 1325-1330.
y Shear rate (s q) Ayazi Shamlou P, Pollard DJ, lson AP, and Lilly MD (1994), Gas holdup
Sparger hole or nozzle diameter (m) and liquid circulation rate in concentric-tubeairlift bioreactors, Chem
ef Gas holdup above upper end of riser (defined by Eng Sci 49, 303-312.
Backhurst JR and Harker JH (1973), Process Plant Design, Heinemann,
equation 82) (-) London, p 108.
8G Overall gas holdup (-) Bakker WAM, van Can HJL, Tramper J, and de Gooijer CD (1993), Hydro-
Gas holdup in the downcomer (-) dynamics and mixing in a multiple air-lift loop reactor, Biotech Bioeng
~r Gas holdup in the riser (-) 42,994-1001.
Bakker WAM, Knitel JT, Tramper J, and de Gooijer CD (1994), Sucrose
. oGrm Minimum gas holdup in riser to start circulation of conversion by immobilized invertase in a multiple air-lift loop bioreac-
liquid (-) tor, Biotech Prog 10, 277-283.
eSd Solids holdup in three-phase systems in downcomer Bakker WAM, den Hartog M, Tramper J, and de Gooijer CD (1995), Oxy-
(-) gen transfer in a multiple air-lift loop reactor, Bioproc Eng 12, 167-172.
Bando Y, Nishimura M, Sota H, Toyoda K, Suzuki S, and ldota A (1992a),
6s~ Solids holdup in three-phase systems in riser (-) Flow characteristics of bubble column with perforated draft tube--
kt Dimensionless time (-) Effects of equipment dimensions and gas dispersion, J Chem Eng Japan
~p Apparent viscosity of non-newtonian fluid (Pa.s) 25, 49-54.
Bando Y, Nishimura M, .Sota H, Suzuki S, and Kawase N (1992b), Flow
Viscosity of the gas (Pa.s) characteristics of countercurrent bubble column with perforated draft
Viscosity of tile liquid (Pa.s) tube, Chem Eng Sci 47, 3371-3378.
gSL Viscosity of the slurry (Pa.s) Bando Y, Nishimura M, Hayashi A, Hiura S, Ido S, and Idota A (1995),
vl, v2 Molar volumes of components 1 and 2 (m3.molq) Flow characteristics of bubble column with perforated draft tube in vis-
cous liquid, d Chem Eng Japan 28, 225-227.
Total number of moles of ions per mole of electrolyte Bayer T, Zhou W, Holzhauer K, and Sch0gerl K (1989), Investigation of
(-) cephalosporin C production in an airlift tower loop reactor, Appl Micro-
pd Density of slurry in downcomer (kg.m"3) biol Biotech 30, 26-33.
Booker S, Sokolichin A, and EigenbergerG (19947, Gas-liquid flow in bub-
po Density of gas (kg.m3) ble columns and loop reactors: Part I]. Comparison of detailed experi-
PL Density of the liquid (kg.m"3) ments and flow simulations, Chem Eng Sci 49, 574%5762.
Pp Density of the particles (kg.m3) Belle RA (1981), A characterization study of airlift contactors for applica-
Density of slurry in riser (kg.m3) tion to fermentations,PhD Thesis, Univ of Waterloo, Ontario, Canada.
P~ Birch JR, Lambert K, Thompso PW, Kenney AC, and Wood LA (1987),
.ps Density of solids (kg-m3) Antibody production with airlift fermenters, in Large Scale Cell Cul-
psa Average density of non-settling fraction of solids ture Technology, LydersenBK (ed), Hanser, New York, 1-20.
(kg.m"3)

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Appl Mech Rev vol 51, no 1, January 1998
Blakebrough N, McManamay WJ, and Walker G (1983a), Heat transfer to
fermentation systems in an air-lift fermenter. Part ll: Osmotic effects on
heat transfer, Chem Eng Res Des 61,264-266.
Blakebrough N, Fatile IA, McManamay WJ, and Walker G (1983b), Heat
transfer to fermentation systems in an air-lift fermenter. Part Ill: Mecha-
nism of heat transfer, Chem Eng Res Des 61,383-387.
Blenke H (1979), Loop reactors, Adv Biochem Eng 13, 121-214.
Bonnarme P and Jeffries TW (1990), Selective production of extracellular
peroxidases form Phanerochaete chrysosporium in an air-lift bioreactor,
J Ferment Bioeng 70, 158-163.
Bonnarme P, Delattre M, Drouet H, Corrieu G, and Asther M (1993), To-
ward a control of lignin and manganese peroxidases hypersecretion by
Phanerochaete chrysosporium in agitated vessels: Evidence of the supe-
riority of pneumatic bioreactors on mechanically agitated bioreactors,
Biotech Bioeng 41,440-450.
Brandes L, Wu X, Bode J, Rhee Jl, and SchOgerl K (1993a), Fed-batch cul-
tivation of recombinant Escherichia coli JM103 and production of the
fusion protein SPA::EcoRI in a 60-L working volume airlift tower loop
reactor, Biotech Bioeng 42,205-214.
Brandes L, Wu X, Maschke HE, Jtirgens H, Reinhardt B, and Sch~gerl K
(1993b), Production of fusion protein SpA::EcoRI in batch culture in a
60-L airlift tower loop reactor, Biotech Prog 9, 122-127.
Brandes L, Wu X, Maschke HE, and Sch~igerl K (1993c), Overproduction
Of recombinant proteins by Escherichia coli---employing the combina-
tions of two and three different types of multicopy plasmids---in a 60-
litre working volume airlift tower loop reactor, Process Biochem 28,
161-169.
Brown KE (1967), Gas Lift Theory and Practice, Prentice-Hall, Englewood
Cliffs.
Bfich GW, Enders L, Berger RG, and Drawert F (1991), Modelling the
growth of airlift cultured plant cells: Role of inoculum, invertase activ-
ity, and pO2, in Biochemical Engineering--Stuttgart, Reuss M, Chmiel
H, Gilles E-D, and Knackmuss H-J (eds), Gustav Fischer, New York,
pp. 232-235.
Buchholz H, Buchholz R, Lt~cke J, and Schtlgerl K (1978), Bubble swarm
behaviour and gas absorption in non-Newtonian fluids in sparged col-
umns, Chem Eng Sci 33, 1061 - 1070.
Butler M (1987), Animal Cell Technology, Taylor and Francis, New York,
p5.
Byun SY and Pedersen H (1994), Two-phase airlift fermentor operation
with elicitation for the enhanced production of benzophenanthridine al-
kaloids in cell suspensions of Escherichia californica, Biotech Bioeng
44, 14-20.
Cai J and Nieuwstad TJ (1991), Hydrodynamic modelling and blank run-
ning of a pilot-scale internal loop airlift reactor for wastewater treat-
ment, Env Tech 12,637-653.
Cai J and Nieuwstad TJ (1992), Modelling the three-phase flow in a pilot-
scale airlift internal-loop reactor for wastewater treatment, Env Tech 13,
101-113.
C~i J, Nieuwstad TJ, and Kop JH (1992a), Fluidization and sedimentation
of carrier material in a pilot-seale airlift internal-loop reactor, Wat Sci
Tech 26, 2481-2484.
Cai Z, Yang S, Kang R, Sun X, Liu W, Li G, and Wang Y (1992b), Studies
on the performance of bioreactors used for the fermentation process
with highly viscous broth, in Proc of Asia-Pacific Biochemical Engi-
neering Conf1992, Furusaki S, Endo I, and Matsuno R (eds), Springer-
Verlag, Tokyo, 429-432.
Calderbank PH and Moo-Young MB (1961), Continuous phase heat and
mass-transfer properties of dispersions, Chem Eng Sci 16, 39-54.
Capuder E and Koloini T (1984), Gas holdup and interfacial area in aerated
suspensions of small particles, Chem Eng Res Des 62,255-260.
Chakravarty M, Singh HD, Baruah JN, Iyengar MS, and Sinha AP (1971),
Heat transfer in gas lift reactor, Paper presented at the Ist Natl Heat and
Mass Transfer Conf, Madras, India. Paper no HMT-53-71.
Chang C-M, Lu WJ, Own KS, and Hwang S-J (1993), Operation conditions
affecting the performance of airlift bioreactors for immobilized enzyme
reactions, Biotech Techniques 7, 317-322.
Chang C-M, Lu WJ, Own KS, and Hwang S-J (1994), Comparison of airlift
and stirred tank reactors for immobilized enzyme reactions, Process
Biochem 29, 133-138.
Chaplin MF and Bucke C (1990), Enzyme Technology, Cambridge Univ
Press, Cambridge, 167-196.
Chatte0jee S, Grethlein A J, Worden RM, and Jain MK (1996), Evaluation of
support matrices for an immobilized cell gas lift reactor for fermentation
of coal derived synthesis gas, .1Ferment Bioeng 81,158-162.
Chen J, Hu X, and Li Z (1994), Innovation of air-lift bioreactor for animal
cell culture, in Better Living Through Biochemical Engineering, Teo
WK, Yap MGS, and Oh SKW (eds), Univ of Singapore, Singapore,
405-407.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 107
Cheng C-M, Lu WJ, Own KS, and Hwang S-J (1992), Application of airlift
reactors for immobilized enzyme reactions, in Proc of Asia-Pacific Bio-
chemical Engineering Conf 1992, Furusaki S, Endo I, and Matsuno R
(eds), Springer-Verlag, Tokyo, 437-440.
Cherry RS (1993), Animal cells in turbulent fluids: Details of the physical
stimulus and the biological response, Biotech Adv 11,279-299.
Chiou T-W, Murakami S, Wang DIC, and Wu W-T (1991), Fiber-bed
bioreactor for anchorage-dependent animal cell cultures: Part I. Biore-
actor design and operation, Biotech Bioeng 37, 755-761.
Chisti Y (1989), Airl~ Bioreactors, Elsevier, London.
Chisti Y (1992a), Build better industrial bioreactors, Chem Eng Progress
88(1), 55-58.
Chisti Y (1992b), Assure bioreactor sterility, Chem Eng Progress 88(9), 80-
85.
Chisti Y (1993), Animal cell culture in stirred bioreactors: Observations on
scale-up, Bioproc Eng 9, 191 - 196.
Chisti Y and Moo-Young M (1986), Disruption of microbial cells for intra-
cellular products, Enzyme Microb Technol 8, 194-204.
Chisti Y and Moo-Young M (1987a), Airlift reactors: Characteristics, ap-
plications and design considerations, Chem Eng Commun 60, 195-242.
Chisti Y and Moo-Young M (1987b), Pneumatically agitated bioreactor de-
vices: effects of fluid height and flow area shape on hydrodynamic and
mass transfer performance, in Bioreactors and Biotransformations,
Moody GW and Baker PB (eds), Elsevier, London, 335-349.
Chisti Y and Moo-Young M (1988a), Prediction of liquid circulation ve-
locity in airlift reactors with biological media, .1 Chem Tech Biotech 42,
211-219.
Chisti Y and Moo-Young M (1988b), Gas holdup in pneumatic reactors,
Chem Eng.138, 149-152.
Chisti Y and Moo-Young M (1988c), Gas holdup behaviour in fermentation
broths and other non-newtonian fluids in pneumatically agitated reac-
tors, Chem Eng.139, B31-B36.
Chisti Y and Moo-Young M (1988d), Hydrodynamics and oxygen transfer
in pneumatic bioreactor devices, Biotech Bioeng 31,487-494.
Chisti Y and Moo-Young M (1989), On the calculation of shear rate and
apparent viscosity in airlift and bubble column bioreactors, Biotechnol
Bioeng34, 1391-1392.
Chisti Y and Moo-Young M (1991), Fermentation technology, bioprocess-
ing, scale-up and manufacture, in Biotechnology: The Science and the
Business, Moses V and Cape RE (eds), Harwood Academic Publ, New
York, 167-209.
Chisti Y and Moo-Young M (1993a), Improve the performance of airlift
reactors, Chem Eng Progress 89(6), 38-45.
Chisti Y and Moo-Young M (1993b), Airlift bioreactors with packed beds
of immobilized biocatalysts: Theoretical evaluation of liquid circulation
performance, Trans I Chem E 71C, 209-214.
Chisti Y and Moo-Young M (1994a), Anchorage dependent animal cell
culture in packed beds with airlift driven liquid circulation: A theoreti-
cal analysis of oxygen transfer and comparison with stirred tank micro-
carrier culture system, Trans 1 Chem E 72C, 91-94.
Chisti Y and Moo-Young M (1994b), Airlift bioreactors for treatment of
hydrocarbon contaminated wastes, in Better Living Through Biochemi-
cal Engineering, Teo WK, Yap MGS, and Oh SKW (eds), Univ of Sin-
gapore, Singapore, 771-776.
Chisti Y and Moo-Young M (1994c), Liquid circulation in external-loop
airlift reactors, .1 Chem Technol Biotechno159, 101 - 103.
Chisti Y and Moo-Young M (1994d), Cleanzin-place systems for industrial
bioreactors: Design, validation and operation, .1 Ind Microbiol 13, 201-
207.
Chisti Y and Moo-Young M (1996), Bioprocess intensification through
bioreactor engineering, Trans I Chem E 74A, 575-583.
Chisti Y, Fujimoto K, and Moo-Young M (1987), Hydrodynamic and oxy-
gen mass transfer studies in bubble columns and airlift bioreactors, in
Biotechnology Processes: Scale-up and Mixing, Ho CS and Oldshue JY
(eds), American Institute of Chemical Engineers, New York, 72-81.
Chisti Y, Halard B, and Moo-Young M (1988), Liquid circulation in airlift
reactors, Chem Eng Sci 43, 451-457.
Chisti Y, Kasper M, and Moo-Young M (1990), Mass transfer in external-
loop airlift bioreactors using static mixers, Can .1 Chem Eng 68, 45-50.
Chisti Y, Wenge F, and Moo-Young M (1995), Relationship between riser
and downcomer gas holdup in internal-loop airlift reactors without gas-
liquid separators, Chem Eng.1 57, B7-BI3.
Choi KH and Lee WK (1990a), Comparison of probe methods for meas-
urement of bubble properties, Chem Eng Commun 91, 35-47.
Choi KH and Lee WK (1990b), Recirculation and flow structures of gas in
downcomer section of a concentric cylindrical airlift reactor, .1 Chem
Tech Biotech 48, 81-95.
Choi KH and Lee WK (1993), Circulation liquid velocity, gas holdup and
volumetric oxygen transfer coefficient in external-loop airlift reactors, .1
Chem Tech Biotech 56, 51-58.
 108 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Choi KH, Chisti Y, and Moo-Young M (1995a), Influence of the gas-liquid
separator design on hydrodynamic and mass transfer performance of
split-channel airlift reactors, J Chem Tech Biotech 62,327-332.
Choi KH, Chisti Y, and Moo-Young M (1995b), Split-channel rectangular
airlift reactors: Enhancement of performance by geometric modifica-
tions, Chem Eng Commun 138, 171-181.
Choi KH, Chisti Y, and Moo-Young M (1996), Comparative evaluation of
hydrodynamic and gas-liquid mass transfer characteristics in bubble
column and airlift slurry reactor, Chem Eng J 62,223-229.
Coulson JM and Richardson JF (1977), Chemical Engineering, vol 2, 3rd
ed, Pergamon Press, Oxford, 126-132.
Dai X, Ouyang F, and Qi Y (1994), Preliminary investigation on anchorage-
dependent cell culture by air-lift suspending microcarriers, in Better
Living Through Biochemical Engineering, Ten WK, Yap MGS, and Oh
SKW (eds), Univ of Singapore, Singapore, 408-410.
de Boer JP, Teixeira de Mattos MJ, and Neijssel OM (1990), D(-)Lactic
acid production by suspended and aggregated continuous cultures of
Bacillus laevolacticus, Appl Microbiol Biotech 34, 149-153.
de Jong RJ, Breure AM, and van Andel JG (1991), Enhanced biodegrada-
tion of o-cresol by activated sludge in the presence of powdered acti-
vated carbon, Appl Microbiol Biotech 34, 683-687.
Deckwer W-D (1985), Bubble column reactors, in Biotechnology, vol. 2,
Rehm H-J and Reed G (eds), VCH, Weinheim, 445-464.
Deckwer W-D (1992), Bubble Column Reactors, John Wiley, New York.
Deckwer W-D and Schumpe A (1984), Transport phenomena in three-phase
reactors with fluidized solids, Ger Chem Eng 7, 168-177.
Deckwer W-D and Schumpe A (1987), Bubble columns---the state of the
art and current trends, lnt Chem Eng 27, 405-422.
Deckwer W-D and Schumpe A (1993), Improved tools for bubble column
reactor design and scale-up, Chem Eng Sci 48, 889-911.
Deckwer W-D, Burckhart R, and Zoll G (1974), Mixing and mass transfer
in tall bubble columns, Chem Eng Sci 29, 2177-2188.
Dedek I (I 990), Effect of bubbling regime and mode of mechanical agita-
tion on the course of hydrogenation of vegetable oil, Coil Czech Chem
Commun 55, 1721-1729.
Doran PM (1993), Design of reactors for plant cells and organs, Adv Bio-
chem Eng Biotech 48, 115-168.
Douek RS and Livingsto AG (1994), Hydrodynamics and instability phe-
nomena in an external-loop three phase air-lift (TPAL) reactor, in Better
Living Through Biochemical Engineering, Ten WK, Yap MGS, and Oh
SKW (eds), Univ of Singapore, Singapore, 447-449.
Douek RS, Hewitt GF, and Livingston AG (1995), A hydrodynamic study
of the riser zone in a three phase airlift (TPAL) reactor, Trans 1 Chem E,
73A, 336-340.
Dreveton E, Monot F, Lecourtier J, Ballerini D, and Choplin L (1996), In-
fluence of fermentation hydrodynamics on gellan gum physico-chemical
characteristics, J Ferment Bioeng, 82, 272-276.
Dunn IJ (1994), High-rate biofilm fluidized bed reactors for specialized
wastewater treatment, in Advances in Bioprocess Engineering, Galindo
E and Ramirez OT (eds), Kluwer, Dordrecht, 365-378.
Fair JR, Lambright AJ, and Andersen .IW (1962), Heat transfer and gas
holdup in a sparged contactor, lnd Eng Chem Process Des Dev 1, 33-
36.
Ferschl A, Loidl M, Ditzelm~ller G, Hinteregger C, and Streichsbier F
( 1991 ), Continuous degradation of 3-chloroaniline by calcium alginate-
entrapped cells of Pseudomonas acidovorans CA28: Influence of addi-
tional substrates, Appl Microbiol Biotech 35, 544-550.
Fleishaker RJ and Sinskey AJ (1981), Oxygen demand and supply in cell
culture, Eur J Appl Microbiol Biotechnol 12, 193- 197.
Fouhy K (1992), Wastewater treatment reaches new heights, Chem Eng
99(12), 101-102.
Fraser RD and Hill GA (1993), Hydrodynamic characteristics of a spinning
sparger, external loop airlift bioreactor, Can J Chem Eng 71, 419-425.
Fraser RD, Ritchie BJ, and Hill GA (1994), Dynamic mixing and oxygen
transfer in small, airlift loop bioreactors: Model and experimental verifi-
cation, Biotech Prog 10, 543-547.
FrOhlich S, Lotz M, Korte T, L~bbert A, SchOgerl K, and Seekamp M
(1991a), Characterization of a pilot plant airlift tower loop reactor: I.
Evaluation of the phase properties with model media, Biotech Bioeng
38, 43-55.
FrShlich S, Lotz M, Larson B, Ltibbert A, SchCtgerl K, and Seekamp M
(1991b), Characterization of a pilot plant airlift tower loop reactor: IlL
Evaluation of local properties of the dispersed gas phase during yeast
cultivation and in model media, Biotech Bioeng 38, 56-64.
Ganzeveld K3, Chisti Y, and Moo-Young M (1995), Hydrodynamic behav-
iour of animal cell microcarrier suspensions in split-cylinder airlift
bioreactors, Bioproc Eng 12,239-247.
Garcia-Calvo E and Let6n P (1996), Prediction of gas hold-up and liquid
velocity in airlift reactors using two-phase flow friction coefficients, J
Chem Teeh Biotech 67, 388-396.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
Gamier A, Chavarie C, Andre G, and Klvana D (1990), Inverse fluidization
airlift bioreactor, part I: Hydrodynamic studies, Chem Eng Commun 98,
31-45.
Garrido F, Banerjee UC, Chisti Y, and Moo-Young M (1994), Disruption of
a recombinant yeast for the release of 13-galactosidase, Bioseparation 4,
319-328.
Garrido JM, van Benthum WAJ, van Loosdrecht MCM, and Heijnen JJ
(1997), Influence of dissolved oxygen concentration on nitrite accumu-
lation in a biofilm airlift suspension reactor, Biotech Bioeng 53, 168-
178.
Gaspillo PD and Goto S (1991), Mass transfer in bubble slurry column with
static mixer in draft tube, d Chem Eng Japan 24, 680-682.
Gavrilescu M and Tudose RZ (1995), Study of the liquid circulation veloc-
ity in external-loop airlift bioreactors, Bioproc Eng 14, 33-39.
Ghirardini M, Donati G, and Rivetti F (1992), Gas lift reactors: Hydrody-
namics, mass transfer, and scale up, Chem Eng Sci 47, 2209-2214.
Ghosh TK and Bhattacharyya BC (1993), Effect of inlet oxygen concentra-
tion on continuous cultivation of baker's yeast in a converging-
diverging tube airlift fermentor, J Ferment Bioeng 76, 153-156.
Ghosh TK, Maiti BR, and Bhattacharyya BC (1993a), Studies on mass
transfer characteristics of a modified airlift fermenter, Bioproc Eng 9,
239-244.
Ghosh TK, Maiti BR, and Bhattacharyya BC (1993b), Gas hold-up in con-
verging-diverging tube airlift fermenter, Biotech Techniques 7, 301-307.
Ghosh TK, Bhattacharyya BC, and Maiti BR (1994a), Residence time dis-
tribution in an external loop air-lift fermenter with a wavy riser, lndian
Chem Eng, Section A, 36, 53-57.
Ghosh TK, Maiti BR, and Bhattacharyya BC (1994b), Performance char-
acteristics of yeast growth in a modified airlift fermenter, Bioproc Eng
10, 241-250.
Gjaltema A, Tijhuis L, van Loosdrecht MCM, and Heijnen JJ (1995), De-
tachment of biomass from suspended nongrowing spherical biofilms in
airlift reactors, Biotech Bioeng 46, 258-269.
Gjaltema A, Vinke JL, van Loosdrecht MCM, and Heijnen JJ (1997), Abra-
sion of suspended biofilm pellets in airlift reactors: Importance of shape,
structure, and particle concentrations, Biotech Bioeng 53, 88-99.
Glasgow LA, Hua J, Yiin T-Y, Erickson LE (1992), Acoustic studies on
interracial effects in airlift reactors, Chem Eng Commun, 113, 155-181.
Glennon B, AI-Masry W, MacLoughlin PF, and Malone DM (1993), Hy-
drodynamic modelling in an air-lift loop reactor, Chem Eng Commun
121,181-192.
Godbole SP, Schumpe A, Shah YT, and Carr NL (1984), Hydrodynamics
and mass transfer in non-Newtonian solutions in a bubble column,
AIChE d 30, 213-220.
Goto S and Gaspillo PD (1992), Effect of static mixer on mass transfer in
draft tube bubble column and in external loop column, Chem Eng Sci
47, 3533-3539.
Goto S, Matsumoto Y, and Gaspillo P (1989), Mass transfer and reaction in
bubble column slurry reactor with draft tube, Chem Eng Commun 85,
181-191.
Griffiths JB (1992), Animal cell culture processes-batch or continuous, J
Biotech 22, 21-30.
GOnzel B, Yonsel S, and Deckwer W-D (1991), Fermentative production of
1,3-~ropanediol from glycerol by Clostridium butyricum up to a scale of
2 m , Appl Microbiol Biotech 36, 289-294.
Guo Y, Lou F, Peng Z-Y, Yuan Z-Y, and Korus RA (1990), Kinetics of
growth and a-amylase production of immobilized Bacillus subtilis in an
airlift bioreactor, Biotech Bioeng 35, 99-102.
Hamada T, lshiyama T, and Motai H (1989), Continuous fermentation of
soy sauce by immobilized cells of Zygosaccharomyces rouxii in an air-
lift reactor, App.l Microbiol Biotech 31,346-350.
Hamada T, Sugishita M, Fukushima Y, Fukase T, and Motai H (1991),
Continuous production of soy sauce by a bioreactor system, Process
Biochem 26, 39-45.
Hano T, Matsumoto M, Kuribayashi K, and Hatate Y (1992), Biological
nitrogen removal in a bubble column with a draught tube, Chem Eng Sci
47, 3737-3744.
Haque MW, Nigam KDP, and Joshi JB (1986), Hydrodynamics and mixing
in highly viscous pseudo-plastic non-newtonian solutions in bubble col-
umns, Chem Eng Sci 41,2321-2331.
Heck J and Onken U (1986), Hysterese-Effekte beim Suspendieren yon
Feststoff~artikeln in Blasens~iule und Airlift-Schlaufenreaktor, Chem-
lng-Tech 58, 131-133.
Heck J and Onken U (1987), Hysteresis effects in suspended solid particles
in bubble columns with and without draft tube, Chem Eng Sci 42, 1211-
1212.
Heck J and Onken U (1988), Einfluss tier Partikelform auf die Feststoffsus-
pendierung in Blasens~.ule und Airlift-Schlaufenreaktor, Chem-lng-Tech
60, 400-401. ,
 Appl Mech Rev vol 51, no 1, January 1998
Hegarty PK, Smart N J, Scragg AH, and Fowler MW (1986), Aeration of C.
roseus L. G. Don suspension cultures in airlift bioreactors: The inhibi-
tory effect at high aeration rates on culture growth, dExp Bot 37, 1911-
1920.
Heijnen SJ, Mulder A, Weltevrede R, Hols PH, and van Leeuwen HLJM
(1990), Large-scale anaerobic/aerobic treatment of complex industrial
wastewater using immobilized biomass in fluidized bed and air-lift sus-
pension reactors, Chem Eng Tech 13,202-208.
Heijnen JJ, van Loosdrecht MCM, Mulder A, and Tijhuis L (1992), Forma-
tion of biofllms in a biofilm air-lift suspension reactor, Wat Sci Tech 26,
647-654.
Highfill JG, Haley SD, and Kompala DS (1996), Large-scale production of
murine bone marrow cells in an airlift packed bed bioreactor, Biotech
Bioeng 50, 514-520.
Hikita H, Asai S, Tanigawa K, Segawa K, and Kitao M (1980), Gas hold-up
in bubble columns, Chem Eng d 20, 59-67.
Hills JH (19765, Operation of a bubble column at high throughputs I: Gas
holdup measurements, Chem Eng J 12, 89-99.
Hong YC, Labuza TP, and Harlander SK (1989), Growth kinetics of straw-
berry cell suspension cultures in shake flask, airlift, stirred-jar, and
roller bottle bioreactors, Biotech Prog 5, 137-143.
Hopf NW, Yonsel S, and Deckwer W-D (1990), Ambrutiein S production
in air-lift and stirred tank bioreactors, Appl Microbiol Biotech 34, 350-
353.
Hsiun D-Y and Wu W-T (1995), Mass transfer and liquid mixing in an air-
lift reactor with a net draft tube, Bioproc Eng 12, 221-225.
HOppe P, Hrke H, and Hempel DC (1990), Biological treatment of effluents
from a coal tar refinery using immobilized biomass, Chem Eng Tech 13,
73-79.
Husemann W (1982), Photoautotrophic growth of cell susper~sion cultures
from Chenopodium rubrum in an airlift reactor, Protoplasma 113, 214-
220.
Ichii T, Takehara S, Konno H, lshida T, Sato H, Suzuki A, and Yamazumi
K (1993), Development of a new commercial-scale airlift fermentor for
rapid growth of yeast, d Ferment Bioeng 75, 375-379.
Immich M and Onken U (1992), Prediction of minimum gas velocity in
suspended bubble columns and airlift reactors, Chem Eng Sci 47, 3379-
3386.
Jadhav SV and Pangarkar VG (1988), Particle-liquid mass transfer in three
phase sparged reactor, Can d Chem Eng 66, 572-578.
• J/iger AG and Wandrey C (19905, Immobilization of basidiomycete Phan-
erochaete chrysosporium on sintered glass: Production of lignin peroxi-
dase, in Physiology of Immobilized Cells, de Bont JAM, Vissert J, Mat-
tiasson B, and Tramper J (eds), Elsevier, Amsterdam, 433.-438.
Janssen AJH, Sleyster R, van der Kaa C, .lochemsen A, Bontsema J, and
Lettinga G (19955, Biological sulphide oxidation in a fed-batch reactor,
Biotech Bioeng 47, 327-333.
Jones AJ (1985), Liquid circulation in a draft-tube bubble column, Chem
Eng Sci 40, 449-462.
.long J-Z, Hsiun D-Y, Wu W-T, and Tzeng Y-M (1995), Fed-batch culture
of Bacillus thuringiensis for thufingiensin production in a tower type
bioreactor, Biotech Bioeng 48, 207-213.
Joshi JB, Ranade VV, Gharat SD, and Lele SS (1990), Sparged loop reac-
tors, Can d Chem Eng 68, 705-741.
Karamanev DG, Nagamune T, and Endo 1 (19925, Hydrodynamic and mass
transfer study of a gas-liquid-solid draft tube spouted bed bioreactor,
Chem Eng Sci 47, 3581-3588.
Karamanev DG, Chavafie C, and Samson R (19965, Hydrodynamics and
mass transfer in an airlift reactor with a semipermeable draft tube, Chem
Eng Sci 51, 1173-1176.
Kato A, Kawazoe S, Iijima M, and Shimizu Y (1976), Continuous culture
of tobacco cells, d Ferment Tech 54, 82-87.
Kautola H (1990), Itaconic acid production from xylose in repeated-batch
and continuous bioreactors, Appl Microbiol Biotech 33, 7-11.
Kautola H, Rymowicz W, Linko Y-Y, and Linko P (1991), Production of
citric acid with immobilized Yarrowia lipolytica, Appl Microbiol Bio-
tech 35, 447-449.
Kawasaki H, Hirano H, and Tanaka H (1994), Effects of multiple draft
tubes with perforated plates on gas holdup and volumetric mass transfer
coefficient in a bubble column, d Chem Eng Japan 27, 669-670.
Kawase Y (1990), Liquid circulation in external-loop airlift bioreactors,
Biotechnol Bioeng 35, 540-546.
Kawase Y (19945, Design and scale-up of external-loop airlift bioreactor, in
Advances in Bioprocess Engineering, Galindo E and Ramirez OT (eds),
Kluwer, Dordrecht, 13-19.
Kawase Y and Moo-Young M (1986), Influence of non-newtonian flow be-
haviour on mass transfer in bubble columns with and without draft
tubes, Chem Eng Commun 40, 67-83.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 109
Kawase Y and Moo-Young M (1987), Theoretical prediction of gas hold-up
in bubble columns with newtonian and non-newtonian fluids, lnd Eng
Chem Res 26, 933-937.
Kawase Y and Moo-Young M (1989), Mixing time in bioreactors, d Chem
Tech Biotech 44, 63-75.
Kawase Y and Moo-Young M (1990), Mathematical models for design of
bioreactors: Applications of Kolmogoroff's theory of isotropic turbu-
lence, Chem Eng J 43, B19-B41.
Kawase Y, Omori N, and Tsujimura M (1994), Liquid-phase mixing in ex-
ternal-loop airlift bioreactors, J Chem Tech Biotech 61, 49-55.
Kawase Y, Tsujimura M, and Yamaguchi T (1995), Gas hold-up in exter-
nal-loop airlift bioreactors, Bioproc Eng 12, 21-27.
Kelkar BG, Phulgaonkar SR, and Shah YT (1983), Effects of electrolyte
solutions on hydrodynamic and backmixing characteristics in bubble
columns, Chem Eng d 27, 125-133.
Kemblowski Z, Przywarski J, and Diab A (1993), Average gas hold-up and
liquid circulation velocity in airlift reactors with external loop, Chem
Eng Sci 48, 4023-.4035.
Keshavarz T, Eglin R, Walker E, Bucke C, Holt G, Bull AT, and Lilly MD
(1990), Large-scale immobilization of Penicillium chrysogenum: Batch
and continuous operation in an air-lift reactor, Biotech Bioeng 36, 763-
770.
Kikuchi K-l, Takahashi H, and Takeda Y (1995), Measurement of local
residence time of particles in a draft-tube slurry bubble column using
magnetic tracer particles, d Chem Eng Japan 28, 483-486.
Koch B, Ostermann M, Hrke H, and Hempel D-C (19915, Sand and acti-
vated carbon as biofilm carriers for microbial degradation of phenols
and nitrogen-containing aromatic compounds, Wat Res 25, 1-8.
Koide K, Sato H, and Iwamoto S (1983a), Gas holdup and volumetric liq-
uid-phase mass transfer coefficient in bubble column with draught tube
and with gas dispersion into annulus, J Chem Eng Japan 16, 407-413.
Koide K, Kurematsu K, lwamoto S, lwata Y, and Horibe K (1983b), Gas
holdup and volumetric liquid-phase mass transfer coefficient in bubble
column with draught tube and with gas dispersion into tube, J Chem
Engdapan 16, 413-419.
Koide K, Iwamoto S, Takasaka Y, Matsuura S, Takahashi E, Kimura M,
and Kubota H (1984a5, Liquid circulation, gas holdup and pressure drop
in bubble column with draught tube, d Chem Eng Japan 17, 611-618.
Koide K, Horibe K, Kawabata H, and Ito S (1984b), Critical gas velocity
required for complete suspension of solid particles in solid-suspended
bubble column with draught tube, J Chem Eng Japan 17, 368-374.
Koide K, Horibe K, Kitaguchi H, and Suzuki N (1984c), Contributions of
annulus and draught tube to gas-liquid mass transfer in bubble columns
with draught tube, d Chem Eng Japan 17, 547-549.
Koide K, Horibe K, Kawabata H, and Ito S (1985), Gas holdup and volu-
metric liquid-phase mass transfer coefficient in solid-suspended bubble
column with draught tube, d Chem Engdapan 18, 248-254.
Koide K, Terasawa M, and Takekawa H (1986), Critical gas velocity re-
quired for complete suspension of multicomponent solid particle mix-
ture in solid-suspended bubble columns with and without draught tube,
J Chem Eng Japan 19, 341-344,
Koide K, Kimura M, Nitta H, and Kawabata H (1988), Liquid circulation in
bubble column with draught tube, d Chem Engdapan 21,393-399.
Koide K, Shibata K, Ito H, Kim SY, and Ohtaguchi K (1992), Gas holdup
and volumetric liquid-phase mass transfer coefficient in a gel-particle
suspended bubble column with draught tube, J Chem Eng dapan 25, 11-
16.
Kracke-Helm HA, Rinas U, Hitzmann B, and Schtlged K (1991), Cultiva-
tion of recombinant E. coli and production of fusion protein in 60-1
bubble column and airlift tower loop reactors, Enzyme Microb Tech 13,
554-564.
Kramer H W and Bailey JE (1991), Mass transfer characterization of an air-
lift probe for oxygenation and mixing cell suspensions in an NMR
spectrophotometer, Biotech Bioeng 37, 205-209.
Kumar A and Shuler ML (1995), Model of a split-flow airlift bioreactor for
attachment-dependent, baculovirus-infected insect cells, Biotech Prog
11,412-419.
Kushalkar KB and Pangarkar VG (1994), Particle-liquid mass transfer in a
bubble column with a draft tube, Chem Eng Sci 49, 139-144.
Kwong C-W, Hsieh J-H, Syu M-J, and Chou S-S (1989), Anti-hepatitis B
surface antigen monoclonal antibody lgM production in suspension and
immobilized cell bioreactors, Biotech Lett 11,377-382.
Laugero C, Sigoillot J-C, Moukha S, Frasse P, Bellon-Fontaine M-N, Bon-
narme P, Mougin C, and Asther M (1996), Selective hyperproduction of
manganese peroxidases by Phanerochaete chrysosporium 1-1512 im-
mobilized on nylon net in a bubble-column reactor, Appl Microbiol
Biotech 44, 717-723.
LemcoffNO , Pereira Duarte SI, and Martinez OM (1990), Heat transfer in
packed beds, Rev in Chem Eng 6, 229-292.
 110 SPECIALISSUE:FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY
Li G-Q, Yang S-Z, Cai Z-L, and Chen J-Y (1993), Liquid phase axial
backmixing in an airlift bioreactor with non-newtonian fluids, Chem
Eng Commun 125, 13-26.
Li G-Q, Yang S-Z, Cai Z-L, and Chen J-Y (1995), Mass transfer and gas-
liquid circulation in an airlift bioreactor with viscous non-Newtonian
fluids, Chem Eng J 56, B 101 -B 107.
Linko S (1992), Production of Phanerochaete chrysosporium lignin per-
oxidase, Biotech Adv 10, 191-236.
Livingston AG and 'Zhang SF (1993), Hydrodynamic behaviour of three-
phase (gas-liquid-solid) airlift reactors, Chem Eng Sci 48, 1641-1654.
Lu W-J, Hwang S-J, and Chang C-M (1993), Liquid mixing in two- and
three-phase airlift reactors, Paper 706 presented at the AIChE 1993
Summer Natl Meeting, Seattle, Aug 15-18.
Lu W-J, Hwang S-J, and Chang C-M (1994), Liquid mixing in internal loop
airlift reactors, lnd Eng Chem Res 33, 2180-2186.
Lu W-J, Hwang S-J, and Chang C-M (1995), Liquid velocity and gas
holdup in three-phase internal loop airlift reactors with low-density par-
ticles, Chem Eng Sci 30, 1301 - 1310.
Ltlbbert A and Larson B (1990), Detailed investigations of the multiphase
flow in airlift tower loop reactors, Chem Eng Sci 45, 3047-3053.
L~bbert A, Larson B, Wan LW, and BrOring S (1990), Local mixing be-
haviour of airlift multiphase chemical reactors, 1 Chem E Symp Ser 121,
203-213.
Maiorella B, lnlow D, Shauger A, and Harano D (1988), Large-scale insect
cell-culture for recombinant protein production, Biotech 6, 1406-1410.
Mao HH, Chisti Y, and Moo-Young M (1992), Multiphase hydrodynamics
and solid-liquid mass transport in an external-loop airlift reactor--A
comparative study, Chem Eng Commun 113, 1-13.
Matthijs HCP, Balke H, Van Hes UM, Kroon BMA, Mur LR, and Binot RA
(1996), Application of light-emitting diodes in bioreaetors: Flashing
light effects and energy economy in algal culture (Chlorella pyrenoi-
dosa), Biotech Bioeng 50, 98-107.
Merchuk JC (1986), Hydrodynamics and hold-up in air-lift reactors, in En-
cTclopedia of Fluid Mechanics, Cheremisinoff NP (ed, vol 3, Gulf Publ,
Houston, 1485-1511.
Merchuk JC and Berzin I (1995), Distribution of energy dissipation in airlift
reactors, Chem Eng Sci 50, 2225-2233.
Messing RA (ed (1975), Immobilized Enzymes for Industrial Reactors,
Academic Press, New York.
Metkin VP and Soko!ov VN (1982), Hydraulic resistance to flow of gas-
liquid mixtures having non-Newtonian properties, J Appl Chem USSR
55, 558-563.
Miller WM, Wilke CR, and Blanch HW (1987), Effects of dissolved oxygen
concentration on hybridoma growth and metabolism in continuous cul-
ture, J Cell Physiol, 132,524-530.
Miyahara T, Hamaguchi M, Sukeda Y, and Takehashi T (1986), Size of
bubbles and liquid circulation in a bubble column with draught tube and
sieve plate, Can J Chem Eng 64, 718-725.
Mojovic L, ~iler-Marinkovic S, Kukic G, Bugarski B, and Vunjak-Novako-
vic G (1994), Rhizopus arrhizus lipase-catalysed interesterification of
palm oil midfraction in a gas-lift reactor, Enzyme Microb Tech 16, 159-
162.
Molina Grima E, Sfinchez Ptrez JA, Garcia Camacho F, Garcia S/mchez JL,
Acitn Fernfindez FG~ and L6pez Alonso D (1994), Outdoor culture of
lsochrysis galbana ALII-4 in a closed tubular photobioreactor, J Bio-
tech 37, 159-166.
Molina Grima E, Sfinchez Ptrez JA, Garcia Camacho F, Fermindez Sevilla
JM, Acitn Fern~indez FG, and Urda Cardona J (1995), Biomass and
icosapentaenoie acid productivities from an outdoor batch culture of
Phaeodactylum tricornutum UTEX 640 in an airlift tubular photobiore-
actor, Appl Microbiol Biotech 42,658-663.
Molina Gfima E, Chisti Y, and Moo-Young M (1997), Characterization of
shear rates in airlift bioreactors for animal cell culture, J Biotech 54,
195-210.
Moo-Young M and Chisti Y (1988), Considerations for designing bioreac-
tors for shear-sensitive culture, Biotech 6(11), 1291-1296.
Moo-Young M and Chisti Y (1994), Bioreactor applications in waste treat-
ment, Res Cons Recycl 11, 13-24.
Moo-Young M, Halard B, Allen DG, Burrell R, and Kawase Y (1987),
Oxygen transfer to mycelial fermentation broths in an airlift fermentor,
Biotech Bioeng 30, 746-753.
Moo-Young M, Chisti Y, and Vlaeh D (1994), Fermentation of cellulosic
materials to mycoprotein foods, Biotechnol Adv, 11,469-479.
Mota MJ and Teixeira JA (1990), Utilization of an external loop bioreactor
for the isolation of a flocculating strain of Kluyveromyces marxianus,
Curr Microbio120, 209-214.
Murakami S, Chiou T-W, and Wang DIC (1991), A fibre-bed bioreactor for
anchorage-dependent animal cell cultures: Part II. Scaleup potential,
Biotech Bioeng 37, 762-769.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Appl Mech Rev vol 51, no 1, January 1998
Murdin AD, Thorpe JS, Kirkby N, Groves DJ, and Spier RE (1987), Immo-
bilization and growth of hybridomas in packed beds, in Bioreactors and
Biotransformations, Moody GW and Baker PB (eds), Elsevier, London,
99-110.
Murhammer DW and Goochee CF (1988), Scaleup of insect cell cultures:
Protective effects of Pluronic F-68, Biotech 6, 1411-1418.
Murhammer DW and Goochee CF (1990), Sparged animal cell bioreactors:
Mechanism of cell damage and Pluronic F-68 protection, Biotech Prog
6, 391-397.
Nishii K, Sode K, and Karube I (1989), Microbial conversion of dihy-
drooxoisophorone (DOIP) to 4-hydroxy-2,2,6-tfimethylcyclohexanone
(4-HTMCH) by thermophilic bacteria, J Biotech 9, I 17-128.
Nishikawa M, Kato H, and Hashimoto K (1977), Heat transfer in aerated
tower filled with non-Newtonian liquid, lnd Eng Chem Process Des
Develop 16, 133-137.
Ohashi H, Sugawara T, and Kikuchi K (1981), Mass transfer between parti-
cles and liquid in solid-liquid two-phase upflow in the low-velocity re-
gion through vertical tubes, J Chem Eng Japan 14, 489-491.
Ohashi H, Sugawara T, Kikuchi K, and Takeda M (1982), Mass transfer
between particles and liquid in solid-liquid two-phase flow through
horizontal tubes, J Chem Eng Japan 15, 311-313.
Okada K, Shibano S, and Akagi Y (1993), Turbulent properties in bubble-
flow region in external-loop airlift bubble column, J Chem Eng Japan
26, 637-643.
Onken U and Weiland P (1983), Airlift fermenters: Construction behaviour
and uses, in Mizrahi A and van Werzel L (eds), Advances in Biotech-
nological Processes, vol. 1, Liss, New York, 67-95.
Opie R (1993), Repeat order, World Water and Environmental Engineer,
Jan/Feb, pp 16-17.
Ouyoung PK, Chisti Y, and Moo-Young M (1989), Heat transfer in airlift
reactors, Trans I Chem E 67, 451-456.
Papoutsakis ET (1991), Fluid-mechanical damage of animal cells in biore-
actors, Trends Biotech 9, 427-437.
Park YS, Ohta N, and Okabe M (1994), Neomycin production by partial
immobilization Streptomyces fradiae on cellulose beads in an air-lift
bioreactor, d Ferment Bioeng 78, 265-268.
Pavlica RT and Olson JH (1970), Unified design method for continuous-
contact mass transfer operations, lndEng Chem 62(12), 45-48.
Petrovie D, Po~arac D, and Skala D (1989), Hysteresis effects of minimum
fluidization velocity in a draft tube airlift reactor, Chem Eng Sci 44,
996-998.
Petrovic DL, Po~arac D, Dudukovic A, and Skala D (1990), Mixing time in
gas-liquid-solid draft tube airlift reactors, Chem Eng Sci 45, 2967-2970.
Petrovic DL, Po~arac D, Dudukovic A, and Skala D (1991), Hydrodynamics
and mass transfer in a draft tube bubble column, J Serb Chem Soc 56,
227-240.
Petrovic DL, Po~arac D, Dudukovic A, and Skala D (1993), Minimum flu-
idization velocity of large particles in a draft tube airlift reactor, Chem
Eng Sci 48, 2663-2667.
Petrovic DL, Po~arac D, and Dudukovic A (1995), Prediction of mixing
time in airlift reactors, Chem Eng Commun 133, 1-9.
Pigache S, Trystram G, and Dhoms P (1992), Oxygen transfer modeling and
simulations for an industrial continuous airlift fermentor, Biotech Bio-
eng 39, 923-931.
Pollard DJ, Ison AP, Ayazi Shamlou P, and Lilly MD (1994), Examination
of bioreactor heterogeneity with rheological different fermentation
broths, in Advances in Bioprocess Engineering, Galindo E and Ramirez
OT (eds), Kluwer, Dordrecht, 163-170.
Po~arac D and Petrovic D (1988), Experimental study of the minimum flu-
idization velocity in a three-phase external-loop air-lift reactor, Chem
Eng Sci 43, 1161-1165.
Po~arac D and Petrovic DL (1991), Prediction of the gas holdup and liquid
recireulation velocity in an external loop airlift reactor, d Serb Chem Soc
56, 359-369.
Potucek F (1990), Specific interfacial area in an airlift tower reactor, Coil
Czech Chem Commun 55, 981-986.
Redman J (1987), Deep shaft treatment for sewage, Chem Eng. (Lond) 441,
12-13.
Reinhard E, Kreis W, Barthlen U, and Helmbold U (1989), Semicontinuous
cultivation of Digitalis lanata cells: Production of ~-methyldigoxin in a
300-L airlift bioreactor, Biotech Bioeng 34, 502-508.
Reuss M, Debus D, and Zoll G (1982), Rheological properties of fermenta-
tion fluids, Chem Eng (Lond) 381,233-236.
Reuveny S (1990), Microcarrier culture systems, in Large-Scale Mammal-
ian Cell Culture Technology, Lubiniecki AS (ed), Marcel Dekker, New
York, 271-341.
Rhodes M and Birch J (1988), Large-scale production of proteins from
mammalian cells, Biotech 6, 518-523.
Ritchie BJ and Hill GA (1995), Biodegradation of phenol-polluted air using
an external-loop airlift bioreactor, d Chem Techn Biotech 62, 339-344.
 Appl Mech Rev vol 51, no 1, January 1998
Roca E, Ghommidh C, Navarro JM, and Lema JM (1995), Hydraulic model
of a gas-lift bioreactor with flocculating yeast, Bioproc Eng 12, 269-
272.
Rodriguez-Mendiola MA, Stafford A, Cresswell R, and Arias-Castro C
(1991 ), Bioreactors for growth of plant roots, Enzyme Microb Tech 13,
697-702.
Roy PK, Khan AW, Kumar J, Chopra SDK, and Basu SK (1992), Steroid
transformation in a laboratory-scale glass air-lift fermenter, World d Mi-
crobiol Biotech g, 399-401.
Rafter HM, Wan L, Lilbbert A, and SchOgerl K (1994), Interpretation of gas
residence time distributions in large airlift tower loop reactors, Bioproc
Engll, 153-159.
Rafter HM, Brrring S, and Schagerl K (1995), Fluiddynamic characteriza-
tion of airlift tower loop bioreactors with and without motionless mix-
tures, Bioproc Eng 12, 119-130.
Russell AB, Thomas CR, and Lilly MD (1994), Influence of vessel height
and top-section size on the hydrodynamic characteristics of airlift fer-
mentors, Biotech Bioeng 43, 69-76.
Russell AB, Thomas CR, and Lilly MD (1995), Oxygen transfer measure-
ments during yeast fermentations in a pilot scale airlift fermenter, Bio-
proc Eng 12, 71-79.
Sajc L, Vunjak-Novakovic G, Grubisic D, Kovacevie N, Vukovic D, and
Bugarski B (1995a), Production of anthraquinones by immobilized
Frangula alnus Mill. plant cells in a four-phase air-lift bioreactor, dppl
Microbiol Biotech 43,416-423.
Sajc L, Obradovic B, Vukovic D, Bugarski B, Grubisic D, and Vunjak-
Novakovic G (1995b), Hydrodynamic and mass transfer in a four-phase
external loop air lift bioreactor, Biotech Prog 11,420-428.
Sano Y, Yamaguchi N, and Adachi T (1974), Mass transfer coefficients for
suspended particles in agitated vessels and bubble columns, .1 Chem
EngJapan 7, 255-261.
Schmidt J, Nassar R, and LCtbbert A (1992), Influence of the wakes in bub-
ble driven multiphase flow systems, Chem Eng Sci 47, 2295-2300.
Schorr-Galindo S, Ghommidh C, and Guiraud JP (1995), Inulin enrichment
by fermentation in a flocculating yeast reactor, Biotech Lett 17, 1305-
1310.
Schiigerl K (1990), Comparison of the performance of stirred tank and air-
lift tower loop reactors, ,1 Biotech 13, 251-256.
Sch~gerl K, Liicke J, and Oels U (1977), Bubble column bioreactors, Adv
Biochem Eng 7, 1-84.
Schumpe A and Deckwer W-D (1987), Viscous media in tower bioreactors:
Hydrodynamic characteristics and mass transfer properties, Bioproc
Eng 2, 79-94.
Scragg AH, Cresswell RC, Ashton S, York A, Bond PA, and Fowler MW
(1989), Growth and alkaloid production in bioreactors by a selected
Catharanthus roseus cell line, Enzyme Microb Tech 11,329-333.
Sharp JM and Doran PM (1990), Characteristics of growth and tropane al-
kaloid synthesis in Atropa belladonna roots transformed by Agrobacte-
rium rhizogenes, JBiotech 16, 17 I- 186.
Shah YT (1979), Gas-Liquid Solid Reactor Design, McGraw-Hill, New
York, 351.
Shah YT, Kelkar BG, Godbole SP, and Deckwer W-D (1982), Design pa-
rameter estimations for bubble column reactors, AIChEJ, 28, 353-379.
Shetty SA, Kantak MV, and Kelkar BG (1992), Gas-phase backmixing in
bubble-column reactors, AIChE.1 38, 1013-1026.
Shi LK, Riba JP, and Angelino H (1990), Estimation of effective shear rate
for aerated non-Newtonian liquids in airlift bioreactor, Chem Eng
Commun 89, 25-35.
Smart NJ and Fowler MW (1984), An airlift column bioreaetor suitable for
large scale cultivation of plant cell suspensions, .1 Exp Bot 35, 531-537.
Smith BC and Skidmore DR (1990), Mass transfer phenomena in an airlift
reactor: Effects of solids loading and temperature, Biotech Bioeng 35,
483-491.
Snape JB and Thomas NH (1992), Modeling of particle transport by bub-
bles for performance guidelines in airlift fermentors, Biotechnol.Bioeng
40, 337-345.
Snape JB, Fialova M, Zahradnik J, and Thomas NH (1992), Hydrodynamic
studies in an external loop airlift reactor containing aqueous electrolyte
and sugar solutions, Chem Eng Sci 47, 3387-3394.
Sokolichin A and Eigenberger G (1994), Gas-liquid flow in bubble columns
and loop reactors: Part 1. Detailed modelling and numerical simulation,
Chem Eng Sci 49, 5735-5746.
Sokolov VN and Metkin VP (1981), Turbulent flow of pseudoplastic liq-
uids in tubes, Zh Prikl Khim, 54, 1321-1326; J Appl Chem USSR 54,
1103-1107.
Spieler H, Alferrnann W, and Reinhard E (1985), Biotransformation of 13-
methyldigitoxin by cell cultures of Digitalis lanata in airlift and stirred
tank reactors, Appl Microbiol Biotech 23, 1-4.
Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
Chisti: Pneumatically mixed bioreactors 111
Spier RE and Whiteside JP (1976), Production of foot-and-mouth disease
virus from BHK 21 C 13 cells grown on the surface of glass spheres,
Biotech Bioeng lg, 649-657.
Srivastava P and Kundu S (1995), Laboratory air lift bioreactor for cepha-
losporin-C production using Cephalosporium acremonium, Indian
Chem Eng, Section A, 37, 136-139.
Stasinopoulos SJ and Seviour RJ (1992), Exopolysacchafide production by
Acremonium persicinum in stirred-tank and air-lift fermentors, Appl Mi-
crobiol Biotech 36, 465-468.
Stejskal J and Potucek F (1985), Oxygen transfer in liquids, Biotechnol
Bioeng 27, 503-508.
Steiner R (1987), Operating characteristics of special bubble column reac-
tors, Chem Eng Process 21, 1-8.
Suh I-S, Schumpe A, Deckwer W-D, and Kulicke W-M (1991), Gas-liquid
mass transfer in the bubble column with viscoelastic liquid, Can J Chem
Eng 69, 506-512.
Sub I-S, Schumpe A, and Deckwer W-D (1992), Xanthan production in
bubble column and air-lift reactors, Biotech Bioeng 39, 85-94.
Sukan SS and Vardar-Sukan F (1987), Mixing performance of air-lift fer-
menters against working volume and draft tube dimensions, Bioproc
Eng 2, 33-38.
Suzuki T, Matsuo T, Ohtaguchi K, and Koide K (1995), Gas-sparged biore-
actors for CO2 fixation by Dunaliella tertiolecta, J Chem Tech Biotech
62, 351-358.
Svendsen HF, Jakobsen HA, and Torvik R (1992), Local flow structures in
internal loop and bubble column reactors, Chem Eng Sci 47, 3297-3304.
Tan WS, Dai GC, Ye W, and Shen JP (1995), Local flow behavior of the
liquid phase in an airlift bioreactor for potential use in animal cell sus-
pension cultures, Chem Eng J 57, B31-B36.
Tijhuis L, Zwols H, van Meekeren K-H, van Loosdreeht MCM, and Heijnen
JJ (1994a), Dynamics of biofilm formation on small suspended particles
in airlift reactors, in Progress in Biotechnology, vol 9, Proc of 6th
European Congress on Biotechnology, Alberghina L, Frontali L, and
Sensi P (eds), Elsevier, Amsterdam, 953-956.
Tijhuis L, van Benthum WAJ, van Loosdrecht MCM, and Heijnen JJ
(1994b), Solids retention time in spherical biofilms in a biofilm airlift
suspension reactor, Biotech Bioeng 44, 867-879.
Tijhuis L, van Loosdrecht MCM, and Heijnen JJ (1994c), Formation and
growth of heterotrophic aerobic biofilms on small suspended particles in
airlift reactors, Biotech Bioeng 44, 595-608.
Tijhuis L, Huisman JL, Hekkelman HD, van Loosdrecht MCM, and Hei-
jnen JJ (1995a), Formation of nitrifying biofilms on small suspended
particles in airlift reactors, Biotech Bioeng 47, 585-595.
Tijhuis L, van Loosdrecht MCM, and Heijnen JJ (1995b), Dynamics of
biofilm detachment in biofilm airlift suspension reactors, Biotech Bio-
eng 45, 481-487.
Townsley PM and Webster F (1983), Recycling airlift transfer fermenter for
plant cells, Biotechnol Let, 5, 13-18.
Trilleros JA and Recio A (1990), Mixing phenomena in airlift loop reactors,
1 Chem E Syrup Set 121,215-239.
Tyagi RD, Couillard D and Tran FT (1991), Comparative study of bacterial
leaching of metals from sewage sludge in continuous stirred tank and
air-lift reactors, Process Biochem 26, 47-54.
Tzeng Y-M and Young Y-H (1996), Production of thufingiensin from Ba-
cillus thuringiensis using a net-draft-tube, modified airlift reactor,
World d Microbiol Biotech 12, 32-37.
Ulonska A, Deckwer W-D, and Hecht V (1995), Degradation of quinoline
by immobilized Comamonas acidovorans in a three-phase airlift reac-
tor, Biotech Bioeng 46, 80-87.
van Benthum WAJ, Garrido-Femfindez JM, Tijhuis L, van Loosdrecht
MCM, and Heijnen JJ (1996), Formation and detachment of biofilms
and granules in a nitrifying biofilm airlift suspension reactor, Biotech
Prog 12, 764-772.
van Houten R, Hulshoff Pol LW, and Lettinga G (1994), Biological sul-
phate reduction using gas-lift reactors fed with hydrogen and carbon di-
oxide as energy and carbon source, Biotech Bioeng 44, 586-594.
van Loosdrecht MCM and Heijnen SJ (1993), Biofilm bioreactors for
waste-water treatment, Trends Biotech 11, 117-121.
van Sonsbeek HM, Verlaan P, and Tramper J (1990a), Hydrodynamics,
mixing and oxygen transfer in liquid-impelled loop reactors, 1 Chem E
Syrup Ser 121,259-268.
van Sonsbeek HM, Verdurmen REM, Verlaan P, and Tramper J (1990b),
Hydrodynamic model for liquid-impelled loop reactors, Biotech Bioeng
36, 940-946.
van Sonsbeek HM, van der Tuin SP, and Tramper J (1992), Mixing in liq-
uid-impelled loop reactors, Biotech Bioeng 39, 707-716.
Verlaan P, Tramper J, Van't Riet K, and Luyben KChAM (1986), Hydro-
dynamic model for an airlift-loop bioreactor with external loop, Chem
Eng J 33, B43-B53.
 112 SPECIALISSUE: FLUIDMECHANICSPROBLEMSIN BIOTECHNOLOGY Appl Mech Rev vol 51, no 1, January 1998

Verlaan P, Vos J-C, and van't Riet K (1989a), Axial dispersion and oxygen
transfer in the transition from bubble column to an airlift-loop reactor, d
Chem TechBiotech 45, 181 - 190.
Verlaan P, Vos J-C, and van't Riet K (1989b), Hydrodynamics of the flow
transition from a bubble column to an airlift-loop reactor, J Chem Tech
Biotech 45, 109-121.
Verlaan P, Van Eijs AMM, Tramper J, Van't Riet K, and Luyben KChAM
(1989e), Estimation of axial dispersion in individual sections of an air-
lift-loop reactor, Chem Eng Sci 44, 1139-I 146.
Vicente AA and Teixeira JA (1995), Hydrodynamic performance of a three-
phase airlift bioreactor with an enlarged degassing zone, Bioproc Eng
14, 17-22.
Vilchez C and Vega JM (1995), Nitrate uptakeby immobilized Chlamydo-
monas reinhardtiicells growing in airlift reactors, EnzymeMicrob Tech
17, 386-390.
Wachi S, Jones AG, and Elson TP (1991), Flow dynamics in a draft-tube
bubble column using various liquids, ChemEng Sci 46, 657-663.
Wagner K and Hempel DC (1988), Biodegradation by immobilized bacteria
in an airlift-loop reactor--Influence of biofilm diffusion limitation,
Biotech Bioeng 31,559-566.
Weiland P (1984), Influence of draft tube diameter on operation behaviour
of airlift loop reactors, Ger Chem Eng 7, 374-385.
Wenge F, Chisti Y, and Moo-Young M (1995), New method for the meas-
urement of solids holdup in gas-liquid-solid three-phase systems, lnd
Eng Chem Res 34, 928-935.
Wenge F, Chisti Y, and Moo-Young M (1996), Split-cylinder airlift reac-
tors: Hydraulics and hydrodynamics of a new mode of operation, Chem
Eng Commun 155, 19-44.
Whiteside JP and Spier RE (1981), Scale-up from 0.1 to 100 liter of a unit
process system based on 3-mm-diameter glass spheres for the produc-
tion of four strains of FMDV from BHK monolayer cells, Biotech Bio-
eng 23, 551-556.
Whitney PJ (1992), Novel bioreactors for the growth of roots transformed
by Agrobacterium rhizogenes, Enzyme Microb Tech 14, 13-17.
Winkler MA (1981), Biological Treatmentof Waste-Water,Ellis Horwood,
Chichester, 146-163.
WOhler W, Riidiger HW, and Passarge E (1972), Large scale culturing of
normal diploid cells on glass beads using a novel type of culture vessel,
Exptl Cell Res 74, 571-573.
Wu J, King G, Daugulis AJ, Faulkner P, Bone DH, Goosen MFA (1990),
Adaptation of insect cells to suspension culture, J Ferment Bioeng 70,
90-93.
Wu J-T and Jong J-Z (1994), Liquid-phase dispersion in an airlift reactor
with a net draft tube, Bioproc Eng 11, 43-47.
Wu J-Y and Wu W-T (1991), Fed-batch culture of Saccharomyces cere-
visiae in an airlift reactor with net draft tube, Biotech Prog 7, 230-233.
Wu J-Y and Wu W-T (1992), Glutamic acid production in an airlift reactor
with net draft tube, Bioproc Eng 8, 183-187.
Wu W-T and Wu J-Y (1990), Airlift reactor with net draught tube, d
FermentBioeng70, 359-361.
Wu W-T, Wu J-Y, and Jong J-Z (1992), Mass transfer in an airlift reactor
with a net draft tube, Biotech Prog8, 465-468.
Yamashita F (1993), Effects of partition plates on gas holdup in a bubble
column with and without a draught tube, J Chem Eng Japan 26, 742-
744.
Yonsel S, Langer B, and Deckwer W-D (1991), Bioreactor compafisons---
Equi-inocular cultivation of Trichosporon cutaneum in stirred tank and
airlift reactors, in Biochemical Engineering--Stuttgart, Reuss M,
Chmiel H, Gilles E-D, and Knackmuss H-J (eds), Gustav Fischer, New
York, 192-195.
Yuan N, Gu X, Li 1, Zhu D, and Ding F (1994), Study on an airlift loop re-
actor with membraneous gas distributor and its application in fermenta-
tions of baker's yeast, in Better Living Through Biochemical Engi-
neering, Teo WK, Yap MGS, and Oh SKW (eds), Univ of Singapore,
Singapore, 426-428.
Zenz FA (1993), Explore the potential of air-lift pumps and multiphase,
Chem Eng Progress 89(8), 51-56.
Zenz FA and Othmer DF (I 960), Fluidizationand Fluid-ParticleSystems,
Reinhold, New York, 427-433; 467-470.
Zhang Z, Chisti Y, and Moo-Young M (1995), Effects of the hydrodynamic
environment and shear protectants on survival of erythrocytes in sus-
pension, J Biotech 43, 33-40.
Zhao M, Niranjan K, and Davidson JF (1994), Mass transfer to viscous liq-
uids in bubble columns and air-lift reactors: Influence of baffles, Chem
Eng Sci 49, 2359-2369.
Zhou W, Holzhaur-Rieger K, and Schtlgerl K (1993), Cephalospofin C pro-
duction by a highly productive Cephalosporium acremonium strain in
an airlift tower loop reactor with static mixers, J Biotech 28, 165-177.
Zhou SQ, Tang LM, Lai GZ, Feng PS, and Yao RH (1994), Fields of flow,
pressure and shear stress in downcomer of an internal airlift loop reactor
based on computational hydrodynamics, in Better Living Through Bio-
chemical Engineering, Teo WK, Yap MGS, and Oh SKW (eds), Univ of
Singapore, Singapore, 477-479.

Yusuf Chisti is with the Chemical Engineering Department, University of Almeria, Spain. He holds
a BSc in chemistry from the University of dos, Nigeria, an MSc in biochemical engineering from the
University of London, United Kingdom, and a PhD in chemical engineering from the University of
Waterloo, Canada. He is a Chartered Engineer, CEng, in the UK. His previous appointments were
with The Polytechnic, Ibadan, Nigeria, CHEMBIOMEDLtd (Edmonton, Canada), and the University
o f Waterloo, Canada. Chisti is the author of the highly cited book Airlift B i o r e a c t o r s (Elsevier, Lon-
don, 1989). He has over 80 other publications in areas such as bioreactor design, downstream bio-
separations, microbial fermentations, enzyme technology, animal cell culture, bioremediation and
bioconversion of wastes, hydrodynamics and transport phenomena in multiphase systems, and
pharmaceutical bioprocessing. He is Associate Editor o f B i o t e c h n o l o g y A d v a n c e s , a member of the
Editorial Board of the I n d i a n C h e m i c a l Engineer, and has been a guest editor of other books and
journals. He is listed in W h o ' s W h o in E n g i n e e r i n g (American Association o f Engineering Societies,
Washington, 1994). He has been a consultant to the United Nations Industrial Development Organization (UNIDO), Vienna,
the Centro de Quimica Farmaceutica, Cuba, Apotex Fermentation lnc., Canada, Hoogovens Groep B V, the Netherlands, and
other institutions and companies worldwide. He is a member of the Institution of Chemical Engineers (UK), the Society of
Chemical Industry (UK), the New York Academy of Sciences, and the American Institute of Chemical Engineers.

Downloaded From: http://appliedmechanicsreviews.asmedigitalcollection.asme.org/ on 01/29/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use