THE JOURNAL OF CHEMICAL PHYSICS 134, 184508 (2011)

Magnetophoresis, sedimentation, and diffusion of particles in concentrated

magnetic fluids
Alexander F. Pshenichnikov,1 Ekaterina A. Elfimova,2,a) and Alexey O. Ivanov2
1
2

Institute of Continuous Media Mechanics UB RAS, 1, Korolyov St., 614013, Perm, Russia
Ural State University, 51, Lenin Ave., 620083, Ekaterinburg, Russia

(Received 5 November 2010; accepted 8 April 2011; published online 13 May 2011)
A dynamic mass transfer equation for describing magnetophoresis, sedimentation, and gradient diffusion of colloidal particles in concentrated magnetic fluids has been derived. This equation takes
into account steric, magnetodipole, and hydrodynamic interparticle interactions. Steric interactions
have been investigated using the Carnahan-Starling approximation for a hard-sphere system. In order to study the effective interparticle attraction, the free energy of the dipolar hard-sphere system is
represented as a virial expansion with accuracy to the terms quadratic in particle concentration. The
virial expansion gives an interpolation formula that fits well the results of computer simulation in a
wide range of particle concentrations and interparticle interaction energies. The diffusion coefficient
of colloidal particles is written with regard to steric, magnetodipole and hydrodynamic interactions.
We thereby laid the foundation for the formulation of boundary-value problems and for calculation
of concentration and magnetic fields in the devices (for example, magnetic fluid seals and acceleration sensors), which use a concentrated magnetic fluid as a working fluid. The Monte-Carlo methods
and the analytical approach are employed to study the magnetic fluid stratification generated by the
gravitational field in a cylinder of finite height. The coefficient of concentration stratification of the
magnetic fluid is calculated in relation to the average concentration of particles and the dipolar coupling constant. It is shown that the effective particle attraction causes a many-fold increase in the
concentration inhomogeneity of the fluid if the average volume fraction of particles does not exceed
30%. At high volume concentrations steric interactions play a crucial role. 2011 American Institute
of Physics. [doi:10.1063/1.3586806]
I. INTRODUCTION

Magnetic fluids are stable colloidal suspensions of ferroand ferrimagnetic nanoparticles in a nonmagnetic liquid
carriers.1 The small size of colloidal ferroparticles (typically
of the order of 1020 nm) provides the particle with a permanent magnetic moment. It is well known that in the course of
time an initially homogeneous magnetic fluid, filling a cavity
of arbitrary shape, becomes spatially inhomogeneous with respect to the magnetic phase concentration due to gravitational
sedimentation and magnetophoresis (the motion of particles
under the action of a nonuniform magnetic field). In the absence of convective motion, the only factor that prevents the
concentration stratification of the fluid is the gradient diffusion of particles. The concentration profile in a cavity can
be obtained at some arbitrary time from the solution of the
boundary-value problem including Maxwells equations for
the magnetic field and the dynamic mass transfer equation
with consideration for the terms attributable to magnetophoresis and sedimentation of particles.
Up to now, this boundary-value problem has been solved
using a dilute solution approximation (the volume fraction of
particles is small compared to unity), which makes it possible
to study the magnetic and diffusion parts of the problem separately and to write the diffusion equation correctly.24 Howa) Author to whom correspondence should be addressed. Electronic mail:

Ekaterina.Elfimova@usu.ru.
0021-9606/2011/134(18)/184508/9/$30.00

ever, this approach is not sufficient when we deal with the

high particle concentrations, i.e., with the range of parameters
of special interest from research and application viewpoints.
In the case of high particle concentrations, the magnetic and
diffusion problems are strictly interrelated, and the concentration profile depends markedly on steric, magnetodipole, and
hydrodynamic interparticle interactions, whose counting is a
problem of great concern.
The influence of interparticle interactions on the diffusion processes taking place in magnetic fluids was considered
by a number of researchers, who focused the attention on the
calculation of the diffusion coefficient. In particular, steric
and hydrodynamic interactions in the linear (with respect
to the particle volume concentration ) approximation were
taken into account using the Batchelor formula for the
gradient diffusion coefficient of particles in low concentrated
suspensions:5
D = b0 kT (1 + 1.45) = D0 (1 + 1.45),

(1)

where b0 is the mobility of particles in the carrier fluid, kT is

the thermal energy, and D0 is Einsteins value of the diffusion
coefficient for dilute solutions. The coefficient of gradient
diffusion (1) appears to be the increasing function of particle
concentration, while the particle mobility decreases with
concentration. This effect was explained5, 6 by the influence
of excluded volume, stimulating the transition of particles
from the high concentration region to the low concentration

134, 184508-1

2011 American Institute of Physics

184508-2

Pshenichnikov, Elfimova, and Ivanov

J. Chem. Phys. 134, 184508 (2011)

region. Also this effect was confirmed by experimental data.7

The region of applicability of the formula (1) is usually
restricted by the condition 102 . Biben and Hansen
considered the non-dilute case.8 They investigated the sedimentation of colloidal particles in a monodispersed charged
colloids. Calculations were based on a simple free energy
functional with account for excluded volume and Coulomb
contributions. This approach was extended by Biesheuvel
and Lyklema9 for the case of binary mixture of charged
colloids, and the influence of the gravity, the hard sphere nonpenetration and the ion pressure was studied. The formula
for the chemical potential of the magnetic fluid, describing
the excluded volume effect (as in the case of the van der
Waals gas) was derived in the work of Cerbers.10 Significant
progress in the problem of taking into account steric interactions was achieved in Refs. 11 and 12. The authors derived the
formula for the gradient diffusion coefficient of particles in
the CarnaganStarling approximation for the system of hard
spheres13 and, besides, introduced the correction, linear in
concentration, for the effective attraction of spherical dipoles:


4
8 2
0 m 2

=
D = D0 K () 1 + 2
(1 )4
3
4 d 3 kT
(2)
Here K() = b/b0 is the relative mobility of particles in the
magnetic fluid, b is the mobility of particles in the magnetic
fluid, 0 = 4 107 H/m, is the coupling constant, m, d are
the magnetic moment and diameter of the particle, respectively. In the absence of an external magnetic field the relative
mobility K () of ferroparticles in a magnetic fluid is a scalar
function of the particle volume concentration. Mobility K ()
could be expanded in power series over particle volume
concentration and the linear term was given by Batchelor:5
K () = (1 6.55).

(3)

The second term in square brackets in Eq. (2) takes into

account steric interactions in the full range of particle concentrations quite accurately and remains unchanged below.
Morozov14, 15 has focused attention on mass transfer
anisotropy in a system of interacting dipolar particles under the presence of an external magnetic field. He has suggested the conception of anisotropic diffusion coefficients and
has calculated these renormalized diffusion coefficients taking into account magnetodipole and steric interactions for the
magnetic fluid in the infinite plane layer placed in the uniform
applied field H0 . The gradient of particle concentration n is
directed across the plane layer so that the magnetic part of the
problem has a trivial solution B = 0 (H + M) = const or
H = const depending on the direction of the applied magnetic field (M is fluid magnetization). Two field orientations,
namely, along the concentration gradient and across it, were
considered. In these limiting cases, the diffusion coefficients
in the linear concentration approximation are, respectively,
equal to
D|| = b0 kT [1 + 1.45 + 2.23 L L 2 (0 )], for H||n,
(4)

D = b0 kT [1 + 1.45 1.112 L L 2 (0 )], for H n.

(5)
Here L = 0 m2 n/3kT is the initial susceptibility of the
fluid, obtained using the Langevin approximation, L( )
= coth( ) 1/ is the Langevin function, and 0
= 0 m H0 /kT is the Langevin parameter.
There are two independent reasons of mass transfer
anisotropy in magnetic fluid.14, 15 First, the particle mobility
might be anisotropic due to field-dipole interaction. The second one is the anisotropy of thermodynamic forces caused
by demagnetizing fields (i.e., cavity shape). The lasts terms
in Eqs. (4) and (5) take into account both of these effects.
However, it was shown by Morozov15 that the influence of
particle mobility anisotropy is much weaker (by an order of
magnitude) than the one of thermodynamic forces anisotropy.
This fact has allowed Morozov to neglect the anisotropy of
particle mobility when studying the diffusion in concentrated
magnetic fluids.15 The obtained results agree well with experimental data,16 but the range of its application is limited by
the case of the uniform applied magnetic field and the plane
layer of magnetic fluids, when the gradient of particle concentration is directed across the layer. Any changes of the cavity
shape and/or magnetic field orientation make the expressions
(4) and (5) incorrect.
The goal of the present study is to derive a dynamic mass
transfer equation for describing magnetophoresis, sedimentation, and gradient diffusion of particles with consideration for
steric, magnetodipole, and hydrodynamic interactions in cavities of arbitrary shape and in magnetic field of arbitrary configuration. Following Refs. 11 and 15, we do not take into account the weak anisotropy of particle mobility in an external
magnetic field. This uncrude approximation leads to a great
simplification of the mass transfer equation. Unlike Refs. 14
and 15, we suggest to consider the strong anisotropy of thermodynamic forces with the help of additional term in the
mass transfer equation, responsible for magnetophoresis. In
this way, the anisotropic diffusion coefficients need not to be
defined. The coefficient of gradient diffusion remains a scalar
function and does not depend on cavity shape and magnetic
field configuration.
II. STATIONARY DISTRIBUTION FOR DILUTE
MAGNETIC FLUIDS

Let us first consider the problem of stationary distribution of particles in a dilute magnetic fluid, contained in a
closed cavity with fixed isothermal boundaries in the absence
of hydrodynamic flows. The particle concentration is low,
the dipolar coupling constant is small, and hence the interparticle interactions are inessential. This case is of interest
because the spatial distribution of particles in the cavity can
be found without formulating and solving the boundary-value
problem and the obtained solution can be further used to
analyze situations that are more complicated. The solution
of the problem is simplified because the force fields are
potential and the cavity walls are impermeable to particles.
Under these conditions, the stationary state of the system is
at the same time the thermodynamic-equilibrium state (fluxes

184508-3

Magnetophoresis, sedimentation, and diffusion of particles in fluids

of mass, momentum, and energy are absent), and the system

is described by the Boltzmann distribution. The potential
energy U of a single-domain colloidal particle is
U = 0 m H (x, y, z) cos Vs g(z cos + x sin ),
(6)
and the distribution function is


U (r, )
f (r, ) = exp
.
kT

(7)

Here is the angle between the particle magnetic moment m

and the field intensity H,  is the density difference between
the solid kernel of the particle and the disperse medium, Vs
is the volume of the solid particle kernel, is the angle
between the z-axis and the gravity vector g, lying in the xOz
plane. The last term in expression (6) describes the influence of the Archimedean force. The number concentration
n(x,y,z) of particles at the arbitrary point of the space can be
determined by averaging Eq. (7) over the magnetic moment
orientations:
n(x, y, z) = A

sinh
exp[G (z cos + x sin )],

G = Vs g/kT.

(8)

Here the Langevin parameter = 0 m H/kT is the known

function of coordinates, and G is the gravitational parameter
(the reciprocal height of the barometric distribution). The
normalization constant A is best determined in terms of
the average volume concentration of particles in the cavity,
therefore it is convenient to change the number concentration
of the particles n for the particle volume concentration In
this case, the concentration field in the cavity is described by
the equation
sinh
exp[G (z cos + x sin )]

,
=  
1
sinh
exp[G (z cos + x sin )]dv
V V

(9)
where  denotes the average volume concentration of
particles.
Since Eq. (9) describes only the equilibrium state of
the system, it does not include any kinetic coefficients. It is
applicable to the cavities of arbitrary shape (including 3D
problems) and arbitrary magnetic fields and can be readily
extended to polydisperse suspensions. In this case, we need to
write an Eq. (9) for each fraction and then summarize the leftand right-hand sides of these equations. In the limit of weak
fields (  1), Eq. (9) coincides with the barometric formula.
The main disadvantage of Eq. (9) is that it neglects the
interparticle interactions and, therefore, cannot be used for
describing concentrated systems, including the case when
the high concentration of particles appears only in some
part of the system. The equation similar to Eq. (9) was derived earlier,4, 17 but it did not take into account particle
sedimentation.

J. Chem. Phys. 134, 184508 (2011)

III. MAGNETOPHORESIS AND SEDIMENTATION

OF PARTICLES IN CONCENTRATED SYSTEM

The problem of magnetophoresis of ferroparticles in a

nonuniform magnetic field is solved with account for interparticle magnetodipole interaction in the framework of the
modified model of the effective field (MMEF), whose efficiency has been verified in the early works devoted to the
study of equilibrium magnetization of concentrated magnetic
fluids.1820 Following Refs. 19 and 20, the equilibrium magnetization of magnetic fluids is described by the system of
equations:
M = mn L(e ),

He = H +

e =

0 m He
,
kT



1 d M L (H )
M L (H )
1+
, M L = mn L( ),
3
48 d H
(10)

where H is the magnetic field in the fluid, which differs from

the applied field H0 by the demagnetizing factor, and He is
the effective field. The system of equations (10) agrees well
with the experimental data18 and the results obtained by the
Monte-Carlo methods and molecular dynamics methods.20, 21
A significant difference between the MMEF and the experimental magnetization data were observed only in the case
of magnetic fluids with a very high (several tens) magnetic
susceptibility.22, 23 Based on the MMEF data, in the known
formula for magnetic force F acting on a particle,24 the field
intensity H is replaced by its effective value He , He  H,
F = 0 (m ) He .

(11)

Since the characteristic diffusion time of concentration disturbances D L2 / 2 D for typical magnetic fluids exceeds
the magnetic moment relaxation time B 3V/kT (L is the
characteristic size of the cavity, is the magnetic fluid viscosity) by at least six-seven orders of magnitude, the right-hand
side of Eq. (11) can be averaged over the time satisfying the
two-sided inequality: B   D . Bearing in mind that, in
view of the MMEF, the mean value of the magnetic moment is
equal to m = mL( e )He /He and assuming that the effective
field is potential, we obtain from Eq. (11)
F = 0 m L(e ) He .

(12)

The particle flux density jm is expressed in terms of their

number density n and the average drift velocity v = bF
jm = nv = 0 bnm L(e ) He .
Assuming that the temperature of the magnetic fluid is homogeneous, we write the magnetophoresis density in the
form
jm = n D0 K ()L(e )e .

(13)

The expression for the density js of the sedimentation flux

caused by the gravitational force is written analogously
js = n D0 K () G e.

(14)

184508-4

Pshenichnikov, Elfimova, and Ivanov

J. Chem. Phys. 134, 184508 (2011)

Here e is the unit vector in the direction of the gravitational

acceleration g, and G is the gravitational parameter obtained
in Eq. (8).
At the end of this section, let us remind once again that
generally the mobility of ferroparticles in a magnetic field is
anisotropic and depends on the mutual orientation of the magnetic field intensity and the velocity of particle directed drift.
However this anisotropy is weak and we neglect it. The thermodynamic force anisotropy exceeds mobility anisotropy by
an order of magnitude.14, 15 This anisotropy is rather strong
and plays a decisive role. Below we take into account the thermodynamic force anisotropy with the help of Eq. (13) as a
part of the total particle volume flux density (20). This term
(13) is responsible for magnetophoresis and is proportional to
the effective magnetic field gradient. In this case the coefficient of gradient diffusion remains a scalar function which
leads to great simplification of the specific boundary-value
problems.
IV. EFFECTIVE ATTRACTION OF PARTICLES

The averaging of the magnetodipole interactions over the

magnetic moment orientations gives rise to the effective attraction between particles and is an additional reason for the
drift of particles in the concentration-inhomogeneous magnetic fluid. Because the corresponding term in the particle flux
is proportional to the concentration gradient, it can be considered formally as a correction to the diffusion coefficient, like it
has been done in Ref. 12 during the derivation of Eq. (2). The
last summand of formula (2) takes into account the interparticle magnetodipole interaction in the linear-in-concentration
approximation. For dense magnetic fluids the diffusion coefficient can be obtained using the hard-sphere approximation
and the known relationship between the diffusion coefficient
and the free energy  of the system. According to Refs. 11
and 12, the free energy of hard spherical dipoles can be represented as
 = N0 0 + N 0 + N kT ln() kT ln(Q s )
+ kT N G(, ),

To obtain G(, ), we use the virial expansion in terms

of the volume ferroparticle concentration . Each virial coefficient is an effective potential for interactions between the p
particles averaged over the ferroparticle position and the orientation of magnetic moments. The order in which the virial
coefficients are taken indicates the number of interacting particles: the coefficient p2 characterizes the p-particle correlations. The virial coefficients are defined by the p-particle
cluster integrals using the diagrammatic expansion method.25
In this paper, G(, ) is calculated up to 2 .
G(, ) =

4
4 2
+ 4
3
75



1
10 3
2
4
0.34194
+
2 ln 2 +
3
9
+ 0.967242 2 + . . . .

(18)

Using the known computer simulation data,26 we define

the applicability range of Eq. (18): 1, < 0.3. The slow
convergence of the series and the existence of the alternate
terms in the right-hand side of Eq. (18) initiate the physically
incorrect effects in the range > 2. Therefore, expression
(18) is modified as
G(, ) =

4 2 (1 + 0.042 ) (1 + 1.28972 + 0.72543 2 )

.
3 (1 + 0.3082 )
(1 + 0.83333)
(19)

Approximation (19) coincides with the right-hand side of

Eq. (18) with accuracy to the terms of the order of 2 , and
it provides better agreement with the results of computer simulations (Fig. 1). One can see that formula (19) is in good
agreement with the results corresponding to the parameter
range provided in Ref. 26. It should be mentioned that the
positive value of G(, ), in expressions (18) and (19), means
that the non-central magnetodipole interaction represents itself in ferroparticle ensemble as the effective interparticle
attraction.

(15)

where N0 is the number of molecules of the carrier fluid, N is

the number of ferroparticles in the magnetic fluid, 0 , 0 are
the chemical potentials of the carrier fluid and the molecules
of ferroparticles, respectively, Qs is the configuration integral
of the hard sphere system,13 G(, ) is the function specifying the contribution of a magnetodipole interaction to the free
energy of the dipolar hard sphere system. Calculation of the
chemical potential yields the diffusion coefficient

2 
4
2 G
3 G
D = D0 K () 1+2
2G

4
.

(1)4

2
(16)
and the diffusion flux density


4
2 ( 2 G)
n.

j D = D0 K () 1 + 2
(1 )4
2
(17)

FIG. 1. Function of the contribution of magnetodipole interactions to free

energy as the function of volume particle concentration and dipolar coupling
parameter. Dots correspond to computer simulations (Ref. 26) and solid lines
are approximation (19).

184508-5

Magnetophoresis, sedimentation, and diffusion of particles in fluids

V. MASS TRANSFER EQUATION

Summarizing the left- and right-hand sides of Eqs. (13)

and (14), and (17) and making allowance for the proportion-

J. Chem. Phys. 134, 184508 (2011)

ality between the number density n of particles and their volume concentration , we obtain an expression for the density
of volume flux of particles in the isothermal magnetic fluid





2(4 )
2 ( 2 G)
.
J = D0 K () L(e )(e ) + G e 1 +

(1 )4
2

The first two terms in braces of Eq. (20) characterize the

magnetophoresis and the particle sedimentation in the gravitational field or in the field of centrifugal forces, the first
and second terms in the square brackets describe the gradient diffusion with the correction for steric interactions, and

the last term in the square brackets is responsible for the effect of magnetodipole attraction. The dynamic mass transfer
equation in the absence of a convective flow is derived from
Eq. (20) in a common way (see, for example, Ref. 27) and can
be written as







2 (4 )
2 ( 2 G)

.
= div D0 K () L (e ) (e ) + G e 1 +

t
(1 )4
2

This dynamic mass transfer equation (21) for a magnetic fluid under the action of magnetic field and gravity is more general the known ones.2, 11, 14, 15, 28 First, the
used gradient diffusion term (16), (19) can be applied to
more concentrated magnetic fluid than the expression (2).
Second, in combination with the magnetostatic equations
it can be used for description of the mass transfer processes in magnetic fluids placed in a cavity of arbitrary
shape under the action of a static magnetic field of different
configuration.
Let us compare our results with expressions (4) and (5),
which have been calculated by Morozov14, 15 in linear approximation over concentration for the plane layer of the magnetic
fluid under the condition when the particle concentration gradient is directed across the plane layer and only parallel and
perpendicular orientations of a magnetic field are considered.
The right-hand part of Eq. (20) could be expanded in power
series over the particle concentration:
The expansion of the first two terms in square brackets
is evident.
Using expression (18) the last summand in square
brackets gives the linear term over volume concentration 82 /3.
Expansion of the first term in braces, which is responsible for magnetophoresis, has to be considered
in more details. Assuming the uniformity of an internal magnetic field H, it could be connected with an
applied field H0 and magnetic fluid magnetization M
via the demagnetizing coefficient (H M H0 ):
H = H0 M.

(22)

The coefficient depends on the geometry of the cavity and

the mutual orientation of the cavity and the applied field. Relationship between the corresponding Langevin parameters

(20)

(21)

follows immediately from Eqs. (22) and (10).

= 0 24 L(e ).

(23)

Simplifying the magnetization (10) with the linear correction

in concentration only (first-order MMEF), we get the effective
Langevin parameter
e = + 8 L( ).

(24)

Using combination of Eqs. (23) and (24) and taking into account only linear terms, we obtain
e = 0 + 8(1 3)L(0 ).

(25)

L(e )e = 8(1 3)L 2 (0 )

= L (1 3)L 2 (0 ).
So, using the Batchelor formula for relative mobility (3), the
Eq. (20) takes a form


8 2
2
J = D0 1 + 1.45 L (1 3)L (0 ) .
3
(26)
Hence, the Eq. (20) transforms to the classical Ficks law, and
the gradient diffusion coefficient of colloidal particles is


8
D = D0 1 + 1.45 L (1 3)L 2 (0 ) 2 .
3
(27)
The anisotropy of the renormalized diffusion coefficient (27)
is defined by demagnetizing coefficient . In the case of
spherical cavity ( = 1/3) the anisotropy equals zero. For the
plane layer of the magnetic fluid, when the external field is
directed across the plane layer but along the particle concentration gradient, the demagnetizing coefficient equals one

184508-6

Pshenichnikov, Elfimova, and Ivanov

( = 1) and the renormalized diffusion coefficient is a maximum:



8
D|| = D0 1 + 1.45 + 2 L L 2 (0 ) 2 for H || n.
3
(28)
In the other limit case, when the external field is directed
along the plane layer but across the particle concentration gradient, the demagnetizing coefficient equals zero ( = 0) and
the renormalized diffusion coefficient approaches the minimal
value:


8
D = D0 1 + 1.45 L L 2 (0 ) 2 for H n.
3
(29)
The Eqs. (28) and (29) coincide with Eqs. (4) and (5) with
accuracy to small corrections concerned with the anisotropy
of mobility and the mutual attraction between particles (last
term quadratic in ).
The formulas (28) and (29) hold valid only for the plane
layer of the magnetic fluid with known demagnetizing coefficients, while in Eqs. (20) and (21) there are no restrictions on the cavity shape and demagnetizing fields. Masstransfer equation (21) extends the scope of the examined
problems. Separation of the summands characterizing magnetophoresis and diffusion provides an adequate formulation
of the boundary-value problem on distribution of the magnetic
phase concentration over an arbitrary cavity with the magnetic
fluid. The only requirement is that the solution of the diffusion
problem should conform to the solution of the magnetostatic
problem considering the magnetic field intensity inside the
fluid.
Terminologically, it seems reasonable to distinguish between the diffusion of particles caused by the Brownian motion and magnetophoresis, i.e., the drift of particles in the
magnetic field including demagnetizing field. Magnetophoresis is responsible for the concentration segregation of the
magnetic fluid in the gradient field, and the diffusion of particles equalizes concentration differences. From this standpoint, the question of whether the summand, which takes into
account the effective attraction between magnetic particles,
should be involved in the diffusion coefficient may be debated. However, by analogy with Ref. 11, this summand is
involved in the diffusion coefficient considered in this study,
which allows us to simplify the notation. Hence, by the diffusion coefficient of particles in magnetic fluids we mean the
quantity


2(4 )
2 ( 2 G)
,
(30)

D = D0 K () 1 +
(1 )4
2
which follows from Eq. (16).
VI. INFLUENCE OF INTERPARTICLE INTERACTIONS
ON MAGNETIC FLUID STRATIFICATION
A. Zero applied field

As can be seen in Eqs. (19) and (30), the existence of

magnetodipole interparticle interactions leads to a decrease
in the diffusion coefficient. The influence of these interactions gets stronger as the parameter increases, and when

J. Chem. Phys. 134, 184508 (2011)

FIG. 2. Diagram of spinodal decomposition of the system in zero applied

field.

it reaches rather high values, the diffusion coefficient may be

negative. In this case, the diffusion flux of particles is directed
along the concentration gradient, and the concentration perturbation, caused randomly by fluctuations, will increase with
time. The system becomes thermodynamically unstable and
stratifies into two phases, weakly and strongly concentrated.
The instability gives rise to the phase transition gasfluid.
Equating the diffusion coefficient (30) to zero, we obtain the
condition for thermodynamic instability of the system in the
absence of applied fieldspinodal decomposition caused by
magnetodipole interactions.
A diagram of decomposition is shown as the plot =
() in Fig. 2. To the critical point of the diagram there correspond * = 0.034 and * = 4.22. At < * (i.e., at the
temperature higher than the critical one) to the equilibrium
state of the system there corresponds the homogeneous particle distribution, whereas at > * the system stratifies into
weakly and strongly concentrated phases. Despite the fact that
the parameter values shown in Fig. 2 lie formally beyond the
applicability of approximation (19), the phase diagram turns
out to be quite realistic. The critical values * and * in this
diagram fall in the range of values previously obtained by
other methods. For example, * = 4.08, * = 0.092 in terms
of the effective field model;10 * = 2.82, * = 0.13 in the
framework of the thermodynamic theory of perturbation;12, 28
* = 4.45, * = 0.056 in a mean-sphere approximation;29
and * = 3.0, * = 0.034 in the context of Monte-Carlo
simulations for a restricted system.30 The above-mentioned
analytical models can be used only within the limited range
of concentration and dipolar coupling constant, and therefore a broad scatter in the critical parameters is absolutely
natural.
B. External applied fields

In the external fields (magnetic or/and gravitational),

the static distribution of particles in the cavity is obtained by equating the full particle flux (20) to zero. Thus,
we have
L (e ) (e ) + G e

184508-7

Magnetophoresis, sedimentation, and diffusion of particles in fluids


2 (4 )
2 ( 2 G)
= 0.
1+

(1 )4
2

J. Chem. Phys. 134, 184508 (2011)

(31)

It is easy to verify that this equation is equivalent to

3
( 2 G)

(1 )3


sinh e
G z + const,
= ln
e

ln +

(32)

which, being in its implicit form, defines the spatial distribution of particles in the magnetic fluid subjected to the action
of magnetic and gravitational fields. The integration constant
in the right-hand side of Eq. (32) can be determined through
the concentration of particles at some (reference) point inside
the cavity or on its boundary, or through the average volume
concentration 
We used Eq. (32) to perform test calculations of the static
profile of particle concentration in a vertical cylinder of finite
height z0 placed in the gravitational field. The magnetic field
was absent. The results of calculation were compared with
the data obtained by the Monte-Carlo method. The computer
simulation method was similar to the technique described in
Ref. 31. A colloidal particle is modeled as a hard sphere with
a constant value of magnetic moment. The energy of the ith
particle is the sum of dipolar interactions, magnetic and gravitational potentials:
N

Ui
= G z i 0 cos i
kT
j=1,
j
=i


3(ei Ri j )(e j Ri j ) (ei e j )

.
Ri5j
Ri3j

(33)

Here Rij is the distance between the centers of the ith and jth
particles, and i is the angle between the applied magnetic
field and the magnetic moment of the ith particle, ei is the unit
vector in the direction of the magnetic moment of the ith particle. Steric interactions were taken into account by forbidding
the hard spheres to overlap with each other or with the cylinder wall. To find the stationary particle distribution profile, the
cylinder was divided into 20 horizontal layers of the thickness
equal to the particle diameter. After the establishment of thermodynamic equilibrium, the local concentration profile was
averaged over 105 MC-steps. The data from the top and bottom layers were not taken into consideration because of the
known boundary effect. The mean concentration of particles
in the cylinder was determined using the rest of 18 layers and
appeared to be slightly different from the concentration at the
initial time. Calculations were performed under the assumption G z0 = 5 for the system consisting of 103 particles. The
concentration profiles are given in Fig. 3 for different values
of .
Curve 1 in Fig. 3 corresponds to the non-magnetic particles ( = 0), and the deflection of this curve from the barometric distribution (dashed curve) demonstrates the influence
of steric interactions on the concentration profile. Irregardless of the relatively low average concentration of particles
(slightly higher than 6% by volume), these interactions turned

FIG. 3. Static concentration profile in a vertical cylinder of finite height

placed in the gravitational field. The magnetic field is absent. Curve 1 corresponds to = 0,  = 0.061; curve 2 is = 2,  = 0.062; curve 3 is
= 3,  = 0.06. Dots are computer simulation data and the solid lines
correspond to formula (32).

out to be significant. Initiation of the magnetodipole interactions (curves 2 and 3) markedly strengthens the system stratification if > 1. In particular, at = 3 the stratification
coefficient P (the ratio between the maximum and minimum
values of concentration) becomes three times larger than that
observed for = 0. In general, Fig. 3 demonstrates quite
good agreement between Eq. (32) and the results of the MCsimulation for all examined parameters.
Figure 4 presents the plot of the stratification coefficient
P versus the average volume concentration, obtained from
Eq. (32) under the same conditions as in Fig. 3. The dashed
line corresponds to the barometric distribution in dilute solutions. The deviation from the barometric distribution indicates the influence of interparticle interactions. As the average volume concentration of particles increases from zero to
the maximum possible value m 0.61, the stratification coefficient decreases by four (!) orders of magnitude. The effective attraction of the magnetic dipoles plays an important
role in the stratification of the magnetic fluid in moderately
concentrated fluids when the volume particle concentration
is 3%30%. In this case, the effective attraction is able to
generate a several-fold increase in the inhomogeneity of the
fluid. In strongly concentrated magnetic fluid, where the particle volume fraction is higher than 30%, the influence of effective interparticle attraction on the stratification becomes
inconsiderable.

VII. INFLUENCE OF APPLIED MAGNETIC FIELD

Equations (21) and (32) describe the spatial distribution

of particles in the magnetic fluid. These equations have been
derived without any assumptions concerning the cavity shape.
However, the presence of the effective Langevin parameter in
Eqs. (21) and (32) indicates that for calculation of concentration fields we need to get more information concerning magnetization and intensity of the field inside the fluid. That is the

184508-8

Pshenichnikov, Elfimova, and Ivanov

J. Chem. Phys. 134, 184508 (2011)

FIG. 4. Stratification coefficient P versus the average volume concentration

, obtained from Eq. (32). Dashed line corresponds to the dilute solution
approximation, and the numbers of solid curves show the numerical values
of the dipolar coupling parameter.

reason why Eq. (21) or Eq. (32) should be solved in combination with Maxwells equations
r otH = 0,

div (H + M) = 0,

(34)

and Eq. (10), which show the relation between the magnetization and the field intensity and particle concentration. In
the general case (complex geometry of the cavity), the inhomogeneous distribution of particles in the cavity and the nonlinearity of the diffusion equation extremely complicate the
problem even for the numerical solution. The problem is simplified for the cavity of simple (ellipsoidal) shape. If the concentration difference in the cavity is small enough, then the
magnetostatic problem with Eq. (34) is solved analytically.24
In this case, we can use formulas (22)(24). It allowed us to
derive the effective value of the Langevin parameter e from
the Langevin parameter 0 (Eq. (25)) determined through the
applied field H0 and to calculate the equilibrium distribution
of the magnetic phase concentration using Eq. (32).
VIII. CONCLUSIONS

In this paper we are studying the mass transfer processes in a magnetic fluid caused by three different origins:
the gradient of the ferroparticle concentration, the gravity or
centrifugal forces, and the gradient of nonuniform magnetic
field; the possible convective fluxes are not considered. The
first one results in the gradient Brownian diffusion flux jD
of the known Ficks type (17), and for dense magnetic fluids
the gradient diffusion coefficient D depends both on the particle concentration and the intensity of interparticle magnetodipole interaction (16). The gravity/centrifugal flux jS is
written in a usual way (14). In applied magnetic field the mass
transfer becomes anisotropic due to the anisotropy of the particle hydrodynamic mobility and the anisotropy of the thermodynamic forces. The influence of the last reason is much
higher (by the order of magnitude) than the first one; so we are
neglecting the anisotropy of the particle hydrodynamic mobil-

ity and are considering the mobility as the scalar function of

the particle concentration (3). We suggest to describe the field
induced spatial nonuniformity of the mass transfer fluxes by
means of separate magnetophoreses term jm (13) in the mass
transfer equation. Thus, the dynamic mass transfer equation
(21) for describing magnetophoresis, sedimentation, and diffusion of colloidal particles in concentrated magnetic fluids
has been derived. A general solution (32) has been found for
static particle distribution. This solution is correct for the cavities of arbitrary shape and for the magnetic field of arbitrary
geometry. Steric interactions (excluded volume effects) are
described with the help of Carnagan-Starling approximation
for a hard sphere system. This approximation agrees with the
results of computer simulation of static concentration profile
in gravitational field (Fig. 3, curve 1).
Magnetodipole interactions demonstrate the double effect. On the one side, they increase the intensity of magnetophoreses flux due to the additional term in effective magnetic field (10) acting on each particle magnetic moment. This
term, in the right hand part of Eq. (10), is proportional to
M L (H ) /3 and tends to zero at the limit of strong fields only.
On the other hand, the magnetodipole interactions give rise to
the effective interparticle attraction (17) and (19), which increases segregation in the system. The magnetic fluid thermodynamic instability (spinodal decomposition) can occur for
the case of rather high values of the dipolar coupling constant
. The result of the instability is the system separation in low
and high concentrated phases (Fig. 2).
The effects of interparticle magnetodipole attraction are
described by the last term of Eqs. (20) and (21), which were
found from the free energy virial expansion (18). Obtained
interpolation formula (19) allows to receive qualitatively correct results for big values of ; in particular, the calculated
diagram of spinodal decomposition (Fig. 2) predicts quite
reasonable values of critical parameters. In zero magnetic
fields the gravity induced particle stratification is increasing
greatly with the strengthening of effective interparticle attraction (Fig. 3) for low ferroparticle concentration (volume concentration is less than 20%). At the same time, the interparticle attraction effects are reduced for dense magnetic fluids
( > 30%).
Concerning the hydrodynamic interparticle interactions,
they influence only on the dynamics of the mass transfer process through the relative mobility coefficient K(). The static
particle distribution, defined in Eq. (32), does not depend on
the particle mobility.
We suggest to use the obtained dynamic mass transfer
equation (21) in the boundary-value problems for calculation
of concentration distribution and magnetic field geometry in
the devices, based on the concentrated magnetic fluid as a
working fluid (for example, the magnetic fluid seals and the
acceleration sensors). According to Eqs. (10) and (20), the
density of particle flux is determined by the magnetic field
intensity and the magnetization of fluid inside a cavity. Time
drift of these quantities is correlated with the change in particle concentration. So, the magnetic and diffusion problems
are strictly interrelated and should be solved together. But in
the concentrated magnetic fluids the magnetic moment relaxation time (103 104 s) is much less than the characteris-

184508-9

Magnetophoresis, sedimentation, and diffusion of particles in fluids

tic diffusion time D L 2 / 2 D (L is the characteristic size

of the cavity). The latter exceeds 105 s even for quite small
cavity about 0.1 cm. It means that the magnetic part of the
problem should be solved in quasistatic approximation, for
example, with Maxwells equation (34), for known boundary
conditions on the cavity wall.
ACKNOWLEDGMENTS

The work was supported by the Russian Foundation for

Basic Research (Grant No. 10-01-96038); the Department of
Energetic, Machine building, Mechanics and Control Processes RAS (Project No. 09-T-1-1005); Federal Target Program Scientific and Academic Teaching Staff of Innovative Russia in 20092013 and Grant of President of RF MK1673.2010.2.
1 R.

E. Rosensweig, Ferrohydrodynamics (Cambridge University Press,

Cambridge, 1985).
2 E. Ya. Blums, M. M. Mayorov, and A. O. Cebers, Magnetic Fluids (Walter
de Gruyter, Berlin, 1997).
3 V. G. Bashtovoi, V. K. Polevikov, A. E. Suprun, A. V. Stroots, and S. A.
Beresnev, Magnetohydrodynamics 43(1), 17 (2007).
4 V. G. Bashtovoi, V. K. Polevikov, A. E. Suprun, A. V. Stroots, and S. A.
Beresnev, Magnetohydrodynamics 44(2), 121 (2008).
5 G. K. Batchelor, J. Fluid Mech. 74, 1 (1976).
6 Yu. A. Buevich and A. Yu Zubarev, Colloid J. USSR 51(6), 915 (1989).
7 I. Snook, W. Megen, and R. Tough, J. Chem. Phys. 78(9), 5825 (1983).
8 T. Biben and J.-P. Hansen, J. Phys.: Condens. Matter. 6(23A), A345
(1994).
9 P. M. Biesheuvel and J. Lyklema, J. Phys.: Condens. Matter. 17, 6337
(2005).

10 A.

J. Chem. Phys. 134, 184508 (2011)

O. Tsebers, Magnetohydrodynamics 18(2), 137 (1982).

11 Yu. A. Buevich, A. Yu. Zubarev, and A. O. Ivanov, Magnetohydrodynamics

2, 39 (1989).
O. Ivanov, Doctoral dissertation, Ural State University, Ekaterinburg,
1998.
13 N. Carnagan and K. Starling, J. Chem. Phys. 51(2), 635 (1969).
14 K. I. Morozov, J. Magn. Magn. Mater. 122, 98 (1993).
15 K. I. Morozov, Phys. Rev. E 53(4), 3841 (1996).
16 J.-C. Bacri, A. Cebers, A. Bourdon, G. Demouchy, B. M. Heegard, B.
Kashevsky, and R. Perzynski, Phys. Rev E 52, 3936 (1995).
17 M. I. Shliomis, in Ferrofluids: Magnetically Controllable Fluids and Their
Applications, Lecture Notes in Physics Vol. 594, edited by S. Odenbach
(Springer, Berlin, 2002), p. 85.
18 A. F. Pshenichnikov, V. V. Mekhonoshin, and A. V. Lebedev, J. Magn.
Magn. Mater. 161, 94 (1996).
19 A. O. Ivanov and O. B. Kuznetsova, Phys. Rev. E 64, 041405 (2001).
20 A. O. Ivanov, S. S. Kantorovich, E. N. Reznikov, C. Holm, A. F.
Pshenichnikov, A. V. Lebedev, A. Chremos, and P. J. Camp, Phys. Rev.
E 75, 061405 (2007).
21 A. O. Ivanov, S. S. Kantorovich, E. N. Reznikov, C. Holm, A. F.
Pshenichnikov, A. V. Lebedev, A. Chremos, and P. J. Camp, Magnetohydrodynamics 43(4), 393 (2007).
22 A. F. Pshenichnikov and A. V. Lebedev, J. Chem. Phys. 121(11), 5455
(2004).
23 A. F. Pshenichnikov and A. V. Lebedev, Colloid J. 67(2), 189 (2005).
24 L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media,
2nd ed. (Pergamon, New York, 1984).
25 E. A. Elfimova and A. O. Ivanov, JETP 111(1), 146 (2010).
26 L. Verlet and J.-J. Weis, Molec. Phys. 28(3), 665 (1974).
27 L. D. Landau and E. M. Lifshitz, Fluid Dynamics, 2nd ed. (Pergamon, New
York, 1987).
28 Y. A. Buyevich and A. O. Ivanov, Physica A 190(34), 276 (1992).
29 K. I. Morozov, Izv. Akad. Nauk SSSR, Seriya Fizicheskaya 51(6), 1073
(1987).
30 A. F. Pshenichnikov and V. V. Mekhonoshin, JETP Lett. 72(4), 182
(2000).
31 A. F. Pshenichnikov and V. V. Mekhonoshin, Eur. Phys. J. E 6 399 (2001).
12 A.

The Journal of Chemical Physics is copyrighted by the American Institute of Physics (AIP). Redistribution of
journal material is subject to the AIP online journal license and/or AIP copyright. For more information, see
http://ojps.aip.org/jcpo/jcpcr/jsp