Chemical Engineering Journal 78 (2000) 21–28

Phenomenological model for bubble column reactors: prediction of

gas hold-ups and volumetric mass transfer coefficients
K. Shimizu, S. Takada, K. Minekawa, Y. Kawase∗
Biochemical Engineering Research Center, Department of Applied Chemistry, Toyo University, Kawagoe, Saitama 350-8585, Japan
Received 25 April 1999; received in revised form 8 October 1999; accepted 22 October 1999

Abstract
Based on a phenomenological model for bubble break-up and coalescence, a new simulation model for gas hold-up and gas-liquid mass
transfer in bubble column reactors is proposed. In order to describe bubble movements in a bubble column reactor, a compartment concept
is combined with the phenomenological model for bubble break-up and coalescence. It is assumed that the bubble column reactor consist
of a series of discrete compartments in which bubble break-up and coalescence occur and bubbles move from compartment to compartment
with different velocities. Gas hold-up and gas-liquid mass transfer rate are evaluated on the basis of bubble behaviors, i.e., bubble break-up
and coalescence. Reasonable agreement is found between the model predictions and the present experimental data obtained in two different
size bubble column reactors with air–water system and available correlations in the literature. Simulation results indicate that the proposed
model provides some insight into the transport phenomena in bubble column reactors and furthermore it is useful for improving CFD
predicting gas hold-ups and gas-liquid mass transfer rates in bubble column reactors. © 2000 Elsevier Science S.A. All rights reserved.
Keywords: Bubble column reactor; Bubble break-up; Bubble coalescence; Gas hold-up; Volumetric mass transfer coefficient

1. Introduction Recently numerical simulation has been recognized as an

ultimate tool for scale-up and design of chemical reactors
Bubble column reactors have been widely adopted in the and several studies have been conducted to understand hy-
chemical and biochemical industries. However, the design drodynamics in bubble column reactors using computational
and scale-up of bubble column reactors are still very difficult fluid dynamic (CFD) models [3–5]. In the numerical simu-
and their approaches are largely empirical [1,2]. Although lations proposed for hydrodynamics in bubble column reac-
a number of empirical correlations have been used for bub- tors, bubble phenomena have not been sufficiently taken into
ble column reactor design, they can shed little light on the account [4]. In order to provide comprehensive pictures of
mechanism of physical processes occurring in bubble col- bubble phenomena in bubble column reactors, therefore, an
umn reactors. In bubble column reactors, the size of bubbles establishment of simulation model based on bubble break-up
is the most important parameter deciding their performance. and coalescence phenomena is required.
It determines the bubble rising velocity and the gas residence Population balance models have been successfully used
time and governs the gas hold-up, the interfacial area and as a to describe liquid–liquid dispersion properties [6–9]. An
result gas–liquid mass transfer rate. In gas–liquid two-phase advantage of the population balance approach is that the
systems, bubble break-up and coalescence can profoundly details of the break-up and coalescence processes can be in-
influence the overall performance, by altering the interfacial cluded. Bapat et al. [10] and Ribeiro et al. [11] employed a
area available for mass transfer between the phases. There- population balance approach to predict the dispersed phase
fore, it is essential to understand the bubble behavior for ra- drop size distribution and mass transfer in liquid–liquid dis-
tional design of bubble column reactors and there have been persion systems. Unfortunately, few studies have focused
extensive studies related to this aspect [2]. It is expected to on gas–liquid systems and the understanding of gas dis-
obtain some insight into the phenomena in bubble column persion in liquids is poor. Mihail and Straja [12] discussed
reactors through simulation models based on the mechanism bubble size distributions in bubble columns by applying a
of bubble phenomena, i.e., bubble break-up and coalescence. population balance concept. They did not examine the influ-
ence of bubble size distributions on the overall performance
∗ Corresponding author. Tel.: +81-492-39-1377; fax: +81-492-31-1031. of bubble columns, such as gas hold-up and mass trans-
E-mail address: [email protected] (Y. Kawase) fer. Prince and Blanch [13] proposed a phenomenological

1385-8947/00/$ – see front matter © 2000 Elsevier Science S.A. All rights reserved.
PII: S 1 3 8 5 - 8 9 4 7 ( 9 9 ) 0 0 1 6 5 - 5
22 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28

model for the rates of bubble break-up and coalescence in The total break-up rate of all bubbles can be estimated by
turbulent gas–liquid dispersions. They compared the model taking account of all possible pairings of bubbles and turbu-
and the experimental data for the rates of bubble break-up lent eddies. The range of turbulent eddy sizes, which must
and coalescence and for the average bubble size and bubble be considered, is from the Kolmogoroff scale to the bubble
size distribution. They found good agreement between the size [9]. The turbulent collision rate, θ ie , for a bubble of
model predictions and the experimental data. This implies diameter, dbi , with an eddy of diameter, de , is
that the proposed model for the rates of bubble coalescence
and bubble break-up in gas–liquid dispersions is reasonable. θie = ni ne Sie (u2ti + u2te )1/2 (2)
However, their model was not applied to predict important The velocities of a bubble having diameters of dbi and an
design parameters for bubble column reactors such as gas eddy having diameters of de , uti and ute , may be written as
hold-ups and gas–liquid mass transfer rates. In this study, we [14–16]
modify and extend their approach to discuss the overall per-
1/3
formance of a bubble column reactor. Utilizing the bubble uti = 1.4ε1/3 dbi (3)
population balance model coupled with appropriate models
of break-up and coalescence functions and the compartment and
model for bubble movements in a bubble column reactor, 1/3
ute = 1.4ε1/3 de (4)
gas hold-ups and gas–liquid volumetric mass transfer rates
are analyzed. In the proposed model, the bubble column is respectively. The eddy diameter, de , may be evaluated using
assumed to consist of a series of discrete compartments in Kolmogoroff’s theory of isotropic turbulence [17,18].
which bubble break-up and coalescence occur. Bubbles are !1/4
considered to move from compartment to compartment with νl3
different velocities. The capability of the model is examined de = (5)
ε
using the present experimental results obtained in two dif-
ferent size bubble columns and available correlations in the where ν l is the kinematic viscosity of the liquid phase. An
literature. energy dissipation rate, ε, in a bubble column reactor is
written as [1]:

ε = Ug g (6)
2. Model development
Here g is the gravitational acceleration. The collision
2.1. Bubble break-up and coalescence cross-section area, Sie , in Eq. (2) may be written as:
π
Bubble size depends on a balance of coalescence and Sie = (dbi + de )2 (7)
4
break-up rates in the bubble column reactor. In the simula-
tion, the bubble behavior is considered by simply observ- The number of eddies per unit mass of the fluid, ne , in Eq. (2)
ing the bubbles one by one. We develop a model for bub- is evaluated by the following relationship derived by Azbel
ble behavior on the basis of the phenomenological model [19]
for the rates of bubble break-up and coalescence used by dne (k) k2
Prince and Blanch [13]. Their approach is based on the phe- = 0.1 (8)
dk ρl
nomenological model for liquid–liquid dispersions proposed
by Coulaloglou and Tavlarides [7]. The eddy wave number, k, is defined as an inverse of the
radius of the eddy, de /2. Integration of Eq. (8) yields the
2.1.1. Bubble break-up number of eddies.
Bubble break-up is modeled by a two-step mechanism; the The fraction of eddies having sufficient energy to cause
deformation of a bubble due to interactions with the turbulent rupture, F(u), in Eq. (1) is given by [7]:
flow field and the break-up of the deformed bubbles. ( !)
u2ci
The break-up rate of bubbles in a turbulent flow is given F (u) = exp − (9)
by the product of the eddy-bubble collision frequency and u2te
the break-up efficiency. In this work, for lack of adequate in which uci is the critical eddy velocity for break-up of a
theoretical and experimental knowledge, it is assumed that bubble of diameter dbi and may be written as:
a bubble breaks into two equal size daughter bubbles [9,14].
 1/2
The resulting equation for the total break-up rate of all σ
bubbles, β T , may be given as [9,13]: uci = (10)
dbi ρl
XX
βT = F (u)θie (1) where σ is the surface tension and ρ l is the density of the
i e liquid phase. Prince and Blanch [13] used a relationship
 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28 23

for uci based on the expression for the maximum stable The collision efficiency for bubble coalescence, λij , in
bubble size derived by Bhavaraju et al. [20]. Instead of their Eq. (11) is
somewhat indistinct relationship, Eq. (10) is used [21]. Using  
tij
the above equations, the total break-up rate for bubbles, β T , λij = exp − (17)
can be estimated. τij
where τ ij is the contact time for the two bubbles and given
2.1.2. Bubble coalescence by [23]
In a turbulent flow field, the bubbles first collide and then
(dij /2)2/3
remain in contact for sufficient time so that the processes τij = (18)
of film drainage, film rupture and coalescence occurs. The ε1/3
coalescence rate is written as the product of a collision rate The time required for coalescence of two bubbles having
and a coalescence efficiency or probability. diameter dbi and dbj , tij , is estimated to be the time required
It is assumed that bubble collisions occur due to turbulent to film drainage between the bubbles.
fluctuation of liquid phase, buoyancy forces of bubbles and ( )1/2  
shear stresses of liquid phase [13]. The overall coalescence (dij /2)3 ρl h0
tij = ln (19)
rate, χ T , is given by 16σ hf
1 XX n o
The equivalent radius for bubble coalescence, dij , is [24]
χT = λij (θijT + θijB + θijS ) (11)
2  
i j
2 2 −1
dij = + (20)
The overall coalescence rate is obtained by considering all dbi dbj
possible pairings of the bubbles. The turbulent collision rate,
Here, h0 is the initial film thickness with hf which is the crit-
θijT , for two bubbles of diameter dbi and dbj is given as a
ical film thickness at which rupture occurs. This relationship
function of bubble size, concentration and velocity:
for bubble–bubble coalescence was derived by Kirkpatrick
θijT = ni nj Sij (u2ti + u2tj )1/2 (12) and Lockett [25]. As well as the work of Prince and Blanch
[13], the initial and final film thickness in air-water systems
Here, ni and nj are the concentrations of bubbles of diameter are assumed to be 1×10−4 and 1×10−8 m, respectively. The
dbi and dbj , respectively. equations given above are used to estimate bubble coales-
The buoyant collision rate, θijB , is cence rates, β T .

θijB = ni nj Sij (uri − urj ) (13) 2.2. Gas hold-up

The rising velocity of the bubble in the turbulent flow regime
The size distribution is obtained as a result of bubble
uri in Eq. (13) can be evaluated by [5]
break-up and coalescence occurring simultaneously in the
 1/2 hypothetical compartment. Adding simply the volume of all
σ
uri = 2.14 + 0.505gdbi (14) bubbles, Vbi , and dividing by the dispersion volume, VR ,
ρl dbi
consisting of the total volume of gas, Vg , and that of liquid,
The collision rate due to shear, θijS , is given as [13]: Vl , the gas hold-ups, φ, can be calculated.
P
 3   Vg Vbi
4 dbi dbj dUl φ= = i (21)
θijS = ni nj + (15) Vg + Vl VR
3 2 2 dr
The shear rate for the shear stress exerted on the bubble, 2.3. Volumetric mass transfer coefficient
(dUl /dr), is postulated to be characterized as:
The specific surface area, a, is related to gas hold-up by
dUl Ul 0.787(gDT Ug )1/3 the following equation [1,26].
≈ = (16)
dr DT /2 DT /2 6φ
a= (22)
The liquid recirculation induced due to radial distribution dVS
of gas hold-up can be observed in a bubble column reactor. where dVS is the Sauter mean diameter which is a measure
Kawase and Moo-Young [22] analyzed the liquid phase mix- of the interfacial area per unit volume of the gas phase.
ing using an energy balance and the mixing length theory. The Higbie penetration theory [1,26] gives the following
As the characteristic liquid velocity, Ul , in a bubble column relationship for liquid-phase mass transfer coefficient kL :
having diameter of DT at superficial gas velocity of Ug in  
Eq. (16), the liquid velocity at the column axis derived by 4DL ur 1/2
kL = (23)
Kawase and Moo-Young [22] is applied. π db
24 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28

where DL is diffusivity. Combining Eqs. (22) and (23) yields of 1t. This time segment is also somewhat arbitrary but it
the expression for kL a in a bubble column reactor must be larger than the time required for bubble break-up
X  4DL uri 1/2  6φi 
and coalescence. If it is too large, the accuracy of the calcu-
kL a = (24) lation is poor. In this study, we used a fixed time increment
πdbi dbi for an occurrence of an event. The typical time increment
i
used in this study was 0.05 s. The number of compartment
During the repeated break-up and coalescence of bubbles, and time increment were selected by varying them to en-
hydrodynamics in the vicinity of a gas–liquid surface may sure that the results were independent of them. An event, if
affect mass transfer rate. However, its effects in turbulent it occurs, is assumed to occur at the beginning of the time
flows are still not quantitatively described or even fully un- increment and only one break-up or coalescence is allowed
derstood. In this study, therefore, no influence of break-up in one observation time 1t [10].
and coalescence of bubbles and of interactions among ad- Bubbles move from lower compartment to upper com-
jacent bubbles on mass transfer is assumed as well as the partment according to bubble rising velocity. When a gas
works for liquid–liquid dispersions of Bapat et al. [10] and is sparged into a bubble column reactor, the gas fraction is
Ribeiro et al. [11]. higher at the center of the reactor than at the wall. As de-
scribed above, the resulting radial density gradient induces a
2.4. Simulation procedure buoyancy-driven liquid circulation, with upflow in the core
and downflow in the outer annulus [27]. Bubbles of different
A bubble column is assumed to be separated into hypo- sizes travel at different velocities represented by Eq. (14). It
thetical compartments being completely mixed stage. The is assumed, therefore, that the actual rising motion of bub-
schematic flow diagram is illustrated in Fig. 1. The volume bles, u0r , can be represented by the following relationship:
of the bubble column reactor having height of HL is di-
vided into a series of M discrete compartments, each hav- u0r = ur + cUl (25)
ing a same height 1H (=HL /M). In a compartment bubble
where ur is a bubble velocity in a still liquid estimated us-
break-up and coalescence occur simultaneously and the bub-
ing Eq. (14). The second term in the RHS of Eq. (25) is
ble size distribution changes. The compartment height 1H
introduced to consider the influence of liquid recirculation
or the number of compartments M is somewhat arbitrary but
on bubble rising velocity. The proportionality constant c in
it must be larger than the maximum bubble size and smaller
Eq. (25) was temporarily assumed to be 2/3. This constant
than the distance of bubble rising. The typical value of M
is introduced to replace liquid circulations with characteris-
in this study was 500. The time is divided into segments
tic liquid upflows. More discussion for this simplification is
required.
The number of compartments 1N which the bubble moves
during 1t is given by
 
1t
1N = u0r (26)
1H
1N is integer by raising fractions to unit. Consequently, a
new compartment of the bubble Nt+1t can be obtained as:

Nt+1t = Nt + 1N (27)

Gas is injected to the bottom compartment. It is assumed

that equal-sized bubbles are continuously injected to the bot-
tom of the bubble column reactor or the first compartment
(N=1). The incoming gas phase is assumed to consist of
uniform size of bubbles and the number of bubbles formed
by the sparger at the reactor bottom is determined from the
gas flow rate and the initial bubble size.
At the start of a simulation, bubbles having a uniform
size exist only in the bottom compartment. From the bot-
tom compartment to top compartment, the new bubble size
distribution is calculated by taking account of the rates for
bubble break-up and coalescence calculated by Eqs. (1) and
(11) and then their new locations are evaluated from their
rising velocities using Eq. (27). The same simulation pro-
Fig. 1. Schematic flow diagram of the compartment model. cedure for a number of simulation times is repeated. The
 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28 25

scale of eddies effective to bubble break-up is ranged from

the smallest possible eddy, demin , to the largest eddy, demax .
The smallest eddy is estimated using Eq. (5) multiplied by
2 [13] and the largest eddy is equal to bubble size [9,28].
Therefore, the eddy size intervals used to calculate the bub-
ble break-up rate is (demax −demin )/100 m. The bubble size
interval between the minimum bubble size of 0.001 m and
the maximum bubble size of 0.015 m is 0.0005 or 0.001 m.
In other words, the number of bubble size intervals used was
15 or 30 in the present simulations.
It is postulated for simplicity that bubbles cannot go back
into the lower compartments. In other words, only forward
movements of bubbles are considered. The effects of the
presence of liquid backflow accompanying small bubbles
on gas hold-up and volumetric mass transfer coefficient is
taken into account by introducing the characteristic liquid
flow rate. In bubble column reactors, an upward liquid flow
is observed in the center portion and a downward liquid
flow near the reactor wall. Some small bubbles are dragged
by the downward liquid flow [2,22]. The effects of liquid
backflow and radial distribution of axial liquid velocity on
bubble movement are approximately considered using the
characteristic liquid velocity evaluated by introducing the
proportionality constant c into Eq. (25) instead of the direct
use of the actual or local liquid velocity.
The simulation runs are continued till the steady state is
attained. The steady-state solution can be obtained when the
change in gas hold-up becomes negligible. Under this con-
dition, bubble break-up and coalescence are in balance and
the gas hold-up is therefore time-independent. Several simu-
lation passes are required before a steady-state solution can
be obtained. After steady state had been attained, we de-
termined gas hold-ups and volumetric mass transfer coeffi-
cients through the bubble column.

Fig. 2. (a) Simulated bubble size histogram in the 0.20 m i.d. bubble
3. Experimental
column reactor at Ug =0.01 ms−1 . (b) Simulated bubble size histogram in
the 0.20 m i.d. bubble column reactor at Ug =0.04 ms−1 .
Experiments were carried out in a column 0.20 m in di-
ameter and 1.7 m in height and a bubble column 0.155 m in
diameter and 0.834 m in height. A ring sparger with 12 holes
of 1.0 mm diameter was used in the 0.20 m i.d. column. Per- 4. Results and discussion
forated plates 57 holes of 1.0 mm diameter and 89 holes of
1 mm diameter were used in the 0.155 m i.d. column. Tap 4.1. Bubble size distribution
water and air were used as the liquid phase and gas phase,
respectively. The density, viscosity and surface tension of Simulation results for bubble size distributions in the
water were measured with a pycnometer, a Cannon–Fenske 0.20 m i.d. bubble column reactor are shown in Fig. 2a and b.
viscometer and the du-Nouy ring method, respectively. The The bubble size distributions are results of bubble break-up
operation was batchwise with respect to liquid phase. The and coalescence occurring simultaneously. In illustrating the
rate of air-flow sparged continuously was measured with a simulation results, the bubble column is redivided to 10
precalibrated rotameter. The gas hold-ups were obtained by stages. In the figures the Sauter mean diameters calculated
the volume expansion method. The volumetric mass transfer from the proposed model are given. It is seen from the data
coefficients were determined by the dynamic method. The for Ug =0.01 m s−1 in Fig. 2a that at the middle of the column
change in the dissolved oxygen concentration was moni- (5th stage) the somewhat steep increase and decrease in num-
tored using a fast response dissolved oxygen electrode (YSI ber density can be obtained for smaller and larger bubbles,
model 57, Yellow Springs Instrument Co.). respectively, as compared with the top of the column (10th
26 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28

Fig. 3. Gas hold-ups in the 0.2 m i.d. bubble column reactor. Fig. 4. Gas hold-ups in the 0.155 m i.d. bubble column reactor.

stage). This trend can be seen more clearly from the data for Fig. 5 shows axial distributions of gas hold-up in the
Ug =0.04 m s−1 in Fig. 2b. The feature of these data sets in 0.20 m i.d. bubble column predicted by the proposed model.
shown Fig. 2a and b is that with increasing gas flow rate the It can be seen that gas hold-up distributions are almost uni-
peak in the number density shifts to the right and the bubble form at low and high gas flow rates.
size distribution become lower and wider. At higher gas flow
rates bubble break-up and coalescence more frequently oc- 4.3. Gas–liquid mass transfer
cur as compared with those at lower gas flow rates and as a
result the bubble size distribution becomes wider. The Sauter Fig. 6 depicts that the simulation model predictions are
mean diameter at the top of the column somewhat increases fairly close to the experimental data for kL a in the 0.2 m i.d.
with gas flow rate as well as the results of Jamialahmadi bubble column reactor.
and Muller-Steinhagen [29]. This may be attributed to the An empirical correlation for gas–liquid volumetric mass
enhancement of bubble coalescence along the column axis. transfer coefficient was developed by Akita and Yoshida
[32].
4.2. Gas hold-up !0.62 !0.31
 
kL aD2T νl 0.5 gD2T ρl gD3T
=0.6 φ 1.1 (30)
It is seen from Fig. 3 that the gas hold-ups in the 0.20 m i.d. DL DL σ νl2
bubble column are in agreement with the values predicted
from the proposed simulation. Shah et al. [1] obtained a simple correlation for kL a in a
Hughmark [30] proposed an empirical correlation for gas bubble column with air–water system which may be written
hold-up. It may be written as: as
1
φ= (28)
2 + (0.35/Ug )(ρl σ/72)1/3
The following theoretical correlation was derived by Kawase
et al. [31].
!1/4
φ Ug3
= 0.0625 (29)
1−φ νl g

For reference, the predictions of Eqs. (28) and (29) are also
shown in Fig. 3. They agree well with the present experi-
mental data and the proposed simulation model.
In Fig. 4, the simulation model is compared with the ex-
perimental data in the 0.155 m i.d. bubble columns besides
the predictions of Eqs. (28) and (29). The proposed simu-
lation model can fit the data reasonably. It is also seen that
the agreement between the model predictions and the pre- Fig. 5. Axial distribution of gas hold-up in a bubble column reactor
dictions of Eqs. (28) and (29) is satisfactory. (DT =0.20 m).
 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28 27

Fig. 7. Gas–liquid mass transfer rates in the 0.155 m i.d. bubble column
reactor.
Fig. 6. Gas–liquid mass transfer rates in the 0.2 m i.d. bubble column
reactor.

kL a = 0.467Ug0.82 (31)

Kawase et al. [33] proposed a semi-theoretical correlation

for kL a in a bubble column. Their correlation for Newtonian
fluids is written as:
    !3/5
kL aD2T νl 1/2 DT Ug 3/4 gD2T ρl
= 0.452
DL DL ν σ
!7/60
Ug2
× (32)
DT g

For reference, the above correlations are given in Fig. 6. Eq.

(30) predicts rather smaller kL a values as compared with Fig. 8. Axial distribution of gas–liquid mass transfer rate in a bubble
the experimental data. The predictions of the proposed sim- column reactor (DT =0.20 m).
ulation model and Eqs. (31) and (32) fit the data reasonably
well.
Fig. 7 compares the predictions of the proposed simulation is not uniform. Larger kL a values are predicted near the
model with the experimental data in the 0.155 m i.d. bubble sparger. This coincides with the data reported in the literature
column. It can be seen that the proposed model agrees rea- [1]. At the low gas flow rate of Ug =0.01 m s−1 , the kL a
sonably with the experimental data. The predictions of Eqs. coefficients slightly decrease from the bottom to the top of
(31) and (32) lie somewhat below the experimental data. It the bubble column reactor. In the high gas flow rate region,
is also seen from Fig. 7 that Eq. (30) underestimates the kL a a decrease in kL a is significant. An increase in the number
coefficients in the 0.155 m i.d. bubble column. of larger bubbles near the top of the bubble column shown
In bubble column reactors, there are three flow regimes in Fig. 2 is responsible for the significant decrease in kL a. It
depending on reactor diameter and gas flow rates, bubbly is seen from Figs. 5 and 8 that the gas–liquid mass transfer
flow, churn turbulent flow and slug flow. The hydrodynam- rate is more sensitive to the bubble behaviors than the gas
ics and heat and mass transfer depend strongly on the flow hold-up.
regime. As well as the study of Prince and Blanch [13],
the proposed simulation model may be applicable to bubbly
flow and churn turbulent flow regimes which are observed 5. Conclusions
in the 0.155 and 0.20 m i.d. bubble columns. Additional ex-
aminations are in progress to clarify the flow conditions in The simulation model based on the bubble behaviors in
which the proposed model is applicable. bubble columns has been developed. The bubble break-up
In Fig. 8, axial distributions of kL a in the 0.20 m i.d. and coalescence which are the primary phenomena in bub-
bubble column reactor are plotted. The distribution of kL a ble column reactors are taken into account to evaluate gas
28 K. Shimizu et al. / Chemical Engineering Journal 78 (2000) 21–28

hold-ups and gas–liquid volumetric mass transfer rates. [7] C.A. Coulaloglou, L.L. Tavlarides, Chem. Eng. Sci. 32 (1977) 1289–
The present simulation is built on the phenomenological 1297.
model for bubble break-up and coalescence proposed by [8] M.A. Hsia, L.L. Tavlarides, Chem. Eng. J. 20 (1980) 225–236.
[9] C. Tsouris, L.L. Tavlarides, AIChE J. 40 (1994) 395–406.
Coulaloglou and Tavlarides [7] and Prince and Blanch [13]. [10] P.M. Bapat, L.L. Tavlarides, G.W. Smith, Chem. Eng. Sci. 38 (1983)
The compartment model is applied to describe the bubble 2003–2013.
movements. Reasonable agreement between the simulation [11] L.M. Ribeiro, P.F.R. Regueiras, M.M.L. Guimaraes, C.M.N.
model and experimental results in the two different size Madureira, J.J.C. Cruz-Pinto, Compters & Chem. Eng. 21 (1997)
bubble column reactors indicates that the model based on 543–558.
[12] R. Mihail, S. Straja, Chem. Eng. J. 33 (1986) 71–77.
the physical picture for bubble dynamics could be used to [13] M.J. Prince, H.W. Blanch, AIChE J. 36 (1990) 1485–1499.
predict bubble size distributions, gas hold-up and gas–liquid [14] G. Narsimhan, J.P. Gupta, D. Ramkrishna, Chem. Eng. Sci. 34 (1979)
mass transfer rate. Due to the stochastic interaction of the 257–265.
bubble swarms some assumptions have been introduced. For [15] J.C. Rotta, Turbulente Stromungen. B.G. Teubner, Stuttgart, 1972.
[16] R. Kuboi, I. Komasawa, T. Otake, J. Chem. Eng. Japan 5 (1972)
instance, we assumed a binary bubble breakage. However,
349–355.
it is somewhat questionable that the two daughter bubbles [17] A. Kolmogoroff, C.R. (Doklady) Acad. Sci. USSR 30 (1941) 301–
have the same diameter. In this study, furthermore, the influ- 305.
ence of liquid recirculation was treated only approximately. [18] R.S. Cherry, E.T. Papoutsakis, Bioprocess Eng. 1 (1986) 29–41.
Although the proposed model cannot provide a complete [19] D. Azbel, Two-Phase Flows in Chemical Engineering, Cambridge
University Press, Cambridge, 1981.
description of bubble behaviors, it gives satisfactory insights
[20] S.M. Bhavaraju, T.W.F. Russell, H.W. Blanch, AIChE J. 24 (1978)
of the phenomena in bubble column reactors. In order to 454–466.
improve the model, the effects of liquid recirculation in a [21] M.A. Delichatsios, R.F. Probstein, Ind. Eng. Chem. Fundam. 15
bubble column reactor to bubble behaviors should be more (1976) 134–137.
precisely included. This study should be regarded as the [22] Y. Kawase, M. Moo-Young, Chem. Eng. Sci. 41 (1986) 1969–1977.
[23] V.G. Levich, Physicochemical Hydrodynamics, Prentice-Hall,
first step to elucidate physical processes occurring in bubble
Englewood cliffs, NJ, 1962.
column reactors through theoretical analyses based on the [24] A.K. Chesters, G. Hofman, Appl. Sci. Res. 38 (1982) 353–361.
mechanism of bubble break-up and coalescence. [25] R.D. Kirkpatrick, M.J. Lockett, Chem. Eng. Sci. 29 (1974) 2363–
2373.
[26] K. Van’t Riet, Trends in Biotechnology 1 (1983) 113–119.
References [27] S.B. Kumar, N. Devanathan, D. Moslemian, M.P. Dudukovic, Chem.
Eng. Sci. 49 (1994) 5637–5652.
[1] Y.T. Shah, B.G. Kelkar, S.P. Godbole, W.D. Deckwer, AIChE J. 28 [28] H. Luo, H.F. Svensen, AIChE J. 42 (1996) 1225–1233.
(1982) 353–379. [29] M. Jamialahmadi, H. Muller-Steinhagen, Trans. I Chem. E. 68 (1990)
[2] W.D. Deckwer, Bubble Column Reactors, Wiley, New York, 1992. 202–204.
[3] A. Lubbert, T. Paaschen, A. Lapin, Biotechnol. Bioeng. 52 (1996) [30] G.A. Hughmark, Ind. Eng. Chem. Process Des. Dev. 6 (1967) 218–
248–258. 220.
[4] A. Sokolichin, G. Eigenberger, A. Lipin, A. Lubbert, Chem. Eng. [31] Y. Kawase, S. Umeno, T. Kumagai, Chem. Eng. J. 50 (1992) 1–7.
Sci. 52 (1997) 611–626. [32] K. Akita, F. Yoshida, Ind. Eng. Chem. Process Des. Dev. 12 (1973)
[5] T.-J. Lin, J. Reese, T. Hong, L.-S. Fan, AIChE J. 42 (1996) 301–319. 76–80.
[6] K.J. Valentas, N.R. Amundson, Ind. Eng. Chem. Fundam. 5 (1966) [33] Y. Kawase, B. Halard, M. Moo-Young, Chem. Eng. Sci. 42 (1987)
533–542. 1609–1617.