Transient movement of fluid spheres using Lattice

Boltzmann method
Lus Orlando Emerich dos Santosa, , Savio Leandro Bertolib , Antonio
Gaspar Gerent J
uniorb, Guilherme Eller Haverrotha
a

Centro de Engenharia da Mobilidade, Universidade Federal de Santa Catarina,

89219-710, Joinville, Santa Catarina, Brasil
b
Departamento de Engenharia Qumica, Universidade Regional de Blumenau, FURB,
SC, C.P. 1507, 89010-971 Blumenau, Brasil

Abstract
In this study, an immiscible lattice Boltzmann is applied to simulate bubble
dynamics at low Reynolds number. Several simulations were performed, considering, specially, the influence of the boundary conditions on the terminal
bubble velocity. The results of the simulations are compared with analytical
solutions and the differences encountered are evaluated, showing the conditions under with the Lattice Boltzmann method can be applied to simulate
bubble dynamics.
Key words: Bubble dynamics, Lattice Boltzmann, Two-phase flow
PACS: 47.11.-j, 47.55.dd
1. Introduction
The lattice Boltzmann method (LBM) has increasingly been applied as
an effective method to simulate fluid dynamics, especially when the focus is
Corresponding author, Tel.: +55 48 3721-6452
Email addresses: [email protected] (Lus Orlando Emerich dos Santos),
[email protected] (Savio Leandro Bertoli)

Preprint submitted to Applied Mathematical Modelling

July 28, 2011

on complex geometries or complex fluids. Unlike the traditional approach

of solving the macroscopic Navier-Stokes equations, the LBM is based in a
mesoscopic kinetic equation, e.g., a discretization of the Boltzmann transport equation. This approach is particularly useful when the fluid flow involves macroscopic behaviors of microscopic origins, like interfaces growth
and breakup, capillary flows and phase transition. The treatment of these
problems by traditional numerical methods is very difficult, usually involving
the use of empirical models, or it is even impossible [1]. In the LBM, however,
the microscopic interactions that originate interfaces and capillary flow can
be incorporated into the model in a natural manner, and many studies has
been published addressing these phenomena (for a review see [2]). Various
LBMs designed to simulate immiscible fluids were proposed [3, 4, 5, 6, 7, 8]
and bubble dynamic simulations were performed using some of them [9, 10].
These simulations are compared with experimental observations[9] or with
other numerical results [10]. In this work, we use the immiscible fluid LBM
proposed by Santos, Facin and Philippi (SFP) [8] to simulate bubble ascension dynamics and compare the results with recent analytical results obtained
by Bertoli[11] and with the experimental results of Haberman e Sayre[12], focusing, especially, under the influence of the walls and the periodic boundary
conditions. The paper is organized as follows. First, in section 2, we revisited
the SFP model, describing its dynamics and its main characteristics. In the
sequence, section 3, we discuss simulation aspects as the boundary conditions
and the way we introduce the forcing term, and presented a comparison of
SFP model and Shan-Chen (SC) model [5]. Section 4 presents simulation
results compared with analytical ones. Finally, in section 5, the results are

discussed concluding the paper.

2. The Santos-Facin-Philippi model
The SFP model was introduced in [8], where a complete analysis id also
presented, including the Chapman-Enskog analysis and the derivation of the
interfacial tension. Little improvements were introduced in [13], however,
the main aspects of the model were not changed. Three distribution functions are used, Ri , Bi and Mi , the first two representing the fluids r and b.
The third distribution, namely, the mediator?s distribution function Mi , is
used to model the long range interaction, carrying out neighborhood information, and being responsible for the segregation between the fluids. These
distributions are updated by two steps, namely:
(a) The local step, which is the particle collision process and the emission/
annihilation of the field mediators;
(b) The non-local step, i.e., the propagation step.
In what follows, X is the position vector and ci is a discrete velocity. In
the local step, particle distributions are updated at each time step T by a
collision process:
0 r ~b
Ri0 (r , ur ) Ri
b Ri ( , ) Ri
+

,
rr
m
0 b ~r
B 0 (b , ub ) Bi
r Bi ( , ) Bi
Bi = Bi + b i
+

,
bb
m

Ri = Ri + r

where
k

bm
X

bm
1 X
u = k
Ki c i ,
i=1
k

Ki ,

i=0

(1)
(2)

(3)

are, respectively, the macroscopic density and the velocity of component k,

k = r, b. Ri0 and Bi0 are the equilibrium distributions (see ref. [14]). The s
3

are the mass fractions, k = k /. The s are the local velocities modified
by the action of mediators,
~ r = ur + Ab

um ,
and

~ b = ub Ab

um ,

(4)

P
Mi ci
b = Pi
u
.
| i Mi ci |

(5)

Mi = Mi + r ,

(6)

The mediators distribution is updated locally by an emission/annihilation

step, defined by

where and are weights used for settling the interaction length[8]. The
propagation step is the only non-local step and is identical for all distributions,
Ki (X + ci , T + 1) = Ki (X, T )

(7)

where K = R, B or M.
The best known and most applied LBM for immiscible fluids is the SC
model. The SFP has some similarities with SC model, which are important
to be mentioned. Both try to reproduce macroscopic behavior using a potential or intermolecular force, without reference to a thermodynamic potential
and in both models the long range interaction acts modifying the velocities
entering in the collision process. The main differences between models and
their importance are described in the sequence.
a) The use of mediators - although, in the Lattice Boltzmann context, the
SFP model be the only one to apply this concept, mediators could be introduced without major changes in other models, not representing an essential
difference between the models. On the other hand, the use of mediators has
4

consequences in the behavior of the model, since the information velocity

becomes finite when using mediators.
b) Splitting of the collision operator - in SC model only one operator is
used in the collision process for each distribution. This process is divided in
the SFP model by the use of two collision operators for each distribution,
using, therefor, three relaxation times, instead of two.
c) Relaxation times depending on the mass fraction - the splitting of the
collision operator permits the use of variable relaxation times, it becoming
possible to consider properly the mass fraction gradients that are present in
the immiscible fluids.
3. Validation: a comparison between SFP and SC models
Considering that the SC model is a well-established method to simulate
immiscible fluids, we compare simulation of bubble ascension using both
methods (the 3D implementation of SC model is described in [15]). Although
this comparison was made simulating in two dimensions, the lattice used was
the D3Q19 one. The reason to do so is that the same computational codes
were used in the sequence to perform 3D simulations. Fig. 1 presents a
schematic view of the simulation, the bubble diameter and dimensions of the
lattice used.
Half-way bounce back boundary conditions were imposed on the solid
surfaces and periodic boundary conditions were imposed on the inlet and
outlet. These boundary conditions will be focused more detailed in comparisons with analytical and experimental results in the next section, here
it is only necessary that the same boundary conditions are impose in both
5

Figure 1: Schematic view of a 2D simulation

simulations. A buoyant force, f = g, is applied changing the momentum

in each lattice site at every time step by the amount r g, where r is the
mass fraction of the bubble fluid.
The ascension can be characterized the dimensionless numbers, Eotvos number (Eo), and Morton number (Mo), defined as:
gD 2
,

(8)

g4L
Mo = 2 3 ,
L

(9)

Eo =
and

where g is the buoyancy force imposed in the bubble, L is the bulk

viscosity of the liquid and is the interfacial tension. These numbers were
6

set Mo = 7.5 107 and Eo = 0.025 in both simulations presented in Fig.

2. This figure shows the agreement between both models confirming again
the capability of the SFP model to simulate multiphase flows. In the next
sections this model will be applied in 3D simulations.
0.02

0.01

0.01

0.01

0.01
0.01
SFP model
SC model

0.00

0.00

0.00
5000

7000

9000

11000

13000

15000

17000

19000

21000

23000

25000

Figure 2: Bubble ascension,comparison between SC model and SFP model.

4. Results
In addition to the boundary conditions used in 2D and 3D simulations,
the specular reflection boundary condition was used, as well. Given the
symmetries, it is not necessary to simulate the entire domain involved in
the problem, applying the specular reflection boundary condition in faces A
and B (see Fig. 3) only one quarter is effectively simulated. This kind of

boundary condition, although not appearing frequently in LB literature, has

already been applied successfully by one of the authors in ref. [13].

Figure 3: Simulation domain in the 3D simulations.

In order to examine the influence of the periodic boundary condition

applied in the inlet and outlet, the simulated results of terminal velocity were
compared with the experimental results considering wall effects obtained by
Haberman and Sayre [12]. They found that spheric fluid particles ascending
(or descending), at low Reynold numbers, inside cylindrical tubes, have the
terminal velocities, UT , affected by the wall in a way that the ratio between
8

the terminal velocity without wall effects, UT , and the terminal velocity,
K = UT /UT , can be expressed by

K=

UT
=
UT

(10)
1+

1+

2.27575 2.27575
2+3

0.70172.1051+2.08653 +1.13785 1.70675 0.726036 +0.726036

1+

,

where = /L is the viscosity ratio, and = D/Dcylinder is the ratio

between the diameter of the fluid sphere (D) and the diameter of the cylinder
(Dcylinder ). The influence of the periodic boundary condition will depend on
the cylinder height, H, therefore this influence can be evaluated simulating
various heights, as presented in Fig. 4. In all these simulations = 2/9
(D = 22 and Dcylinder = 99), and = 0.5. As expected, the terminal velocity
is lower when the cylinder is higher, that is, when the influence of the walls
preponderates over the influence of the periodic boundary conditions. This
influence can be quantified using the correction in the terminal velocity given
by Eq. 10. According to this expression, we should have K = 1.455, while,
considering the case in which H = 360, we have K = 1.396. That means, even
when the cylinder height is many times greater then the bubble diameter,
the effect of the periodic boundary conditions is still significant, increasing
the expected terminal velocity. In these simulations, the Reynolds number
varies from Re = 0.99 (H = 360) to Re = 1.25 (H = 240).
As a matter of comparison, we present in Fig. 5 the results obtained
through the analytical solution of the transient movement of a fluid sphere
in a infinite newtonian fluid in the Stokes regime developed, originally, by
Bertoli [11], using Laplace transforms and a functional solution proposed by
9

0.0100
0.0090
0.0080

U (velocity)

0.0070
0.0060
0.0050
0.0040
0.0030

H=240
H=280
H=320
H=360

0.0020
0.0010
0.0000
0

5000

10000

15000

20000

25000

30000

step

Figure 4: Results of bubble ascension for different heights H.

Hackenberg[16]. The interested reader will find other details in the formal
solution of Chisnell [17], expressed in terms of improper integrals.
To minimize the effects of the periodic boundary condition in order to analyze
the wall effects in the simulations we set the ratio D/H = 0.061 (D =
22, H = 360). As would be expected the simulated results tend to the
analytical solution as the cylinder radius increases. The Reynolds number in
this simulations is about one.
The behavior of the bubble in the analytical and the simulated results
seems to be different, as it can be noted in the figures 5 and 4, the bubbles
in the simulations reaching a steady state earlier. But in fact, this is a

10

1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30

Simulated (r=99)
Simulated (r=89)

0.20

Simulated (r=79)
Theoretical

0.10
0.00
0

10

15

20

25

30

35

40

t*

Figure 5: Comparison between theoretical and simulated results.

consequence of the presence of the walls as can be seen in Fig. 6. When we

increase the cylinder radius, r, the behavior of the bubble in the simulations
tends toward the analytical result.
5. Conclusion
In this work we present diverse simulations of bubble ascension using the
lattice Boltzmann method. These simulations are compared with analytical results from Bertoli [11] and considering the empirical corrections from
Haberman & Sayre [12]. Special attention is given in the influences of boundary conditions: the influence of the walls and periodic boundary conditions
in the inlet and outlet. The differences between theoretical and analytical
11

1.20

1.00

0.80

0.60

0.40

0.20

r=40
r=120

0.00
0

2000

4000

6000

8000

10000

12000

14000

16000

18000

20000

steps

Figure 6: The bubble gets a steady state earlier ascending in cylinder with small r.

results are entirely explicable taking into account the boundary conditions
used and can be made smaller increasing the simulation domains.
6. Acknowledgments
The autors would like to acknowledge the finacial support of CNPq (Conselho Nacional de Desenvolvimento Cientfico e Tecnologico) and to the program PiPe /Art. 170.

12

References
[1] Y. C. Chang, T. Y. Hou, B. Merriman, and S. Osher, A level set formulation of eulerian interface capturing methods for incompressible fluid
flows, Jounal of Computational Physics 124(2), pp. 449464, 1996.
[2] S. Chen and G. D. Doolen, Lattice boltzmann method for fluid flows,
ANNUAL REVIEW OF FLUID MECHANICS 30, pp. 329364, 1998.
[3] A. K. Gunstensen, D. H. Rothman, S. Zaleski, and G. Zanetti, Lattice
boltzmann model of immiscible fluids, Phys. Rev. A 43, p. 4320, 1991.
[4] X. Shan and H. Chen, Lattice boltzmann model for simulating flows
with multiple phases and components, Phys. Rev. E 47, p. 1815, 1993.
[5] X. Shan and H. Chen, Simulation of nonideal gases and liquid-gas
phase-transitions by the lattice boltzmann equation, Phys. Rev. E 49,
p. 2941, 1994.
[6] M. R. Swift, W. R. Osborn, and J. M. Yeomans, Lattice boltzmann
simulation of nonideal fluids, Phys. Rev. Lett. 75(5), pp. 830833, 1995.
[7] M. R. Swift, E. Orlandini, W. R. Osborn, and J. M. Yeomans, Lattice
boltzmann simulations of liquid-gas and binary fluid systems, Phys.
Rev. E 54(5), pp. 50415052, 1996.
[8] L. O. E. Santos, P. C. Facin, and P. C. Philippi, Lattice-boltzmann
model based on field mediators for immiscible fluids, Physical Review
E 68(5), p. Art. No. 056302 Part 2, 2003.

13

[9] X. Frank, D. Funfschilling, N. Midoux, and H. Z. Li, Bubbles in a viscous liquid: lattice boltzmann simulation and experimental validation,
J. Fluid Mech. 546, p. 113?122, 2005.
[10] T. Inamuro, T. Ogata, and F. Ogino, Numerical simulation of bubble
flows by the lattice boltzmann method, Future Generation Computer
Systems 20, p. 959?964, 2004.
[11] S. L. Bertoli, Solucoes exatas para o movimento transiente de objetos
esfericos em fluido Newtoniano infinito no regime de Stokes. PhD thesis,
Federal University of Santa Catarina - Brazil, 2003.
[12] W. L. Haberman and R. M. Sayre, Motion of rigid and fluid spheres in
stationary and moving liquids inside cylindrical tubes, tech. rep., Dept.
of the Navy, David Taylor Model Basin, 1958.
[13] L. O. E. dos Santos, F. G. Wolf, and P. C. Philippi, Dynamics of interface displacement in capillary flow, Journal of Statistical Physics 121,
pp. 197207, 2005.
[14] Y. H. Qian, D. D?Humieres, and P. Lallemand, Lattice bgk models for
navier-stokes equation, Europhys. Lett. 17, pp. 479484, 1992.
[15] N. S. Martys and H. Chen, Simulation of multicomponent fluids in complex three-dimensional geometries by the lattice boltzmann method,
Physical Review E 53(1), pp. 743750, 1996.
[16] C. M. Hackenberg, On the Unsteady Resistance of Submerged Spherical
Bodies. PhD thesis, University of Florida, 1969.
14

[17] R. F. CHISNELL, The unsteady motion of a drop moving vertically

under gravity, J. Fluid. Mech. 176, pp. 443464, 1987.

15