Chemical Engineering Science 56 (2001) 6359–6365

www.elsevier.com/locate/ces

Numerical simulation of multiphase #ow in bubble column reactors.

In#uence of bubble coalescence and break-up
E. Olmos, C. Gentric ∗ , Ch. Vial, G. Wild, N. Midoux
Laboratoire des Sciences du Genie Chimique, 1, rue Grandville BP451, 54001 Nancy Cedex, France

Abstract
Population balance equations have been combined to a classical hydrodynamic Euler=Euler simulation to investigate the operation
of a cylindrical bubble column. The MUSIG (mutiple-size-group) model implemented in the CFX 4.3 commercial software has
been used. Hydrodynamic experimental variables, i.e. local axial liquid velocity and local gas hold-up, have been compared to the
corresponding calculated values, showing a quite good agreement, except for the gas hold-up when the column is no more operating
in the homogeneous regime. Bubble sizes have been investigated, showing that two domains of super;cial gas velocities can be
distinguished. In the ;rst domain, coalescence occurs predominantly, Sauter diameter increases with the super;cial gas velocity,
bubble size distribution is narrow and Sauter diameter is continuously evolving along the column axis. In the second domain,
break-up becomes more intensive and compensates coalescence, bubble size distribution becomes wider, since more small bubbles
are formed, an equilibrium Sauter diameter appears when the super;cial gas velocity increases. Furthermore an equilibrium Sauter
diameter appears along the column axis, and it can be noticed that this phenomenon appears lower in the column when the gas #ow
rate is increased. In these two domains the characteristics of the bubbles are typical of those of the homogeneous and transition
regimes. ? 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Bubble column; Euler–Euler; Simulation; Gas–liquid #ow; Coalescence; Break-up; CFD

1. Introduction properties of the liquid phase, the gas sparger, the bub-
ble properties (size and shape) and the hydrodynamics.
1.1. Bubble columns hydrodynamics The type of #ow regime and consequently the nature of
the dispersion in#uences particularly the bubble-bed be-
Bubble column reactors are widely used in bio- and haviour.
petro-chemical industries. They are known as excellent Three regimes (Zahradnik et al., 1997) generally occur
reactors for processes which require large interfacial area in bubble columns. The homogeneous regime is obtained
for gas–liquid mass transfer and e@cient mixing for react- at low gas super;cial velocities. Its bubble size distribu-
ing species. A better knowledge of the local hydrodynam- tion is monomodal, narrow and is only in#uenced by the
ics to increase the predictability of the reactor design and type of gas sparger used since coalescence-break-up phe-
to improve the e@ciency of the processes appears now nomena can be neglected. Gas hold-up and bubble diam-
necessary. The use of the computational #uid dynamics eter pro;les are uniform. As super;cial gas velocity in-
(CFD) should be able to improve this knowledge, by pro- creases, the heterogeneous regime is obtained, in which
viding a complete description of the local hydrodynamics coalescence and break-up occur. Large bubbles with sizes
if an adequate model is used. However, the complex inter- depending on the column diameter are formed. Usually,
actions between the gas and the liquid phases cause many an intermediate region, called the transition regime, ex-
modelling problems still to be solved. Numerous stud- ists between these two regimes.
ies (Shah, Kelkar, Godbole, & Deckwer, 1982; Camarasa The study of these regimes shows that the major phe-
et al., 1999) have shown that strong interactions exist nomenon that governs the transition between the homo-
between the operating conditions, the physico-chemical geneous and the heterogeneous regimes is the occurrence
of the large bubbles described previously. Therefore it
∗ Corresponding author. clearly appears that the description of the #ow regimes

0009-2509/01/$ - see front matter ? 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 0 1 ) 0 0 2 0 4 - 4
6360 E. Olmos et al. / Chemical Engineering Science 56 (2001) 6359–6365

is improved when bubble coalescence and break-up are related to #ow regimes. Bubble size distributions obtained
accounted for in the model. by the model are compared to experimental distributions
Detailed investigations on break-up and coalescence obtained by photography.
mechanisms in air-sparged reactors have been proposed
(Prince & Blanch, 1990; Lee, Erickson, & Glasgow,
1987) but they were limited to the development of mod-
2. Flow equations
els for the prediction of bubble coalescence and break-up
frequencies. More recently, coalescence and break-up
2.1. Bubble coalescence and break-up model
have been studied numerically in bubbly #ows by solv-
ing two-group (a spherical=distorted bubble group and a
The implementation of population balance equations
cap=slug bubble group) interfacial area transport equa-
into CFX 4.3 developed by Lo (1996) is used to take into
tions (Hibiki & Ishii, 2000), or an additional balance
account the non-uniform bubble size distribution. Bub-
equation for the average bubble volume derived from a
bles from 1 to 10 mm diameter equally divided into 10
population balance in bubble columns (Millies & Mewes,
classes are considered (see Table 1). Instead of consid-
1999) and in vertical #ows (Wu, Kim, Ishii, & Beus,
ering 11 diJerent complete phases, each bubble class is
1998).
assumed to move at the same mean algebraic velocity for
signi;cant reduction of computing time. A system of 11
coupled equations (1 momentum and 10 continuity equa-
1.2. Objectives
tions) is therefore to be solved for the gas phase coupled
to 1 momentum and 1 continuity equations for the liquid
For the past few years, two main approaches have
phase.
been used for numerical investigations of bubbly #ows.
The break-up model is taken from Luo and
The ;rst one is the Euler–Euler approach considering
Svendsen (1996) and assumes isotropic turbulence. The
the two phases as two continuous media. Its low com-
coalescence model from Prince and Blanch (1990) is
putational load has allowed to develop many studies
used. It describes the coalescence process as occurring in
(Sokolichin & Eigenberger, 1994; Van den Akker, 1998;
three steps: a ;rst step where the bubbles collide and trap
Pan, Dudukovic, & Chang, 2000). The second one is
a layer of liquid between them, a second step where this
the Eulerian–Lagrangian discrete bubble model. This
liquid layer drains until it reaches a critical thickness, and
approach is more realistic but the track of a su@cient
a last step during which this liquid ;lm disappears and the
number of particles, required for accurate modelling,
bubbles coalesce. The collisions between bubbles may be
demands high computational memory and speed (Lapin
caused by turbulence, buoyancy or laminar shear. Only
& LIubbert, 1994; Delnoij, Lammers, Kuipers, & van
the ;rst cause of collision (turbulence) is considered in
Swaaij, 1997).
the present model. Indeed collisions caused by buoyancy
All these studies based on CFD are most of time limited
cannot be taken into account as all the bubbles from each
to the homogeneous regime because they consider only
class move at the same speed. Moreover, calculations
one bubble size. As a matter of fact, in the heterogeneous
showed that laminar shear collisions are negligible in the
regime, the bubble size distribution becomes bimodal
range of super;cial gas velocities considered.
with the apparition of large bubbles (De Swart, van Vliet,
The general form of the population balance equation is
& Krishna, 1996) inducing di@culties to describe the
#ow correctly. For this purpose, Krishna, Urseanu, van @ni
Baten, and Ellenberger (1999) considered a three-phase +  · (ug ni ) = PB + PC − DB − DC ; (1)
@t
continuum (liquid, small and large bubbles phases) and
applied two distinct interphase momentum terms on the PB ; PC ; DB ; DC are, respectively, the production rates due
small and on the large bubbles. to break-up and coalescence and the death rate to break-up
The aim of this work is to combine a coalescence and and coalescence of bubbles of diameter di . To ;t our data,
break-up model with a complete #ow numerical simula- coalescence and break-up rates are calibrated by a factor
tion. To achieve this objective, the Euler–Euler two-#uid of 0.075.
approach is used. The #ow equations are solved using The number density of the ith class ni is then related
a ;nite volume technique proposed by the commercial to the gas hold-up by
package CFXTM 4.3 from AEA Technology. Hydrody-
namic variables such as liquid velocity, global and local ni vi = g fi : (2)
gas hold-up are calculated and compared to experimen-
tal data. The pro;le of the mean bubble diameter in the 2.2. Flow equations
column and its variation with the super;cial gas veloc-
ity are studied and enable us to emphasize the in#uence Euler=Euler volume-averaged continuity and momen-
of the primary gas distribution. This evolution is also tum transport equations are written for each phase as
 E. Olmos et al. / Chemical Engineering Science 56 (2001) 6359–6365 6361

Table 1
Diameter of each bubble class
Class no. 1 2 3 4 5 6 7 8 9 10

Central class diameter di (mm) 1.58 2.43 3.41 4.2 5.09 5.98 6.88 7.78 8.67 9.57

follows: turbulent kinetic energy and of the gradient of the volume

fraction of liquid:
@
(l l ) +  · (l l ul ) = 0 for the liquid phase; (3) t t t
@t Flg = − CTD l k  l and Fgl = − Flg (7)

@(g g fi ) with the recommended value CTD = 0:1.

+  · (g ug g fi ) = Si for the gas phase; Consequently the interphase momentum exchange
@t
D t
term is given by: Flg = Flg + Flg .
(4)
@
(k k uk ) +  · (k k uk uk )
@t 3. Experimental set-up
2
= − k p + k k g + Fkm + k k  u (k; m = l; g):
3.1. Experimental apparatus
(5)
The experimental set-up is a 10 cm diameter bubble
In Eq. (4) Si is a source term due to coalescence and ◦
column ;lled with tap water (H = 1:35 m, T = 20 C). No
break-up. The right term of Eq. (5) describes all the forces mass and no heat transfer are considered. The gas phase
acting on phase k in each control volume: the pressure (air) is introduced at the bottom of the column through
gradient, gravity, the interphase momentum exchange be- two possible spargers: either a porous plate (5 mm thick,
tween phase k and phase m Fkm and the viscous stress 10 –16 m mean pore diameter) or a multiple-ori;ce
term. Lift and added mass forces are neglected. The in- nozzle (62 1 mm circular holes uniformly spaced). The
terphase momentum transfer between gas and liquid due super;cial gas velocity varies from 0.5 to 9:6 cm s−1
to drag force is given by allowing homogeneous and transition regimes to be
3 1 investigated. Axial liquid velocity has been measured
D
Flg = CD g l |ul − ug |(ul − ug ) by laser Doppler velocimetry, the local gas hold-up by
4 dS
optical probes and the bubble size distributions using
and photography . All measurements were done at the middle
height of the column. Further details are given in Vial,
D D
Flg = − Fgl : (6) LainNe, Poncin, Midoux, and Wild (2001).

The drag coe@cient is chosen constant to 0.8 which cor-

3.2. Numerical methods
responds to the ellipsoidal regime.
Axial symmetry and steady state are assumed. The
2.3. Turbulence model grid is rectangular, structured and uniform, the calcula-
tion cells are 5 × 5 mm2 large. The population balance
For the continuous liquid phase, a k– model is ap- equations are solved every 20 hydrodynamic equations in
plied with its standard constants: C1 = 1:44, C2 = 1:92, order to reduce the computational time and only bubbles
C = 0:09; k = 1,  = 1:3. No turbulence model is ap- of the 5th class (5:1 mm) enter the column (estimated
plied on the dispersed phase but the in#uence of the dis- bubble size formed at the sparger). The convergence is
persed phase on the turbulence of the liquid phase is obtained between approximately 5000 and 10000 itera-
taken into account with Sato’s additional term (Sato & tions which correspond to 3–6 h of calculation on a com-
Sekoguchi, 1975) with its constant chosen to 0.6 as best puter provided with a Pentium III 1 GHz processor.
;t of our data.

2.4. Additional force 4. Hydrodynamics: experiments vs. calculations

t
An additional interphase force Flg taking into account First, predicted and experimental hydrodynamics are
the turbulence induced dispersion (Kurul & Podowski, compared. It can be observed in Fig. 1 that the agree-
1990) is considered. This force is a function of the ment between the experimental and the calculated liquid
6362 E. Olmos et al. / Chemical Engineering Science 56 (2001) 6359–6365

Fig. 3. Comparison between calculated and experimental global gas

Fig. 1. Comparison between calculated and experimental (obtained hold-up. In#uence of coalescence and break-up.
with the multiple ori;ce nozzle) radial liquid velocity pro;les
(h=H = 0:5 m).

Fig. 4. Evolution of the calculated global Sauter diameter with Ug .

heterogeneous regimes are characterised by large bubbles

Fig. 2. Comparison between calculated and experimental local gas (d ¿ 20 mm) with high velocities. That entails a plateau
hold-up pro;les (h=H = 0:5 m).
in the global gas hold-up. The issue probably originates
in the choice of the maximum diameter (10 mm) con-
velocities is quite good in tendency. The model without sidered for the modelling. This narrow distribution only
coalescence gives similar results (data not shown). leads to global Sauter mean diameters of 7 mm (see
The local gas hold-up measured and calculated at mid Fig. 4) at these super;cial gas velocities, so bubbles are
column are shown in Fig. 2: radially uniform gas hold-up still in the theoretical domain of constant terminal veloc-
is obtained at low gas velocities corresponding to the ho- ity and the large bubbles with high velocities appearing
mogeneous regime. For higher gas velocities, these pro- in the transition regime are not taken in account. These
;les become parabolic which is characteristic of the tran- very large diameters have not been considered in order
sition and the heterogeneous regimes. to ;nd a compromise between calculation time and a rel-
The in#uence of the super;cial gas velocity on the ative ;ne description of bubble size distribution.
global gas hold-up (Fig. 3) shows that the numerical
method allows a good prediction of the experimental val-
ues as long as the super;cial gas velocity is low to mod- 5. Bubble size distributions: experiments vs.
erate (UG ¡ 7 cm s−1 ). Beyond this critical velocity, the calculations
transition regime is well established and the stagnation of
g  is not predicted neither with nor without the imple- To get a better understanding of the gas–liquid #ow
mented coalescence and break-up model. When the tran- structure, the photographic technique has been used and
sition regime is established, the calculations overestimate its results compared with the numerical simulation re-
the experimental values. This may be due to the choice of sults. Two main simulated evolutions of variables are in-
the bubble size distribution. Indeed, the transition and the vestigated. They consist in the evolution of the global
 E. Olmos et al. / Chemical Engineering Science 56 (2001) 6359–6365 6363

Fig. 6. Sauter diameter global distribution.

Fig. 5. Comparison between the experimental and calculated evolu-
tions of global mean diameter with Ug .

number mean diameter d and the global Sauter mean

diameter d
S . These two diameters
can be expressed as:
dncells = cells d and dS ncells = cells dS where ncells
is the number of
grid cells,

d is the local number aver-
age diameter d i fi =vi =
i (fi =vi )di and ds is the local
Sauter diameter 1=dS = fi =di . They, respectively, in-
form us on the coalescence and breakage frequencies and
on the in#uence of size distribution on the global hydro-
dynamics.
The evolution of d with the super;cial gas veloc-
ity is compared to the experimental values (see Fig. 5)
and shows a good prediction of the experimental data. Fig. 7. Axial pro;les of volume fraction of the 5th class bubbles
Fig. 4 shows the evolution of dS  with the super;cial gas (r=R = 0).
velocity. We can notice that two distinct super;cial gas
velocity domains can be made out by the calculations : narrow distribution centred on the 5th class to a quite
a ;rst sensible increase of the Sauter diameter from 5 to larger distribution centred on the 6th class (see Fig. 6).
6:5 mm, and a second one where this diameter tends to a These distributions are composed essentially by the rest
constant value. These domains (A, B) can be related to of the bubbles of the 5th class that have not disappeared
the homogeneous and transition #ow regimes as follows. and by the 6th and 7th formed by coalescence. Moreover,
we can notice that smaller bubbles (di ¡5 mm) do not
5.1. Domain A appear and break-up phenomena observed earlier may be
negligible when Sauter diameters are considered in this
In domain A (Ug ¡ 4:8 cm s−1 ), the Sauter diameter range of super;cial gas velocities.
increases quickly from 5 to 6:5 mm. The Sauter diameter In addition to this, it can be noticed that no equilibrium
distribution (see Fig. 6) shows that the entering bubbles Sauter diameter is reached before the top of the column in
(5th class) disappear by coalescence to form essentially this domain of super;cial gas velocities (see Fig. 8) since
larger bubbles. This observation is con;rmed by the 5th dS increases continuously. Thus, the sparger governs es-
class volume fraction pro;le along the column axis (see sentially the #ow in the column. This can be explained by
Fig. 7). The observed evolution of the Sauter mean di- coalescence and break-up kinetics. Indeed, for this low
ameter with Ug can be linked to the evolution of d super;cial gas velocities, the input energy coming from
(Fig. 5). It shows that ;rstly, i.e. for very low super;cial the isothermal gas expansion is low. Consequently, coa-
velocities (Ug ¡ 1:5 cm s−1 ), the coalescence phenom- lescence and break-up kinetics is slow and does not allow
ena is slightly predominant. And then, as Ug increases to to reach an equilibrium state before the top of the column.
4:7 cm s−1 , break-up becomes more intensive since d These three observations (narrow distributions, in-
is decreasing. The consequence of the apparition of this crease of the global Sauter diameter, eJect of the sparger)
break-up phenomenon is the in#ection of dS  increase. let us think that the domain A can be linked to the ho-
Furthermore, in domain A, the Sauter diameter distri- mogeneous #ow regime for which these characteristics
butions evolve quickly from an initial monomodal and are known similar.
6364 E. Olmos et al. / Chemical Engineering Science 56 (2001) 6359–6365

nozzle spargers respectively (Camarasa et al., 1999). Our

numerical transition super;cial velocity (4:7 cm s−1 ) is
therefore in good agreement with these experimental
values.
All these observations show that the model simulates
the bubbles properties of the transition regime (predomi-
nance of break-up, wide size distributions) satisfactorily.
By still increasing Ug , which has not been considered
in the present study, the heterogeneous regime is about
to appear. At the same time, the bubble size distribution
should be stable and the gas distributor eJects negligible,
which indeed already occurs for our highest super;cial
gas velocities.
Fig. 8. Evolution of the axial pro;le of dS with Ug .
6. Summary and conclusion
5.2. Domain B
Euler–Euler simulations of gas–liquid #ows in a bub-
In domain B (Ug ¿ 4:7 cm s−1 ), dS  reaches a ble column have been coupled with a study of population
quasi-constant value of approximately 6:5 mm. This balance. Bubbles have been distributed into 10 diameter
equilibrium diameter can be interpreted by the fact that classes, and for each of them coalescence and break-up
break-up becomes non-negligible and compensates coa- phenomena have been taken into account. Calculated bub-
lescence eJects. This trend is well con;rmed by Fig. 5 ble size distributions, liquid velocity and gas hold-up
showing that d reaches a stable value around 3:5 mm, pro;les have been compared with experimental results
which demonstrates the equality of bubble coalescence obtained by photography, laser Doppler velocimetry and
and break-up frequencies. The evolution to this equilib- optical probes. This comparison has shown that, as long as
rium value observed numerically is characteristic of the the homogeneous regime exists, the hydrodynamic vari-
transition and the heterogeneous regimes and has already ables are still in good agreement with the experimental
been observed experimentally by Camarasa et al. (1999) data but an underestimation of the global gas hold-up ap-
and by Bhavaraju, Russel, and Blanch (1978). pears when the transition regime begins. The evolution
Fig. 6 shows that less than 10% of the bubbles of the of dS  reveals two distinct domains in which some bub-
entering class are still present in the column for these su- ble properties (size and distribution) have been linked
per;cial gas velocities. They are essentially transformed to those encountered in the homogeneous and transition
into larger bubbles (dS = 7–7:5 mm) and smaller bubbles regimes. In the ;rst domain, the in#uence of the sparger
(dS = 3:5–4:5 mm). The sensible point is that the largest on the #ow is predominant. In the second one, break-up
bubbles (dS ≈ 7:2 mm) are almost no more aJected by and wider size distributions prevail. These size distribu-
coalescence while Sauter diameter distribution widens tions evolve till an equilibrium which characterises the
toward the small bubbles (dS ≈ 3–4 mm). A stable bubble properties in the heterogeneous regime.
diameter distribution seems to appear for the highest su- This study has shown the in#uence of hydrodynamics
per;cial gas velocities studied. No bimodal distributions on bubble size distribution and has allowed a charac-
are highlighted but our calculations show a clear wide terisation of the #ow regimes as well as of the dis-
distribution from the largest bubbles which essentially tributor regimes as far as bubble sizes are concerned.
determine the hydrodynamics of the reactor to small Nevertheless, as soon as the #ow transition regime
ones formed by break-up but which in#uence slightly begins, all the other #ow variables are less precisely
this hydrodynamics. estimated. Coalescence and break-up frequencies are
This equilibrium state can be observed along the col- essentially determined by the turbulent energy dissipa-
umn axis too (see Fig. 8). Indeed, a stable Sauter diame- tion rate and the bubble diameter. Moreover the total
ter exists in the column, this equilibrium state is obtained dissipation rate is only related to the energy input due
lower in the column when the super;cial gas velocities to the isotherm gas expansion which only depends on
increases. Thus, the distributor eJects on the bubbles the super;cial gas velocity. Thus, it remains coher-
properties can be assumed negligible beyond a su@cient ent to obtain a realistic size distribution without the
height of liquid. In this case, this equilibrium state is pos- expected global hydrodynamics. The in#uence of the
sible since the high input energy allows a quick kinetics bubble size on the reactor hydrodynamics should be
of coalescence and break-up. obtained with a broader and more precise bubble di-
The experimental transition velocities are observed at 3 ameter distribution containing some very large bubbles
and 4 cm s−1 for the porous plate and the multiple-ori;ce with signi;cantly larger velocities and speci;c drag
 E. Olmos et al. / Chemical Engineering Science 56 (2001) 6359–6365 6365

laws but would require a much higher computational De Swart, J. W. A., van Vliet, R. E., & Krishna, R. (1996). Size,
load. structure and dynamics of large bubbles in a 2-D slurry bubble
column. Chemical Engineering Science, 51, 4619–4629.
Hibiki, T., & Ishii, M. (2000). Two-group interfacial area transport
equations at bubbly-to-slug #ow transition. Nuclear Engineering
Notation and Design, 202, 39–76.
Krishna, R., Urseanu, M. I., van Baten, J. M., & Ellenberger, J.
CD drag coe@cient (1999). In#uence of scale on the hydrodynamics of bubble columns
d global number mean diameter (averaged operating in the churn-turbulent regime: Experiments vs. Eulerian
on the whole column), m simulations. Chemical Engineering Science, 54, 4903–4911.
dS  global Sauter mean diameter (averaged Kurul, N., & Podowski, M. Z. (1990). Multi-dimensional eJects in
sub-cooled boiling. Proceedings of the 9th heat transfer conference.
on the whole column), m Lapin, A., & LIubbert, A. (1994). Numerical simulation of the
d local mean diameter, m dynamics of two-phase gas–liquid #ows in bubble columns
di diameter of bubbles of class i, m reactors. Chemical Engineering Science, 49, 3661–3674.
dS Sauter mean diameter, m Lee, C. H., Erickson, L. E., & Glasgow, L. A. (1987). Bubble break-up
fi volume fraction of bubbles of class i, fi = i =g and coalescence in turbulent gas–liquid dispersions. Chemical
Engineering Communications, 59, 65–84.
h axial position, m Lo, S. (1996). Application of population balance to CFD modelling
k turbulent kinetic energy, m2 s−2 of bubbly #ow via the MUSIG model. AEA Technology.
ni number density of bubbles of class i AEAT-1096.
r radial position, m Luo, H., & Svendsen, H. (1996). Theoretical model for drop and
R column radius, m bubble break-up in turbulent dispersions. A.I.Ch.E. Journal, 42,
1225–1233.
uk velocity of phase k, m s−1 Millies, M., & Mewes, D. (1999). Interfacial area density in bubbly
Ug super;cial gas velocity, m s−1 #ow. Chemical Engineering and Processing, 38, 307–319.
vi volume of bubbles of class i, m3 Pan, Y., Dudukovic, M. P., & Chang, M. (2000). Numerical
investigation of gas-driven #ow in 2D bubble columns. A.I.Ch.E.
Greek letters Journal, 46(3), 434–449.
Prince, M. J., & Blanch, H. W. (1990). Bubble coalescence and
break-up in air-sparged bubble columns. A.I.Ch.E. Journal, 36,
k local fraction of phase k 1485–1499.
g  global average gas hold-up Sato, Y., & Sekoguchi, K. (1975). Liquid velocity distribution in
 turbulent energy dissipation rate, m2 s−3 two-phase bubbly #ow. International Journal of Multiphase Flow,
k viscosity of phase k, Pa s 2, 79–95.
Shah, Y. T., Kelkar, B. G., Godbole, S. P., & Deckwer, W.-D.
k density of phase k, kg m−3 (1982). Design parameters estimations for bubble column reactors.
 surface tension, N m−1 A.I.Ch.E. Journal, 28, 353–379.
Sokolichin, A., & Eigenberger, G. (1994). Gas–liquid #ow in bubble
columns and loop reactors: Part I, Detailed modeling and numerical
simulation. Chemical Engineering Science, 49, 5735–5746.
References Van den Akker, H. (1998). The Euler–Euler approach applied to
dispersed two-phase #ow in the turbulent regime. ERCOFTAC
Bhavaraju, S. M., Russel, T. W. F., & Blanch, H. W. (1978). The Bull, 36, 30.
design of gas sparger devices for viscous liquid systems. A.I.Ch.E. Vial, C., LainNe, R., Poncin, S., Midoux, N., & Wild, G. (2001).
Journal, 24, 454–466. In#uence of gas distribution and regime transitions on liquid
Camarasa, E., Vial, C., Poncin, S., Wild, G., Midoux, N., & Bouillard, velocity and turbulence in a 3-D bubble column. Chemical
J. (1999). In#uence of coalescence behaviour of the liquid and Engineering Science, 56, 1085–1093.
of gas sparging on hydrodynamics and bubble characteristics in Wu, Q., Kim, S., Ishii, M., & Beus, S. G. (1998). One-group
a bubble column. Chemical Engineering and Processing, 38, interfacial area transport in vertical bubbly #ow. International
329–344. Journal of Heat and Mass Transfer, 41(8), 1103–1112.
Delnoij, E., Lammers, F. A., Kuipers, J. A. M., & van Swaaij, W. P. Zahradnik, J., Fialova, M., Ruzicka, M., Drahos, J., Kastanek,
M. (1997). Dynamic simulation of dispersed gas–liquid two-phase F., & Thomas, N. H. (1997). Duality of the gas–liquid #ow regimes
#ow using a discrete bubble model. Chemical Engineering in bubble column reactors. Chemical Engineering Science, 52,
Science, 52, 1429–1458. 3811–3826.