International Journal of Heat and Mass Transfer 48 (2005) 2443–2451

www.elsevier.com/locate/ijhmt

Mesoscale simulation of ferrofluid structure

Yimin Xuan *, Meng Ye, Qiang Li
School of Power Engineering, Nanjing University of Science and Technology, Nanjing 210094, China

Received 20 August 2004; received in revised form 13 December 2004

Available online 29 March 2005

Abstract

By means of the lattice-Boltzmann (LB) method, the mesoscaled structure of ferrofluids consisting of magnetic nano-
particles and a carrier fluid as well as some surfactant is investigated. The ferrofluid is a complicated system and its
morphology is affected by a number of internal and external forces including gravitational force, Brownian force,
van der Waals attraction potential, and dipole–dipole interaction potential. All these factors are included in the lat-
tice-Boltzmann model. The distribution of suspended magnetic nanoparticles and morphology of the ferrofluid are sim-
ulated in both cases of the absence and the presence of an external magnetic field. The effects of the dipole–dipole
interaction energy and the thermal energy on the aggregation structures of the magnetic nanoparticles are discussed.

2005 Elsevier Ltd. All rights reserved.

1. Introduction the effect of an external magnetic field, such properties

of the ferrofluid will be strongly anisotropic, which
Ferrofluids are composed of magnetic nanoparticles means that by means of an external magnetic field one
(3–15 nm) and carrier fluid. Such ultrafine particles can make flow and energy transport processes of the fer-
may be coated by a suitable surfactant (activator or dis- rofluid controllable.
persant) to keep a stable suspension state and they can Since the ferrofluid is a complicated ferromagnetic
be treated as particles of single magnetic domain. Ferro- colloidal system, the suspended magnetic nanoparticles
fluids exert some unique performances under the influ- experience actions from a number of forces such as grav-
ence of external magnetic fields, i.e. an applied itational force, Brownian force, Stokes drag force, van
magnetic field can be used to control physical and flow- der Waals attraction potential, and dipole–dipole inter-
ing properties of the ferrofluid. Ferrofluids behave as a action potential in the presence of an external magnetic
smart or functional fluid and has been finding more field. All these acting forces and potentials dominate
and more applications in a variety of fields such as elec- behavior and structure of the ferrofluid. To get insights
tronic packing, mechanical engineering, aerospace, bio- into the morphology and performance of the ferrofluid,
engineering, and thermal engineering [1–4]. The it is necessary to investigate its microstructure and the
distribution structure of the suspended nanoparticles relevant affecting factors. The main method of investi-
remarkably affects the transport properties and heat gating the microstructure of the ferrofluid is Brownian
transfer characteristic of the ferrofluid. With respect to dynamics and molecular dynamics simulation [5,6].
These publications simulated the cluster aggregation of
*
Corresponding author. Tel.: +86 25 84315700; fax: +86 25 the particles and the structure of the ferrofluid by focus-
84431339. ing attention on each particle and considering some
E-mail address: [email protected] (Y. Xuan). forces (or potentials) acting on the particle. Obviously,

0017-9310/$ - see front matter
2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2004.12.039
2444 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451

Nomenclature

a particle radius, m N/V number density of the particles

dp particle diameter, m T absolute temperature, K
kB Boltzmann constant, J/K x Lorentz number, x ¼ 13
m magnetic dipole moment of single particle, l0 vacuum permeability, l0 = 4p · 107 H/m
A m2 l fluid viscosity, Pa s
Md bulk magnetization of the material, A/m g relaxation time

such a method is time consuming and is limited by a fi- 2, some features and magnetization property of the ferro-
nite number of the suspended particles. The lattice- fluid are analyzed. The LB model for a multiple-compo-
Boltzmann method originating from the Boltzmann nent system is briefly introduced in Section 3. Section 4 is
equation became popular over the past few years [7]. It contributed to description of the interparticle potentials,
bridges the gap between the microscopic world and the Stokes drag force, Brownian force, and gravitational
macroscopic phenomenology and provides a possible force are. Finally, some simulation results of a ferrofluid
approach for simulating flow and energy transport pro- system with homogeneous ferromagnetic nanoparticles
cesses of a flow system at the mesoscaled level without under the influence of all these forces are presented.
particle limitation. As regards the multiscaled simula-
tion algorithm for some system, the lattice-Boltzmann
method is an intermediate between the molecular 2. Magnetic features and magnetization property
dynamics simulation for each microparticle and the
Navier–Stokes equations for macroscaled simulation. It The ferromagnetic nanoparticles suspended in the
simulates fluid motion and energy transport by follow- liquid carrier can generally be considered as a single
ing the evolution of a prescribed Boltzmann equation in- magnetic domain because of their ultrafine sizes. The
stead of solving the Navier–Stokes equations. One of the macroscopic features of the ferrofluid depend not only
important advantages of this method is that microscopic upon the properties of single suspended particles but
physical interactions among the fluid particles can be also on the interactions among these particles. In the
conveniently incorporated into the model. This method presence of an external magnetic field, the ferrofluid will
has been found recent successes in a lot of fluid dynamic be magnetized in the direction of the field and may easily
problems, including flow in porous media [8], thermal approach saturation magnetization. Once the external
two-phase flow [9] and diffusion in multicomponent flu- field is removed, the ferrofluid magnetization will imme-
ids [10]. Sofonea and Frueh [11] proposed a lattice- diately vanish with a randomly oriented domain and
Boltzmann model to investigate the competition relation show no macroscaled magnetism. It is for this reason
between surface tension and dipolar interaction in mag- that the ferrofluid is generally considered as being super-
netic fluids. By taking into account the effects of several paramagnetic. Such magnetization of the ferrofluid is re-
forces and potentials acting on the suspended nanopar- lated with the arrangement structure of the suspended
ticles, this method has recently been introduced to sim- magnetic nanoparticles. The external field forces attract
ulate nanoparticle behavior in nanofluids [12]. the magnetic nanoparticles in a regular alignment, but
From the point of microscopic view, there exist a the Brownian force always puts the nanoparticles in ran-
number of forces and potentials acting on ferrofluids, dom motion and tries to keep them in an irregular struc-
for example, gravitational force, Brownian force, van ture. The external field will align the magnetic moment
der Waals attraction potential, magnetic field gradients, of the suspended particles along the field and lead to
and dipole–dipole interaction potential. Some of them solid particle-carrier liquid separation. Among all the
may bring magnetic nanoparticles together and form forces acting on the ferrofluid, the dipole–dipole interac-
clusters, but the others may make nanoparticles move tion potential and the Brownian force are the two most
apart. It is these internal and external forces that affect important external factors of affecting the microstruc-
and control the morphology of the ferrofluid and make ture and magnetization process of the ferrofluid. In the
the fluid have some unique features. In this paper, we case when an external magnetic field exists, the two main
apply the lattice-Boltzmann method to simulate the meso- parameters of describing the features and performances
scaled morphology of the ferrofluid by accounting for of the ferrofluid are the volume fraction / of the sus-
the effects of the acting forces and potentials on the fer- pended magnetic nanoparticles and the ratio k of the
rofluid structure, especially the effects of the magnetic di- dipole–dipole interaction energy of two contacting parti-
pole–dipole interaction energy and the kinetic energy. cles to the thermal energy kBT [5]. These two dimension-
The paper is organized in the following way. In Section less parameters are given as follows:
 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451 2445

3
N pd p with the velocity ei at lattice site x and time t is denoted
/¼ ð1Þ by fir ðx; tÞ, where {ei; i = 1, . . . , b} is the set of vectors
V 6
pointing from x to its neighboring sites. In order to sim-
m2 ulation the two-dimensional morphology and perfor-
k¼ ð2Þ
4pl0 k B Td 3p mance of the ferrofluid, one gets S = 2 and b = 8.
The evolution of fir ðx; tÞ for a multicomponent sys-
With respect to the fact that the suspended magnetic tem is described by the well-known lattice-Boltzmann
nanoparticles may aggregate under the influences of all equations:
the acting forces or potentials and there may exist some
particle clusters, the conception of magnetic aggregation fir ðx þ ei Dt; t þ DtÞ  fir ðx; tÞ
is put forward to deal with the interaction between two 1 r
¼ ðf ðx; tÞ  fireq ðx; tÞÞ; r ¼ 1; . . . ; S ð7Þ
adjacent clusters or between the particles located in two sr i
adjacent lattices. For this purpose, the magnetic dipole
where sr = gr/Dt is the dimensionless collision–relaxa-
moment of a cluster aggregation can also be considered
tion time constant of the rth component. On the right
as its magnetization and is treated with WeissÕ theory
side, the Bhatnagar–Gross–Krook (BGK) single-relaxa-
[13]. A single-domain nanoparticle is subjected to an
tion time collision term [14] is adopted and the equilib-
effective field instead of the applied field:
rium distribution functions could be express as follows:
H e ¼ H þ H inter ð3Þ

3 9 3
fir;eq ¼ qr wi 1 þ 2 ðei  ur Þ þ 4 ðei  ur Þ2  2 ur  ur
where Hinter = xM, the Lorentz value x is equal to 1/3, c 2c 2c
H is the applied field, and M is the magnetization of ð8Þ
cluster aggregation. Here Hinter is introduced to treat
8
the nanoparticles distributed on a lattice as a cluster. i¼0
< 4=9
>
In ideal conditions, the cluster aggregation possesses
wi ¼ 1=9 i ¼ 1; 2; 3; 4 ð9Þ
the saturation magnetization as >
:
1=36 i ¼ 5; 6; 7; 8
M sat ¼ l0 M s;b V ð4Þ
The equilibrium distribution function fir;eq ðx; tÞ is se-
where Ms,b denotes the bulk saturation magnetization of
lected to ensure that each of the components obeys the
the material, and V is the volume of nanoparticle
macroscopic Navier–Stokes equations. The following
aggregation.
expressions describe the macroscaled quantities such as
According to the terms of the Langevin function
the mass density (or the number density) and the veloc-
L(a) = coth(a)  1/a [5], one can obtain
ity of the rth component and the microscaled quantities
M ¼ M sat LðaÞ ð5Þ defined on the basis of lattice sizes:
X
where a = mHe/kBT. qr ¼ fir ðx; tÞ ð10Þ
Since the relation among expressions (3)–(5) is non- i
linear, expression (5) is handled with TaylorÕs expansion X
near the point M = Msat by ignoring the o[(M  Msat)] qr ur ðx; tÞ ¼ fir ðx; tÞei ðx; tÞ ð11Þ
i
term. Thus, the magnetic dipole moment of the nano-
particle aggregation in a lattice is calculated. These expressions are derived under the assumptions
Thus, the ratio of dipole–dipole interaction energy of that there exist no internal or external forces and poten-
two contacting aggregation cluster to the thermal energy tials acting on the particles, so that the total momentum
kBT is similarly defined as: of the particles of all components should be conserved
M2 by the scattering term (i.e. the collision term) at each lat-
k0 ¼ ð6Þ tice. For such a case, Shan and Doolen [14] assumed that
4pl0 D3 k B T
the equilibrium velocity ueqr of each component is equal
where D is the equivalent diameter of the magnetic to a common velocity ueq
,
aggregation cluster. Here the aggregation cluster is con- X qr ur X qr
eq
sidered as a sphere for the sake of simplicity. u ¼ ð12Þ
r
sr r
sr

For the actual multiphase fluid system such as a mag-

3. Lattice-Boltzmann model form multicomponent system netic fluid, there exist some forces and/potentials acting
on the particles. The impacts and influences from these
Consider the motion of particles of S different com- forces and potentials cause an extra momentum change.
ponents in a regular lattice in D-dimensional space. As mentioned in a previous paper [12], the momentum
The population of the particles of the rth component supplement method and the modified collision-function
2446 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451


method are two main approaches to deal with the A 2a2 2a2 L2  4a2
V¼ þ þ ln ð16Þ
momentum variation induced by the forces and 6 L2  4a2 L2 L2
potentials. The first is the commonest and the most fre-
quently used method by introducing an auxiliary where L is the distance between two adjacent particle
momentum change srFrDt of the rth component to take cores and A is the Hamaker constant (approximately
the vector sum Fr of all the acting forces and potentials 4 · 1020 J for iron oxides in a solvent).
into account. Thus, the equilibrium velocity is modified For two adjacent spheres of radius as and ap, Armin
as et al. [16] directly gave an expression for the van der
Waals attraction force as follows:
ðueq  eq " #
r Þ ¼ u þ sr Fr Dt=mr ð13Þ
Að16RÞ3 s
Fw ¼  er
where mr is the mass density of the magnetic aggrega- 3ðap þ as Þ ð1 þ RÞ2 ½s2 ð1 þ RÞ2  4ð1  RÞ2 2
tion of the rth component. ð17Þ
It has been proven that formula (10) leads to a veloc- 2ðLþap þas Þ
ity field that is a solution of the Navier–Stokes equation. where s ¼ ap þas
,
R = as/ap.
Here the kinetic viscosity mr of the rth component and The dipolar interactions among the suspended mag-
the mean kinetic viscosity of the ferrofluid are respec- netic nanoparticles and/or agglomeration clusters imply
tively given as the interaction of fields produced by two magnetic par-
ticles and/or aggregations themselves and should be
2sr  1 ðDxÞ2 taken into account in the presence of an external mag-
mr ¼ ð14Þ
6 Dt netic field. Generally, the magnetic interaction between
P two magnetic nanoparticles and/or aggregation clusters
2 ur sr  1
ðDxÞ2 with magnetic dipole moments M and M 0 at a distance
m¼ r ð15Þ
6 Dt r is given as
where theP concentration of each component
V MM 0 ¼ M0  H
ur ¼ qr = r qr . Obviously, both expression (14) and

(15) determine the relation between the kinetic viscosity, 1 M  M0 3 0
¼  ðM  rÞðM  rÞ ð18Þ
the relaxation constant, lattice length, and time step. For 4pl0 r3 r5
a real fluid system with a known kinetic viscosity, deter-
mination of the other parameters is confined by these For the uniform external magnetic field, the magnetic
expressions. force induced by the external field is equal to zero. How-
ever, there still is a force acting on each particle as a re-
sult of the sum of the magnetic fields of all the other
4. Forces and potentials in a ferrofluid particles in the surrounding carrier liquid [17]. This
internal force induced by such magnetic dipole moments
Among all the possible factors, as mentioned before, potential is expressed as
the mutual interaction among single particles and the ac- FMM0 ¼ rV MM 0
tion due to the external fields may be most important in
3
controlling the phase behavior and morphology of the ¼ ½ðM  M0 Þ  3ðM  r0 ÞðM0  r0 Þr0 ð19Þ
ferrofluid. Such actions can be described by a series of 4l0 pr4
forces and potentials. Although the forces and potentials where r0 denotes a unit vector.
are considerably complicated, they can be sorted into These potentials are originally three dimensional. For
two types: the continuous action and the transient im- the sake of simplicity, one may assume that they are
pact. The gravitational force, Stokes drag force, the applicable to the two-dimensional problem.
buoyancy, the magnetic force, and van der Waals attrac-
tive force belong to the first category and the Brownian 4.2. Other internal forces and external force
force is of the second.
The other internal and external forces may include
4.1. Interaction potentials the Brownian force, drag force, gravitational force and
buoyant force. All these actions also lead to momentum
The ferrofluid structure is greatly affected by a num- change inside the system.
ber of acting forces and potentials. For the ferrofluid, StokesÕ law is applied to describe the drag force due
the London-van der Waals attraction and the dipole– to the slip velocity difference Du between the aggregation
dipole interaction exist between the magnetic colloids. cluster and the carrier fluid as follows:
Generally, the van der Waals attraction potential be-
tween two adjacent spheres is expressed as [15] FD ¼ 6plaDu ð20Þ
 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451 2447

Since the lattice-Boltzmann method assumes that the Since Brownian motion is generally simulated as a
particles of the fluid are distributed in all lattices and Gaussian white-noise process, the algorithm for simulat-
the particle distribution in a given lattice possesses the ing the Brownian force is similar to that for generating a
same features, one may equate this particle distribution white noise process modeled as a Gaussian white noise
in the lattice as a fictitious particle while handling the process [18],
drag force. rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
12palk B T
The sum of the buoyant force and the gravitational FB ¼ 1 ð24Þ
force is simply expressed as Dt
where the parameter 1 is a Gaussian random number
4
FH ¼ pa3 gDq ð21Þ with zero mean and unit variance.
3 The total force as F acting on the ferrofluid system is
As for the irregular Brownian force, it can be considered defined as the vector sum of all these forces, so that the
as the comprehensive effect of the actions exerted by the extra momentum change induced by these internal and
surrounding fluid molecules. The Brownian force will external forces is obtained as DP = FsrDt. Thus, the
lead to abrupt alteration of both velocity magnitude new equilibrium velocity is determined by the modifica-
and direction of suspended particles or clusters. This tion expression (13).
random uncorrelated force statistically satisfies
hFi ðtÞi ¼ 0 ð22Þ
5. Results and discussion
0 0
hFi ðtÞFj ðt Þi ¼ Cdij dðt  t Þ ð23Þ
The above-derived model can be used to investigate
where i and j denote the components in the Cartesian the structure of the magnetic particles suspended in a
coordinates, d(t  t 0 ) is the Dirac function, C is a carrier liquid and simulate the morphology of the ferro-
constant. fluid. The two-dimensional numerical simulations are

Fig. 1. Distribution of the particles d in the absence of an external magnetic field.

2448 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451

performed in a region consisting of 102 · 102 lattice of various forces and potentials the suspended nanopar-
grids, with periodic boundary conditions imposed on ticles may tend to aggregate or flocculate, and then
the x-axis and y-axis. Additional lattice sites are re- aggregation clusters may form. The larger sizes of nano-
quired at each end of the simulation region to implement particle clusters accelerate the sedimentation process. On
the boundary conditions. The nominal diameter of the the other hand, the thermal kinetic motion will break
suspended Fe3O4 nanoparticles is 10 nm and the volume stable structures of the clusters, so that the clusters will
fraction of the solid magnetic particles is 10%. Since the be in an unstable state and aggregation and breakup of
time step Dt is set to be 107 s and sr is set to be equal to the clusters will coexist, which makes the distribution
0.8 which is assumed to be the same for both compo- structure of the suspended nanoparticles more irregular.
nents to get the correct statistical result, the spatial lat- Another example of the magnetic nanoparticle distri-
tice length is determined with expression (14) or (15). bution in the presence of a perpendicular uniform field
It is assumed that the magnetic nanoparticles are evenly of 6 · 104 A m1 is illustrated in Fig. 2 with the param-
distributed all the lattices at the initial state. eter k 0 = 4970. Compared with Fig. 1, it is evident that
Fig. 1 shows the distribution structures of the mag- the distribution structure of the suspended magnetic
netic nanoparticles suspended in the carrier liquid whose nanoparticles under the effect of the external magnetic
the temperature is 298 K in the absence of an external field quite differs from that corresponding to no applied
magnetic field, in which the vertical coordinate on the magnetic field. Long chainlike structures of the magnetic
right-hand side denotes the number density of the parti- nanoparticles along the direction of the applied mag-
cles. It can be found that when the macroscopic flow netic field form and such structures become more stable
velocity of the ferrofluid is zero, the nanoparticles tend and clearer with increasing time. This phenomenon can
to flocculate and the aggregation becomes more evident be explained by the mechanism that the magnetic mo-
as the time increases. The Brownian force becomes the ment of the suspended magnetic nanoparticles quickly
primary reason for keeping the suspended nanoparticles turns and remains identical to the direction of the ap-
in random motion that provides a large probability of plied field. Interaction potentials of these particles make
nanoparticle collision. Due to the sophisticated actions the particles form chainlike structures along the mag-

Fig. 2. Distribution of the particles under a uniform perpendicular field.

 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451 2449

netic field. The Brownian motion may break larger clus- from the random distribution of the suspended magnetic
ters into smaller segments and tends to make the parti- nanoparticles to short clusters or string-like alignments
cles in irregular arrangements rather than along the along the direction of the external applied magnetic field
direction of the external field, which will reduce the mag- with increasing k 0 . According to numerical simulations,
netization ability of the ferrofluid. the dependence of magnetic nanoparticle distribution
It is expected that in the presence of an external mag- on an external magnetic field is rather weak if
netic field, the dipole–dipole interaction energy and the k 0 < 4029. In such cases, the Brownian force due to the
thermal energy are two of the dominant factors for con- thermal impulse plays a preponderant role on the
trolling the morphology and magnetization ability of the arrangement structure of the suspended magnetic nano-
ferrofluid. As mentioned before, the magnitude of particles as well as the distribution of the magnetic mo-
parameter k 0 indicates the comprehensive effect of both ment. But if k 0 P 4577, the nanoparticles appear in a
these two types of energy on the microscale structure chainlike arrangement along the external field and
of the ferrofluid. For example, consider the case that short-chain aggregation clusters emerge.
the diameter of a magnetic aggregation cluster is The effects of a horizontal magnetic field on the dis-
100 nm, the bulk saturation magnetization of the mate- tribution structures of the suspended magnetic nanopar-
rial is 4.8 · 105 A/m, and the fluid temperature is 298 K, ticles are illustrated in Fig. 4, in which the other
one can get k 0 = 1530. A higher value of k 0 means a parameters remain the same as those in Fig. 3. Com-
stronger effect of the dipole–dipole magnetic energy. pared with the simulation results corresponding to the
Note that the parameter k 0 increases with the magnetiza- vertical magnetic field, the horizontally applied magnetic
tion intensity and decreases with the fluid temperature. field leads to a quite different morphology of the ferro-
Fig. 3 provides some comparison examples about the fluid. Of course, the distribution structures of the sus-
distribution structures of the suspended magnetic nano- pended nanoparticles depend upon the integrated effect
particles corresponding to different values of the para- of all the forces and potentials acting on the ferrofluid
meter k 0 after the same timestep (timestep = 1000). As system. Such effects are distinctly indicated by two
shown, the morphology of the ferrofluid system varies extreme cases (as shown in Fig. 4): one is that the

Fig. 3. Distribution of the particles with an applied uniform perpendicular field corresponding to different k 0 .
2450 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451

Fig. 4. Distribution of the particles with an applied uniform horizontal field corresponding to different k 0 .

dipole–dipole interaction among the magnetic nanopar- three-dimensional simulation. For this purpose, the suit-
ticles is ignored (k 0 = 0) and the other is that the effect of able density distribution functions of artificial particles
the Brownian force is neglected (k 0 = 1). In the first corresponding to the three-dimensional structures such
case, the Brownian force may play a dominant role in the as the D3Q15 or D3Q19 model should be substituted
distribution of the nanoparticles, so that the morphol- for the D2Q9 model and the relevant variations involved
ogy appears as a random state. As the value of parame- in relations between the microscopic parameters and the
ter k 0 increases, the effect of the external magnetic field macroscopic ones should be taken into account.
through the dipole–dipole interaction becomes more
important. When k 0 = 1 and the Brownian force is
small enough to be neglected, the magnetic moments 6. Conclusions
of the suspended magnetic nanoparticles turn to the
direction of the applied magnetic field and the nanopar- In this paper, a lattice-Boltzmann model for simulat-
ticles horizontally appear in a chainlike arrangement ing the distribution structure of the suspended magnetic
along the external horizontal magnetic field. The inter- nanoparticles and the morphology of the ferrofluid has
mediate states between these extreme cases reveal the been developed. Some possible acting forces and poten-
competitive and opposite functions of the Brownian tials such as the gravitational force, the Stokes drag
force and the applied magnetic field. The Brownian force, the Brownian force, van der Waals attraction po-
force always promotes the random motion of the sus- tential, and the dipole–dipole interaction potential have
pended nanoparticles and tries to prohibit the nanopar- been incorporated into the model. The momentum sup-
ticles to be aligned in the external magnetic field, so it plement method has been used to deal with the momen-
will reduce the magnetization ability of the ferrofluid. tum variation induced by the vector sum of all these
It should be mentioned that the D2Q9 model has forces and potentials by introducing an auxiliary momen-
been introduced for two-dimensional simulation of the tum change of the rth component. By means of the
morphology of a magnetic fluid. In principle, the proposed model the structure of the ferrofluid at the
above-described approach can directly be extended to mesoscale level can be simulated without the limitation
 Y. Xuan et al. / International Journal of Heat and Mass Transfer 48 (2005) 2443–2451 2451

of nanoparticle number, compared with the direct [5] Z.W. Wang, C. Holm, A.W. Mueller, Molecular dynamics
numerical simulation method of a flow system. study on the equilibrium magnetization properties and
Several examples have been computed and the structure of ferrofluids, Phys. Rev. E 66 (2002) 021405.
numerical simulation results have revealed the effects [6] H. Morimoto, S. Masuda, T. Maekawa, Cluster–cluster
aggregation of directional and symmetric potential parti-
of the forces and potentials on the distribution micro-
cles, in: Proceedings of the 5th ASME/JSME Joint Thermal
structure of the magnetic nanoparticles. Among all the Engineering Conference, SanDiego, 1999, AJTE99-6516.
forces and potentials the thermal fluctuation force and [7] D.A. Wolf-Gladrow, Lattice-gas Cellula Automata and
the external allied magnetic field are two dominant fac- Lattice Boltzmann Models, Springer-Verlag, Berlin, 2000.
tors of controlling the morphology of the ferrofluid. [8] B. Manz, L.F. Gladden, P.B. Warren, Flow and disper-
Furthermore, they exert quite opposite effects on the sion in porous media: Lattice-Boltzmann and NMR
aggregation clusters structures and the magnetization studies, Fluid Mech. Transport Phenom. 45 (1999)
ability of the ferrofluid. The applied magnetic field en- 1845–1854.
hances the tendency of coalignment with the field direc- [9] L. Vahala, D. Wah, G. Vahala, Thermal lattice Boltzmann
tion and aggregation of the suspended magnetic simulation for multispecies fluid equilibration, Phys. Rev.
E 62 (2000) 507–514.
nanoparticles, but the Brownian force keeps the nano-
[10] X. Shan, G. Doolen, Diffusion in a multicomponent lattice
particles in random motion and suppresses the effect of Boltzmann equation model, Phys. Rev. E 54 (1996)
the external magnetic field. 3614–3620.
[11] V. Sofonea, W.G. Frueh, Lattice Boltzmann model for
magnetic fluid interfaces, Euro. Phys. J. B 20 (2001)
Acknowledgment 141–149.
[12] Y. Xuan, Q. Li, Z.P. Yao, Application of lattice
This work was supported by the National Natural Boltzmann scheme to nanofluids, Sci. China Ser. E 47
Science Foundation of China (Grant no. 50376023 and (2004) 129–140.
50436020). [13] M. Rasa, Magnetic properties and magneto-birefringence
of magnetic fluids, Euro. Phys. J. E 2 (2000) 265–275.
[14] X.W. Shan, G. Doolen, Multicomponent lattice-Boltz-
mann model with interparticle interaction, J. Stat. Phys. 81
References (1995) 379–393.
[15] W.B. Russel, D.A. Saville, W.R. Schowalter, Colloidal
[1] R.E. Rosensweig, Ferrohydrodynamics, Cambridge Uni- Dispersion, Cambridge University Press, Cambridge, 1989.
versity Press, 1985. [16] D.E. Armin, J.A. Ritter, H.J. Ploehn, Magnetic hetero-
[2] R. Hiegeister, W. Andra, N. Buske, R. Hergt, I. Hilger, flocculation of paramagnetic colloidal particles, J. Coll.
U. Richter, W. Kaiser, Application of magnetite ferrofluids Interface Sci. 225 (2000) 39–46.
for hyperthermia, J. Mag. Mag. Mater. 201 (1999) [17] M. Mohebi, N. Jamasbi, J. Liu, Simulation of the
420–422. formation of nonequilibrium structure in magnetorheolog-
[3] Stefan Odenbach, Ferrofluids, Springer-Verlag, Berlin, ical fluids subject to an external magnetic field, Phys. Rev.
2002. E 54 (1996) 5407–5413.
[4] K. Nakatsuka, B. Jeyadevan, S. Neveu, H. Koganezawa, [18] C.H. He, G. Ahmadi, Particle deposition in a nearly
The magnetic fluid for heat transfer applications, J. Mag. developed turbulent duct flow with electrophoresis, J.
Mag. Mater. 252 (2002) 360–362. Aerosol Sci. 30 (1999) 739–758.