You are on page 1of 9

Chemical Engineering Science 63 (2008) 1941 1949

www.elsevier.com/locate/ces

Hydrodynamics and mass transfer in a novel multi-airlifting


membrane bioreactor
Zhe Xu, Jian Yu
Hawaii Natural Energy Institute, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, 1680 East-West Road,
Honolulu, HI 96822, USA
Received 16 March 2007; received in revised form 15 December 2007; accepted 18 December 2007
Available online 20 February 2008

Abstract
A novel multiple-airlifting membrane bioreactor is built with four sintered stainless steel tubular lters as the risers and downcomers. This
work investigates the hydrodynamics including gas holdup, liquid velocity, liquid circulation and mixing times by aerating different number
of risers (one to three) at supercial gas velocities of 0.020.07 m/s The mass transfer phenomena, including oxygen mass transfer (kL a) and
effective molecular diffusivity of lactic and acetic acids through the walls of tubular lters, are also investigated. It is found that gas holdup
in individual risers increases linearly with the supercial gas velocity, and performs independently under multiple-airlifting conditions. The
vessel-based gas holdup and liquid velocity in downcomer(s) increase with aeration rate of individual risers as well as the number of risers.
The liquid velocity in downcomers reaches an upper limit (about 0.6 m/s), because of ow resistance or energy loss of liquid circulation. The
oxygen mass transfer coefcient (kL a) is primarily affected by gas holdup and the number of risers, and to some extent inuenced by liquid
velocity. The novel airlifter conguration results in good liquid mixing in the bioreactor that quickly reaches new steady state in response to a
sudden pH change from acid addition.
2008 Elsevier Ltd. All rights reserved.
Keywords: Multiphase reactors; Bioreactors; Airlifter; Hydrodynamics; Mass transfer

1. Introduction
Since the rst airlifter was invented by Lefrancois et al.
(1955) about 50 years ago airlifting reactors have found increasing applications in bio-industry, because of many advantages
such as simple structure, low energy consumption, good mixing and low shear stress to cells (Chisti and Moo-Young, 1987;
Chisti, 1989). Airlifters usually have two compartmentsa
riser where gas is sparged to give an upow of uids and a
downcomer where little or no gas exists to give a net downow of liquid. The difference of gas holdup between the two
compartments is the driving force for a stable and dened uid
circulation and mixing in the reactors. The most common conguration of airlifting is an internal airlifter with a draft tube
or an external airlifter with separated riser and downcomer.
The latter has better degassing and hence smaller gas holdup
Corresponding author. Tel.: +1 808 956 5873; fax: +1 808 956 2336.

E-mail address: jianyu@hawaii.edu (J. Yu).


0009-2509/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2007.12.026

in the downcomer than the former does (Chisti, 1989). These


conventional airlifting bioreactors are used for single culture or
fermentation under aerobic conditions.
Progress in biotechnology has generated new biosystems
that require novel bioreactors for new metabolic products
or improved yield and productivity. For instance, membrane
bioreactors have been used for simultaneous cell cultivation
and removal of the spent medium or metabolic wastes (Wang
and Zhong, 2006). In some applications, aerobic and anaerobic
cultures should work together for a complete biosynthesis or
bioconversion (Yu and Si, 2001). Although conventional bioreactors could be modied by adding external circulation lines
equipped with pump, lter or centrifuge to meet the demands
(Yu et al., 1999), the facility cost is high and continuous operation may have problems such as contamination, cell damage
and so on. Different airlifting congurations have been designed and investigated, but most of them are based on conventional structure of one riser with one downcomer (Bendjaballah
et al., 1999; Benyahia and Jones, 1997; Camarasa et al., 2001;

1942

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949

Choi, 2002; Freitas et al., 2000; Kawase and Hashimoto, 1996;


Mohanty et al., 2006; Nikakhtari and Hill, 2005; Vial et al.,
2002). A novel cascaded airlifting reactor was investigated by
Bakker et al. (1994) for hydrolysis of sucrose to glucose and
fructose on immobilized invertase. The advantage of this new
conguration, in comparison with multiple separate units, is
the elimination of interconnecting pipe and pumps. Liu et al.
(2000) designed an airlifting structure with two risers and
two downcomers that were connected by a gasliquid separator on the top and a liquid container at the bottom. These
new airlifting congurations are actually two airlifters that
are connected in series to increase the length of riser and
downcomer.
In this work, a multiple-airlifting membrane bioreactor
(MAMBR) was designed, fabricated and investigated for operational hydrodynamics and mass transfers. The bioreactor
could be operated with different numbers of risers (one to
three) with corresponding downcomers (three to one) to give
a control on overall mixing and gasliquid mass transfer. In
addition, the multiple risers and downcomers were microltration lter tubes, allowing molecular diffusion of metabolic
products through the wall membranes. We report the results
on hydrodynamics and mass transfer in this novel bioreactor,
including gas holdups in risers and downcomers, supercial
liquid velocities, liquid circulation and mixing times, oxygen mass transfer (kL a) in aerobic compartment and effective
molecular diffusivities of lactic and acetic acids across the
membranes of microltration lters.
2. Materials and methods
2.1. Reactor conguration
Fig. 1 is a schematic diagram of the MAMBR made from
acrylic plastics. The vessel had a diameter of 0.28 m and an
overall height of 1.10 m. Four sintered stainless steel lter tubes
(GKN Sinter Metals, IL, USA) were installed between the bottom (0.088 m high) and top (0.25 m high) sections, separating
the vessel into two compartments, a tube-side aerobic compartment of 0.016 m3 and a shell-side anaerobic compartment
of 0.032 m3 . Air and water were the working media for the
study, and all the experiments were conducted at 23 1 C. Air
could be sparged into the individual tubes (risers) at the bottom
through a jet (orice diameter of 1 mm). Air bubbles rose up
in the risers and left from water in the top degassing section.
The degassed water owed into the tubes that were not aerated
(downcomers) forming overall liquid circulations as shown in
Fig. 1. Three operation modes were investigated by introducing air into one, two and three risers, respectively, with the rest
tube(s) as the downcomer(s). The number of airlifting in the
MAMBR, therefore, could be easily changed for studies on hydrodynamics and mass transfer. For two risers, there were two
arrangements of the risers and downcomers, a parallel pattern
(P1) and a diagonal pattern (P2). The stainless steel lter tubes
had an external diameter of 6.5 cm and a wall thickness of
2.5 mm with a porosity of 20% and average pore size of 10 m.
In addition to oxygen mass transfer in the aerated tube-side

Fig. 1. Schematic diagram of multi-airlifting membrane bioreactor (MAMBR).


Symbols: Air (N2 ), inlets of air or nitrogen gas; DO, dissolved oxygen
probe; pH, pH probe; Pa and Pb , ports for pressure measurement; T1 .T4 ,
stainless steel membrane tubes; U1 and U2 , inclined manometer for pressure
measurement; S1 and S2 , liquid sampling ports.

compartment, molecular diffusion of substances such as nutrients and metabolic products occurs across the porous walls of
lter tubes under concentration gradients This novel bioreactor, therefore, could efciently integrate aerobic and anaerobic
cultures in one single vessel.
2.2. Experimental measurements
2.2.1. Gas holdup and liquid velocity
The air ow rate in individual risers was controlled and measured with a needle valve and ow meter (Cole-Parmer, IL,
USA). In order to measure the liquid velocity and gas holdup
in the risers and downcomers, one lter tube was replaced with
a transparent acrylic tube of the same size as shown in Fig. 1.
Without aeration, the tube performed as a downcomer. The liquid velocity at the tube center was measured as pressure drop
against the wall based on the working principal of Pitot tube.
Two inclined manometers were used to measure the pressure
difference. For turbulent pipe ow, the average velocity is about
80% of the measured maximum velocity (Geankoplis, 2004).
With aeration, the tube performed as a riser. The pressure difference between two ports (Pa and Pb , H = 38.1 cm apart) was
measured for difference of uid density and used to estimate
the gas holdup in the riser (or downcomer). The gas holdup of
individual risers (r ) or downcomers (d ) is determined with
r(d) =

h
,
H

(1)

where h is the pressure difference between the two ports.


An overall gas holdup (v ) based on total aerobic volume of

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949

tube-side compartment is estimated with


V =

n  r Vt
,
n  r Vt + VL

(2)

where n is the number of risers and Vt and VL are the


volumes of individual tube and tube-side compartment,
respectively.
2.2.2. Calibration of liquid velocity measurement
For liquid velocity measurement in downcomer, a calibration
facility of similar structure was used to calibrate the water ow
rate with the inclined manometers. A pair of Pitot tube was
mounted on a pipe (25 mm in diameter) to measure the pressure
difference when tap water ew upward at different volumetric
rates. The supercial liquid velocity (UL,t ) was correlated with
the pressure drop (P ):

2P
UL,t = Cp
,
(3)
L
where Cp is a dimensionless coefcient and w is the water
density. At a liquid velocity ranging from 0.13 to 0.47 m/s, the
value of Cp is 0.901, in agreement with the reported values of
0.981.0 (Geankoplis, 2004).
2.2.3. Measurement of circulation time
Acid tracer was used to measure the overall mixing time (tm )
for 95% homogeneity and liquid circulation time (tc ) for one
complete cycle between riser and downcomer. Five milliliters of
hydrochloric acid solution ( 5 M) was injected into the liquid
above a downcomer, and a pH probe (model UP-10, Denver,
CO, USA) located above a riser was used to monitor the change
of solution pH with time. The circulation time was estimated by
taking an average value of consecutive time intervals between
two pH peaks (Chisti, 1989).
2.2.4. Oxygen mass transfer coefcient kL a
The value of kL a under different operation conditions was
determined by a dynamic method (Chisti, 1989). A dissolved
oxygen (DO) probe (BIOFLO 110, New Brunswick Scientic Co., Edison, NJ, USA) was mounted in the top section
right above a downcomer as shown in Fig. 1. DO in water was rst removed by sparging nitrogen gas till the DO
concentration reached near zero. Aeration was then started
with different number of risers at different supercial gas velocities, and a time course was recorded till a steady-state
DO concentration was reached. By assuming good mixing and
negligible time delay of oxygen sensor, the following equation describes the dynamic response of DO concentration with
time:
dCL
= kL a(C CL ),
dt

(4)

where t refers to time and C and CL the saturated and instantaneous concentration of DO in water, respectively. Since the
time constant of the oxygen sensor is about 30 s (Mettler-Toledo

1943

GmbH, Ockerweg, Germany), a rst-order model (Eq. 5) is


used to predict the signal delay in the measurement of dynamic
response (Philichi and Stenstrom, 1989; Letzel et al., 1999):
CL C p
dCp
=
,
dt


(5)

where Cp is the instantaneous concentration of DO indicated


by sensor and  is the sensors time constant. Simultaneously
solving Eqs. (4) and (5) gives
Cp
=1+
C

ekL at kL ae(t/)
kL a 1


.

(6)

The volumetric oxygen mass transfer coefcient kL a is estimated by tting Eq. (6) on re-aeration curves. An average value
of kL a with 20% was obtained with duplicate or triplicate
measurements.
2.2.5. Molecular diffusion of organic acids through lter walls
The shell-side compartment was lled up with a fermentation solution containing lactic acid (16.8 g/L) and acetic acid
(6.3 g/L). An external circulation (10 L acid solution as shown
in Fig. 1) was provided for good mixing. The tube-side compartment was lled with a mineral solution containing no organic acids. Solution samples were taken at an interval of 1 h
to monitor the change of organic acid concentrations in the
two compartments. The organic acids were determined with an
HPLC (Shimadzu, Chiyoda-ku, Tokyo, Japan) equipped with
an organic acid column maintained at 65 C. They were eluted
by a watersulfuric acid solution (pH 2.5) at 0.8 mL/min, and
detected at UV 210 nm. The acid concentrations were calibrated
against standard solutions.
Assuming a quasi-steady-state diffusion through the porous
walls, the following equation gives a mass balance and concentration change with time for lactic acid:
NLA = A DLA,e

(CLA,s CLA,t ) dCLA,t


=
 VL
dt

(7)

and
DLA,e =

 DLA
,


(8)

where NLA is the moles of lactic acid transferred from the shell
side to the tube side per volume per second, A is the total membrane surface area (0.62 m2 ), DLA,e is the effective diffusivity
of lactic acid (m2 /s),  is the wall thickness of the lter tubes
(m), VL is the liquid volume of tube-side compartment (m3 )
and CLA,s and CLA,t refer to the lactic acid concentrations in
the shell side and tube side (mol/m3 ), respectively. The effective diffusivity of lactic acid depends on the diffusivity in water
(DLA , m2 /s), the open area fraction of porous wall () and a
tortuous path that is greater than the thickness of tube wall ()
by a factor of .

1944

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949

The value of DLA,e is obtained from a time course given by


the following equation that is derived from Eq. (7):
VL
A(V  +1)

ln

0.12

0
0
CLA,s
CLA,t

0
0
+CLA,s
V  CLA,t CLA,t
V  CLA,t

0.1


,

(9)

0.08
r

DLA,e t=

0.14

0.06

where V  is the ratio of tube-side volume to shell-side volume.


The same method is also used for the effective molecular diffusivity of acetic acid.

1-riser
2-riser
3-riser

0.04
0.02

3. Results and discussion

0
0

3.1. Gas holdup in riser(s)

UG,r

4
(10-2

m/s)

0.05
0.045

1-riser
2-riser
3-riser

0.04
0.035
0.03
v

The gas holdup in risers is the driving force of liquid circulation in airlifting reactors. Linear increase of gas holdup
(r ) with the riser-based supercial gas velocity (UG,r ) was observed under three operation conditions as shown in Fig. 2a.
At low gas velocities (UG,r < 0.04 m/s), the difference of gas
holdup with different number of risers was quite small. It indicates that the individual risers performed independently under
multiple-airlifting conditions and the gas holdup in one riser
was determined by its aeration rate, not by aeration in other risers. After the supercial gas velocity exceeded 0.06 m/s, however, a clear trend of higher gas holdup was observed when
three risers were operated, in comparison with one or two risers. This observation indicates that at high aeration rate the
gas holdup was to some extent affected by the number of risers and downcomers. When three risers with one downcomer
were used, the ratio of downcomer-to-riser cross-sectional area
(Ad /Ar ) was the smallest (0.33) and the liquid velocity in the
risers was slowest (data not shown here). Under these conditions, the air bubbles might have a relatively long retention time
in the risers, giving a relatively high gas holdup. The effect of
liquid velocity on gas holdup in riser has also been observed
by other investigators (Bendjaballah et al., 1999; Choi, 2002;
Vial et al., 2002).
The vessel-based gas holdup (v ), as shown in Fig. 2b, increases with the aeration rate of individual risers as well as the
number of risers. When the operation mode was shifted from
one riser to three risers, the measurement error also increased
because of the high turbulence and uctuation of uid ow.
The result shows that the oxygen mass transfer in this bioreactor can be controlled according to the desired respiration rate
of biological systems by adjusting the aeration rate of individual risers, the number of risers or both. In comparison, a
conventional internal or external airlifting bioreactor with one
riser and one downcomer has a xed Ad /Ar , and air ow rate
is the only operational parameter to meet the demands of good
mixing, less shear stress and different respiration rates. Many
times, it cannot meet all these demands.
In conventional internal airlifter, investigators have observed
that the overall gas holdup increases and levels off with gas
velocity (Chisti, 1989; Vial et al., 2002). This phenomenon is

0.025
0.02
0.015
0.01
0.005
0

UG,r (10-2 m/s)


Fig. 2. Effect of riser-based supercial gas velocity (UG,r ) on the gas holdups
with one, two and three risers, respectively: (a) gas holdup (r ) in individual
risers and (b) the overall gas holdups (v ) in the aerobic compartment.

attributed to the increased gas holdup in downcomer at high


gas ow rate (Chisti, 1989; Choi, 2002). In contrast, the gas
holdup of this multiple airlifter increases almost linearly with
supercial gas velocity. This fact suggests an efcient degassing
and a low gas holdup in downcomer(s) because of the separated
risers and downcomers. Gas holdup in downcomers is a direct
evidence of degassing efciency and described as follows.
3.2. Gas holdup in downcomer(s)
Fig. 3 shows the gas holdup in downcomer(s) when the reactor was operated with two and three risers, respectively. The
gas holdup in downcomer(s) was mainly caused by dissolved
air or tiny air bubbles that were brought into the downcomers
by liquid circulation before they left from the degassing zone.
The tiny bubbles, once they were formed, hardly escaped from
the downcomers because of the downward liquid ow, particularly when the reactor was operated with three risers at high
aeration rates. The gas holdup in downcomer could adversely

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949


0.7

1.2
2-riser P1
2-riser P2
3-riser

1-riser
2-riser P1
2-riser P2
3-riser

0.6
0.5
UL, d (m/s)

0.8
d10-2

1945

0.6

0.4
0.3

0.4

0.2
0.2

0.1

0
0

3
4
5
UG,r (10-2 m/s)

Fig. 3. The gas holdup (d ) in individual downcomer(s) with two and three
risers at different riser-based supercial gas velocities (UG,r ). For two risers,
the two downcomers had two positions relative to the risers, parallel position
(P1) and diagonal position (P2) (see Fig. 1).

affect the liquid circulation, mixing and mass transfer of airlifting bioreactors. In the worst conditions, however, it is less than
10% of the gas holdup in risers as shown in Figs. 2a and 3.
The liquid degassing in the top degassing zone could be affected by the distance and ow direction between risers and
downcomers. For two risers, the downcomers could be arranged
in two positions relative to the risers, a parallel position (P1)
and a diagonal position (P2) as shown in Fig. 1. Fig. 3 reveals
that the two positions (P1 vs P2) had statistically little inuence
on the downcomer gas holdup. This observation implies a high
degassing efciency of the multiple risers and downcomers that
are separated from each other. In contrast, a conventional internal airlifter often has a high gas holdup in downcomer at high
aeration rates because of relatively poor degassing efciency
(Chisti, 1989).
3.3. Supercial liquid velocity in downcomer(s)
The overall liquid circulation in this reactor was estimated
by measuring the supercial liquid velocity (Ul,d ) in downcomer(s). Fig. 4 shows the measured liquid velocity at different supercial gas velocities (UG,r ) with one, two and three
risers, respectively. It is interesting to note that with one or
two risers, the liquid velocity in downcomers increased almost
linearly with gas velocity. Under three risers, however, it approached a maximum value around 0.6 m/s. In addition, larger
increase in liquid velocity was observed when the number of
risers was increased from one to two than two to three, even
though each riser was operated at the same gas velocity. It
should be pointed out that with increasing of riser numbers (n),
the number of downcomers (m) was correspondingly reduced
(n + m = 4). As the ratio of cross-section areas of downcomers
to risers (Ad /Ar ) was linearly reduced from 3 to 1 and further
to 0.33, the liquid velocity in downcomers at a given gas velocity would also increase linearly, because of an almost constant
gas holdup difference as pointed out above (Figs. 2a and 3).

0
0

3
UG,r

4
(10-2

m/s)

Fig. 4. Supercial liquid velocities in downcomer(s) (UL,d ) at riser-based gas


velocity (UG,r ) with one, two and three risers, respectively. For operation of
two risers, the liquid velocities were measured in two downcomers that were
located at different positions relative to the riser(s), parallel position (P1) and
diagonal position (P2) (see Fig. 1).

The non-linear increase of liquid velocity with number of risers, therefore, might reect the increasing energy loss or ow
resistance under high turbulence and liquid velocity. It is also
interesting to note that the liquid velocity in individual risers
was substantially lower with three risers than those with one or
two risers (data not shown here).
Fig. 4 also shows that with two risers and two downcomers,
the parallel arrangement (P1) gave lower liquid velocity than
the diagonal arrangement (P2), particularly at high aeration
rates. Compared to the diagonal position, the parallel position
of risers and downcomers had not only longer ow path but
also stronger mixing of uids in the top and bottom zones
(Fig. 1), which resulted in a higher energy loss.
Liquid circulation velocity in airlifter is a crucial operational
parameter that affects the performance of bioreactors, such
as mixing time (Merchuk et al., 1996), oxygen mass transfer
(Sotiriadis et al., 2005), downcomer gas holdup (Chisti, 1989),
ow regime transition (Bendjaballah et al., 1999) and adhesive
strength of biolms (Chen et al., 2005). The liquid velocity
plateau under three risers, observed at high gas velocities, may
be attributed to the energy loss of liquid circulation, or to an
increased gas holdup in downcomer (Chisti, 1989). The former, as shown above, may be the major factor because the gas
holdup difference is actually increased linearly from 0 to 0.1
(compare Figs. 2a and 3). In other words, the liquid velocity
can be restricted due to energy loss of liquid circulation.
3.4. Circulation time and mixing time
Liquid mixing in airlifters depends to a great extent on the
overall liquid circulation between the risers and downcomers.
Fig. 5 is the time course of solution pH after a pulse acid
tracer was introduced into the multiple airlifter that was aerated
steadily at three rates with one, two and three risers, respectively. As expected, the time duration required to reach new

1946

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949


Table 1
Liquid circulation times (tc ) estimated from the tracer response curves (Fig.
5) and with Eq. (10)

pH

1-riser UG,r = 0.014 m/s


1-riser UG,r = 0.028 m/s

1-riser UG,r = 0.054 m/s


4

Number of
riser(s)

Gas velocitya
UG,r (102 m/s)

From tracer response


Fig. 5, tc (s)

With Eq. (10),


tc (s)

1
1
1
2
2
2
3
3
3

1.35
2.75
5.43
1.35
2.75
5.43
1.35
2.75
5.43

55
42
33
36
30
23
b
b
b

27
21
18
24
13
10
16
11
8

a The
b Not

2
0

25

50

75

100

125

150

175

200

same supercial gas velocities in individual risers for multiple risers.


measured.

225

time (s)

gas velocities, while with three risers the response curves became quite smooth. It demonstrates that this multiple-airlifting
bioreactor has a quick and smooth control on solution pH.
From the time gap between two peaks in the response curves
(Figs. 5a and b), the circulation time of overall liquid ow is
estimated and given in Table 1. The circulation time is reduced
from 55 s with one riser at a low gas velocity (UG,r =0.014 m/s)
to 23 s with two risers at a high gas velocity (UG,r =0.054 m/s).
An ideal circulation time is also calculated by using the liquid
velocity in the downcomers (UL,d ) with

tc = ((m + n) Vt + Vtop + Vbot )

2-riser U G,r = 0.014 m/s

pH

2-riser U G,r = 0.028 m/s


2-riser U G,r = 0.054 m/s

2
0

25

50

75

100

125

150

175

200

time (s)
8

7
3-riser U G,r = 0.014 m/s

pH

3-riser U G,r = 0.028 m/s


3-riser U G,r = 0.054 m/s
5

2
0

25

50

75

100

125

150

175

200

time (s)

Fig. 5. The response curves of solution pH to a pulse input of acid tracer at


three aeration rates: (a) one riser, (b) two risers and (c) three risers.

steady state of solution pH became short with increase in gas velocity and the number of risers. With one or two risers, the peaks
of solution pH could be clearly observed, particularly at low

1
,
UL,d Ad

(10)

where Ad refers to the total cross-section area of downcomers and Vt , Vtop and Vbot the volumes of individual tube, top
degassing zone and bottom distribution zone, respectively.
Eq. (10) assumes an ideal plug ow of a liquid circulation loop
without backmixing, local turbulence, overall dispersion and
so on. As given in Table 1, the ideal circulation time has a similar trend with regard to the effects of number of risers and gas
velocity. Generally speaking, the circulation time from tracer
response curves is about 2 times longer than the ideal one. In
actual liquid circulation, the time is extended because of nonplug ow conditions such as local backmixing, and dead zones
near the sudden changes of cross-section and ow direction.
The mixing performance of an airlifting bioreactor is often
characterized by a mixing time that is dened as the time required to reach a steady-state concentration (e.g. 95% homogeneity) after a pulse of tracer is introduced. Verlaan et al.
(1989) suggested a correlation between the mixing time (tm )
and liquid circulation time (tc ):
U L Lc
tm
= 0.093Pe = 0.093
,
tc
EL

(11)

where Pe is an overall Peclet number, UL the mean liquid circulation velocity, Lc the average length of one circulation loop
and EL the overall axial dispersion coefcient. The overall
Peclet number of liquid circulations ranged between 40 and 60
and increased with gas ow rate (Verlaan et al., 1989). For the

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949

1947

Lactic acid experimental data

1-riser

3.5
3

Fitting curve for lactic acid


Fitting curve for acetic acid

3-riser
2.5
y ( 10-6 m 2)

KLa (10-2 s -1)

Acetic acid experimental data

2-riser

2
1.5

4
3
2

1
0.5

0
0

UG,r (10-2 m/s)

10

time (10 s)

Fig. 6. Oxygen mass transfer coefcient (kL a) at different riser-based gas


velocities (UG,r ) with one, two and three risers, respectively.

turbulent liquid and gas ow in this study, Eq. 12 is recommended by Joshi et al. (1990) to estimate the value of EL
that depends on the diameter (dt ) and liquid velocity (UL,d ) of
downcomers:
EL,d(r) = 0.25 dt UL,d(r) .

(12)

At supercial gas velocities from 0.014 and 0.054 m/s, the mixing time of this multiple airlifter is reduced from 10.8 to 6.4 min
with one riser and from 6.2 to 4.2 min with two risers. These
estimated mixing times, however, are much longer than those
found with the tracer response curves (Fig. 5). Actually, it took
less than 3 min to reach a new steady pH level after an aliquot
of acid solution was added.
3.5. Oxygen mass transfer (kL a)
The specic volumetric oxygen mass transfer rate (kL a) in
this multiple airlifter depends on the number of risers and their
aeration rates as shown in Fig. 6. The oxygen mass transfer
from air to water mainly occurred in the risers because of the
very low gas holdup in downcomers (compare Figs. 2a and 3).
The mass transfer coefcient of whole airlifter is, therefore,
proportional to the number of risers. At a supercial gas velocity from 0.013 to 0.061 m/s, kL a was increased by 3-folds with
the increase of risers from one to three. In addition, its increase
with supercial gas velocity (UG,r ) is also in good agreement
with the linear increase of gas holdup as shown in Fig. 2a. Usually, liquid circulation velocity is also considered an important
factor on the value of kL a. High liquid circulation velocities,
for example, may help disperse air bubbles, reduce their coalescence and improve the interfacial contact between gas and
liquid. This positive effect of liquid circulation on oxygen mass
transfer, however, is often associated with high gas velocities.
For a conventional airlift with a xed riser/downcomer ratio,
it is impossible to identify the individual effects of gas holdup
and liquid circulation, because both of them are determined by
gas velocity. Our result shows that the effect of liquid velocity

Fig. 7. Measurement of diffusivity of lactic and acetic acids through the wall
of stainless steel lter tubes. The parameter y represents the right-hand side
of Eq. (9).

is not signicant compared to the number of risers and their gas


holdups. First, the value of kL a is increased in proportional to
the number of risers, rather than the non-linear increase of liquid
velocity as shown in Fig. 4. Second, when the liquid circulation
velocity approached a plateau at high gas velocities (Fig. 4),
the value of kL a increased almost linearly with gas velocity
(Fig. 6), implying a minor inuence of liquid velocity on oxygen mass transfer. Actually, a bubble column without a large
loop of liquid circulation may provide a higher kL a than an airlift does. Table 2 compares the values of kL a of representative
pneumatically agitated bioreactors including a bubble column,
an internal airlifter (IL-ALR), two external airlifters (EL-ALR)
and this novel multiple airlifter (MAMBR). The data show that
the oxygen mass transfer in this multiple airlifter can be controlled quite conveniently in a relatively broad range to meet
different respiration rates of cells.
3.6. Molecular diffusivity of lactic and acetic acids
Four stainless steel lter tubes were used as the multiple
riser(s) and downcomer(s) as shown in Fig. 1. Their porous
surface area (0.62 m2 ) also let molecules diffuse through the
walls under concentration gradients. This MAMBR, therefore, can integrate two different biological systems together.
For instance, simultaneous aerobic and anaerobic fermentations could be performed in the tube-side and the shell-side
compartments, respectively. Fig. 7 is the time course for
measurement of diffusivities of two fermentative acids, lactic and acetic acids, under concentration gradients from the
shell-side compartment toward the tube-side compartment (see
Fig. 1). The effective diffusivities are obtained from the slopes
of the linearized curves given by Eq. (9). For lactic acid,
DLA,e = 6.77 1011 m2 /s (R 2 = 0.9974), and for acetic acid,
DAC,e = 5.74 1011 m2 /s (R 2 = 0.9974). A semi-theoretical
equation (Geankoplis, 2004) is used to give the diffusivities of lactic and acetic acids in diluted aqueous solutions of

1948

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949

Table 2
Comparison of kL a among pneumatically agitated bioreactors in waterair system
Bioreactor

VL (102 m3 )

Ar (102 m2 )

Ad /Ar ()

Gas sparger

vvm (min1 )

kL a (102 s1 )

Reference

Bubble column

6.96

4.64

1.0.1.3

2.4.3.0

Chisti (1989)

IL-ALR

1.63

0.54

0.49

0.5.0.6

1.3.2.1

Fadavi and Chisti (2005)

EL-ALR (1)

2.60

1.89

0.46

1 mm, 106 holes


perforated plate
Jet-based
static
mixer
1 mm, 57 holes
perforated plate

0.4.0.8

0.2.0.5

Kawase and Hashimoto (1996)

EL-ALR (2)

2.30
1.20

0.62

0.20
0.28

0.4.0.8
With packing bed
0.3.0.4
Without packing
bed 0.2.0.3
0.4.0.6
1.1.1.3
1.7.1.9

0.9.1.9
1.7.1.9

Nikakhtari and Hill (2005)

MAMBR

1.63

0.28
0.57
0.85

3.00
1.00
0.33

1.6 mm, 6 holes


plate sparger

1 mm jet
2 1 mm jet
3 1 mm jet

9.964 1010 and 1.289 109 m2 /s, respectively. According to the manufacturers information, the open area of lter
walls () is about 20%, and the tortuous factor  is estimated
from the diffusivities of two organic acids. It is found that
the molecules of organic acids actually traveled 2.94.5 times
longer path than the wall thickness.

Cp
CP
C
CLA,s

4. Conclusions

CLA,t

The gas holdup and liquid circulation velocity in this


MAMBR are determined by the supercial gas velocity or
aeration rate as observed in conventional internal or external
airlifters. This novel airlifter, however, provides another operational parameter, the numbers of risers and downcomers,
to control the overall gas holdup and liquid circulation. It
is convenient to control the liquid mixing and oxygen mass
transfer for requirements of different biological systems, such
as different respiration rates during a batch culture and sensitivity to gas or liquid share stress. In addition, the multiple
risers and downcomers made from porous stainless steel lter
tubes provide a means of integrating two different biological
systems such as aerobic and anaerobic fermentations. Their
metabolites can be exchanged via molecule diffusion through
the porous membranes under concentration gradients while the
cells are maintained under different conditions. The information provided above could be used to design a novel MAMBR
that exhibits good liquid mixing, moderate shear stress to cells,
appropriate oxygen supply and enough membrane area for
metabolite exchange.

0
CLA,s
0
CLA,t
dt
DLA,e

Notation
Ad
Am
Ar
CL

the total cross-sectional area of downcomers, m2


overall surface area of four membrane tubes, m2
the total cross-sectional area of risers, m2
instantaneous concentration of dissolved oxygen in
water, kg/m3

DO
EL
EL,d
EL,r
h
H
kL a
Lc
m
n
NLA
P
Pe
t
tc
tm
UG,r

0.4.0.7
0.8.1.0
1.5.1.8
2.0.2.4

This work

instantaneous concentration of dissolved oxygen


indicated by oxygen sensor, kg/m3
dimensionless coefcient of Pitot tube calibration,
dimensionless
saturated concentration of dissolved oxygen in
water, kg/m3
instantaneous concentration of shell-side lactic
acid, kg/m3
instantaneous concentration of tube-side lactic acid,
(kg/m3 )
initial concentration of shell-side lactic acid, kg/m3
initial concentration of tube-side lactic acid, kg/m3
diameter of porous tubes, m
effective diffusivity of lactic acid in the porous wall
of stainless steel lter tubes, m2 /s
dissolved oxygen in percentage of saturation,
dimensionless
overall axial dispersion coefcient, dimensionless
axial dispersion coefcient in individual downcomers, dimensionless
axial dispersion coefcient in individual risers,
dimensionless
reading difference between two incline manometers, m
distance between two Pitot tube ports, m
volumetric mass transfer coefcient of oxygen in
water, s1
average length of one circulation loop, m
the number of downcomer(s), dimensionless
the number of riser(s), dimensionless
molar ux of lactic acid from shell side to tube
side, mol/m3 s
pressure drop between two measuring ports, Pa
Peclet number, dimensionless
time, s
circulation time in the tube-side compartment, s
mixing time in the tube-side compartment, s
supercial gas velocity based on the cross-sectional
area of individual risers, m/s

Z. Xu, J. Yu / Chemical Engineering Science 63 (2008) 1941 1949

UL
UL,d
UL,r
UL,t
VL
Vt
Vbot
Vtop
V

mean liquid circulation velocity, m/s


supercial liquid velocity in downcomer(s), m/s
supercial liquid velocity in riser(s), m/s
supercial liquid velocity in calibration device, m/s
the liquid volume of tube-side compartment, m3
volume of individual tubes, m3
liquid volume in the bottom section, m3
liquid volume in the top section, m3
volumetric ratio of tube side to shell side,
dimensionless

Greek letters

d
r
v


w


the wall thickness of the lter tubes (2.5 mm),


dimensionless
gas holdup in individual downcomers, dimensionless
gas holdup in individual risers, dimensionless
overall gas holdup in the tube-side compartment,
dimensionless
time constant of oxygen sensor, s
the open void fraction of the porous wall of lter
tubes (20%), dimensionless
water density, kg/m3
tortuosity of the porous wall of stainless steel tubes,
dimensionless

Acknowledgments
The authors appreciate the supports from I-PHA Biopolymers Ltd., EGI Technologies LLC and the School of Ocean and
Earth Science and Technology at University of Hawaii. Z.X.
is enrolled in the postgraduate program of bioengineering at
University of Hawaii.
References
Bakker, W., et al., 1994. Sucrose conversion by immobilized invertase
in a multiple airlift-loop bioreactor. Biotechnology Progress 10 (3),
277283.
Bendjaballah, N., et al., 1999. Hydrodynamics and ow regimes in external
loop airlift reactors. Chemical Engineering Science 54 (21), 52115221.
Benyahia, F., Jones, L., 1997. Scale effects on hydrodynamic and mass
transfer characteristics of external loop airlift reactors. Journal of Chemical
Technology and Biotechnology 69 (3), 301308.
Camarasa, E., et al., 2001. Development of a complete model for an air-lift
reactor. Chemical Engineering Science 56 (2), 493502.
Chen, M.J., et al., 2005. Effects of operating conditions on the adhesive
strength of Pseudomonas uorescens biolms in tubes. Colloids and
Surfaces BBiointerfaces 43 (2), 6171.

1949

Chisti, M.Y., Moo-Young, M., 1987. Airlift reactors: characteristics


applications and design considerations. Chemical Engineering
Communications 60, 195242.
Chisti, Y., 1989. Airlift Bioreactors. Elsevier, London.
Choi, K.H., 2002. Effect of unaerated liquid height on hydrodynamic
characteristics of an external-loop airlift reactor. Chemical Engineering
Communications 189 (1), 2339.
Fadavi, A., Chisti, Y., 2005. Gasliquid mass transfer in a novel forced
circulation loop reactor. Chemical Engineering Journal 112 (13), 7380.
Freitas, C., et al., 2000. Hydrodynamics of a three-phase external-loop airlift
bioreactor. Chemical Engineering Science 55 (21), 49614972.
Geankoplis, C.J., 2004. Transport Processes and Separation Process
Principles (Includes Unit Operations). Prentice-Hall, Englewood Cliffs, NJ,
pp. 90446.
Joshi, J.B., et al., 1990. Sparged loop reactors. Canadian Journal of Chemical
Engineering 68 (5), 705741.
Kawase, Y., Hashimoto, N., 1996. Gas hold-up and oxygen transfer in threephase external-loop airlift bioreactors: non-Newtonian fermentation broths.
Journal of Chemical Technology and Biotechnology 65 (4), 325334.
Lefrancois, M.H., et al., 1955. Effectiomements aux Proceds de Cultures
forgiques et de Fermentations Industrielles. Brevet dInvention, No. 1,
pp. 102200.
Letzel, H.M., Schouten, J.C., Krishna, R., van den Bleek, C.M., 1999. Gas
holdup and mass transfer in bubble column reactors operated at elevated
pressure. Chemical Engineering Science 54 (1314), 22372246.
Liu, Y.M., et al., 2000. Liquid circulation in a multi-tube air-lift loop reactor.
Chinese Journal of Chemical Engineering 8 (3), 267271.
Merchuk, J.C., et al., 1996. Liquid ow and mixing in concentric tube airlift reactors. Journal of Chemical Technology and Biotechnology 66 (2),
174182.
Mohanty, K., et al., 2006. Hydrodynamics of a novel multi-stage external
loop airlift reactor. Chemical Engineering Science 61 (14), 46174624.
Nikakhtari, H., Hill, G.A., 2005. Hydrodynamic and oxygen mass transfer
in an external loop airlift bioreactor with a packed bed. Biochemical
Engineering Journal 27 (2), 138145.
Philichi, T.L., Stenstrom, M.K., 1989. Effects of dissolved oxygen probe lag
on oxygen transfer parameter estimation. Journal of the Water Pollution
Control Federation 61 (1), 8386.
Sotiriadis, A.A., et al., 2005. Bubble size and mass transfer characteristics
of sparged downwards two-phase ow. Chemical Engineering Science 60
(22), 59175929.
Verlaan, P., et al., 1989. Estimation of axial-dispersion in individualsections of an airlift-loop reactor. Chemical Engineering Science 44 (5),
11391146.
Vial, C., et al., 2002. Experimental and theoretical analysis of the
hydrodynamics in the riser of an external loop airlift reactor. Chemical
Engineering Science 57 (2223), 47454762.
Wang, S.J., Zhong, J.J., 2006. Bioreactor engineering. In: Yang, S.T. (Ed.),
Bioprocessing for Value-Added Products from Renewable Resources: New
Technologies and Applications. Elsevier, Amsterdam, p. 131.
Yu, J., Si, Y.T., 2001. Dynamic study and modeling on formation of
polyhydroxyalkanoates coupled with treatment of high strength wastewater.
Environmental Science and Technology 35, 35843588.
Yu, J., et al., 1999. A three-phase uidized bed reactor in the
combined anaerobic/aerobic treatment of wastewater. Journal of Chemical
Technology and Biotechnology 74 (7), 619626.