Computers & Fluids 39 (2010) 11911207

Contents lists available at ScienceDirect

Computers & Fluids

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c o m p fl u i d

Single bubble rising dynamics for moderate Reynolds number using Lattice
Boltzmann Method
Luz Amaya-Bower, Taehun Lee *
Department of Mechanical Engineering, City College of City University of New York, New York, NY 10031, USA

a r t i c l e

i n f o

Article history:
Received 26 May 2009
Received in revised form 9 November 2009
Accepted 4 March 2010
Available online 18 March 2010
Keywords:
Bubble rising
Lattice Boltzmann Method
Large viscosity ratio

a b s t r a c t
Dynamics of a single rising gas bubble is studied using a Lattice Boltzmann Method (LBM) based on the
CahnHilliard diffuse interface approach. The bubble rises due to gravitational force. However, deformation
and velocity of the bubble depend on the balance of other forces produced by surface tension, inertia, and
viscosity. Depending on the primary forces acting on the system, bubble dynamics can be classied into different regimes. These regimes are achieved computationally by systematically changing the values of Morton number (Mo) and Bond number (Bo) within the following ranges 1  105 < Mo < 3  104 and
1 < Bo < 1  103 . Terminal shape and Reynolds number (Re) are interactive quantities that depend on
size of bubble, surface tension, viscosity, and density of surrounding uid. Accurate simulation of terminal
shape and Re for each regime could be satisfactorily predicted and simulated, since they are also functions of
Mo and Bo. Results are compared with previous experimental results.
2010 Elsevier Ltd. All rights reserved.

1. Introduction
Bubble rising under the inuence of gravitational force is one of
the most common gasliquid ow phenomena. Understanding the
dynamic interaction between the phases is an important key for
the design and operation of industrial applications such as gasliquid column reactors [13]. Dynamic behavior of a rising bubble
has been a subject of both experimental and numerical studies
for many years. Experimental studies and correlations were reviewed by Clift et al. [4]. Later, a more comprehensive study was
presented by Bhaga and Weber [5], introducing experimental data
and correlations for bubble rise velocity and shape regimes. Rising
of a bubble in a viscous uid under the inuence of gravitational
forces can be generally grouped in three different regimes: Spherical, ellipsoidal, and spherical cap as described in [4]. (Description
of each regime will follow.) These regimes are achieved depending
on the values of two important non-dimensional quantities, as dened below:
Bond number:
2

Bo

gMqd0

Morton number:

Mo

gMqg4l

r3 q2l

* Corresponding author. Tel.: +1 212 650 6122.

E-mail address: thlee@ccny.cuny.edu (T. Lee).
URL: http://www-me.engr.ccny.cuny.edu/faculty/lee.html (T. Lee).
0045-7930/$ - see front matter 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compuid.2010.03.003

where d0 is the bubble diameter, ql is liquid density, gl is liquid viscosity, Mq ql  qg is density difference between continuous medium and the dispersed uid, g is gravitational acceleration, and r is
the surface tension. Bo is the ratio of the body forces (effective gravitational forces) and the surface tension, but it could also be considered as a dimensionless size value of the bubble [6]. Mo provides a
description of the properties of the surrounding uid, mainly focusing in viscosity and surface tension. Using these non-dimensional
parameters, the dynamics of a single bubble could be predicted.
Clift et al. [4] introduced a bubble diagram to predict the terminal
velocity and shape of a rising bubble at low density ratios, as could
be seen in Fig. 1. This diagram uses Bo, Mo and Reynolds number,
which is dened as follows:
Reynolds number:

Re

ql U t d0
;
gl

where U t is the terminal velocity of the bubble. Terminal velocity is

dened as the steady velocity that the bubble reaches when there is
a balance between buoyancy and drag forces. A general classication of the bubble is usually done depending on nal shape or by
the primary forces acting on the system. However, the transition
areas from regime to regime are still not well dened. A brief
description of each of the three main regimes: Spherical, ellipsoidal,
and spherical cap is provided as follows [4]:
 Spherical regime: This regime is dominated by surface tension
and viscous forces. Original size of the bubble is small, usually
less than 1.3 mm. The shape of the bubble remains spherical

1192

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

Fig. 1. Shape regime map [4].

or nearly-spherical. Usually the ratio of minor to major axis is

less than 10%. In this regime, the terminal velocity is proportional to size of the bubble. For pure systems, the evolution of
terminal rising velocity can be described by Hadamard [7] and
Rybczynski [8] as follows:
2

Ut

gd0 Dq 1 k1
;
6gl 2 3k1

where k1 gg =gl is viscosity ratio.

 Ellipsoidal regime: This regime is mainly dominated by surface
tension. Bubble size is intermediate, typically from 1.3 to
6 mm, and the range for Bo is 0:25 < Bo < 40. Final shape is usually oblate with convex interface if looking from inside. Generally, there is no fore-and-aft symmetry. In this regime, there is
little viscous resistance to internal circulation, therefore drag
and terminal velocity are very sensitive to contamination. Terminal velocity for pure systems is closely approximated by a
correlation suggested by Mendelson [9] based on wave theory

s
2:14r
Ut
0:505gd0 :
ql d0

 Spherical cap regime: This regime is governed by inertia force.

Bubble size is large, usually bigger than 6 mm. Bo > 40 and
Re > 1:2. In this regime, terminal velocity of bubble is proportional to size. In addition, large inertia creates higher deformation by creating high pressure on the front and rear and while
low at the sides. Rising velocity can be described by a theory
proposed by Davies and Taylor [10]

Ut

2
3

s
gd0 Dq
:
2 ql

Computational studies for multiphase ows have been performed using various numerical methods see [11]. Specically for
bubble rising, numerical methods used include but are not limited
to following computational studies. Chen et al. [12] evaluated
dynamics of transformation from spherical to toroidal bubble using
Volume-Of-Fluid (VOF) method. Tomiyama et al. [13] studied dy-

namic behavior in the regime of surface tension dominance. However, more recent studies have expanded evaluation to different
shape regimes. Van Sint Annaland et al. [14] presented a 3-dimensional (3-D) VOF algorithm using a new interface reconstructing
technique based on piecewise linear interface representation. Later, Van Sint Annaland et al. [15] introduced a 3-D front tracking
model using new surface force to avoid the explicit computation
of interface curvature. Bonometti and Magnaudet [16] explored
the dynamics of a single rising bubble using a 3-D algorithm which
utilizes interface-capturing method for two-phase ows with high
density and viscosity ratios. Smolianski et al. [17] proposed a
numerical method to simulate single bubble and bubble swarms
in ellipsoidal and spherical ow regimes, which accurately captures the change and the deformation of the interface. Hua et al.
[18] used a front tracking algorithm to simulate dynamics of single
bubble in a 3-D model.
The study of dynamics of a rising bubble using LBM have been
explored by Sankaranarayanan et al. [19] using a multi-component
model as described by Shan and Chen [20], Shan and Doolen [21],
and Shan and He [22] as a benchmark study to gauge the capabilities of LBM to simulate bubbly ows, with density ratio of 100,
Bo < 5 and Mo > 1  106 . Later, same authors introduced a new
implicit formulation in order to increase the range of dimensionless simulation parameters to Bo < 10 and Mo > 1  108 [23].
Takada et al. [24] provided a comparison study for single and multiple bubble rising between LBM and VOF. LBM simulation was
performed using free energy based on the model by Swift et al.
[25], and VOF simulation was done based on a model presented
by Hirt and Nichols [26]. This study showed that LBM exhibited
higher deformation than VOF, which was attributed to the uniformly applied buoyancy force in the LBM simulation. The free energy model described by Swift et al. [25] was also employed by
Frank et al. [27], for a single bubble rising at Re < 1:8. They reported inadequate agreement with experimental results, largely
because the simulations were performed in two-dimensions. A
projection method was introduced by Inamuro et al. [28], where
simulations with density ratios up to 1000 were achieved. Kurtoglu
and Lin [6] presented a LBM based on phase-eld approach [29]
and total variation diminishing (TVD) discretization [30], and the
simulation results were presented in the range of Bo < 100 and
the density ratio of 2.45. Considering that gasliquid systems in
industrial practice typically have values of Mo and Bo in the range
1012 < Mo < 106 , and 102 < Bo < 104 , and the density ratio of
up to 1000, studies of a single bubble rising using LBM thus far,
have been limited to a few shape regimes and low ranges of Bo,
Mo, and density ratio used in the simulations.
The aim of this paper is to present a comprehensive study on
dynamics of a bubble using a LBM based on the CahnHilliard diffuse interface approach [31]. This study will present eight different
shape regimes, which are achieved by changing the values of Bo
and Mo within the following ranges 1  105 < Mo < 3  104
and 1 < Bo < 1  103 . In addition, simulations are performed
using large density and viscosity ratios, 1000 and 100 respectively.
This study will evaluate the topological changes, velocity and pressure proles, and drag coefcient of the bubble while rising at different regimes, depending on the forces acting on the system. In
addition, terminal shape and Re dependence on numerical parameters such as interface thickness, surface tension, viscosity and
density of surrounding are explored. Simulation results are compared with experimental results for each regime.

2. Lattice Boltzmann Equations

Discrete Boltzmann Equation (DBE) with external force in the
potential form described by Lee and Fischer [32] can be written as

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207


@fa
1
1
ea  rfa  fa  faeq 2 ea  u  F Gfaeq ;
k
@t
qc s

where fa is the particle distribution function in the a direction, ea is

the microscopic particle velocity, q is the macroscopic density, u is
the macroscopic velocity, cs is a constant, k is the relaxation time, F
represents the intermolecular force, and G represents gravitational
force. The equilibrium distribution function faeq is given by

"

eq

fa

#
ea  u ea  u2 u  u
;
ta q 1 2

cs
2c2s
2c4s

to yield Ptot p0 P  jC r2 C 12 jjrCj2 , where and j is the gradient parameter.

The behavior of a binary system is governed by the advective
CahnHilliard equation for the composition C, see [36].

@C
u  rC r  M rl;
@t

F rqc2s  rp1  C rl;

where p1 is the dynamic pressure used to enforce incompressibility,

and l is the chemical potential. The gravitational force takes the
form as G qg rh, with h being the elevation from a reference
point so that @ y h 1, but @ x h @ z h 0.
Eq. (7) recovers mass and momentum equations. However, a
transformation of DBE is necessary to eliminate acoustic waves at
low frequency limit and to improve stability. This transformation
is to recover pressure evolution and momentum equations, rst introduced by He et al. [33]. We use a particle distribution function by
including both dynamic pressure and static pressure of the liquid
phase, ql gh, in the denition of equilibrium distribution function.

g a fa c2s p1 ql gh  qc2s Ca 0;

10

@g a
1
ea  rg a  g a  g eq
a ea  u
k
@t


 rqc2s Ca  Ca 0  C rl qg rhCa
ea  ql g rhCa 0;
where the new equilibrium g eq
a is

"

2
g eq
a t a P qc s

ea  u ea  u2 u  u


c2s
2c2s
2c4s

11
!#
;

12

ha

ga x ea dt; t dt  ga x; t

1

g  geq j
s 0:5 a  a x;t

dtea  u  rqc2s Ca  Ca 0  C rl qg rhCa x;t

17

qg Ca 0
faeq

;
ql  qg ql  qg

18

and
eq

ha

such that the mixing rule introduced in Eq. (16) is recovered when
taken zeroth moment. Taking total derivative Dt of the new distribution function ha and taking into account Eq. (15) produces
@ha
1
eq
ea  rha  ha  ha ea  u  rC Ca r  M rlCa
k
@t
1
ea  u  rP @ C p2 rC C rl q  ql g rh

Ca :
ql  qg
c2s

In Eq. (19), an articial pressure gradient @ C p2 rC is introduced to

avoid negative value of the composition C by imposing @ C p2 > 0
for C < 0 and @ C p2 0 otherwise. As can be seen in the last term,
introduction of the static pressure into the denition of P is equivalent to modifying the gravitational acceleration G q  ql g rh.
 eq can be dea and h
LBE for modied particle distribution function h
a
rived similarly to Eq. (13).
Calculating the zeroth and rst moments of the modied particle distribution functions yields the composition, momentum, and
a combination of static and dynamic pressures.

13

14

A total pressure P tot can be calculated as the sum of the thermodynamic part, the dynamic and the static parts, and the curvature part

a dt r  Mrl;
h
2
X
1
dt
qu 2
ea ga  C rl;
cs a
2
X
dt
P
ga u  rqc2s :
2
a
C

In Eq. (13), the modied particle distribution function ga and equilibrium distribution function geq
a are introduced (see Appendix A).
The macroscopic equation recovered from Eq. (11) using the
ChapmanEnskog expansion is

@P
qc2s r  u 0:
@t

qg Ca 0
fa

;
ql  qg ql  qg

19

in which P p1 ql gh. Using dimensionless analysis, it could be

shown that in the case of low Mach number (Ma) the ratio of the dynamic and static pressures to the thermodynamic pressure p0 is
P=p0  OMa2 , indicating that all thermodynamic quantities can
be assumed to be independent of the dynamic pressure at low Ma
[35]. In the derivation of Eq. (11), the terms with u  rP are omitted
since they are OMa3 . LBE is obtained by discretizing Eq. (11) along
characteristics over the time step dt:

dtea  ql g rhC0a jx;t :

16

The mixing energy density of an isothermal system is related to C by

Emix C; rC E0 C jjrCj2 =2, where E0 bC 2 1  C2 is the bulk
energy and it is related to classical part of the chemical potential
by l0 @ C E0 , and to the thermodynamic pressure by @ C p0
C@ C l0 . Equilibrium prole is obtained when the mixing energy is
minimized, i.e., l l0  jr2 C. The interface prole in a plane at
equilibrium is then given by Cz 1 tanh2z=D=2, where D is
the interface thickness. Surface tension r is a measure of free energypper
unit area at constant temperature, and dened as
r 2jb=6.
A new particle distribution function is necessary to recover the
CahnHilliard equation of the composition C. The new particle distribution function could be obtained as

eq

in which Ca u fa =q. Taking the total derivative Dt @ t ea  r

of the new variable g a gives

15

where M > 0 is the mobility. C and q have a linear relationship given by

q ql C qg 1  C:

t a being a weighting factor. F represents the intermolecular attraction by using the mean-eld approximation, the exclusion-volume
of molecules, as described by He et al. [33] and Rowlinson and Widom [34], and the dynamic pressure gradient, yielding

1193

20

21
22

In order to obtain composition C using Eq. (22) an implicit calculation is required since chemical potential l is a function of C. Our
model takes this value from the previous time step explicitly [31].
For the discretization of forcing terms, see [32].

1194

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

3. Validation test

3.1. Grid resolution dependence

A grid resolution analysis was performed using a grid size as a
function of the diameter of the bubble 4d0  7d0 . The size of the grids
used in this analysis were 160  280, 320  560, and 640  1120. As
shown in Fig. 2, the effect of the grid resolution in the nal shape is
not very signicant. Comparing grid sizes 320  560, and 640 
1120, the nal shape for each case is almost overlapping each other.
However, the nal shape in the coarse grid, 160  280, has a slight
difference at the sides of the bubble, but overall the shape is properly
maintained. In addition, Fig. 3 presents the effect of grid resolution in
terminal Re indicating a percent difference no greater than 5% between the results. Considering the results provided by this analysis,
it is determined that using the coarse grid in this study, corresponding to d0 40 would provide reasonable accuracy and avoid high
computational cost.
3.2. Domain size dependence
Simulations are performed in a conned domain therefore
boundary conditions are no-slip in all directions and bounce-back
conditions are applied for both g a and ha . Analysis of domain size
dependence is performed to determine the inuence of the radial
and vertical walls on bubble rising. In the radial direction, as domain increases, the viscous effect created by wall becomes less signicant in the behavior of the bubble. Viscous effect caused by wall
will constrain the radial deformation and slow down the bubble.
Therefore, it is critical to determine the minimum radial size of
the computational domain that practically eliminates the wall effect in the dynamics of bubble rising.
In order to determine the minimum radial size for 3-D calculations, tests are performed by systematically increasing the radial

12

Terminal Re

In order to gauge the capabilities of the LBM algorithm, a series

of 2-D analysis are performed to evaluate the effect of important
numerical parameters in the simulation. These parameters included grid resolution, domain size, density ratio, viscosity ratio,
interface thickness, and surface tension. Unless otherwise
mentioned, the evaluation is performed for xed values of
Bo 10; Mo 2  102 , and d0 40 in lattice unit.

3201120
160560
80280

Fig. 3. Effect of grid resolution in terminal Re.

size by multiples of d0 . A comparison on terminal shape and velocity between different tests is presented. As shown in Fig. 4, the
deformation of the bubble is less affected by wall when radial
direction is 4d0 . In addition wall effect in terminal velocity is reduced signicantly when the domain is 6d0 , as could be seen from
Fig. 5. However, use of this dimension will become computational
costly in 3-D calculations. When domain size is less than 4d0 , bubble velocity is decreased due to higher wall viscous effect. In addition, this will reduce the vertical translation of the bubble and
minimize the deformation radially. Therefore, to reduce this viscous effect a radial domain size of 4d0 is selected, since it has a difference of less than 10% and provides a nal shape very close to
that in larger domain. This domain size provides a compromise between computational cost and simulation accuracy. Note that plots
presented
pin this study use non-dimensional time T t=t n , where
tn d0 =g . The inuence of the size in the vertical direction is
not very signicant as long as the bubble does not reach the top
wall. The vertical size is different in each case, since the bubble displacement depends on the value of Re. For cases of Re  1, vertical
size is around 4d0 . However, for Re  100 size could be as high as
9d0 . Similar validation was performed by Hua and Lou [37], resulting in a radial domain of 4d0 and vertical of 12d0 .

---- 3201120
_._. 160560
_.._
80280

14

-..- 2d0
-.-. 4d0
---- 6d0
_______
8d0

12

10
10

-2

x
Fig. 2. Effect of grid resolution in terminal shape at T 10.

10

8
-2

x
Fig. 4. Effect of radial domain size in terminal shape at T 10.

1195

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

---- l/g = 1000

_._. l/g = 100
_.._ l/g = 10

12
4

Terminal Re

8d0
7d0
6d0
5d0
4d0
3d0
2d0

10
-2

0
0

10

Fig. 7. Effect of density ratio in terminal shape at T 10.

Fig. 5. Effect of radial domain size in terminal Re.

3.3. Density ratio dependence

Effect of density ratio is evaluated by comparing the terminal Re
and shape of three different cases. A density ratio is dened as
ql =qg and the test cases have density ratios of 10, 100, and 1000,
while maintaining all other parameters constant. As it is shown
in Fig. 6, density ratio has a small effect in the nal Re of the bubble. The difference between cases of the density ratios of 10 and
1000 is about 10%. The rising Re is inuenced by the numerical value of gravity. Using non-dimensional numbers, it could be seen
that gravity is inversely proportional Dq. Assuming ql 1; Dq
for each case is 0.9, 0.99 and 0.999 respectively. Therefore, the difference between the numerical values of gravity in each case is also
around 10%. Fig. 7 shows the nal shape for all the cases at T 10.
As could be seen from the gure, change in a density ratio has a
small effect on nal shape. Deformation mainly depends on the value of surface tension, which remains unchanged for all the test
cases. However, the density ratio has an effect on nal position.
A higher density ratio provides a lower value of terminal Re,
therefore the bubble will reach smaller distances in the same

amount of time. Study presented by Hua and Lou [37] investigated

the effect of density ratio on terminal velocity and shape. Their
conclusion indicates that effect of density ratio is more signicant
in terminal velocity than in terminal shape. The results of their
study agree with the results obtained in current validation.
3.4. Viscosity ratio dependence
In order to evaluate the effect of viscosity ratio, dened as gl =gg ,
in the dynamic behavior of bubble rising, the six cases evaluated
are gl =gg 1; 10; 100; 200; 500 and 1000. As viscosity ratio decreases, interface becomes more immobile and shear stress in the
system becomes more signicant. These factors make the bubble
behave more like a rigid object and enhance the effect of the wall
on the rising bubble. In addition, reduction of viscosity ratio has increased the internal circulation of the bubble, which delays the
wake formation behind the bubble [4]. Therefore, when the viscosity ratio is at the lowest value, 1, the terminal Re is considerably
reduced and the bubble exhibits little deformation. The effect of
modifying viscosity ratio can be seen in Fig. 8 for terminal Re. It

8
8

Terminal Re

Terminal Re

l/g = 1000
l/g = 100
l/g = 10

0
0

T
Fig. 6. Effect of density ratio in terminal Re.

l/g
l/g
l/g
l/g
l/g
l/g

10

T
Fig. 8. Effect of viscosity ratio in terminal Re.

= 1000
= 500
= 200
= 100
=
10
=
1
10

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

is noted that as the value of viscosity ratio is increased, terminal Re

tends to collapse into a single curve. Effect of viscosity ratio for terminal shape is presented in Fig. 9. These results show that there is
very little change in the nal shape is viscosity ratio is P500. However, terminal shapes for viscosity ratios P100 maintain same
characteristics and prole. In general, using a high viscosity ratio
is needed to achieve a high and accurate value of Re and reach a
consistent nal shape. Study by Hua and Lou [37] also concluded
that change in viscosity ratio has a strong effect in the terminal
velocity as well as terminal shape, viscosity ratio of 100 is used
in their results.

Terminal Re

1196

Cn
Cn
Cn
Cn
Cn

3.5. Interface thickness dependence

2

EU y 1 

Rv
Rh

0.200
0.175
0.150
0.125
0.100

10

T
Fig. 10. Effect of interface thickness in terminal Re at T 10.

---_._.
_.._
____
....

Cn
Cn
Cn
Cn
Cn

=
=
=
=
=

0.200
0.175
0.150
0.125
0.100

12

Interface thickness D is a numerical value used in the diffuse

interface method to improve stability and accuracy in the simulation. A dimensionless number that relates D and a characteristic
length scale is the Cahn number (Cn), dened as Cn D=d0 . As
the value of D increase, a higher level of accuracy and stability in
terms of numerical discretization is achieved while the description
of phase interface deviates from the sharp interface limit. As can be
seen in Figs. 10 and 11, change of Cn value does not have a significant effect in the terminal Re or shape of the bubble. Note that
there is a lower stability limit in the value of D  3, which is related
to the number of grid points needed to resolve interface region.
LBM used in this study uses an isotropic discretization suggested
by Lee [31], in order to eliminate the parasitic currents in the
system.
In addition to evaluation of the effect of interface thickness in
terminal Re and shape, numerical tests are also performed to ensure satisfaction of Galilean invariance. Lack of Galilean invariance
is an undesirable feature of some two-phase ow LBM, and is due
to unwanted terms proportional to u  rq in the recover macroscopic governing equations [38]. Galilean invariance can be quantied by a dimensionless error as function of U y , a constant and
homogenous vertical velocity, dened as

=
=
=
=
=

23

where Rv and Rh are the vertical and horizontal radii of a moving

bubble measured after 50; 000 iterations respectively. Numerical
tests are performed in order to determine EU y by varying the values of vertical velocity U y from 5  103 to 8  102 , while maintaining identical numerical conditions: Square computational
domain with 100  100 grid points, bubble diameter with 50 grid
points, density and viscosity ratios at 10, surface tension at

16

14

..-..
__ __
.....
_____
.-.-.
----

l/g
l/g
l/g
l/g
l/g
l/g

= 1000
= 500
= 200
= 100
=
10
=
1

10
-2

x
Fig. 11. Effect of interface thickness in terminal shape at T 10.

1  104 and relaxation time at 0.5. As can be seen in the results

plotted in Fig. 12, as the value of D increases, the dimensionless error remains unchanged and is parallel to second-order line. Therefore, the model used in this study satises Galilean invariance to
second-order of accuracy in velocity, an expected and satisfactory
accuracy for LBM.
3.6. Surface tension dependence

12

10

6
-2

x
Fig. 9. Effect of viscosity ratio in terminal shape at T 10.

Surface tension is a measure of free energy per unit area at constant temperature. Evaluation of the effect of surface tension is
performed by systematically changing the value of Bo. As Bo increases, the effect of surface tension is reduced. Therefore, inertia
force has more inuence in the bubble and higher deformation will
be experienced by bubble. Four different cases are performed at
Bo 1, 10, 100, and 1000. Terminal Re is maintained constant by
changing the values of Mo accordingly. Results for nal shape at
T 10 are presented in Fig. 13. As can be seen in the gure at
Bo 1, nal shape remains almost spherical. When, Bo 10 the

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

case are as follows: test 1, Bo 32:2; Mo 8:2  104 and test 2,

Bo 243; Mo 266. As can be seen in Fig. 14, the difference in
the deformation in both test cases for the three alternatives is almost non-existent. In addition, the difference between the terminal Re is less than 1% in both test cases. This indicates that the
method of application of buoyancy force in the system has no
dependence in the results for deformation and terminal velocity.
This contradicts previous results obtained by Takada et al. [24],
where a comparison study between VOF and LBM was performed
for bubble rising. This study showed that LBM exhibited higher
deformation than VOF, which was attributed to the uniformly applied buoyancy force in the LBM simulation.

100

1-RV/RH

1197

10-2

D=6
D=5
D=4
2nd Order

4. Numerical results
10

-4

0.01

0.1

Uy
Fig. 12. Dimensionless error as function of U y with different interface thicknesses D
at 50,000 iterations. Domain size is 100  100, bubble radius d0 50; ql 1:0;
qg 0:1; sl sg 0:5.

Bo =

Bo = 10

Mo = 310-5

Mo = 210-2

Bo = 100

Bo = 1000

A new shape regime diagram based on Bo, Mo, and Re, was introduced by Bhaga and Weber [5]. This diagram has different locations
for the transition lines and introduces additional shape regimes to
the diagram introduced by Clift et al. [4]. This diagram is a rise velocityvolume relationship, represented as Re vs. Bo and function of
constant Mo. The new eight different shape regimes are: S, spherical;
OE, oblate ellipsoid; OED, oblate ellipsoidal (disk-like and wobbling); OEC, oblate ellipsoidal cap; SCC, spherical cap with closed,
steady wake; SCO, spherical cap with open, unsteady wake; SKS,
skirted with smooth, steady skirt; SKW, skirted with wavy, unsteady
skirt. The dynamics of single bubble rising has been previously explored numerically using LBM, however previous studies have limited analysis on different ow regimes. In this study, numerical
tests are performed by using LBM to simulate a case for each shape
regime. Location of each test case in Bhagas shape regime diagram is
presented in Fig. 15 and Table 1 provides the values of Bo and Mo
used in each case. Five cases were selected from experimental results obtained by Bhaga and Weber. These cases are indicated with
an asterisk  in the table.
4.1. Terminal velocity and shape

Mo = 1.510

Mo = 1.010

Fig. 13. Effect of surface tension in terminal shape at T 10.

bubble is deformed mainly at the bottom, forming a spherical cap.

The bubble becomes skirted when Bo 100 and Bo 1000, however, as can be seen when Bo 1000 the skirt is wider, taller and
thinner.

3.7. Application of buoyancy force dependence

In order to establish the effect of the method of application of
buoyancy force in the model, three different alternatives were explored. First, gravitational force G is dened so that the force is applied uniformly in all domain, G qg rh, as suggested by Takada
et al. [24]. Second, force is dened so that it only affects the gas,
equivalent to G q  ql g rh, as dened in Eq. (11). And, third,
force is dened as it was suggested by Bunner and Tryggvason [40],
G q  qav e g rh, where qav e is the average density, which is
introduced to guarantee that the net momentum ux across the
boundaries in the domain is zero. These three alternatives were
tested in two different 2-D cases, evaluating the possible change
in deformation and terminal velocity. The parameters used in each

Results of 3-D simulation are summarized in Tables 2 and 3. Table 2 provides a comparison between current 3-D simulation and
experimental results from Bhaga and Weber [5]. In this table, the
rst column provides the values of terminal Re obtained from the
simulation and experimental results from Bhagas shape regime
map. The second column provides an illustration of the simulated
nal shape in each case. The third column provides experimental
nal shape in order to quantitatively compare the simulation results. Table 3 shows current results versus previous computational
simulations. In this table, the rst column provides the values of
terminal Re obtained from the current simulation and simulation
results based on Bonometti and Magnaudet [39], Hua et al. [18],
and Hua and Lou [37]. The second column provides an illustration
of the current simulated nal shape in each case. The third column
provides the nal shape from previous simulation results. In general, simulation results for terminal Re tend to be slightly lower
than experimental and previous simulation results. This pattern
can be explained by three important factors. First, simulation domain was reduced in the radial direction to maintain computational cost reasonable. As shown in previous section, smaller
radial size, provides lower terminal Re values. Second, experimental results were obtained based on small density ratio 1.5. However, simulations were carried out with density ratio of 1000.
Higher value of density ratio, can also decrease the value of terminal Re, as it was shown in previous section. Third, experimental results were carried out with large viscosity ratio 104 . Simulation
results were originally performed using a viscosity ratio of 100
producing low terminal Re. As shown in previous section, a high

1198

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

Re = 34.441
-2

Re =

4
2

-2

G=-(-ave)gh

G=-(-l)gh
6

Re = 6.417

Re = 6.410

Re = 6.413
2

G=-gh
6

Re = 34.446
-2

-2

34.435
0

Test 2
y

G=-(-ave)gh

G=-(-l)gh
y

Test 1
y

G=-gh

-2

-2

x
Fig. 14. Effect of the method of application of buoyancy force at T 4:5.

Table 1
Values of Bo and Mo used in current simulations.
Cases

Fig. 15. Shape regime map for bubbles in liquids: S, spherical; OE, oblate ellipsoid;
OED, oblate ellipsoidal (disk-like and wobbling); OEC, oblate ellipsoidal cap; SCC,
spherical cap with closed, steady wake; SCO, spherical cap with open, unsteady wake;
SKS, skirted with smooth, steady skirt; SKW, skirted with wavy, unsteady skirt [5].

Bo

Mo

A spherical

1:0  10

B oblate ellipsoid

1:0  101

1:0  102

C* oblate ellipsoidal disk-like

3:2  101

8:2  104

D* oblate ellipsoidal cap

2:4  102

2:6  102

E* spherical cap closed wake

1:2  102

4:6  103

spherical cap open wake

1:0  102

1:0  103

G* skirted smooth steady skirt

3:4  102

4:3  101

H* skirted wavy unsteady skirt

6:4  102

4:3  101

1:0  105

viscosity ratio is necessary to obtain high terminal Re values.

Unfortunately, increasing the value of viscosity ratio, creates instability in the simulation for cases with high Re and deformation.
Therefore, computations with higher viscosity ratios were performed in most of the cases, while maintaining numerical stability,
from 200 to 1000. It is noted that using a higher the viscosity ratio,
provides a higher terminal Re, this trend can be shown in Table 4.
Cases A and B were performed using a viscosity ratio of 1000,
increasing the terminal Re by 15%. Cases D and G were run with
viscosity ratio of 500 raising terminal Re by at least 10%. Case H
was done using viscosity ratio of 300, improving terminal Re by
20%. Case C was performed with viscosity ratio of 200, increasing
terminal Re by 7.5%. Unfortunately, numerical stability was not
reached for cases E and F. However, it is estimated that if this stability issue is resolved, the terminal Re would be raised in order to
obtain more accurate results. Final simulation shape is compared
using second and third columns in Tables 2 and 3. It is the purpose
of the comparison to gauge the capability of the model in order to
capture the topological changes based on initial values of Bo and
Mo. Variation of Bo and Mo provides a change on deformation
and velocity, respectively. Deformation is directly proportional to
Bo and velocity is inversely proportional to Mo. Cases A and B have
the lowest values of Bo therefore they experience the least amount

1199

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

Table 2
Comparison of 3-D simulation results and experimental results based on study by Bhaga and Weber [5].
Cases

Current stimulation results

Experimental results

C
Reexp = 55.3
Resim = 51.7

D
Reexp = 7.8
Resim = 6.2

E
Reexp = 94.0
Resim = 78.9

G
Reexp = 18.3
Resim = 15.2

H
Reexp = 30.3
Resim = 26.8

of deformation. These cases are governed mainly by viscous and

surface tension forces. On the other hand, cases EH have the highest value of Bo exhibiting greater deformation. These cases are
dominated by inertia. However, cases E and F have lower values
of Mo, therefore terminal Re tends to be higher. In addition, simulated results for these cases show that the bubble is transformed
from a spherical cap to a toroidal bubble. Although these results
are different from an experimental solution, this transformation
has been obtained by other numerical studies, including
[16,37,39]. Bonometti and Magnaudet [16] established that the
transition from spherical cap to a toroidal shape is very sensitive
to grid resolution. Therefore, in order to capture a more accurate
transition process, it is necessary to increase the grid resolution
signicantly. Overall, simulated nal shapes exhibit the desired
topological changes as described experimentally [5] and numerically [18,37,39].
In order to provide a better description of the topological
changes encounter in each case, Figs. 1619 are presented. These
gures provide the evolution of the shape and the position for each
case. As can be seen in Fig. 16, case A remains spherical throughout
the trajectory. On the other hand, case B starts to have a deformation in the lower surface around T 2 until it reaches an oblate
ellipsoidal shape. Fig. 17 shows the evolution for cases C and D.
In case C, bottom surface of the bubble starts to deform pushing
side surfaces outward, creating a disk-like shape. In case D, the bottom surface is deformed until an oblate ellipsoidal cap is formed.

Note that the bubble in case C reaches a higher position than in

case D at the same T, due to a higher value of Re. Fig. 18 presents
the evolution of the spherical cap cases E and F. As mentioned previously, these two cases transform from a spherical cap to a toroidal shape. This transformation takes place sooner in case F, due to
its higher Re and its unsteady wake. Fig. 19 shows the evolution of
the skirted cases G and H. The bottom surface of the bubble in case
H has a higher deformation throughout the trajectory, due to a
higher value of Bo, making inertia the dominating force in the
system.
4.1.1. Velocity and pressure elds
Figs. 2023 present the streamlines for each case, at their nal
T. A closed toroidal wake region is formed at the back of the bubble
as it rises. A recirculation pattern is produced within the wake. This
wake is related to the degree of deformation, terminal Re, and the
size of the bubble. Experimental observation of this wake region
was performed by Bhaga and Weber [5] using H2 tracers. It was
noted that at higher volume, deformation, and Re, a larger wake region was formed. Cases A and B undergo very little deformation.
Therefore, the wake produced behind the bubble is very small.
Streamlines surround the periphery of the bubble, as could be seen
in Fig. 20. On the other hand, cases G and H have high Re and deformation, therefore the wake region is larger. Another factor to quantify the wake size is the value of Bo, since it could be considered as
the dimensionless size of the bubble. As Bo increases, so does the

1200

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

Table 3
Comparison of 3-D simulation results and previous simulation results based on
Bonometti and Magnaudet [39], Hua et al. [18], and Hua and Lou [37].
Cases

Current stimulated results

Case A

Case B

10

10

Previous stimulation results

T=7
A
Resim = 8.9

T=7

T=5

Case E [39]
Resim  10

T=5
B
Resim = 9.5

T=3

T=2

Case E [39]
Resim  10

T=0

2
C
Resim = 51.7

Case A2 [18]
Resim  54.8

-2

T=0

-2

x
D
Resim = 6.2

Fig. 16. Shape and position evolution for cases A and B.

Case A3 [18]
Resim  7.6

E
Case A8 [18]
Resim  88.7

Resim = 78.9

Case C

Case D

T=5
T=4
6

T=4
F
Case K [39]
Resim  100

Resim = 106

T=3

T=2

T=2
T=1

G
Resim = 15.2

-2

H
Resim = 26.8

-2

Fig. 17. Shape and position evolution for cases C and D.

Case A5 [37]
Resim  29.9

toroidal wake. Therefore, when comparing wake size between case

E and F, and G and H, cases E and H will have higher wake regions
since their Bo values are also higher, see Figs. 22 and 23.
As can be seen from Fig. 21, case C experiences an open wake,
unlike other cases. It is believed that this case has crossed the transition point from closed to open wake. Different experimental

T=0

T=0

Case A5 [18]
Resim  17.8

studies have been performed to determine the transition from a

closed, steady to an open wake region. Bhaga and Weber [5] indicate that for high Mo cases, the transition is believed to occur at
Re 110. However, for low Mo, this transition is a function
of other
p
p
factors such as the Weber number, dened as We Re2 Mo= Bo.
In their study, the critical value to describe high or low Mo is given
by Mo 4  103 . Case C has a Mo 8:2  104 , which is considered as low Mo case, therefore it is possible that at its terminal
Re, the bubble has crossed the transition point.

Table 4
Comparison of 3-D simulation terminal Re results with different viscosity ratios.
Cases

Resim at
Resim at

lratio

lratio 100

A
Reexp = 9.9

B
Reexp = 10. 1

C
Reexp = 55.3

D
Reexp = 7. 8

G
Reexp = 18.3

H
Reexp = 30.3

Resim = 7.6
Resim = 8.9
lratio 1000

Resim = 8. 2
Resim = 9. 5
lratio 1000

Resim = 47.8
Resim = 51.7
lratio 200

Resim = 5. 6
Resim = 6. 2
lratio 500

Resim = 13.5
Resim = 15.2
lratio 500

Resim = 21.7
Resim = 26.8
lratio 300

1201

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

10

Case C
T = 5

10

Case E

Case D
T = 4

Case F
8
10

6
8

T=2.5
T=2
4

T=2

T=1

T=1
2

T=0

T=0
2
-2

-2

-2

-2

Fig. 21. Streamlines for cases C and D.

Fig. 18. Shape and position evolution for cases E and F.

10

10
Case G

Case F
T = 2

Case E
T = 2.5

Case H

T=3
T=2.5

T=2

2
T=1

T=1

T=0

T=0

-2
-2

-2

-2

x
Fig. 22. Streamlines for cases E and F.

Fig. 19. Shape and position evolution for cases G and H.

Case A
T = 7

Case B
T = 7

Case H
T = 2.5

Case G
T = 3

12
10
6
6
10

8
4
4
8
6
2
2
6
4
-2

-2

x
Fig. 20. Streamlines for cases A and B.

-2

-2

x
Fig. 23. Streamlines for cases G and H.

1202

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

12

12

12

P tot

p 0 and p c

10

10

10
-1
0

-10

-5

-15

0
2

-10

3
6

-5

10

-20

-15

-20

-25

8
7
6
5

4
3

-2

-2

-2

Fig. 24. Pressure contours for case A. P tot is total pressure, P is sum of the static and the dynamic parts, p0 is the thermodynamic part, and pc is the curvature part.

P tot

12

P
12

12

p 0 and p c

10

1
0.5

-1

-1.5
0

1.2
0.8

-2

-0.5

10

-0.2
-0.6

-1.5

0.4
0.6
1.2

1.4

-2.5

-0.5

10

-2.5

-1

38
-2

-2

-2

x
Fig. 25. Pressure contours for case B. Description same as Fig. 24.

Figs. 2431 provide three different pressure contours: Total

pressure Ptot , sum of the static and the dynamic parts (P) and
sum of the thermodynamic part p0 and the curvature part pc .
These gures have been non-dimensionalized using

Pi P top
,
0:5ql U 2c

where

Pi is the pressure in each contour, Ptop is the pressure value at highest point in the domain, far away from the bubble and U c is the
p
characteristic velocity of the system dened as U c gd0 . These
results emphasize the idea that thermodynamic properties are
independent of the dynamic pressure, since p1 =p0  OMa2 . This
can also be seen in the contours of Ptot , which are very similar to
the ones of P, indicating that p0 and pc have very little effect in
the total pressure of the system. Another important observation
can be done by comparing the contour values between the cases.
Cases AC, have high range values of p0 and pc , indicating the primary forces acting on the system are surface forces. This corre-

sponds to low values of Bo, where surface tension is higher.

These three cases have a range of Bo < 35, see Figs. 2426. On
the other hand, cases G and H, have low range values p0 and pc ,
indicating that surface tension is not the primary force in the system. These cases have large Bo, which makes inertia the driving
force in the system. In general, the range level of P remains very
close in all the cases. In addition, it could be noticed that higher
values of P are concentrated at the top surface of the bubble. This
indicates that at that point, the bubble is trying to push away liquid
in order to keep rising.

4.1.2. Drag coefcient

As the bubble rises, there are two main forces acting on the system: buoyancy and drag. Buoyancy is the driving force and drag is
the force that opposes the bubble motion. The rising bubble

1203

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

P tot

10

4
6
12

-0.76

6
-2

-0.31

-0.52

-0.2
-0.47

-6

10

12

-6

8
10

p 0 and p c

10

10

-0.31

-2
-4

-4

-0.25

-6

-0.2

-0.15

2
-2

2
-2

-0.1

2
-2

Fig. 26. Pressure contours for case C. Description same as Fig. 24.

10

10

10

P tot

0.1

0.1

0.2
0.4

0.15
0.2
0.3
0.35
0.05

0.05

0
-0.2 -0.1

-0.15

p 0 and p c

-0.04 -0.1

-0.08

-0.1

-0.05

-0.06

-0.04

-0.02

-2

39
0

-2

-2

x
Fig. 27. Pressure contours for case D. Description same as Fig. 24.

reaches a point where these two forces are balanced, reaching a

steady velocity, known as terminal velocity. Studies have been performed to relate the terminal velocity with drag force. Using experimental studies, Bhaga and Weber [5] established a relationship
between the drag coefcient C d and Re, limited to large values of
Mo > 4  103 . The relationship between C d and Re is given by:

"

 0:9 #0:91
16
0:9
C d 2:67
:
Re

24

Using the principle that terminal velocity is reached when drag and
buoyancy forces are balanced, a theoretical relationship can be
established between U t and C d . Buoyancy force can be estimated

as weight of a spherical bubble F b pd0 Dqg=6 and drag force can

be calculated using F d qU 2t AC d =2, where A is the cross-sectional
area normal to the bubble motion. Using these denitions, a theoretical relationship between C d and U t is given by

Cd

4Dqgd0
3U 2t ql

25

Simulation results are compared to experimental correlation Eq.

(24), based on the constraint that Mo > 1  104 . Comparison was
performed by calculating C d using theoretical relationship given in
Eq. (25), using numerical values for U t . As can be seen in Fig. 32,
simulation results are in good agreement with experimental
correlation.

1204

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

10

10

10

P tot

p 0 and p c

-0.2

-0.2

-0.3
-0.2
-0.3
-0.4
-0.6
-0.8
-1
-0.9
-0.7

-0.4
-0.6

-0.4

-0.5
-0.7

-0.1

-0.25

-0.2
-0.15

-2

0
-0.25

-0.3

-2

-2

Fig. 28. Pressure contours for case E. Description same as Fig. 24.

10

10

10

P tot

p 0 and p c

-0.2

-0.2

-0.2

-0.3
-0.4 -0.2

-0.3
-0.6

-0.6
-1

0
-0.05
-0.25

-0.4

-0.1
-0.2

-0.4
-0.7

-0.7
-0.9

4
-0.15

-0.1

2
-0.05

-2

-2

-2

Fig. 29. Pressure contours for case F. Description same as Fig. 24.

4.2. Comparison with correlation

Velocity number:

Rodrigue [41] proposed a simple correlation for the dynamics of

bubble rising in a pure viscous uid, obtaining terminal velocity as
function of physical parameters. This correlation was obtained
using a non-dimensional analysis on experimental results from
19 studies, covering a large range of physical parameters, such as
ql ; gl ; r, Re, and Mo. Bubble motion is described using two nondimensional numbers obtained from this study:
Flow number:

Fg

" 8 #1=3
d0 q5l

rg4l

 2=3
Re
Bo
;
Ca

V Ut

rgl

Re2 Ca1=3 ;

27

where Ca glrU t the capillary number. Rodrigue [41] proposed a new

correlation between these two non-dimensional numbers in the
form of

V
26

" 2 #1=3
d0 q2l

aF b
1 cF d

where a 1=12; b 1; c 49=1000 and d 3=4.

28

1205

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

10

10

10

P tot

0.1

p 0 and p c

0.1

8
0.15

0.15
0.2

0.2

0.3

0.3

0.25

0.25

-0.02

-0.04

0.2

0.15

-0.08

-0.06

0.1
0.05

0.05

0.1

0.15
0.2

0.2

-2

-2

-2

Fig. 30. Pressure contours for case G. Description same as Fig. 24.

P tot

p 0 and p c

6
-0.4

0
-0.5

-0.8

-0.9

-1

-0.8

-0.9

-1

-0.04

-0.03

-0.5

-0.03

-0.02

0
-2

0
-2

0
-2

-0.01

Fig. 31. Pressure contours for case H. Description same as Fig. 24.

A comparison between current numerical results and this correlation is performed to test the ability of LBM to predict V for a large
range of physical parameters, and it is shown in Fig. 33. In general,
there is a reasonable agreement between the numerical results and
the correlation. This indicates that numerical simulation can predict an important non-dimensional number, which is related to
terminal velocity, a factor that describes the dynamic behavior of
a rising bubble.

5. Summary
A comprehensive study of the dynamics of a single bubble rising
is presented using a LBM based on the CahnHilliard diffuse interface approach. 2-D analysis is performed in order to establish the

dependence on physical and numerical parameters such as density

ratio, viscosity ratio, surface tension, interface thickness, and domain size. Analysis shows that density ratio has little effect in terminal shape and velocity. In the cases tested, the difference
between terminal velocities is proportional to the difference in
density ratio, which is less than 10%. In contrast, the effect of
modifying viscosity ratio has a greater effect in the terminal velocity and shape of the bubble. As viscosity ratio decreases interface
becomes more immobile and shear stress is higher in the system.
Therefore, the bubble suffers less deformation and terminal velocity is reduced. The present study shows that the increase of Bo
reduces the surface tension, therefore higher deformation is exhibited by the bubble. In addition, numerical parameters such as
interface thickness and domain size are also tested. It is shown that
varying the interface thickness has very little to no effect in

1206

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

25

Experimental

dynamics of a bubble using a LBM based on the CahnHilliard diffuse interface approach.

Simulation

Acknowledgment

15

Acknowledgment is made to the Donors of The Petroleum Research Fund, administered by the American Chemical Society, for
partial support of this research.

Cd

20

Appendix A. Modied particle distribution functions

10

a and
In Eq. (13), the modied particle distribution function g
eq
equilibrium distribution function g
a are introduced:
5

ga g a

20

40

60

80

1
dt
g  g eq
ea  u
a 
2s a
2


 dt
 rqc2s Ca  Ca 0  C rl qg rhCa  ea
2
 ql g rhCa 0;

100

29

Re
and
Fig. 32. Drag coefcient comparison between experimental results from Bhaga and
Weber [5] and simulation results.

eq
geq
a ga 

dt
ea  u
2

Velocity Number, V


 dt
 rqc2s Ca  Ca 0  C rl qg rhCa  ea
2
 ql g rhCa 0:

10

30

References

100

Correlation
Simulation
0

1000

2000

3000

4000

5000

Flow Number,F
Fig. 33. Comparison of velocity number for correlation and simulation results.

terminal velocity and shape, as long as the value is high enough to

use an isotropic discretization to reduce the parasitic currents in
the system. On the other hand, variation of domain size has a great
effect in the dynamic of the bubble. The effect of wall is reduced, as
domain size is increased, reaching a point where this effect is no
longer signicant. 3-D analysis was performed by simulating a case
for each of the eight different shape regimes introduced by Bhaga
and Weber [5]. Terminal Re and shape for each case are in good
agreement with experimental results. For each case, streamlines
are given, showing the wake behind the bubble. It was shown that
the shape of the wake has a dependence on the size, velocity, and
deformation of the bubble. Also, pressure contours are presented,
showing that pressure concentrates at the top of the bubble, indicating that the bubble is pushing liquid away in order to continue
rising. In addition, drag coefcient is calculated using simulation
results and it is compared satisfactorily with experimental results.
In conclusion, a comprehensive study was presented for the

[1] Deckwer W-D. Bubble column reactors. Chichester: Wiley; 1992.

[2] Blanch HW, Clark DS. Biochemical engineering. New York: Dekker; 1996.
[3] Moshtari B, Babakhani EG, Moghaddas JS. Experimental study of gas hold-up
and bubble behavior in gas liquid bubble column. Petrol Coal 2009;1:2732.
[4] Clift R, Grace JR, Weber M. Bubbles, drops, and particles. New York: Academic
Press; 1978.
[5] Bhaga D, Weber ME. Bubbles in viscous liquids: shapes, wakes and velocities. J
Fluid Mech 1981;105:6185.
[6] Kurtoglu IO, Lin C-L. Lattice Boltzmann study of bubble dynamics. Numer Heat
Transfer, Part B 2006;50:33351.
[7] Hadamard JS. Mouvement permanent lent dune sphre liquide et visqueuse
dans une liquide visqueuse. CR Acad Sci 1911;152:17358.
[8] Rybczynski W. On the translatory motion of a uid sphere in a viscous
medium. Bull Int Acad Pol Sci Lett, Sci Math Nat Ser A 1911:406.
[9] Mendelson HD. The prediction of bubble terminal velocities from wave theory.
AIChE J 1967;13:2502.
[10] Davies RM, Sir Taylor GI. The mechanics of large bubbles rising through
extended liquids in tube. Proc R Soc Ser A 1950;200:37590.
[11] Prosperetti A, Tryggvason G. Computational methods for multiphase
ow. Cambridge: Cambridge University Press; 2007.
[12] Chen L, Garimella S, Reizes J, Leonard E. The development of a bubble rising in
a viscous liquid. J Comput Phys 1999;387:6196.
[13] Tomiyama A, Celata GP, Hosokawa S, Yoshida S. Terminal velocity of single
bubble in surface tension force dominant regime. Int J Multiphase Flow
2002;28:1479519.
[14] Van Sint Annaland M, Deen NG, Kuipers JAM. Numerical simulation of gas
bubbles behavior using a three-dimensional volume of uid method. Chem
Eng Sci 2005;60:29993011.
[15] Van Sint Annaland M, Dijkhuizen W, Deen NG, Kuipers JAM. Numerical
simulation of behavior of gas bubbles using a 3D front-tracking method. AIChE
J 2006;52:99110.
[16] Bonometti T, Magnaudet J. Transition from spherical cap to toroidal bubbles.
Phys Fluids 2006;33:052102.
[17] Smolianski A, Haario H, Luuka P. Numerical study of dynamics of single
bubbles and bubble swarms. Appl Math Mod 2008;32:64159.
[18] Hua J, Stene JF, Lin P. Numerical simulation of 3D bubble rising in viscous
liquids using a front tracking method. J Comput Phys 2008;227:335882.
[19] Sankaranarayanan K, Shan X, Kevrekidis K, Sundaresan IG. Bubble ow
simulations with the lattice Boltzmann method. Chem Sci Eng
1999;54:481723.
[20] Shan X, Chen H. Lattice Boltzmann model for simulating ows with multiple
phases and components. Phys Rev E 1993;47:18154327.
[21] Shan X, Doolen G. Multicomponent lattice Boltzmann model with interparticle
interaction. J Stat Phys 1995;81:37993.
[22] Shan X, He X. Discretization of the velocity space in the solution of the
Boltzmann equation. Phys Rev Lett 1998;8:658.

L. Amaya-Bower, T. Lee / Computers & Fluids 39 (2010) 11911207

[23] Sankaranarayanan K, Shan X, Kevrekidis K, Sundaresan IG. Analysis of drag and
virtual mass forces in bubbly suspensions using an implicit formulation of the
lattice Boltzmann method. J Fluid Mech 2002;452:6196.
[24] Takada N, Misawa M, Tomiyama A, Fujiwara S. Numerical simulation of twoand three-dimensional two-phase uid motion by lattice Boltzmann method.
Comput Phys Commun 2000;129:23346.
[25] Swift MR, Orlandini E, Osborn WR, Yeomans JM. Lattice Boltzmann simulations
of liquidgas and binary uid systems. Phys Rev E 1996;54:504152.
[26] Hirt CW, Nichols BD. Volume of uid (VOF) method for the dynamics of free
boundaries. J Comput Phys 1981;39:20125.
[27] Frank X, Funfschilling D, Midoux N, Li HZ. Bubble in a viscous liquid: lattice
Boltzmann simulation and experimental validation. J Fluid Mech
2006;546:11322.
[28] Inamuro T, Ogata T, Tajima S, Konishi N. A lattice Boltzmann method for
incompressible two-phase ows with large density differences. J Comput Phys
2004;198:62844.
[29] Jacqmin D. An energy approach to the continuum surface tension method.
AIAA 1996:960858.
[30] Lee T, Lin C-L, Chen L-D. A lattice Boltzmann algorithm for calculation of the
laminar jet diffusion ame. J Comput Phys 2006;215:13352.
[31] Lee T, Effects of incompressibility on the elimination of parasitic currents in
the lattice Boltzmann equation method for binary uids. Comput Math Appl
2009;58:98794.

1207

[32] Lee T, Fischer PF. Eliminating parasitic currents in the lattice Boltzmann
equation method for nonideal gas. Phys Rev E 2006;74:046709.
[33] He X, Chen S, Zhang R. A lattice Boltzmann scheme for incompressible
multiphase ow and its application in simulation of RayleighTaylor
instability. J Comput Phys 1999;152:64263.
[34] Rowlinson J, Widom B. Molecular theory of capillarity. New York: Dover
Publications, Inc.; 2002.
[35] Day MS, Bell JB. Numerical simulation of laminar reacting ows with complex
chemistry. Combust Theor Model 2000;4:53556.
[36] Jacqmin D. Calculation of two-phase NavierStokes ows using phase-eld
modeling. J Comput Phys 1999;155:96127.
[37] Hua J, Lou J. Numerical simulation of bubble rising in viscous liquid. J Comput
Phys 2007;222:76995.
[38] Holdych DJ, Rovas D, Georgiadis JG, Bukius RO. An improved hydrodynamics
formulation for multiphase ow lattice-Boltzmann models. Int J Mod Phys C
1998;9:1393.
[39] Bonometti T, Magnaudet J. An interface-capturing method for incompressible
two-phase ows. Validation and application to bubble dynamics. Int J
Multiphase Flow 2007;33:10933.
[40] Bunner B, Tryggvason G. Dynamics of homogeneous bubbly ows Part 1. Rise
velocity and microstructure of the bubbles. J Fluid Mech 2002;466:1752.
[41] Rodrigue D. Generalized correlation for bubble motion. AIChE J
2001;47:3944.