148 Ind. Eng. Chem. Res.

2009, 48, 148–158

Uniform Flow in Bubble Columns

R. F. Mudde,* W. K. Harteveld,† and H. E. A. van den Akker
Kramers Laboratorium Voor Fysische Technologie and J.M. Burgers Center for Fluid Mechanics, Delft
UniVersity of Technology, Pr. Bernhardlaan 6, 2628 BW Delft, The Netherlands

Uniform bubbly flow in a 15 cm bubble column is investigated. We use a special needle sparger consisting
of 559 separate needles, uniformly distributed over the bottom. With this sparger, we can ensure that all
bubbles generated are of the same size and that the bubble injection is very uniform over the entire bottom
of the column. Detailed experiments are reported, using optical glass fibers to measure the local gas fraction
and bubble size and velocity and using laser Doppler anemometry to measure the liquid axial velocity field.
We find that the homogeneous flow regime extends up to a gas fraction of 55% well beyond the predictions
of theory. The superficial gas velocity at which the homogeneous regime looses its stability depends on the
water quality: fresh water looses its stability much earlier than old water. However, the gas fraction as a
function of the superficial gas velocity is in the homogeneous regime independent of the water quality. The
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

overall gas fraction can be described by a Richardson and Zaki type of relation or by the proposal by Garnier
et al. We have indications that, at the point of instability, the bubble size has increased to a critical value at
Downloaded via NATL CHENG KUNG UNIV on January 13, 2023 at 03:06:46 (UTC).

which the lift force reverses sign. This causes the radial gas fraction to change from flat with a small wall
peaking to core peaking provoking the instability as suggested by Lucas et al. [Chem. Eng. Technol. 2005,
60, 3609]. Alternatively, at higher gas fractions, the swarm gets denser and the bubble wakes get suppressed.
According to Fox and co-workers [Chem. Eng. Sci. 2007, 62, 3159], this causes the flow to lose its stability.

1. Introduction size. Furthermore, the critical gas fraction peaks at db ≈ 3 mm.

In bubble columns, two basic flow regimes are possible: the
homogeneous and the heterogeneous regime (see, e.g., refs 3–6).
The latter prevails at the higher gas flow rates and is generally
characterized by transient flow, a net circulation, and a wide
bubble size distribution. Bubble columns operate in most cases
in the heterogeneous regime. The homogeneous regime is found
at low superficial gas velocities. In this regime, the bubble size
has a narrow distribution and the liquid motion is restricted to
small scales induced by the passing bubbles. At some point,
the homogeneous regime becomes unstable and a transition sets
in to the heterogeneous one. Numerous papers have appeared
that deal with one or more of the aspects of this transition from
homogeneous to heterogeneous.
One way to study this transition is by constructing a gas
fraction (R) vs superficial gas velocity (Usup) plot from
experimental data,4,7–11 and find the inflection point in the plot.
Here the transition to the heterogeneous regime sets in.
Linear stability analysis is also frequently employed to
investigated the stability of the homogeneous regime, see e.g.,
1, 2, and 12–16. An extensive review is presented by Joshi et
al.17 In their review, they discuss the one-dimensional linear
stability of the homogeneous flow in various dispersed systems,
like fluidized beds and bubble columns. The authors present a
relation for the maximum gas fraction in the bubble column in
the homogeneous regime. This maximum is reached when the
gas fraction versus superficial gas velocity is at its maximum.
However, the system may become unstable before this point is
reached. The stability analysis provides the critical gas fraction
at which small disturbances of the homogeneous regime get
amplified. Various parameters influence the critical gas fraction
calculated from the stability analysis. For bubble sizes above 3
mm, the critical gas fraction is found to decrease with increasing
* To whom correspondence should be addressed. E-mail address:
r.f.mudde@tudelft.nl.
†
Present address: Shell Global Solutions, Amsterdam, The Netherlands.
10.1021/ie8000748 CCC: $40.75  2009 American Chemical Society
Published on Web 09/06/2008
The predicted peak values range from as small as 10% to 100%
(i.e., always stable), depending on the closure relations used in
the hydrodynamic description. In the same paper, also the
experimental data from the literature are collected. For air-water
columns at ambient conditions, the maximum critical gas
fraction reported is 36.5%. However, the majority of data are
around a value of 20-25% with critical superficial velocities
of about 3-4 cm/s.
León-Becerril et al.13,14 report that, according to their stability
analysis, the homogeneous flow becomes unstable at R ≈ 30%
for spherical bubbles. This critical gas fraction reduces to 10%
to 15% for ellipsoidally shaped bubbles. Lucas and co-workers1,15
investigated the role of the lift force and concluded that the
influence of the lift force on stability is much higher than that
of the turbulent dispersion force. Moreover, for small bubbles
that have a positive lift coefficient, the lift force is stabilizing,
whereas for the larger bubbles, that have a negative lift
coefficient, this force is destabilizing. Bhole and Joshi16 collected
a lot of literature data for pure systems as well as for systems
with additives. A wide range of critical gas fractions was
reported. For water, the maximum critical gas fraction is as high
as 45.8%.11 The theoretical result of the work of Bhole and
Joshi gave a good prediction of the transition from homogeous
to heterogeneous flow for a wide range of systems. Joshi12 also
conducted a linear stability analysis for a 2-dimensional
air-water bubble column. From this work, it follows that the
critical gas fraction follows from a simple relation: Rcrit ) 1/(n
+ 1) with n bing the Richardson and Zaki coefficient. According
to the authors, a typical value for n is 1.4, leading to a critical
gas fraction of 42%. Finally, Monahan and Fox2 inspected
instability from both horizontal and vertical modes in the bubbly
flow. Different from e.g. Lucas et al. they find that the instability
is associated with the bubble induced turbulence production.
A lot of work on the transition between homogeneous and
heterogeneous flow in the bubble columns has been done by
Drahos and co-workers. Zahradnik9 already in 1979 investigated

this problem and found that for a perforated plate the critical
gas fraction was higher for smaller hole diameter. They reached
a value of 39% for a perforated plate with a hole size do ) 0.5
mm. Locket and Kirkpatrick8 also experimentally investigated
the stability of bubble flows. They operated in a countercurrent
mode such that the swarms could be kept stationary. This way
they managed, in a pipe of 76 mm diameter, to obtain a critical
gas fraction of 66%. However, this is not directly comparable
to the homogeneous bubble column, as now there is a distinct
liquid velocity profile close to the wall creating a lift force on
the bubbles. Furthermore, the liquid flow is turbulent and thus
also a turbulent dispersion force is action on the bubbles.
Drahos et al.18 studied pressure fluctuations in a bubble
column. They report a maximum hold up of 30% in their
equipment. Zahradnik et al.4 found in a 14 cm diameter bubble
column from residence time distribution measurements and from
the gas fraction that for a perforated plate the homogeneous
flow could easily be obtained for superficial gas velocities up
to 4 cm/s. They also reported that the stability of the homoge-
neous flow could be significantly enhanced by adding surface
active solutes to the liquid. They used a 14 cm diameter bubble
column with different perforated plates for sparging the gas.
For water, they found a critical gas fraction of 32%, but this
could easily be increased to 63% by adding 0.5% of ethanol.
Also for electrolytes, the critical value was higher, e.g. they
report 38% for KI.
Ruzicka and co-workers11 made an extensive study of the
transition from homogeneous to heterogeneous flow. They
provided an extensive literature review on the topic and gave a
model description for the critical prediction. From this, it
followed that the maximum gas fraction for the homogeneous
regime has a value of 55%. However, this was found for shallow
columns with a height of diameter ratio of 3. In a subsequent
paper in 2001,19 the same group of researchers investigated the
effects of the diameter and of the mixture height on the stability.
They reported that both an increase in the diameter and the
height had a destabilizing effect on the homogeneous regime.
In most cases, the critical gas fraction was 30% or below. Also
this research group reported20 the effect of the liquid viscosity
on the stability: for µl ) 1-3 mPa s, no big difference was
found, but for the range of 3-22 mPa s, the stability decreased
with increasing viscosity. Finally, in a recent paper,21 the effect
of surfactants was investigated. It was found that addition of
e.g. CaCl2 at low concentrations stabilized the homogeneous
regime. However, for larger concentrations a destabilizing was
observed. In this paper, the authors conclude that the important
problem of the prevailing flow regime in bubble column reactors
has been paid only very little attention. The flow regimes are
in most studies not specified at all, or mentioned only margin-
ally. If mentioned, so usually vaguely, based on the authors
intuition and experience. If specified, so usually based only on
the visual inspection of experimental data, mainly on the shape
of the e(q) graph.
Virtually all the above-mentioned papers used rather course
experimental techniques. The number of papers on the homo-
geneous bubble regime that reported detailed hydrodynamic
experiments have been rather limited. Yao et al.22 preformed
experiments using a conductivity probe for the bubble size and
velocity and for the gas fraction and an ultrasonic Doppler
anemometer also for the bubble velocity (i.e., for 2 components
of the velocity vector). Furthermore, they employed a hot wire
anemometer for the liquid velocity. They report that a large
scale liquid circulation, characteristic for the heterogeneous
regime, sets in above a superficial gas velocity of 4 cm/s. Garnier
Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009 149
23
et al. measured, using a hot wire, the liquid velocity in a bubble
column of 8.0 cm diameter equipped with a needle sparger
consisting of 271 needles. Very uniform flows could be obtained.
They also measured the local gas fraction using a double optical
probe. A small cocurrent liquid flow was used in the experi-
ments; hence, this is not literally a bubble column. They found
very flat gas fraction profiles for gas fractions as high as 40%.
They also gave a scaling law for the bubble rise velocity in the
swarms. It scales with R1/3, like sedimenting particles in the
dense regime. Harteveld and co-workers24 also used a needle
sparger to achieve very uniform inlet conditions. They reported
experiments using the laser Doppler anemometry (LDA) and
optical glassfibers and showed very uniform flow at gas fractions
of 10%. In a later paper25 these authors reported in the same
column very uniform flow up to a gas fraction of 25%. Kulkarni
and Joshi26 performed 2D laser Doppler anemometry measure-
ments in a 15 cm diameter column using two different spargers:
a single-point and a multipoint one. The superficial gas velocity
was kept constant at 2.0 cm/s during all experiments. The results
were compared to CFD (computational fluid dynamics) simula-
tions of the flow. The flow showed large scale circulation already
at a gas fraction of about 7%. More importantly for the present
paper, the authors made an attempt to incorporate a real sparger
in the CFD simulations. They conclude that “still [a] lot of scope
exists for the development of generalized codes, possibly by
understanding the physics of the flow”. The present paper aims
at providing detailed, reliable data on the hydrodynamics in the
homogeneous regime. CFD simulations should be able to capture
this type of flow relatively easy as the modeling of phenomena
such as turbulence does not play a crucial role here. Monahan
et al.27 performed CFD simulations of a bubble column and
paid special attention to the homogeneous regime. They showed
that indeed the simulations can reproduce the homogeneous
regime at relatively high superficial gas velocities.
From the above, it does not become clear where the border
between the two regimes is exactly located. In general, it is
found that the homogeneous flow goes through a transition
regime toward the heterogeneous regime at gas fraction between
15-25%. Theoretical work28 suggest that at gas fractions
beyond 30% the homogeneous flow is inherently unstable and,
in smaller diameters, a transition to slug flow sets in. However,
stability analysis also resulted in higher predections.
Although for industrial practice the heterogeneous regime is
the most important one, proper experimental data on the
homogeneous regime are relevant. Nowadays, bubbly flows are
also studied via CFD. In particular, for bubble columns, CFD
simulations of the heterogeneous regime have been reported.27,29–32
A good test of the codes and models used is to see if a code
can predict the right homogeneous flow, when the gas is
introduced uniformly over the bottom of the bubble column. It
is an open question up to which gas fractions the flow can be
homogeneous and whether or not CFD can predict this.
Moreover, the onset of the instability triggering the change from
homogeneous to heterogeneous flow is not completely understood.
The number of detailed experiments with well-defined and
controlled inlet conditions is rather limited. In almost all cases,
porous plates or perforated plates have been used. With these
spargers, the inlet conditions are not accurately controlled, nor
are they known with high accuracy. Therefore, the present paper
revisits the homogeneous bubbly flow. We aim at providing
detailed data of the gas fraction distribution (using glass fibers)
and liquid velocity (using LDA) that can, e.g., be used to test
CFD simulations of the homogeneous regime. Our sparger is
capable of providing a very uniform inlet condition to the bubble
150 Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009
Figure 1. System used to provide the bubble column with air.
Figure 2. Top view of needle sparger (left) and picture showing bubble formation with one-third of the needles in operation (right, gas fraction about 15%).
column, something that is easily implemented in computer
simulations and can serve as a good test as discussed in the
work of Monahan et al.27 Furthermore, the sparger is capable
of generating a very narrow bubble size distribution and maintain
that over a relatively wide range of superficial gas velocities.
Experiments like this are very limited in the literature but, in
our opinion, provide a well defined “test case”. Moreover, we
estimate the bubble size and measure the bubble velocity using
a four point optical probe. We have carried out experiments in
a 15 cm diameter column and found that the uniform flow can
be present up to 55% volume fraction of the bubbles. Finally,
at the onset of instability, we find an intriguing duality in the
flow that may give a clue to the instability we observed.
2. Experimental Setup
All experiments are carried out in a cylindrical column of
15 cm inner diameter, filled with tap water to an ungassed height
of 130 cm. A special bubble distributor is used for the injection
of the bubbles. It consists of 559 needles, with an inner diamet-
er of 0.8 mm and a length of 20 cm where each needle exit is
5 mm above the column bottom. More details can be found in
ref 25. The needles are placed in a triangular pattern, with a
pitch of 6 mm. The flow through the needles is controlled in
small groups. In total, the needles are grouped in eleven groups.
The number of needles operating is a function of the required
gas fraction. In order to obtain a uniform bubble size, the gas
flow rate through each needle needs to be between 1 and 3 mL/
s. Therefore, for gas fractions below 15%, 187 needles are used,
i.e. one out of every three. For larger gas fractions, all needles
are in operation. This split ensures that the bubbling charac-
teristics of the needles stay the same during the experiments.
Eleven electronic mass flow meters (VPInstruments VPFlow-
mate with a range of 0-5 slm) have been used in combination
with metering valves to obtain accurate control over the flow
to the groups. For an accurate measurement of the total flow
rate, another electronic mass flow meter (VPInstruments VP-
Flowmate with a range of 0-100 slm) has been used. A
schematic representation of the setup is given in Figure 1. Note
that upstream of each group of needles a so-called group-needle
(inner diameter 0.6 mm, length 10 cm) is placed, in order to
obtain a very uniform flow to each of the groups of needles.
A picture of the sparger operating with one out of every three
needles is given in Figure 2.
3. Experimental Techniques
Glass Fiber Probes. Optical glass fibers are used to probe
the bubble phase. The probe’s working principle is the difference
in refraction index between glass (the probe material), water,
and air. Light is sent into the glass fiber. At the tip, it will reflect
back when the tip (size 200 µm) is surrounded by air. Most of
the light will propagate into the water phase if the tip is
surrounded by water. By detecting the reflected light, the phase
present at the probe tip is detected (see, e.g., refs 33–35). Gas
fraction profiles are measured using a single point optical probe.
First, the time series recorded is made binary via an appropriate
threshold (thR). The latter is set at 10% of the difference between
the output value of the tip in air (sa) and that in water (sw).
b(tj) ) { 1
0
if s(tj) > thR ) 0.1(sa - sw) + sw
else
(1)

Figure 3. Typical signal of the piercing of a single bubble. The noise levels
of the probe are very low. The binary version of the signal is also given.
Figure 4. Effect of small departures from vertical alignment on the axial
liquid velocity profile (R ) 8.1%).
Next the (local) gas fraction is obtained as follows:
N 200 000 data points, whereas at a radial position r/R ≈ 0.5 at a
∑ b(t ) · ∆t
j s
j)1
R) (2)
N∆ts
with N the total number of samples, and ∆ts ) 10 µs, the sample
time. The threshold should not be too high as this would cause
bubbles that touch the tip rather than being pierced to be filtered
out, thus underestimating the gas fraction. Obviously, it can not
be too close to 0 as then unwanted noise would cause an
overestimate of the gas fraction. An example of the signal (and
its binary) of a single bubble passage is given in Figure 3. Note
that when the tip of the probe leaves the bubble the response is
very fast: the original signal and the binary step almost
completely overlap. The measuring time per data point was set
at 900 s. The number of bubbles detected scales roughly linearly
with the gas fraction and is in all cases above 5000.
The experimental error of measurements with the glass fibers
is made of two parts. The first part is the statistical error
associated to the measuring time. It is proportional to the square
root of the measuring time. Repeated experiments show that in
our case the absolute uncertainty in the gas fraction is
0.001-0.003 for the gas fraction range studied. The second part
is a bias error due to the difficulties of piercing a bubble at the
bubble edge. This is caused by the surface tension force that
makes the bubble deviate from its vertical path upon colliding
with the glass fiber. This error depends on the approach velocity
of the bubble and on the swarm density surrounding the bubble:
the higher the velocity, the smaller the bias; the denser the
swarm, the less room there is for the bubble to flow around the
glass fiber tip. In the literature,36,35 it is reported that this bias
caused an underestimate of 5-10% of the actual gas fraction.
In the present case, the swarm gets much denser than used in
the studies mentioned. We expect that the bias will be smaller
and estimate it to be 5%. Note that this effect causes a systematic
underprediction of the gas fraction.
Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009 151
LDA. The LDA equipment consists of a 4W Spectra-Physics
Ar+ laser and a TSI 9201 colorburst multicolor beam separator.
The beam pairs are focused using a backscatter probe with a
lens of 0.132 m focal length. Detected light is sent to a TSI
9230 colorlink. The Doppler bursts have been processed by a
TSI IFA-750 processor.
The application of LDA in bubbly flows is far from
straightforward. Upon measuring deeper in the bubbly flow, the
data rate gets strongly reduced due to blockage of the laser
beams by the bubbles. This reduction is approximately propor-
tional to exp(-c[Rl/db]), where c is a constant in the range of
1.5-3, l is the distance of the measuring volume in the bubbly
flow, R is the void fraction, and db is the bubble diameter (see,
e.g., the work of Groen et al.37).
Apart from blocking the laser beams, the bubbles can also
cause Doppler bursts that will lead to a velocity measurement.
From a given burst, it is very difficult to tell whether it is caused
by a seeding particle or by a bubble. In the present experiments,
we used neutrally buoyant hollow glass particles with a diameter
of about 8 µm as seeding. It is discussed by Groen et al.37 that
under these conditions the number of velocity measurements
that actually comes from a bubble is negligible. It is therefore
safe to assume that the LDA measures the liquid velocity.
Typical data rates for void fractions near 10% are 150-400
Hz near the wall and 0.8 Hz in the column center. This shows
the difficulty of experiments deeper inside the column, especially
at the higher void fractions, where experiments in the center
are no longer possible. The measuring time per point was set
at 900 s. Close to the wall a single data set contains more than
gas fraction of 30% this number has dropped to less than 5000.
LDA data are not bias free. First, due to increased noise levels
caused by the bubbles multiple validation, i.e. measuring the
burst of a single seeding particle multiple times, occurs more
often than in single phase flow. The multiple data points are
removed by rejecting those data points that are outside 4 times
the standard deviation of the data set. Next, for all two data
points within a time interval of 1 ms, the second data point is
removed. Second, a velocity bias correction is employed. This
correction is based on the so-called 2D+ weighing: the weighing
factor is inversely proportional to the magnitude of the velocity
vector that is estimated from two of the three velocity
components. For a full discussion, see, e.g., the work of
Tummers.38 These corrections amount to about 2 cm/s, reducing
the velocity. It brings the measured net cross-sectional flow of
the liquid down from 1 or 2 cm/s to values of 0.3 cm/s or less.
Column Alignment. Rice and co-workers39,40 have shown
that very small deviations from vertical alignment result in the
creation of a large scale liquid circulation in the column and an
increase in the axial dispersion of the bubble phase. Tinge and
Drinkenburg41 discuss the maximum inclination angle before
this circulation sets in. Extrapolation of their results gives a
maximum inclination of 0.05° in our case. Careful alignment
made it possible to have a horizontal deviation at a position of
1.5 m above the bottom of the column with respect to the vertical
of less than 1 mm, which indicates a maximum inclination angle
of 0.04°. We have run a test with an average gas fraction of
8.1% (measured from the liquid level increase) to see the
influence of the inclination angle on the axial liquid velocity as
measured by the LDA. Results are shown in Figure 4. It is clear
that even angles as small as 0.04° are visible. Note further that,
for “perfect” alignment, we measure a slightly nonzero mean
flow. This is a consequence of a small bias in the LDA
measurements.
152 Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009
Figure 5. Void fraction as a function of the superficial gas velocity with
time t after filling the bubble column with fresh tap water. Note that both
theoretical lines (RZ and Garnier) almost overlap.
4. Gas Fraction and Liquid Velocity
4.1. Overall Gas Fraction. The overall gas fraction is plotted
in Figure 5. For low superficial velocities, the bubbly flow is in
the homogeneous regime and the gas fraction raises more or
less linearly with the gas flow rate. However, at a certain point,
this linearity breaks down and the gas fraction only slowly
increases with the superficial gas velocity. This marks the change
from homogeneous to heterogeneous flow.11 Where exactly this
happens, depends on the purity of the water. As can be seen in
Figure 5, for fresh tap water, we find the transition around
30-35%. But as the water stays longer in the column, this point
shifts until we finally can reach gas fractions of 55% before
the uniform flow becomes unstable. This is surprising as all
reports so far on bubble columns have this transition much
lower: at or below 30%. Moreover, linear stability analysis of
the uniform flow predicts instability from 25-40% for spherical
bubbles or as low as 10-15% for ellipsoidal bubbles (see, e.g.,
refs 28 and 13). We have measured the surface tension of the
water in the column as a function of time, using the Du Noüy
Ring method. We find, that according to this method the surface
tension stays constant: σ ) (71.6 ( 0.5) × 10-3 N/m. No
decreasing trend with time has been found.
The data, presented in Figure 5, can be analyzed based on
the slip velocity between the gas and liquid phase. Both the
Richardson and Zaki (RZ) relation for the slip velocity and the
proposal by Garnier et al.23 are used.
The slip velocity (Vs), defined as the bubble velocity (Vd)
minus the velocity of the continuous phase (Vc), is connected
to the superficial velocities:
Ug Uliq
Vs ≡ Vd - Vc f Vs(R) ) - (3)
R 1-R
The superficial liquid velocity can be set to zero as we are in
the homogeneous regime. Hence, no liquid circulation is induced
and the local, averaged liquid velocity is zero. Thus, we obtain
a direct relation between the void fraction and the superficial
gas velocity:
{
V∞R(1 - R)n-1 RZ
Ug ) R · Vs(R) ) (4)
V∞R(1 - CµR1⁄3) Garnier
with V∞ being the slip velocity of a single bubble, n the
Richardson and Zaki power () 2.39 for bubbles of the present
size in the original RZ model), and Cµ a parameter set to 1 in
Table 1. Fit Parameters
V∞(m/s) n(-) Cµ(-)
RZ 0.195 1.40
Garnier 0.280 0.60
Garnier’s model.23 Both models are fitted to the data of Figure
5. The results are given in Table 1.
Clearly, we find a much lower power for the Richardson and
Zaki slip velocity than according to the original model for
particles with a single particle Reynolds number above 500.
However, the value of n ) 1.4 is in agreement with the one
mentioned by Joshi.12 Moreover, the single bubble rise velocity
needs to be set at 0.195 m/s. This is too small: the bubbles in
the present experiment have an equivalent diameter of 4 mm
(see section 4.4). These bubbles have a terminal rise velocity
of 0.25 m/s, significantly higher than the value needed for a
proper fit of the Richardson and Zaki relation to our data. If we
use the power n ) 1.4, the maximum possible gas fraction of
the homogeneous regime (Rmax ) 1/n) is 0.71,8,17 and the critical
value12 is Rcrit ) 1/(n + 1) ) 0.42. This underestimates our
findings. Joshi17 provides a more extensive analysis and arrives
at a different expression for the critical gas fraction:
Vs2Cv(1 + Cv) ) Ddbg(Rcrit + Cv) (5)
in which Vs(R) is the slip velocity between the two phases, Cv
is the virtual mass coefficient, D is a dispersion constant given
a value of 1.2, db is the bubble size, and g is gravity’s
acceleration. If in eq 5, the Richardson and Zaki relation is
substituted with n ) 1.4 and the virtual mass is set to 0.5, then
for 3 mm sized bubbles the critical gas fraction is 0.19. For
larger bubbles, this value decreases. Again, we have an
underestimate of our maximum stable gas fraction found in the
experiment.
The Garnier model uses an R1/3 dependence. This reflects
the bubble-bubble distance in the swarm. Garnier et al. report
that their bubbles have an equivalent diameter ranging from
3.4 to 5.5 mm. This is in the same range as the bubble size in
our case. For the model of Garnier, we find a value of 0.28 m/s
for the slip velocity of a single bubble which is in reasonable
agreement with the actual velocity of a single bubble (≈ 0.25
m/s) and is the same value as reported by Garnier. However,
the coefficient, Cµ, in the Garnier model is found to be 0.60 in
our experiments, rather than close to 1 as Garnier et al. report.
A difference with the Garnier experiments is that in our case
the liquid velocity is 0, whereas in their case, it is not. From
their results, it can be observed that their findings are different
for different liquid velocities: the lower the liquid velocity, the
quicker the slip velocity drops as a function of the gas fraction.
In contrast, our data show a slower decrease. It is not clear why
we observe such a difference. The single bubble rise velocity
is almost the same in both Garnier and our experiment.
Furthermore, the bubble size is quite comparable.
If we use the Joshi criterion of eq 5 with the same parameters
as given above, we find for the critical gas fraction a value of
0.2. This is almost equal to the prediction using the Richardson
and Zaki expression for the slip velocity. Again, this value is
much smaller than our experimental one and it decreases with
increasing bubble size. Joshi et al.17 performed a parameter
variation study on all variables in the criterion of eq 5. They
show in their paper that the dispersion coefficient, D, needs to
be as small as 0.5 to get a critical gas fraction above 0.5. This
seems to indicate that at the high gas fraction we can reach, the
dispersion of the gas phase is reduced significantly compared
to the more dilute cases. An explanation for this may be that,

Figure 6. Void fraction profiles in the entrance region (Ug ) 0.023 m/s).
for these dense systems, the motion of individual bubbles is
greatly reduced by the presence of all other bubbles. In essence
this is the well-known mutual hindrance that restricts the bubble
mobility.
4.2. Radial Gas Fraction Profiles. Radial void fraction
profiles were determined with single glass fiber probes for
various superficial gas velocities at various heights in the
column. Figure 6 gives an impression of the uniformity of gas
injection that is achieved with the needle sparger. The result
shows that the void fraction profile contains small nonunifor-
mities at z ) 0.07 m, introduced by imperfections in the sparger,
but that these nonuniformities quickly even out: at z ) 0.15 m,
the void fraction distribution is very uniform. This shows that
the needle sparger can provide very uniform gas injection.
Typically, the gas fraction for bubbles of the present size
(equivalent diameter about 4 mm) is underestimated by 5-10%,
see, e.g., refs 33 and 35. This is due to the difficulty of piercing
a bubble at its side: in many cases, this results in a bouncing
off from the tip rather than being pierced. Surface tension forces
are responsible for this bouncing.
Small peaks are observed in the void fraction at distances of
5 mm from the wall (about one bubble diameter). At smaller
distances to the wall, the void fraction drops more or less
linearly. The drop near the wall may be explained with a local
force driving bubbles away from the wall, such as that modeled
by Antal et al.42 This model is employed quite frequently in
modeling studies for the estimation of, e.g., void fraction profiles
for bubbly pipe flow or bubble columns.43,44 It is not clear,
however, if such a force is really present: the derivation by
Antal42 is performed for small bubble Reynolds numbers. The
works by Takemura and Magnaudet45 and De Vries46 suggest
that, instead, for larger bubble Reynolds numbers, bubbles may
be attracted to and bounce against the wall. An alternative
explanation for the drop in void fraction near the wall may be
the fact that bubbles cannot overlap with the wall. This way,
their shape determines the variation in void fraction. A simple
check with the assumption of an ellipsoidal shape shows that
this gives a reasonable approximation for the decrease in void
fraction near the wall.
Next, the shape of the void fraction profile is investigated
for various heights, superficial gas velocities, and water
contamination levels. Since the gas injection is very uniform,
the nonuniformities observed in the profiles are due to hydro-
dynamic effects and not due to the sparger. First, results are
shown for tap water of several days old (large vortical structures
occur for Ug > 0.06 m/s) and an ungassed liquid height of 1.3 m.
This water is still relatively clean. (The water in the column is
in a batch and will get contaminated over time by contamination
from the surroundings and, more importantly, from the air
supply; air needs to be sparged continuously as otherwise water
would seep into some of the needles, blocking air flow through
Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009 153
Table 2. Superficial Gas Velocities for the Various Gas Fractions
Studied
Ug (m/s) R (-) no. of needles
0.015 6.1% 187
0.017 7.6% 187
0.025 11% 187
0.032 16% 559
0.039 20% 559
0.049 25% 559
them. Moreover, some bacterial growth can not be completely
excluded over the long time that the water is in the column.)
The superficial gas velocities that have been used are given in
Table 2. This table also shows the number of needles that have
been used. The flow rates through the needles for these
conditions give equivalent bubble diameters in the size range
3.5-5.0 mm. If the superficial gas velocity is increased from
0.015 to 0.025 m/s, the bubble size increases due to the larger
flow rates through the needles. If the superficial velocity is
increased further, the number of needles is tripled, the flowrate
through the needles decreases, and the bubble diameter drops
again. As a result, the bubble diameter variation over the entire
void fraction range is reduced and attains a local maximum value
for R ) 11%.
Figure 7 shows void fraction profiles for superficial gas
velocities of 0.015 and 0.049 m/s. Figure 8 shows more profiles
for the superficial gas velocities without large scale structures
for z ) 1.2 and 0.6 m. The figures show that the inside of the
bubble column is very uniform for all superficial velocities,
except close to the surface for R ) 25%. The low superficial
gas velocities show a wall peaking behavior for all heights. The
same peaking behavior is observed for Ug > 0.035 m/s (R >
18%) for z < 0.9 m. In higher parts of the column, the peak
disappears. A small dip occurs near x/R ) 0.7 for all experiments
with Ug > 0.03 m/s. This small dip is due to an artifact of the
experiments with the probe. We have used 5 different probes
simultaneous to speed up the measurements. The one positioned
at x/R ) 0.7 systematically slightly underpredicts the void
fraction.
Additional void fraction profiles for more contaminated tap
water of two weeks old are shown in Figure 9. For all the
experiments from this point on, the ungassed liquid height was
1.0 m, unless noted otherwise. The behavior is similar to that
observed in Figures 7 and 8. For the superficial gas velocities,
that are significantly below the critical superficial gas velocity
where the onset of vortical structures occurs (Ugcrit ≈ 0.07 m/s),
again wall peaking is observed for all heights (e.g., Figure 9a
with Ug ) 0.047 m/s). For the superficial gas velocities
somewhat below the critical superficial gas velocity (Figure 9b
with Ug ) 0.066 m/s), the wall peaking disappears near the top
of the bubble column. For the case where dynamic large scale
structures are present (Figure 9c with Ug ) 0.076 m/s), the
familiar core-peaking void fraction profile is found. In this case,
an additional underestimation of the void fraction is present:
bubbles rising downward have a smaller probability of being
pierced.47
Wall peaking behavior has been reported for bubbly pipe
flows by e.g. Serizawa et al.48 Similar void fraction wall peaking
is observed in model predictions for bubble columns and pipe
flow by Guet et al.43 The behavior is explained by the effect of
the lift force, which causes accumulation of small bubbles (i.e.,
smaller than 5 mm) in the negative velocity part of the flow,
which is close to the wall. The position where the peaking is
observed in the current experiments is at 5 mm from the wall.
The lift force acting on the bubbles at this location is significant
154 Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009

Figure 7. Void fraction profiles for 6.1% void fraction (Ug ) 0.015 m/s) (a) and 25% void fraction (Ug ) 0.049 m/s) (b). The tap water is several days old.

Figure 8. Void fraction profiles for various gas fractions at z ) 1.2 m (a) and z ) 0.6 m (b). The tap water is several days old.

Figure 9. Void fraction profiles for tap water which is 2-3 weeks old. Large scale structures are found for Ug > 0.07 m/s.
since these bubbles experience a shear due to the down flow 0.9 with a typical very small velocity around 0.01 m/s.
close to the wall. This downward liquid motion is caused by Downflow occurs for r/R > 0.9 and has a typical velocity of
the slightly lower gas fraction very close to the wall and thus around -0.04 m/s. The downflow is most likely caused by
a higher local mixture density. the low void fraction very close to the wall, which causes a
4.3. Axial Liquid Velocity Profiles. Mean axial liquid density difference between the inner bubble column regions
velocity profiles for various gas fractions are shown in Figures and the wall region. Most publications on bubble columns
10 and 4. The profiles show that for all superficial gas (e.g., refs 49 or 50) report that the inversion point (i.e., the
velocities, upflow occurs in the central region with r/R < point where uax,liq ) 0) is located around r/R ) 0.7, whereas

Figure 10. Mean axial liquid velocity profile for various void fractions.
The tap water is several days old. The ungassed liquid height is 1.3 m.
the present study finds the inversion point near r/R ) 0.9.
This shows the very strong homogeneity for the present flow.
The difference is most likely caused by the absence of any
large scale structure in the flow. For the other studies, the
gas injection may not have been uniform enough to achieve
this. The results close to the wall show no trend with Ug.
This is probably due to the presence of a strong preshift
frequency bleed through in the LDA signal for measurements
near the wall. This affects the mean velocity. In addition,
since the wall location is determined by locating positions
where zero velocity is found, a small error in the radial
alignment is created, which manifests itself strongest close
to the wall where the velocity gradient is largest.
4.4. End of Homogeneous Regime. In the previous sections,
wall peaking of the void fraction profiles was found, which is
generally associated with a lift force driving the smaller bubbles
closer to the wall. Close to the wall a narrow region with liquid
down flow exists, providing a velocity gradient that can produce
a significant lift force. In the higher parts of the column, and
for the higher superficial gas velocities approaching the critical
gas velocity where the first large structures are observed, the
velocity of the down flow increases. Nevertheless, for these
conditions, the wall peaking disappears. This indicates that
the lift force toward the wall is decreasing in magnitude. Since
the actual velocity gradient is increasing, the decrease in the
lift force is probably caused by a decrease of the lift coefficient,
which may possibly even reach a negative value. The results in
ref 51 show that the lift coefficient decreases if the horizontal
bubble diameter approaches a critical value, which is ap-
proximately 5.8 mm for air-water. The results, therefore, could
indicate that the mechanism for the transition that was proposed
by Lucas et al.1 plays an important role. These authors argue
that small, local disturbances of the state of uniform bubble
distribution get positive feedback from the lift force if the bubble
size is larger than the critical size. Consequently, the distur-
bance grows and the homogeneous regime looses its stability.
Fox and co-workers2 proposed a different mechanism. From
their stability analysis, they concluded that the transition from
homogeneous to heterogeneous flow could be caused by an
instability triggered by a reduction of the bubble induced
turbulence. This reduction is a consequence of the suppression
of bubble wakes in the more dense swarms. The authors found
no need to link the transition to a change in lift coefficient as
has been proposed by Lucas and co-workers. The mechanism
proposed by Fox and co-workers is interesting and requires
additional experiments in which the focus should be on
retrieving turbulence data.
Figure 11 shows images of the bubbles close to the wall. It
is evident from these pictures that with increasing gas fraction
the bubble size increases. If this continues, then according to
the Lucas criterion, the uniform flow looses its stability when
Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009 155
the bubble size has increased beyond a critical size. To further
investigate this, measurements of the actual bubble size have
been conducted. However, at the same time, the swarm becomes
denser and the wakes may be suppressed causing a transition
toward homogeneous flow according to the Fox mechanism.
Figure 12 shows the means and standard deviations for the
major (a) and minor (b) axis lengths of the bubbles obtained
with a photographic technique. For the horizontal diameter, clear
differences are observed between the case of “fresh” tap water
and the case of contaminated tap water. For the relatively clean
water, the horizontal bubble size is increasing for Ug < 0.05
m/s, whereas it is more or less constant for the contaminated
water. In this range, the bubble shape for the contaminated water
is more or less ellipsoidal, whereas the bubbles in the relatively
“clean” water have much more irregular shapes and exhibit
strong shape oscillations. As soon as the horizontal diameter
reaches a value around 5.8 mm in the region near the free
surface, the increase levels off and strongly increased liquid
downflow is observed close to the wall in the top regions of
the bubble column (“A” for fresh, “a” for contaminated water).
If the superficial gas velocity is increased slightly further, large
scale instability is observed, first in the top parts of the column
(“B” for fresh water) and for a small further increase in the
entire bubble column (“C” for fresh, “c” for contaminated
water).
The results could be explained by the reversal of the lift-
force direction as proposed by Lucas et al.1 The graph in Figure
13 reproduces the results of the stability analysis by Lucas et
al.1 for a Gaussian bubble size distribution. If the standard
deviation of the bubble diameter exceeds a certain critical value,
the flow becomes unstable. This critical standard deviation is a
function of the mean bubble diameter. The mean and standard
deviation that were observed in the present experiments at the
onset of dynamic large scale structures are also plotted in the
same figure. The agreement with the result of the stability
analysis is good, although the good match is not very sensitive
to the precise value of the standard deviation for the present
conditions. In addition, the result shows that the width of the
bubble diameter distribution is quite small in terms of lift-force
behavior.
When the gas flowrate is increased to about Ug ) 0.08 m/s,
the uniform flow looses its stability. However, at this setpoint,
the flow starts to oscillate between homogeneous and hetero-
geneous, with a period of some 20 s. The gas fraction as a
function of time is for this condition given in Figure 14. After
reaching its highest gas fraction of about 55%, an instability
quickly develops: from the free surface, the liquid starts flowing
rapidly downward in the wall region and the entire flow becomes
“turbulent”. Note that at these high gas fractions it is even close
to the wall difficult to measure with the LDA.
The downward velocity is high enough to drag bubbles
downward. This can be seen in Figure 15 that shows the time
trace of the vertical component of the liquid velocity in a point
close to the wall as obtained with the LDA. After a few seconds,
the gas fraction reaches a minimum and the downward liquid
velocity a maximum. Then, the flow starts to relax, the gas
fraction starts to climb, and the homogeneous flow is restored.
When the gas fraction reaches 55%, the cycle repeats in a
sawtooth manner.
From visual observation, we tend to conclude that the bubble
size is decreased during the “turbulent” stage. This makes sense
as the large scale liquid circulation will strip the bubbles
somewhat earlier from the needles of the sparger. Therefore,
the bubble size decreases. This fits with the picture that the onset
156 Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009
Figure 11. Images of bubbles close to the wall for increasing void fraction (contaminated water of three weeks old).
Figure 12. Evolution of mean and standard deviation of the major and minor
bubble axis lengths for two levels of water contamination. The dimensions
were obtained with a photographic technique.
Figure 13. Comparison of the maximum stable standard deviation of the
bubble diameter according to ref 43 and the conditions in the current
experiment at the onset of transition.
of the instability is caused by a reversal of the direction of the
lift force. The bubbles at the top are slightly larger than in the
Figure 14. Alternating regimes for Ug ) 0.08m/s: time series of the gas
fraction.
Figure 15. Alternating regimes for Ug ) 0.08m/s: time series of the axial
liquid velocity close to the wall.
rest of the column due to the lower hydrostatic pressure. Hence,
we can expect the first instability to occur at the top of the
column, with the bubbles moving away from the wall. The
relatively heavy layer close to the wall then starts moving
downward, creating the large scale instability and moving more
bubbles to the center at the position where it “bends inward”
(for continuity reasons). Finally, when the liquid circulation
reaches the sparger, the bubbles formed are smaller and the lift
force regains its original sign but is now stronger and can restore
the uniform distribution. This takes several seconds as both the
small bubbles will have to move to the top of the column and,
more importantly, the liquid flow will have to decay. Once this
flow has decayed, the uniform gas distribution can be re-
established. However, at the same time, the bubbles formed at
the sparger will get bigger as they no longer feel the drag from
the liquid flow trying to pull them off. Thus slightly larger
bubbles will move upward, reaching the critical size at the top
of the column and the cycle repeats.
The wake suppression (as proposed by Fox et al.) could
equally well explain the above observation. Now the critical
swarm density with the required turbulence reduction also starts
at the top of the column, as due to the reduced hydrostatic
pressure, the bubble size is increased, and the gas fraction is
slightly higher in the top region. Also with this model, the flow

may become in a transitional state. Now due to the circulation,
the swarm gets less dense and the turbulence level may be
restored. Additional experiments are required to sort out the
cause of the instability.
5. Concluding Remarks
In this paper, we have reported experiments on bubbly flow
in a 15 cm-diameter cylindrical column. The continuous phase
(water) is in a batch. The air bubbles (equivalent diameter about
3.5-5 mm) are injected via a special needle sparger that
generates a very uniform input of the bubbles. We employed
optical glass fiber probes and LDA to assess the uniformity of
the flow.
After careful alignment of the column, we obtained flat gas
fraction profiles up to gas fractions as high as 55%, with a small
wall peaking. As a consequence of this flat gas fraction profile,
the liquid shows hardly any circulation: liquid velocities close
to the wall are a few centimeters per second downward, and in
the center, we measure typically 1-2 cm/s upward flow, which
is partly a bias error in the LDA measurements. The downflow
at the wall results in a shear field that will via a lift force push
bubbles smaller than about 5-6 mm toward the wall. This could
be responsible for the wall peak we observe.
The flow is steady: only small scale fluctuations are observed.
Our findings are in contrast to predictions from linear stability
analysis. Generally, the latter predicts a transition from the
uniform regime into a heterogeneous one for gas fractions in
the range of 25-35%. However, if we let the water “age”, we
find stable uniform flow up to 55%, i.e. well beyond the stable
regime predicted by the linear stability analysis for air-water
bubbles of the present size. In the homogeneous regime, the
gas fraction and the superficial gas velocity are coupled by a
Richardson and Zaki (RZ) type of relation, i.e. the slip velocity
can be described by Vs ) V∞(1 - R)n-1 with V∞ ) 0.195 m/s,
which is too low for the present bubbles, and n ) 1.4. However,
also the relation proposed by Garnier et al.23 gives an adequate
fit to the data with Vs ) V∞(1 - CµR1/3) with V∞ ) 0.28 m/s and
Cµ ) 0.6.
At a gas fraction of 55%, the flow becomes unstable.
Instabilities at the top develop and propagate downward. The
nature of the flow changes drastically: the gas profile is no longer
flat, but core peaking is found instead. This causes large density
differences within the mixture and gravity will act, making the
flow unstable. Our findings could be explained by the reversal
of the lift force once the bubble size passes a critical diameter,51
resulting in a flow instability as proposed by Lucas et al.1
However, care should be taken, as at these high gas fractions
the bubble-bubble distance is small and the wakes cannot
develop in a similar way as in the case of isolated bubbles.
Therefore, the instability may also be triggered by the suppres-
sion of the wake as proposed by Fox and co-workers.2
Acknowledgment
This work has been done via a grant from the Dutch
Foundation for Fundamental Research on Matter (FOM) under
the Dispersed Multiphase Flow programme.
Literature Cited
(1) Lucas, D.; Prasser, H. M.; Manera, A. Influence of the lift force on
the stability of a bubble column. Chem. Eng. Technol. 2005, 60, 3609–
3619.
(2) Monahan, S. M.; Fox, R. O. Linear stability analysis of a two-fluid
model for air-water bubble columns. Chem. Eng. Sci. 2007, 62, 3159–3177.
Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009 157
(3) Deckwer, W.-D. Bubble column reactors; Wiley: Chicester, 1992.
(4) Zahradnı́k, J.; Fialová, M.; Røužička, M.; Drahoš, J.; Kaštánek, F.;
Thomas, N. H. Duality of gas-liquid flow regimes in bubble colomn reactors.
Chem. Eng. Sci. 1997, 52 (21/22), 3811–3826.
(5) Shaikh, A.; Al-Dahhan, M. H. A Review of FLow Regime transitions
in Bubble Columns. Int. J. Chem. Reac. Eng. 2007, 5, 1–68.
(6) Mudde, R. F. Gravity-Driven Bubbly Flows. Ann. ReV. Fluid Mech.
2005, 37, 393–423.
(7) Reith, T.; Renken, S.; Israel, B. A. Gas hold-up and axial mixing in
the fluid phase of bubble columns. Chem. Eng. Sci. 1968, 23, 619–629.
(8) Lockett, M. J.; Kirkpatrick, R. D. Ideal bubbly flow and actual flow
in bubble columns. Trans. Inst. Chem. Eng. 1975, 53, 267–273.
(9) Zahradnı́k, J.; Kaštánek, F. Gas holdup in uniformly aerated bubble
column reactors. Chem. Eng. Commun. 1979, 3, 413–429.
(10) Maruama, T.; Yoshida, S.; Mizushina, T. The flow transition in a
bubble column. J. Chem. Eng. Jpn. 1981, 14, 352–357.
(11) Ruzicka, M.; Zahradnı́k, J.; Drahoš, J.; Thomas, N. H. Homogeneous-
heterogeneous regime transition in bubble columns. Chem. Eng. Sci. 2001,
56, 4609–4626.
(12) Shnip, A. I.; Kolhatkar, R. V.; Swamy, D.; Joshi, B. J. Criteria for
the transition from the homogeneous to the heterogeneous regime in two-
dimensional bubble column reactors. Int. J. Multiphase Flow 1992, 18, 705–
726.
(13) León-Becerril, E.; Liné, A. Stability analysis of a bubble column.
Chem. Eng. Sci. 2001, 56, 6135–6141.
(14) León-Becerril, E.; Cockx, A.; Liné, A. Effect of bubble deformation
on stability and mixing in bubble columns. Chem. Eng. Sci. 2002, 57, 3283–
3297.
(15) Lucas, D.; Krepper, E.; Prasser, H. M.; Manera, A. Investigations
on the stability of the flow characteristics in a bubble column. Chem. Eng.
Technol. 2006, 29 (9), 1066–1072.
(16) Bhole, M. R.; Joshi, J. B. Stability analysis of bubble columns:
predictions for regime transition. Chem. Eng. Sci. 2005, 60, 4493–4507.
(17) Joshi, J. B.; Desphande, N. S.; Dinkar, M.; Phanikumar, D. V.
Hydrodynamic Stability of Multiphase Reactors. AdV. Chem. Eng. 2001,
26, 1–130.
(18) Drahoš, J.; Zahradnı́k, J.; Punčochár, M.; Fialová, M.; Bradka, F.
Effect of operating conditions on the characteristics of pressure fluctuations
in a bubble column. Chem. Eng. Process 1991, 29, 107–115.
(19) Ruzicka, M.; Drahoš, J.; Fialová, M.; Thomas, N. H. Effect of
bubble column dimensions on flow regime transition. Chem. Eng. Sci. 2001,
56, 6117–6124.
(20) Ruzicka, M.; Drahoš, J.; Mena, P. C.; Teixeira, J. A. Effect of
viscosity on homogeneous-heterogeneous flow regime transition in bubble
columns. Chem. Eng. J. 2003, 96, 15–22.
(21) Ruzicka, M.; Vecer, M. M.; Orvalho, S.; Drahoš, J. Effect of
surfactant on homogeneous regime stability in bubble column. Chem. Eng.
Sci. 2008, 63, 951–967.
(22) Yao, B. P.; Zheng, C.; Gasche, H. E.; Hofmann, H. Bubble
behaviour and flow structure of bubble columns. Chem. Eng. Process 1991,
29, 65–75.
(23) Garnier, C.; Lance, M.; Marié, J. L. Measurement of local flow
characteristics in buoyancy-driven bubbly flow at high void fraction. Exp.
Therm. Fluid Sci. 2002, 26, 811–815.
(24) Harteveld, W. K.; Mudde, R. F.; Van Den Akker, H. E. A.
Dynamics of a bubble column: influence of gas distribution on coherent
structures. Can. J. Chem. Eng. 2003, 81 (3-4), 389–394.
(25) Harteveld, W. K.; Juliá, J. E.; Mudde, R. F.; Van Den Akker,
H. E. A. Large scale vortical structures in bubble columns for gas fractions
in the range of 5%-25%. Proceedings of the 16th International Conference
on Chemical and Process Engineering, Prague, Czech Republic, Aug 22–
26, 2004.
(26) Kulkarni, A. A.; Ekambara, K.; Joshi, J. B. On the development of
flow pattern in a bubble column reactor: experiments and CFD. Chem. Eng.
Sci. 2007, 62, 1049–1072.
(27) Monahan, S. M.; Vitankar, V. S.; Fox, R. O. CFD predictions for
flow-regime transitions in bubble columns. AIChE J. 2005, 51 (7), 1897–
1923.
(28) Biesheuvel, A.; Gorissen, J. Void fraction disturbances in a uniform
bubbly flow. Int. J. Multiphase Flow 1990, 16, 211–231.
(29) Rampure, M. R.; Kulkarni, A. A.; Ranade, V. V. Hydrodynamics
of bubble column reactors at high gas velocity: Experiments and compu-
tational fluid dynamics (CFD) Simulations. Ind. Eng. Chem. Res. 2007, 46
(25), 8431–8447.
(30) Dhotre, M. T.; Joshi, J. B. Design of a gas distributor: Three-
dimensional CFD simulation of a coupled system consisting of a gas
chamber and a bubble column. Chem. Eng. J. 2007, 125 (3), 149–163.
158 Ind. Eng. Chem. Res., Vol. 48, No. 1, 2009
(31) Lucas, D.; Krepper, E.; Prasser, H. M. Investigations on the stability
of the flow characteristics in a bubble column. Chem. Eng. Technol. 2005,
29 (9), 1066–1072.
(32) Van Baten, J. M.; Krishna, R. Scale effects on the hydrodynamics
of bubble columns operating in the heterogeneous flow regime. Chem. Eng.
Res. Des. 2004, 82 (A8), 1043–1053.
(33) Cartellier, A. Simultaneous void fraction measurement, bubble
velocity, and size estimate using a single optical probe in gas-liquid two-
phase flows. ReV. Sci. Instrum. 1992, 63, 5442–5453.
(34) Mudde, R. F.; Saito, T. Hydrodynamical similarities between bubble
column and bubbly pipe flow. J. Fluid Mech. 2001, 437, 203–228.
(35) Juliá, J. E.; Harteveld, W. K.; Mudde, R. F.; Van Den Akker,
H. E. A. On the accuracy of the void fraction measurements using optical
probes in bubbly flows. ReV. Sci. Instrum. 2005, 76 (3), 1–13.
(36) Barrau, E.; Riviere, N.; Poupot, C.; Cartellier, A. Single and double
optical probes in air-water two-phase flows: real time signal processing
and sensor performance. Int. J. of Multiphase Flow 1999, 25, 229–256.
(37) Groen, J. S.; Mudde, R. F.; Van Den Akker, H. E. A. On the
application of LDA to bubbly flow in the wobbling regime. Exp. Fluids
1999, 27, 435–449.
(38) Tummers, M. Investigation of a turbulent wake in an adverse
pressure gradient using laser Doppler anemometry. PhD theis, Delft
University of Technology, The Netherlands, 1999.
(39) Rice, R. G.; Littlefield, M. A. Dispersion coefficients for ideal
bubbly flow in truly vertical bubble columns. Chem. Eng. Sci. 1987, 42
(8), 2045–2053.
(40) Rice, R. G.; Barbe, D. T.; Geary, N. W. Correlation of nonverticality
and entrance effects in bubble columns. AIChE J. 1990, 36 (9), 1421–1424.
(41) Tinge, J. T.; Drinkenburg, A. A. H. The influence of slight
departures from vertical alignment on liquid dispersion and gas hold-up in
a bubble column. Chem. Eng. Sci. 1986, 41 (1), 165–169.
(42) Antal, S. P.; Lahey, R. T.; Flaherty, J. E. Analysis of phase
distribution in fully developed laminar bubbly two phase flow. Int. J.
Multiphase Flow 1991, 17, 635–652.
(43) Guet, S.; Ooms, G.; Oliemans, R. V. A. Simplified Two-Fluid model
for gas-lift efficiency predictions. AIChE J. 2005, 51 (7), 1885–1896.
(44) Krepper, E.; Reddy Vanga, B. N.; Prasser, H.-M.; Lopez de
Bertodano, M. A. Experimental and numerical studies of void fraction
distribution in rectangular bubble columns. Proceedings of the 3rd
International Symposium on Two-Phase Flow Modelling and Experimenta-
tion, Pisa, Italy, Sept 22-24, 2004.
(45) Takemura, F.; Magnaudet, J. The transverse force on clean and
contaminated bubbles rising near a vertical wall at moderate Reynolds
number. J. Fluid Mech. 2003, 495, 235–253.
(46) De Vries, P. W. G. Path and wake of a rising bubble. PhD thesis,
Twente University, The Netherlands, 2001.
(47) Groen, J. S. Scales and structures in bubbly flows. PhD thesis, Delft
University of Technology, The Netherlands, 2004.
(48) Serizawa, A.; Kataoka, I.; Michiyoshi, I. Turbulent structure of air/
water bubble flow. Int. J. Multiphase Flow 1975, 2, 235–246.
(49) Mudde, R. F.; Groen, J. S.; Van Den Akker, H. E. A. Liquid velocity
field in a bubble column: LDA experiments. Chem. Eng. Sci. 1997, 52,
4217–4224.
(50) Hills, J. H. Radial non-uniformity of velocity and voidage in a
bubble column. Trans. Inst. Chem. Eng. 1974, 52, 1–9.
(51) Tomiyama, A.; Tamai, H.; Zun, I.; Hosokawa, S. Transverse
migration of single bubbles in simple shear flows. Chem. Eng. Sci. 2002,
57, 1849–1858.
ReceiVed for reView January 16, 2008
ReVised manuscript receiVed June 24, 2008
Accepted July 3, 2008
IE8000748