Characterization of Gas Transfer and
Mixing in a Bubble Column Equipped
with a Rubber Membrane Diffuser
Bjarne Rask Poulsen, Jens Jørgen Lønsmann Iversen
Institute of Biochemistry, Odense University, Campusvej 55, DK-5230
Odense M, Denmark; telephone: +45-6615-8600 (ext. 2376/2430);
fax: +45 6557 2467; e-mail: jjl.iversen@biochem.ou.dk
Received 20 June 1997; accepted 31 October 1997
Abstract: Gas transfer and mixing were characterized in
a 32-L bubble column reactor equipped with a commer-
cially available rubber membrane diffuser. The perfor-
mance of the membrane diffuser indicates that the slits
in the membrane are best described as holes with elastic
lids, acting as valves cutting off bubbles from the gas
stream. The membrane diffuser thus functions as a one-
way valve preventing backflow of liquid. Our design of
the bottom plate of the reactor enabled us to optimize the
aeration by changing the tension of the membrane. We
thereby achieved mass transfer coefficients higher than
those previously reported in bubble columns. A strong
dependence of mass transfer on gas holdup and bubble
size was indicated by estimates based on these two vari-
ables. The microalga, Rhodomonas sp., sensitive to
chemical and physical stress, was maintained for 8
months in continuous culture with a productivity identi-
cal to cultures grown in stirred tank reactors. © 1998 John
Wiley & Sons, Inc. Biotechnol Bioeng 58: 633–641, 1998.
Keywords: bioreactor; elastic sparger; bubble size; oxy-
gen transfer efficiency; characteristic mixing time;
wrinkled bubbles
INTRODUCTION
Rubber Membrane Spargers
Membrane diffusers are often used to facilitate aerobic mi-
crobial degradation of organic material in biological waste-
water treatment plants due to their efficient aeration. Dif-
fusers with perforated rubber membranes have several ad-
vantages compared with rigid spargers; small bubbles are
formed with high oxygen transfer rates due to a large inter-
facial area between gas and liquid phases. Also, the pressure
drop across the membrane is low compared with that of
rigid spargers. These characteristics are maintained at high
gas flow rates, giving competitive energy costs. Further-
more, the membrane functions as a one-way valve prevent-
ing backflow of liquid. Liquid is thus avoided in the gas
supply system and, in contrast to rigid spargers, the slits in
Correspondence to: J. J. L. Iversen
Contract grant sponsors: The Danish Research Councils; Novo Nor-
disk A/S
© 1998 John Wiley & Sons, Inc.
the membrane are not clogged by suspended material and
microbial growth.
The aim of this work was to investigate the application of
these advantages of membrane diffusers in laboratory-scale
bioreactors.
Comparative Studies
Reports of similar work are scarce and difficult to find
because a generally accepted name has not been given to
membrane spargers. To our knowledge the use of commer-
cially available membrane diffusers in bioreactors has not
been reported.
Rice et al. (1980) presented results obtained with diffus-
ers of their own construction using an ‘‘oscillating rubber
sheet sparger’’ in a bubble column. The gas holdup was
doubled compared with a traditional ‘‘perforated plate
sparger.’’ More detailed results were reported by Rice et al.
(1981) who used a 1.4-mm-thick vulcanized rubber sheet
with a diameter of 7.1 cm and 3.0 to 7.7 holes z cm−2 (the
diameter of the nails used to pierce the rubber sheet was 2.0
and 2.8 mm, respectively) in a 9.5-cm-inner-diameter
bubble column. They reported that these ‘‘flexible rubber
spargers’’ were self-cleaning due to oscillating hole size,
produced a reduced bubble size, and allowed for signifi-
cantly higher gas rates and voidages (gas holdup) before
flooding occurred. They suggested that this was due to
‘‘elastic holes’’ acting as valves or ‘‘knives’’ cutting off the
air stream during bubble formation, and concluded that the
increased uniformity of bubbles reduced the axial mixing.
Later, Rice and Littlefield (1987) reported that an ideal
bubble flow (small uniform bubbles) was maintained at
relatively high superficial gas velocities with a ‘‘punctured
rubber membrane’’ in a bubble column giving minimal mix-
ing.
Deckwer et al. (1982) measured the volumetric gas–
liquid mass transfer coefficient, kLa (per second), obtained
with a ‘‘rubber plate distributor’’ made from a 2-mm-thick
rubber sheet with 6.5 holes z cm−2 (needle diameter 0.3 mm)
placed in a bubble column (14-cm inner diameter, 42-L
liquid volume). The kLa (0.015 s−1 at superficial gas veloc-
ity vs 4 0.02 m z s−1; method: gassing-in N2 in 0.7% CMC
CCC 0006-3592/98/060633-09
solution) of this construction was approximately twice the
size of the value observed with perforated or sintered plates.
Deckwer’s group also noticed that, with this rubber plate
distributor, clogging was avoided because of its elastic
properties. Medic et al. (1989) used a ‘‘perforated rubber
aeration pad’’ in a bubble column with a 2-m2 cross-
sectional area and from 2 to 5 m of liquid above the mem-
brane. The aeration pad made from reinforced rubber of 1.5
mm thickness with 4 holes z cm−2 (1 mm approx. size, made
with a needle) covered the entire bottom. They achieved
higher kLa values (ca. 0.01 s−1 at vs 4 0.004 m z s−1;
method: sulfite oxidation in 2 mM sodium sulfite solution)
than expected from correlations obtained with other sparger
types. Weber et al. (1990) described a tubular ‘‘elastic
sparger’’ made from 3.2-mm-thick natural latex rubber with
12 holes z cm−2 (0.6 mm approx. slit size, made with a
needle). The kLa values (0.061 s−1 at vs 4 0.03 m z s−1;
method: gassing-out N2 in tap water, 9-L liquid volume)
were twice the values of a similar rigid sparger. They found
no effect of prestretching the membrane in the longitudinal
direction of the slits. They observed that the enlargement of
the orifices as gas pressure and flow rate increased delayed
the onset of jetting, resulting in more uniform bubbles over
a wider range of flow rates than with a rigid sparger. Only
one report (Talbot et al., 1990) has described a ‘‘porous
membrane diffuser’’ inferior to other spargers. Similar kLa
values (0.011 h−1 at vs 4 0.004 m z s−1; method: gassing-in
N2 in distilled water, 50-L liquid volume) were measured
for the porous membrane and the porous plate investigated.
The details of the membrane diffuser were not given, but a
slightly larger average bubble diameter was observed with
the porous membrane as compared with the porous plate, in
contrast to results obtained with other membrane diffusers.
Thus, apart from the results of Talbot et al. (1990) im-
proved mass transfer has been observed in all cases in which
membrane diffusers were compared with traditional rigid
spargers in bubble columns. This improvement is caused by
a reduction in bubble size, maintained at high gas velocities,
thereby increasing the operational range of bubble columns.
However, most of the kLa values reported in these studies
were relatively low and the improvement of gas transfer was
always at the expense of mixing intensity.
Membrane Characteristics
To explain the improved gas-transfer performance of mem-
brane spargers as compared with rigid spargers, Rice and
Howell (1986) developed an ‘‘elastic theory’’ by which the
shear modulus of ‘‘elastic membrane spargers’’ can be cal-
culated from the vertical displacement of the membrane at
the center and the pressure difference across the membrane.
They introduced this as an important element in membrane
sparger design. Geary and Rice (1991) predicted the bubble
size from rigid and ‘‘flexible spargers,’’ but several vari-
ables have to be measured to use their theory.
A more quantitative description of the excellent perfor-
mance of membrane spargers can be found in the flow rate–
634 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 58, NO. 6, JUNE 20, 1998
pressure drop relationship. The Bernoulli relationship is
modified to give the orifice equation (Rice and Howell,
1986):
Qo 4 Co z Ao z (2 z Dp/rg)1/2 (1)
where Qo (m3 z s−1) is the volumetric gas flow rate per ori-
fice, Co (dimensionless) is the orifice coefficient, Ao (m2) is
the orifice area, Dp (Pa) is the pressure difference across the
membrane sparger, and rg (kg z m−3) is the gas density. If Co
is assumed to be constant it is seen from Eq. (1) that
Qo z Dp−1/2 is proportional to the area of the hole. Rice et al.
(1981) observed a linear relationship between Qo z Dp−1/2
and Dp and concluded that the hole area is proportional to
the pressure difference across the membrane. Consequently,
the membrane is self-regulating: the hole area increases
when increasing pressure is applied. The proportionality
between Qo z Dp−1/2 and Dp for elastic holes observed by
Rice et al. (1981) can be rewritten as a proportionality be-
tween Dp and Q2/3 o . From the experimentally observed lin-
earity between Dp and Q2/3 o , Weber et al. (1990) also con-
cluded that there was proportionality between the hole area
and the pressure difference across an ‘‘elastic sparger.’’ In
comparison, Eq. (1) shows that, for rigid holes, Dp is pro-
portional to Q2o because the hole area is constant.
Model Media
Equilibrium between coalescence and dispersion of bubbles
is obtained so rapidly in coalescent liquids that formation of
bubbles smaller than the equilibrium size will have no in-
fluence on the actual bubble size in the reactor (Heijnen and
van’t Riet, 1984). Consequently, when characterization of
spargers is made in coalescent liquids the improved perfor-
mance of a sparger due to formation of bubbles smaller than
equilibrium size will not be observed. This is in agreement
with observations by Bovonsombut et al. (1987), who found
that the improved mass transfer obtained with a porous plate
relative to a perforated ring was more evident in noncoa-
lescent liquid. A minimal medium for growth of heterotro-
phic microorganisms is mostly coalescent, but, during batch
culture, the broth often changes to mainly noncoalescent, as
the result of the accumulation of metabolites excreted by the
microorganisms (Oolman and Blanch, 1986). Therefore,
both coalescent and noncoalescent liquids were used to
characterize the performance of the rubber membrane in this
study.
MATERIALS AND METHODS
Bubble Column Reactor
A perspex prototype bubble column was built in our work-
shop. It was 2 m high with an inner diameter of 15 cm and
insertions for probes at 5 cm, 74 cm, 128 cm, and 197 cm
above the bottom plate. The maximal working volume was
32 L. As shown in Figure 1, the membrane sparger covered

Figure 1. Cross-section of lower part of bubble column.
the entire bottom of the reactor (aerated area Ai 4 170 cm2)
and consisted of a modified commercially available mem-
brane diffuser, Sanitaire 99 (Water Pollution Control Corp,
WI), a 2-mm-thick EPDM rubber plate with 15 holes z cm−2
and a perforated area of 295 cm2. The holes in the mem-
brane are slits of an approximate length of 1 mm cut in the
tangential direction. Before use, growth-inhibiting chemi-
cals in the membrane were removed by autoclaving in 5%
Deconex (Borer Chemie AG, Zuchwil, Switzerland) solu-
tion and rinsed in distilled water. This was repeated three
times.
The central part of the membrane was positioned as
shown in Figure 1. By moving the inlet up and down it was
possible to change the tension of the membrane until the
best performance was observed (i.e., the highest gas
holdup).
Characterization was done at 20°C in distilled water and
0.25 M NaCl as coalescent and noncoalescent liquids, re-
spectively (Oolman and Blanch, 1986). The gas flow rate
was controlled and measured with a mass flow controller
(Bronkhorst Hi-Tec, Type F-201AC-FD-44-V, Ruurlo, The
Netherlands). The liquid volume was between 25 and 31 L
depending on the gas flow rate, which was varied between
5 and 30 L z min−1 corresponding to superficial gas veloci-
ties between 0.006 and 0.033 m z s−1, respectively, and
power inputs per liquid volume between 100 and 650
W z m−3, respectively. The superficial gas velocity is de-
fined as the gas flow rate divided by Ai and is equal to the
gas velocity in the reactor without liquid. The gas holdup, e
(dimensionless), was determined from surface levels of
nonaerated liquid and dispersion as the ratio between the
volume of gas bubbles and the total volume of liquid and
gas (dispersion). The oxygen probe was a polarographic
YSI-5331 oxygen probe with a YSI standard Teflon mem-
brane (YSI Inc., Yellow Springs, OH), giving half-times
POULSEN AND IVERSEN: BUBBLE COLUMN EQUIPPED WITH A RUBBER MEMBRANE DIFFUSER
between 0.5 and 1 s, placed 128 cm above the bottom plate.
pH was measured with two glass electrodes (GK4103C,
Radiometer, Copenhagen, Denmark) placed 5 cm and 197
cm above the bottom plate.
Photographs of Bubbles
Air bubbles were photographed in the bubble column using
a black background and light perpendicular to the direction
of photography from a 300-W video light (Type 8S, Kaiser
Fototechnik GmbH & Co., Buchen, Germany) connected to
an adjustable power supply. The lens used was a Micro
Nikor equipped with a filter (PL-linear M52, Nikon). Pic-
tures were taken using Fuji Neopan 1600 film (lens aperture
8, shutter speed 10−3 s). A ruler was placed on the outside
of the reactor wall for measurement of bubble size. There
was no optical distortion through the liquid and perspex in
the vertical direction and the distortion in the horizontal
direction was small (<2.5%) when bubbles near the reactor
wall were photographed. This was confirmed by placing a
ruler in the liquid inside the bubble column.
Measurement of Gas–Liquid Mass Transfer
The dynamic methods for determination of kLa are based on
an abrupt formation of a difference between the oxygen
concentration in the liquid in equilibrium with the gas phase
CG (mol z m−3) and the oxygen concentration in the liquid
CL (mol z m−3), followed by recording of CL. We deter-
mined the kLa by two different dynamic methods: the stan-
dard dynamic method (SDM), in which the sparger gas is
changed from atmospheric air to nitrogen or vice versa
(Bartholomew et al., 1950); and the pressure step dynamic
method (PDM; Linek et al., 1989), in which the pressure in
the reactor is abruptly changed by 20%. We obtained the
pressure changes by switching the outflow gas through a
column of liquid. Measurements using both SDM and PDM
were automatized using an analog/digital interface (Degn
and Nielsen, 1987) and an IBM-compatible computer run-
ning specially written computer programs.
Measurement of Mixing Intensity
Mixing was characterized using a buffer (2 mM succinate
and malonate) in the model medium giving a linear rela-
tionship between the tracer concentrations (20 mL 1 M HCl
or NaOH ∼1 mM in the reactor) and pH (Poulsen and
Iversen, 1997). Acid and base reservoirs were pressurized to
prevent formation of air bubbles in the tubings and to in-
crease the accuracy of tracer/titrant additions. The differ-
ence between the signals from two pH probes placed at the
top and bottom of the reactor was used as the tracer signal.
The characteristic mixing time, tM, was defined as the time
constant of the exponential equation fitted to the response
curve (Poulsen and Iversen, 1997).
635
RESULTS AND DISCUSSION takes place when the local pressure difference is established
again, giving an oscillating behavior.

Mode of Action
Bubble Morphology
As seen in Figure 2, the membrane sparger described in this
study has a flow rate–pressure drop relationship character- Photographs were taken of air bubbles in coalescent and
istic for rigid spargers. This unexpected result is in contrast noncoalescent liquids (Figs. 3 and 4) in the bubble column.
to the observations of Rice et al. (1981) and Weber et al. As a consequence of coalescence the bubbles in distilled
(1990), who found a linear relationship between Dp and water are larger than the bubbles in the noncoalescent 0.25
Q2/3
o for their ‘‘elastic spargers.’’ Therefore, the holes in the
M sodium chloride solution. If the bubbles are assumed to
membrane used here are obviously not elastic. This can be be spherical the specific gas–liquid interfacial area per unit
explained by the membrane material, which is a hard rubber liquid volume, a (m−1), is given by:
(EPDM) as compared with the latex used by Weber et al.
AG
(1990). Co is assumed constant, but might vary due to a=
changes in the Reynolds number. However, it is unlikely VL
that this would give the exact flow rate–pressure drop rela- AG e
= ?
tionship characteristic of rigid spargers. VG ~1−e!
The critical pressure difference, Dpc (Pa), at which the (4p~db,i / 2!2 e
slits in the EPDM membrane open was 4.6 kPa (Fig. 2). = ?
3 ~1−e !
(4 / 3p~db,i / 2!
Weber et al. (1990) also reported threshold values for slit
opening (20 kPa). However, the total area of the holes did 6(d2b,i e
= ?
not increase significantly when the pressure difference (db,i
3 ~ 1−e !
across the sparger increased (Fig. 2). Therefore, we suggest 6 e
that the membrane slits can be described as holes with elas- = ? (2)
ds ~1−e!
tic lids, rather than elastic holes characterized by increasing
hole area at increasing pressure. This suggestion is in agree- where AG is the total interfacial area of bubbles, VL (m3)is
ment with the idea that the membrane slits act as valves, the liquid volume, VG is the total volume of bubbles, db,i (m)
cutting off bubbles during their birth process, thereby re- is the diameter of the ith bubble, and ds (m) is the Sauter
ducing bubble size (Rice et al., 1981). This ‘‘knife’’ action mean bubble diameter, often used to correlate bubble size
is possibly caused by a local pressure drop after the release with mass transfer:
of one bubble to a value below Dpc, which is, by definition,
insufficient to maintain an open hole/elastic lid. Reopening

Figure 2. Pressure difference across membrane sparger in bubble column

at different liquid volumes: s, 0 L; n, 5 L; ,, 10 L; L, 15 L; +, 25 L
(corresponding to hydrostatic pressures above the membrane of 0, 3.00,
6.15, 9.37, and 15.5 kPa, respectively). The equation Dp 4 c1 z Q2o + Dpc
is fitted to the results (z z z z), giving Dp 4 2.5 z 1014 kPa z m−6 z s2 z Q2o + 4.6
kPa; r2 4 0.96.
Figure 3. Photograph of air bubbles in coalescent liquid (distilled water)
in bubble column at a superficial gas velocity of 0.0153 m z s−1. Bar 4
1 mm.

636 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 58, NO. 6, JUNE 20, 1998

Figure 4. Photograph of air bubbles in noncoalescent liquid (0.25 M
NaCl) in bubble column at a superficial gas velocity of 0.0153 m z s−1. Bar
4 1 mm.
(d3b,i
ds =
(d2b,i
The Sauter mean bubble diameter (ds) is the diameter of a
hypothetical bubble with a volume:area ratio equal to the
average ratio of all bubbles. We also calculated the mean
bubble diameter dm:
(db,i
dm =
N
where N is the number of bubbles. The different definitions
(ds and dm) of average bubble diameter can give very dif-
ferent values. ds is always larger than or equal to dm. In
coalescent and noncoalescent liquids dm was 4.5 mm and
0.52 mm, respectively, and ds was 5.0 mm and 1.1 mm,
respectively (Table I). In a stagnant pool of water, these
Sauter mean bubble diameters were shown to give bubble
rise velocities vb (m z s−1) of around 0.25 m z s−1 and 0.20
m z s−1, respectively (Clift et al., 1978; Table I). The air
bubbles were not photographed at every superficial gas ve-
locity used, but we did not visually observe any difference
in bubble size except from a few large bubbles in the upper
part of the reactor at high gas velocities in the coalescent
liquid. In distilled water the bubbles were of approximately
the same size as in 0.25 M sodium-chloride close to the
membrane surface, but already at 5 cm above the membrane
the bubble size had stabilized to the equilibrium value char-
acteristic of the liquid. The formation of bubbles was there-
fore apparently independent of the liquid coalescence prop-
erties.
It is seen in Figure 3 that the bubbles in the coalescent
liquid are neither spherical nor ellipsoidal as usually as-
sumed. The shape is undefined and the surface is wrinkled,
increasing the surface area. This unusual surface pattern has
to our knowledge not been reported before, which can be
explained by our observation that the wrinkled surfaces
(Fig. 3) are only seen when pictures are taken using light
perpendicularly directed to the focus line, in combination
with the filter as described in Materials and Methods. If the
filter is omitted the bubble surfaces appear normal (smooth)
(3) on the photographs (not shown). The wrinkled bubble sur-
face is probably due to local microturbulence at the bubble
surface caused by the high bubble rise velocity characteris-
tic of large bubbles. Turbulence is also created by high local
densities of bubbles (bubble swarms), resulting in high local
liquid velocities. The height of such local ‘‘airlifts’’ was
visually determined to be in the range of 5 to 20 cm. As a
result of wall effects the upward liquid velocity is highest in
(4) the middle of the column.
In their study of aeration in bubble columns Heijnen and
van’t Riet (1984) distinguished between fine bubble sys-
tems (small bubbles, 0.5 mm < ds < 1 mm) and coarse
bubble systems (large bubbles, 4 mm < ds < 6 mm). Fine
bubble systems are obtained with porous discs, two-phase

Table I. Characteristic variables in coalescent and noncoalescent liquids.

Liquid

Variable Coalescent Noncoalescent Reference

ds 3.0 mm 1.0 mm Burckhart and Deckwer (1976)

ds 5.0 mm 1.1 mm This work
eest 4 s z m−1vs 5 s z m−1vs This work
e 0.6vs0.7 — Heijnen and van’t Riet (1984)
e 4.4 s z m−1vs 5.5 s z m−1vs This work
vb 0.25 m z s−1 0.20 m z s−1 Clift et al. (1978)
vb,ave 0.23 m z s−1 0.18 m z s−1 This work
(kLa)est 2.1 m−1vs z (1–4.4 s z m−1 vs)−1 4.5 m−1 vs z (1–5.5 s z m−1 vs)−1 This work
kLa 0.32 vs0.7 — Heijnen and van’t Riet (1984)
kLa 1.0 vs0.79 1.5 vs0.78 Burckhart and Deckwer (1976)
kLa 0.67 vs0.73 2.5 vs0.84 This work

POULSEN AND IVERSEN: BUBBLE COLUMN EQUIPPED WITH A RUBBER MEMBRANE DIFFUSER 637
ejectors–injectors (gas and liquid are concurrently pumped
through a nozzle), or membrane spargers (as in this study)
in noncoalescent liquids or at relatively low superficial gas
velocities in coalescent liquids. Coarse bubble systems are
obtained in both types of liquid by sparging through holes
larger than 1 mm as in most ring spargers or perforated
plates.
Comparison of Measured and Estimated
Gas Holdup
Gas holdup measured volumetrically as described in Mate-
rials and Methods is increased by 25% in the noncoalescent
liquid (0.25 M NaCl) compared with the coalescent liquid
(distilled water, see Fig. 5). This number is twice the in-
crease reported by Bello et al. (1985) (using a perforated
plate) reflecting the formation of small bubbles by the mem-
brane sparger—a characteristic which is less obvious in the
coalescent liquid. In spite of this, our values for the coales-
cent liquid (coarse bubbles) are high compared with other
systems (Bello et al., 1985; Heijnen and van’t Riet, 1984).
In the homogeneous flow regime gas holdup depends
linearly on gas velocity (Deckwer and Schumpe, 1993) and
can be estimated from Eq. (5) using the bubble rise velocity
predicted from the Sauter mean bubble diameter as the av-
erage bubble rise velocity, vb,ave. The gas holdup estimated
this way, eest (dimensionless), is 4 s z m−1 vs and 5 s z m−1 vs
in the coalescent and noncoalescent liquid, respectively
(Table I), because the bubble rise velocity predicted from
the Sauter mean bubble diameter in a stagnant liquid is 0.25
m z s−1 and 0.20 m z s−1 (Clift et al., 1978), respectively.
e 4 vs/vb,ave (5)
In a heterogeneous flow regime, the correlation between e
and vs is given by e 4 0.6vs0.7 (Heijnen and van’t Riet,
1984). From the direct proportionality between the superfi-
Figure 5. Gas holdup in distilled water (h) and 0.25 M NaCl (,) as a
function of superficial gas velocity. Lines are the equation e 4 c2 z vs fitted
to the results, giving e 4 4.4 s z m−1 vs, r2 4 0.996 (h) and e 4 5.5 s z m−1
vs, r2 4 0.996 (,).
638 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 58, NO. 6, JUNE 20, 1998
cial gas velocity and gas holdup (equal to assuming constant
vb,ave) it is concluded that a homogeneous flow regime is
obtained (Fig. 5), giving vb,ave 4 0.23 m z s−1 and vb,ave 4
0.18 m z s−1 in the coalescent and noncoalescent liquid, re-
spectively.
Gas–Liquid Mass Transfer Coefficient Measured
by Dynamic Methods
Measurements of kLa using two methods are shown in Fig-
ure 6. Values based on pressure change (PDM), where the
change in CG is most abrupt, are expected to be somewhat
higher than the values determined by change in gas com-
position (SDM). When the rate of change in CG is low
compared with the mass transfer, as in the case of SDM, the
time course of CL, which is used for determination of kLa,
will be a result of both the rates of mass transfer and change
in CG. The effect of the latter change is of course more
pronounced at high kLa values, in agreement with the results
in Figure 6.
Heijnen and van’t Riet (1984) derived a power law cor-
relation for coarse bubble systems obtained with different
conventional rigid spargers in water: kLa (20°C) 4 0.32vs0.7
(different methods). The power law correlation obtained
here for the coalescent liquid (kLa 4 0.67vs0.73; Table I)
shows an increase in kLa, but almost the same dependency
on vs.
The kLa values in the noncoalescent liquid (Fig. 6) are
higher than, but in the range of, the values previously pub-
lished in fine bubble systems (Onken and Weiland, 1983;
Heijnen and van’t Riet, 1984; kLa determined with different
methods). Burckhart and Deckwer (1976) found a power
law correlation, kLa 4 1.5vs0.78 (method: gassing-in/out N2,
78-L liquid volume), for a sintered plate sparger in NaCl
Figure 6. Mass transfer coefficient kLa in distilled water and 0.25 M
NaCl measured with the standard dynamic method (SDM) and pressure
step dynamic method (PDM): h, distilled water PDM; L, distilled water
SDM; ,, 0.25 M NaCl PDM; n, 0.25 M NaCl SDM. Solid lines are power
laws fitted to PDM results: kLa 4 0.67vs0.73, r2 4 0.986 (h) and kLa 4
2.5vs0.84, r2 4 0.990 (,). Dashed lines are estimates of kLa from gas
holdup and bubble size [Eqs. (6) and (7)].
solutions. Our correlation in noncoalescent liquid shows an
increase in kLa and a minor increase in the dependency on
vs (kLa 4 2.5vs0.84; Table I). Bello et al. (1985) found kLa
values for a perforated plate (conventional rigid sparger) in
a ‘‘bubble coalescence-inhibiting system’’ of 0.15 M NaCl
(method: gassing-in/out N2, 30-L liquid volume) of only
40% to 50% of the values obtained in this report.
Again, it is noteworthy that the different methods used
make a direct comparison difficult. However, it is obvious
that the PDM is the most reliable method for reactors with
slow exchange of the gas phase in agreement with the sug-
gestion of Linek et al. (1990).
Gas–Liquid Mass Transfer Coefficient Estimated
from Bubble Size and Gas Holdup
The volumetric gas–liquid mass transfer coefficient can be
estimated from bubble size (ds) and gas holdup. The liquid
mass transfer coefficient kL (m z s−1) is approximately
4 z 10−4 m z s−1 at Sauter mean bubble diameters above 2
mm, and around 1.5 z 10−4 m z s−1 at a Sauter mean bubble
diameter of 1 mm (Heijnen and van’t Riet, 1984). However,
it most be noted that the variations in kL determinations are
inherently high. It is obvious from Figure 3 that a is under-
estimated if it is calculated assuming spherical bubbles in
the case of coalescent liquids. However, it seems to be a
good approximation in the case of noncoalescent liquids. A
direct proportionality between vs and e was found in Figure
5. By substitution of e 4 4.4 s z m−1 vs (coalescent liquid)
or e 4 5.5 s z m−1 vs (noncoalescent liquid) into Eq. (2), a
can be estimated as a function of vs. This gives:
(kLa)est (coalescent) 4 2.1 m−1vs z (1 − 4.4 s z m−1vs)−1 (6)
(kLa)est (noncoalescent) 4
4.5 m−1 vs z (1 − 5.5 s z m−1 vs)−1 (7)
Estimated and measured mass transfer coefficients are simi-
lar in the coalescent and noncoalescent liquids (Fig. 6).
Because the interfacial area, a, was underestimated in the
case of coalescent liquid, due to the irregular shape and
surface of the bubbles (Fig. 3), it follows that kL has been
overestimated.
Ayazi Shamlou et al. (1995) found that kLa also is highly
dependent on ds in an airlift reactor and suggested that kLa
is given by:
kLa 4 12(D/p)1/2 z [vb + C z (vs,r + vL,r)]1/2
z ds−3/2 z e(1 − e)−3/2 (8)
where D (m2 z s−1) is the diffusion coefficient, C (dimen-
sionless) is the distribution parameter dependent on the type
of sparger, vs,r (m z s−1) is the superficial liquid velocity in
the riser. We exchanged vs,r with vs and used the values 1,
0, and 2.0 z 10−9 m2 z s−1 for C, vL,r, and D, respectively, and
found kLa values almost identical to our estimated values for
the coalescent liquid (within 8%). However, for the non-
coalescent liquid, Eq. (8) gave highly overestimated values
POULSEN AND IVERSEN: BUBBLE COLUMN EQUIPPED WITH A RUBBER MEMBRANE DIFFUSER
(>400%) even when the value of C was set to zero, which
indicates too strong dependency on ds (ds−3/2).
Oxygen Transfer Efficiency
The mass transfer coefficients reported in Figure 6 are high
compared with literature values, emphasizing the capacity
of the membrane to produce small bubbles. Small bubbles
affect a positively by increasing AG and residence time.
They affect kL negatively via a low vb, which decreases
mixing of gas and liquid interfaces. However, kLa increases
with decreasing bubble diameter in the range of bubble
diameters studied here. The superior performance in terms
of gas transfer is seen more clearly from kLa values than
from gas holdup values. This confirms that the mass transfer
coefficient is the most important variable for description of
aeration. Oxygen transfer efficiency, OTE (kg O2 z kWh−1),
is commonly used in the industry to describe power con-
sumption. When CL 4 0, it attains its maximal value:
kLa ? CG ? VL
OTE = (9)
P
where P (W) is the power consumption of the air pump
calculated from the measured pressure of the incoming gas,
pg (Pa), assuming a pumping efficiency of 63%:
vs ? Ai ? pamb ? ln~pg / pamb!
P= (10)
0.63
where pamb (Pa) is ambient pressure. Our results expressed
as OTE are shown in Figure 7. Only a few values of OTE are
given in the literature. Calculation of OTE based on pub-
lished kLa values is essentially impossible because informa-
tion about the power consumption is not provided or the
term is defined in various ways. Therefore, a comparison
with literature values is difficult. However, it is noteworthy
that the decrease in OTE with increasing vs is small.
Figure 7. OTE in distilled water (h) and 0.25 M NaCl (,). CG 4 240
mM was used.
639
Mixing Intensity teristic of rigid spargers, in contrast to results from other
In parallel to the increase in mass transfer with smaller reports on ‘‘elastic spargers.’’ This can be explained by the
bubble diameters a decrease in mixing efficiency was seen, hard rubber material of the membrane. In spite of this, the
as expected (Fig. 8). The increase in mixing time as a result membrane diffuser produced very small bubbles over a
of the small bubbles characteristic of the noncoalescent liq- wide range of superficial gas velocities. Bubble formation in
uid is an order of magnitude larger compared with the co- this membrane is therefore not dependent on an increasing
alescent liquid. Similar results were described by Rice et al. area of holes with increasing flow rate. A possible expla-
(1981) and Rice and Littlefield (1987). The low mixing nation of this phenomenon is that the elastic holes act as
intensity in the noncoalescent liquid shows that the liquid valves dividing the gas stream into very small bubbles. We
flow is uniform and laminar, as expected from Figure 4, achieved mass transfer coefficients higher than those previ-
whereas it is turbulent in the coalescent liquid, as expected ously obtained in bubble columns, irrespective of aeration
from Figure 3. method. There are probably two reasons for this result: first,
our design enabled us to change the tension of the mem-
brane in a direction transverse to the slits; and second, the
Growth Experiments
intrinsic properties of the membrane were superior to those
The rubber membrane sparger described here has been used used by other workers. kLa could be reasonably well esti-
successfully in a bubble column. The simplicity of the me- mated from measured values of gas holdup and bubble size
chanical construction of the bubble column in combination and literature values of kL. The observed mixing intensity
with the self-cleaning property of the membrane sparger was very low. However, considering the high mass transfer,
allowed us to maintain the microalga Rhodomonas sp. in low energy costs, extremely low physical stress, and low
continuous culture for 8 months without interruption (de- maintenance, this bioreactor is competitive with not only
tails to be described elsewhere). Rhodomonas sp. is sensi- other bubble column reactors, but also stirred tank reactors.
tive to chemical and physical stress, but the yield coeffi-
cients obtained with the bubble column described in this We thank Raymond P. Cox for criticism and reading of the
study were identical to cultures grown in conventional bio- manuscript, and Svend Åge Madsen for technical assistance.
reactors (Eriksen and Iversen, 1995), ruling out any inhibi-
tory effects of the membrane material when treated as de-
scribed in Materials and Methods. NOMENCLATURE
The productivity of the bubble column was 4 z 109
cells z L−1 z day−1. The algae were used as feed for marine a specific gas–liquid interfacial area per unit liquid volume
(m−1)
invertebrates in bioenergetic studies of suspension feeding AG total area of gas bubbles (m2)
and growth (Eriksen and Iversen, 1997; Nielsen et al., 1995; Ai inner area of bubble column reactor (m2)
Thomassen and Riisgaard, 1995). The application of such A0 area of orifice (m2)
studies is important in aquaculture. c1 correlation factor (Pa z m−6 z s2)
c2 correlation factor (s z m−1)
C distribution parameter (dimensionless)
CONCLUSIONS Co orifice coefficient (dimensionless)
With the rubber membrane diffuser used in this study we CL oxygen concentration in liquid (mol z m−3)
CG oxygen concentration in liquid in equilibrium with gas phase
found the gas flow rate–pressure drop relationship charac- (mol z m−3)
db,i bubble diameter of ith bubble (m)
dm mean bubble diameter (m)
ds Sauter mean bubble diameter (m)
D diffusion coefficient (m2 z s−1)
kL liquid mass transfer coefficient (m z s−1)
kLa volumetric gas-liquid mass transfer coefficient (s−1)
(kLa)est estimated volumetric gas-liquid mass transfer coefficient (s−1)
N number of bubbles (dimensionless)
OTE oxygen transfer efficiency (kg O2 z kwh−1)
pamb ambient pressure (Pa)
pg pressure of incoming gas (Pa)
Dp pressure difference across sparger (Pa)
Dpc critical pressure difference across sparger (Pa)
P compressor power input (W)
Qo volumetric gas flow rate per orifice (m3 z s−1)
tM characteristic mixing time (s)
vb bubble rise velocity (m z s−1)
vb,ave average bubble rise velocity (m z s−1)
vL,r superficial liquid velocity in the riser (m z s−1)
vs superficial gas velocity (m z s−1)
Figure 8. Characteristic mixing time, tM, in distilled water (h) and 0.25 vs,r superficial gas velocity in the riser (m z s−1)
M NaCl (,). VG total volume of bubbles (m3)

640 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 58, NO. 6, JUNE 20, 1998
 VL volume of liquid in reactor (m3)
Greek letters
e gas holdup (dimensionless)
eest estimated gas holdup (dimensionless)
rg gas density (kg z m−3)
References
Ayazi Shamlou, P., Pollard, D. J., Ison, A. P. 1995. Volumetric mass trans-
fer coefficient in concentric-tube airlift bioreactors. Chem. Eng. Sci.
50: 1579–1590.
Bartholomew, W. H., Karow, E. O., Sfat, M. R., Wilhelm, R. H. 1950.
Oxygen transfer and agitation in submerged fermentations. Mass trans-
fer of oxygen in submerged fermentation of Streptomyces griseus. Ind.
Eng. Chem. 42: 1801–1809.
Bello, R. A., Robinson, C. W., Moo-Young, M. 1985. Gas holdup and
overall volumetric oxygen transfer coefficient in airlift contactors.
Biotechnol. Bioeng. 27: 369–381.
Bovonsombut, S., Wilhelm, A. M., Riba, J. P. 1987. Influence of gas dis-
tributor design on the oxygen transfer characteristics of an airlift fer-
mentor. J. Chem. Technol. Biotechnol. 40: 167–176.
Burckhart, R., Deckwer, W. D. 1976. Der einfluss der Betribsbedingungen
und die Wirkung von Elektrolytzusätzen auf den Sauerstoffübergang
in Blasensäulen. Verfahrenstechnik 10: 429–433.
Clift, R., Grace, J. R., Weber, M. E. 1978. Bubbles, drops and particles.
Academic Press, New York.
Deckwer, W. D., Schumpe, A. 1993. Improved tools for bubble column
reactor design and scale-up. Chem. Eng. Sci. 48: 889–911.
Deckwer, W. D., Nguyen-Tien, K., Schumpe, A., Serpemen, Y. 1982.
Oxygen mass transfer into CMC solutions in a bubble column. Bio-
technol. Bioeng. 24: 461–481.
Degn, H., Nielsen, J. B. 1987. A general purpose serial interphase for the
laboratory computer. Binary 10: 25–28.
Eriksen, N. T., Iversen, J. J. L. 1995. Photosynthetic pigments as nitrogen
stores in the cryptophyte alga Rhodomonas sp. J. Marine Biotechnol.
3: 193–195.
Eriksen, N. T., Iversen, J. J. L. 1997. On-line determination of respiration
rates of aquatic organisms in a mono-phase oxystat at steady-state
dissolved oxygen tensions. Marine Biol. 128: 181–189.
Geary, N. W., Rice, R. G. 1991. Bubble size prediction for rigid and flex-
ible spargers. AIChE J. 37: 161–168.
POULSEN AND IVERSEN: BUBBLE COLUMN EQUIPPED WITH A RUBBER MEMBRANE DIFFUSER
Heijnen, J. J., van’t Riet, K. 1984. Mass transfer, mixing and heat transfer
phenomena in low viscosity bubble column reactors. Chem. Eng. J. 28:
B21–B42.
Linek, V., Benes, P., Sinkule, J. 1990. Critical assessment of the steady-
state Na2SO3 feeding method for kLa measurement in fermentors. Bio-
technol. Bioeng. 35: 766–770.
Linek, V., Benes, P., Vacek, V. 1989. Dynamic pressure method for kLa
measurement in large-scale bioreactors. Biotechnol. Bioeng. 33:
1406–1412.
Medic, L., Cehovin, A., Koloini, T., Pavko, A. 1989. Volumetric gas–liq-
uid mass transfer coefficients in a rectangular bubble column with a
rubber aeration pad. Chem. Eng. J. 41: B51–B54.
Nielsen, A. M., Eriksen, N. T., Iversen, J. J. L., Riisgaard, H. U. 1995.
Feeding, growth and respiration in the polychaetes Nereis diversicolor
(facultative filter-feeder) and N. virens (omnivorous)—a comparative
study. Marine Ecol. Prog. Ser. 125: 149–158.
Onken, U., Weiland, P. 1983. Airlift fermentors: construction, behaviour,
and uses, pp. 67–95. In: A. Mizrahi and A. L. Wezel (eds.), Advances
in biotechnological processes, vol. I, Alan R. Liss, New York.
Oolman, T. O., Blanch, H. W. 1986. Bubble coalescence in stagnant liq-
uids. Chem. Eng. Commun. 43: 237–261.
Poulsen, B. R., Iversen, J. J. L. 1997. Mixing determinations in reactor
vessels using linear buffers. Chem. Eng. Sci. 52: 979–984.
Rice, R. G., Howell, S. W. 1986. Elastic and flow mechanics for membrane
spargers. AIChE J. 32: 1377–1382.
Rice, R. G., Littlefield, M. A. 1987. Dispersion coefficients for ideal bub-
bly flow in truly vertical bubble columns. Chem. Eng. Sci. 42:
2045–2053.
Rice, R. G., Tupperainen, J. M. I., Hedge, R. M. 1980. Dispersion and
hold-up in bubble columns. Division of Industrial and Engineering
Chemistry. Second Chemical Congress of the North American Conti-
nent. San Francisco, CA, August 25–29, 1980.
Rice, R. G., Tupperainen, J. M. I., Hedge, R. M. 1981. Dispersion and
hold-up in bubble columns—comparison of rigid and flexible spargers.
Can. J. Chem. Eng. 59: 677–687.
Talbot, P., Lencki, R. W., de la Noüe, J. 1990. Carbon dioxide character-
ization of a bioreactor for biomass production of Phormidium bohneri:
comparative study of three types of diffuser. J. Appl. Phycol. 2:
341–350.
Thomassen, S., Riisgaard, H. U. 1995. Growth and energetics of the
sponge Halichondria panicea. Marine Ecol. Prog. Ser. 128: 239–246.
Weber, M. E., Loiselle, E., Charrier, J.-M. 1990. Performance of a tubular
elastic sparger. Can. J. Chem. Eng. 68: 742–747.
641