You are on page 1of 11

China Particuology 5 (2007) 50–60

Ferrofluid magnetoviscous control of wall flow


channeling in porous media
Faı̈çal Larachi ∗ , Damien Desvigne
Department of Chemical Engineering, Laval University, Que. G1K 7P4, Canada
Received 18 April 2006; accepted 5 December 2006

Abstract
We analyzed the phenomenon of ferrofluid magnetoviscosity in high-permeability wall-region non-magnetic porous media of the Müller kind.
After upscaling the pore-level ferrohydrodynamic model, we obtained a simplified volume-average zero-order axisymmetric model for non-Darcy
non-turbulent flow of steady-state isothermal incompressible Newtonian ferrofluids through a porous medium experiencing external constant
bulk-flow oriented gradient magnetic field, ferrofluid self-consistent demagnetizing field and induced magnetic field in the solid. The model was
explored in contexts plagued by wall flow maldistribution due to low column-to-particle diameter ratios. It was shown that for proper magnetic field
arrangement, wall channeling can be reduced by inflating wall flow resistance through magnetovisco-thickening and Kelvin body force density
which reroute a fraction of wall flow towards bed core.
© 2007 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V.
All rights reserved.

Keywords: Ferrofluid; Magnetosviscosity; Porous medium; Volume-average ferrohydrodynamic model; Kelvin body force; Spin-vorticity coupling

1. Introduction the deconstructing thermal agitation nor shear flow, thus main-
taining the nanoparticles afloat and unassociated. In addition,
Ferrofluids are non-naturally occurring magnetic liquid sus- should the ferrofluid motionless state be restored, the non-
pensions obtained by seeding surfactant-wrapped single-domain zero magnetization relaxation time will impede instantaneous
superparamagnetic nanoparticles in appropriate organic or aque- directional readjustment of magnetization vector towards equi-
ous carrier matrices (Odenbach, 2003; Rinaldi & Zahn, 2002; librium.
Rosensweig, 1997). Such stabilized ferrocolloidal suspensions When ferrofluids are brought to motion in shear flows,
exhibit a number of intriguing behaviors when subjected to misalignment between “dynamic” magnetization M and total
external – uniform, spatially inhomogeneous or oscillating – magnetic field H vectors arises due to an asynchrony between
magnetic fields. Owing to a giant single-domain magnetic the nanoparticle spin ␻ and the ferrofluid vorticity, ½ × v
moment, ca. 104 Bohr magnetons, each nanoparticle stands as a (Rosensweig, 1997). For illustration purposes, the total magnetic
small permanent magnetic dipole. Collective response of these field vector and the fluid vorticity vector are assumed orthogonal
nanoparticles to magnetic field stimulation gives rise to strong as depicted in Fig. 1, and in first approximation, both angu-
and coherent magnetization which, for motionless ferrofluids, lar acceleration and diffusive couple dyadic contributions are
aligns along the direction of the total magnetic field, and which ignored. Hence, this spin-vorticity asynchrony gives rise to a
obeys the equilibrium Langevin magnetic-field-locked magne- mechanical torque that will tend to pull the dynamic magnetiza-
tization law. In this study, we will make the assumptions that tion vector out from the direction of equilibrium (or equivalently,
the Brownian relaxation outweighs the Néelian relaxation (i.e., the direction of the local total magnetic field at motionless state).
magnetic moment spatially locked to nanoparticle) and that the Non-collinearity between dynamic magnetization and magnetic
magnetic dipole interaction energy is unable to oppose neither field in return results in a counteracting magnetic torque, which
will try to pull in the magnetization vector back to its equilib-
rium position. The mechanical torque is proportional to a vortex
∗ Corresponding author. Tel.: +1 418 656 3566; fax: +1 418 656 5993. viscosity, which proportionates the magnetic torque response.
E-mail address: faical.larachi@gch.ulaval.ca (F. Larachi). Two magnetoviscous behaviors, as a consequence, could arise

1672-2515/$ – see front matter © 2007 Chinese Society of Particuology and Institute of Process Engineering, Chinese Academy of Sciences. Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.cpart.2006.12.001
F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60 51

Nomenclature V bounded open set representing the averaging


region
a coefficient in the Müller porosity model |V| volume of the averaging region V (m3 )
b coefficient in the Müller porosity model
V␥ open set region within V occupied by the fer-
B magnetic flux density (T)
rofluid
BPw wall by-pass fraction |V␥ | volume of V␥ (m3 )
C symmetric couple stress dyadic (kg/s2 )
V␬ union of a finite number of open sets within V
d grain diameter (m) occupied by porous media (granular)
D column diameter (m) x position vector locating in the macroscale frame,
dh hydrodynamic diameter of coated nanoparticle the centroid of the averaging volume region V (M)
(m) y relative position vector in the microscale frame of
dm nanoparticle diameter of magnetic core (m) a point in V (M)
eθ unit vector in tangential direction z axial coordinate in the column (m)
E-- antisymmetric triadic
f0 Larmor frequency (Hz) Greek letters
G body couple density vector (Pa) Γ ␥␬ Ferrofluid-solid interface inside V
g gravity vector (m/s2 ) Ξ demagnetization LMM Correction
h self-consistent demagnetizing field vector α gradient programming of the applied magnetic
induced inside the magnetized ferrofluid mat- field (A/m2 )
ter and the induced magnetic field in the β offset constant of the applied magnetic field (A/m)
non-magnetic granular phase (A/m) χ0 initial magnetic susceptibility
he equilibrium demagnetizing field vector prevail- χ␣ magnetic susceptibility of ␣-phase
ing in the ferrofluid phase at magnetohydrostatic χ␬ magnetic susceptibility of non-magnetic material
conditions (A/m) making up the granular phase
H dynamic total magnetic field vector (A/m) εb coefficient in the Müller porosity model
He magnetohydrostatic total magnetic field vector ε␥ porous medium porosity
(A/m) φh hydrodynamic volume fraction of the surfactant-
H0 externally applied magnetic field vector (A/m) dressed nanoparticles
I moment of inertia density of the magnetic φm suspension volume fraction of magnetic material
nanoparticles (m2 ) embedded in the nanoparticles
I idemfactor dyadic γ␣ gravitational amplification factor of ␣-phase
k Boltzmann constant (J/K) η dynamic viscosity (Pa s)
K anisotropy constant of the single-domain uniaxial η0 dynamic viscosity of the liquid carrier (Pa s)
ferromagnetic nanoparticle (J/m3 ) η shear spin viscosity (kg m/s/rad)
L column length (m) ␸␥ ␥-phase-side values of the variable ␸ on Γ ␥␬
m̄ magnetic dipole moment of an individual ␸␬ ␬-phase-side values of the variable ␸ on Γ ␥␬
nanoparticle (A/m) ϕ̃ spatial fluctuation of the bounded piecewise con-
M ferrofluid dynamic magnetization vector (A/m) tinuous scalar field ϕ around the intrinsic average
My magnetic stress tensor in ␥-phase in Vγ ∪ Γγκ
n␥␬ normal canonical unit vector associated with con- ϕ̃ spatial fluctuation of the bounded piecewise con-
tour (␥␬ ) tinuous vector field ϕ around the intrinsic average
P pressure (Pa) in Vγ ∪ Γγκ
r radial coordinate in the column (axisymmetry) λ second coefficient of viscosity (Pa s)
(m) λ bulk spin viscosity (kg m/s/rad)
R column radius (m) μ0 absolute magnetic permeability of vacuum,
R position vector locating a point in V (m) 4␲ × 10−7 (N/A2 )
t time (s) ρ density (kg/m3 )
T absolute temperature (K) ρs dressing surfactant density (kg/m3 )
T pressure-viscous-spin/vorticity stress tensor (Pa) ρP magnetic core density (kg/m3 )
U0 ferrofluid superficial velocity (inlet) (m/s) υ␣ kinematic viscosity of ␣-phase (m2 /s)
v ferrofluid local instantaneous velocity vector τ effective relaxation time of magnetization (s)
(m/s) τB Brownian relaxation time (s)
∂V V associated entrance/exit contours τN Néelian relaxation time (s)
∂V␥e ␥-phase associated entrance/exit contours ␻ spin density vector (rad/s)
∂V␬e ␬-phase associated entrance/exit contours ζ vortex viscosity (Pa s)
52 F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60

contrast with preferential flow short-circuiting alongside the


Subscripts high-permeability wall area. The challenge will be to identify
l Langevin which configurations are propitious to creating sufficient mag-
r radial projection of variable netoviscosity near the vessel wall to hammer the bypass and to
z axial projection of variable reroute maximum fraction of fluid flow towards bed core.
␥ phase related to ferrofluid High fluid vorticity is encountered near the vessel wall where
␬ phase related to granular porous media an unobstructed gross wall flow develops. Conversely, bed core
θ azimuthal projection of variable vorticity is the lowest due to porosity distribution damping,
which implies weaker local velocity variations. It is anticipated
Tensor operations that application of magnetic fields to ferrofluids would yield skin
t transpose magnetoviscothickening mainly located in the wall region. It is

|| || Euclidian norm (for vectors): ϕ||ϕ|| = ϕ• ϕ hoped that nanoparticle spin retardation would therefore arti-
⊗ Tensor product ficially increase resistance to flow in this region thus yielding
: double contracted product lower wall bypass fraction. One simple arrangement to induce
• dot product or contracted product magnetoviscosity corresponds to axisymmetric magnetic field
× curl product and geometry to have magnetic torque, vorticity and spin veloc-
 gradient ity orthoradial (Fig. 1). To sharpen the contrasts in the velocity
• divergence profiles owing to the Kelvin body force, a divergenceless exter-
× curl nal magnetic field with positive and negative constant-gradient
 Laplacian streamwise component is chosen.
Magnetoviscosity is analyzed by formulating the general
local ferrohydrodynamic model, the general upscaled volume-
average model in porous media with the closure problem, and
depending on magnetic field and ferrofluid vorticity magnitudes solution and discussion of a simplified zero-order steady-state
and mutual orientations with respect to wall. isothermal incompressible axisymmetric model for non-Darcy
Magnetoviscothickening, a magnetically driven apparent flow of a Newtonian ferrofluid in a Müller porous medium sub-
inflation in local ferrofluid suspension viscosity, occurs when ject to a total magnetic field accounting for bulk-flow oriented
the fluid vorticity and the magnetized nanoparticle spin are in external field H0 , ferrofluid self-consistent demagnetizing field
contrarotation or are in corotation with fluid vorticity faster than h and induced magnetic field h in the non-magnetic granu-
the nanoparticle spin velocity (Fig. 1a). The retarding effect by lar phase. Shliomis (1972) magnetization relaxation equation
the nanoparticle spin velocity results locally in an increased dis- has been written in the fluid compressible form according to
sipation in the flow and, eventually, a slow down in fluid linear Felderhof (2001). Both external and total magnetic fields are
velocity. Magnetoviscothickening has been reported for Hagen- taken to be irrotational (zero free-current density in packed bed,
Poiseuille and turbulent pipe flows with steady and oscillating i.e., non-conducting ferrofluid and porous medium).
spatially homogeneous magnetic fields (Schumacher, Sellien,
Knoke, Cader, & Finlayson, 2003). Similarly, Magnetoviscoth- 2. Problem statement
inning, tantamount to an apparent drop off in local ferrofluid
suspension viscosity, occurs when fluid vorticity and magne- 2.1. Local description
tized nanoparticle spin are in corotation with the nanoparticle
spin velocity faster than the fluid vorticity (Fig. 1b). Part of the Isothermal and incompressible single-phase Newtonian fer-
nanoparticle internal angular momentum is transmitted to the rofluid (␥-phase) upflow in a non-magnetic (␬-phase) rigid
fluid deflating fluid dissipation as in drag reduction reminiscent Müller porous medium (Fig. 2c) is assumed. Also, a pseudo-
of polymeric behaviors. This phenomenon has been observed in Coulombic approach of the coupled magnetostatic problem is
laminar pipe flow with high-frequency oscillating linearly polar- used. The Cauchy (or boundary-value) problem describing flow
ized magnetic fields (Bacri, Perzynski, Shliomis, & Burde, 1995; in the macroscopic region V, needed for spatial averaging, is
Zeuner, Richter, & Rehberg, 1998). An intermediate case, corre- stated in Table 1. The (macroscopic region) open set V, bounded
sponding to classical fluid mechanics situation, also exists when by the ∂V␥e and ∂V␬e phase-associated entrance/exit contours
no spin lag between fluid element and magnetized nanoparticle on V faces, is composed of the ferrofluid part, V␥ , the porous
yields null magnetic and hydrodynamic torques. This cancels medium part, V␬ , and the fluid-solid interface, Γ ␥␬ , inside V
the internal angular momentum equation and eliminates spin (Fig. 2b): V = V␥ + V␬ + Γ ␥␬ . For the sake of brevity, the bound-
coupling with linear momentum balance. ary conditions on ∂V = ∂V␥e + ∂V␬e of the velocity (v␥ ), spin
We propose to explore in this paper potential applications (␻␥ ), magnetization (M␥ ), induced magnetic fields (h␥ , h␬ )
of magnetoviscosity in chemical engineering. The demonstra- and the initial conditions are omitted, even though these are
tion example will highlight the control of wall channeling usually not known a priori. We will postulate, without proof,
in porous media exhibiting low column-to-particle diameter that there exists for our model porous medium, a spatially peri-
ratio. It is well known that such low-ratio columns may occa- odic representative elementary volume REV of the order of the
sion severe maldistribution due to large radial permeability macroscopic volume, such that ignorance of these boundary
F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60 53

Fig. 1. Illustration of the magnetoviscosity effect: (a) magnetoviscothickening and (b) magnetoviscothinning.

conditions does not harm the closure problem of the volume- and vortex viscosities, η and ζ, suspension volume fraction, φh ,
averaged equations (Whitaker, 1999). Table 1 summarizes the scalar moment of inertia density, I, fluid initial magnetic suscep-
continuity, linear momentum, internal angular momentum, mag- tibility, χ0 , linear magnetic material (LMM) approximation for
netization relaxation, Maxwell-flux law and Ampère-Maxwell demagnetizing and induced fields, Ξ, porosity model, ε, external
law local equations in V, along with Γ ␥␬ boundary conditions magnetic field, H0 , volume-average Langevin magnetization,
and constitutive tensor and vector equations. Table 2 gathers the m , under resultant external and equilibrium demagnetizing
auxiliary equations for the relaxation time, τ, ferrofluid dynamic fields, and wall bypass fraction, BPw .

Table 1
Local ferrohydrodynamic model
Continuity ∇ • v␥ = 0 in V␥ (1)

Linear momentum ρv + ∇ • ρv␥ ⊗ v␥ = ∇ • T + ∇ • M + ρg in V␥ (2)
∂t ␥ ␥ ␥

Internal angular momentum ρI␻␥ + ∇ • ρI␻␥ ⊗ v␥ = ∇ • C − E : T + G␥ in V␥ (3)
∂t ␥ ␥

Magnetization relaxation M␥ + ∇ · M␥ ⊗ v␥ = ␻␥ × M␥ − τ −1 (M␥ − Ml ) in V␥ (4)
∂t
Maxwell-flux law ∇ • μ0 (H0 + h␥ + M␥ ) = 0 in Vτ (5)
∇ • μ0 (1 + χ␬ )(H0 + h␬ ) = 0 in V␥ (6)
∇ • μ0 H 0 = 0 in V (7)
Ampère-Maxwell law ∇ × (H0 + h␥ ) = 0 in V␥ (8)
∇ × (H0 + h␬ ) = 0 in V␬ (9)
∇ × H0 = 0 in V (10)
Boundary conditions on fluid-solid interface v␥ = 0 on Γ␥␬ (11)
Γ ␥␬ inside macroscopic region V
(h␬ − h␥ ) × n␥␬ = 0 on Γ␥␬ (12)
(h␬ − h␥ ) • n␥␬ = M␥ • n␥␬ on Γ␥␬ (13)
μ0
Constitutive relations ω␥ = M × (H0 + h␥ )on Γ␥␬ (14)
4ζ ␥
␻␥ × M␥ = τ −1 (M␥ − Ml )on Γ␥␬ (15)
Pressure-viscous stress tensor T = −p␥ I + η(∇v␥ + t ∇v␥ ) + λ∇ • vγ I + ζE · (∇ × vγ − 2ωγ ) (16)

μ0
Magnetic stress tensor M = − ||H0 + h␥ ||2 I + μ0 (H0 + h␥ + M␥ ) ⊗ (H0 + h␥ ) (17)
␥ 2
Couple stress dyadic C = η (∇␻␥ +t ∇␻␥ ) + λ ∇ • ␻␥ I (18)

External–internal exchange angular momentum −E : T = 2ζ(∇ × v␥ − 2␻␥ ) (19)

Body couple density G␥ = μ0 M␥ × (H0 + h␥ ) (20)
  π μ m̄  6 kT
 H + he
3 0 e −1 0 ␥
Langevin equilibrium (e) magnetization Ml = φm m̄ coth dm ||H0 + he␥ || − ||H + h␥ || (21)
6 kT 3 μ m̄
πdm 0
0
||H0 + he␥ ||
54 F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60

from a description using the local ferrohydrodynamic equations.


We have used the spatial averaging theorem and its corollaries,
the theorem of derivative-integral interchange, a spatial decom-
position around the intrinsic averages and truncation at the
zeroth order of the Taylor expansion around the macroscopic
volume centroid (Fig. 2b) of the non-localized intrinsic aver-
ages to build the volume-average ferrohydrodynamic model.
Without going into the details of the procedure, the resulting
volume-average equations contain several integrals of the spa-
tially fluctuating microscale terms that need to be approximated
regarding velocity, spin density, magnetization, external mag-
netic field, ferrofluid demagnetizing field and induced magnetic
field in granular phase. The macroscopic ferrohydrodynamic
model with the unknown closure terms is summarized in Table 3.
Note that this model applies to the points belonging to the packed
bed interior domain excluding the vessel peripheral boundaries,
Fig. 2. Porous medium details. which need appropriate boundary conditions. Formulation of the
15 model closure terms is a task of formidable complexity, and
2.2. Upscaling to the authors’ best knowledge, such closures have not been yet
formulated for porous media flows.
Upscaling is an inescapable prerequisite step allowing the
passage from complex and still out-of-reach pore level solutions 2.3. Closure problem
of the above physical phenomena. Upscaling consists in formu-
lating an equivalent set of volume-average equations describing Assuming as in Schumacher et al. (2003) that spin viscosity
the ferrofluid macroscopic behavior at the porous medium level shear and bulk coefficients are zero relaxes four closure terms

Table 2
Auxiliary equations used in ferrohydrodynamic model
1 1 1 2kT 3
Relaxation time = + = + f0 e−K/kTπdm /6 (22)
τ τB τN πdh3 η0
 

−1
η 5 dn − dm 3 dn − dm 6
Ferrofluid suspension viscosity = 1 − φm 1+ − 1.552φm 1 + (23)
η0 2 dm dm
Hydrodynamic-to-magnetic volume fraction φh dm = φm dh
3 3
(24)
5 + ρ (d 5 − d 5 )
1 ρp dm s h m
Scalar moment of inertia density I= (25)
3
10 ρp dm + ρs (dh3 − dm
3
)
Vortex viscosity ζ = 1.5ηφh (26)
π φm μ0 m̄2 dm3
Initial magnetic susceptibility χ0 = (27)
18 kT
LMM approximation correction Ξ = χ0 (3 + 2χ0 )−1 (28)
a(D − 2r)

Bed porosity radial distribution (Müller model) ε(r) = εb + (1 − εb )J0 e−b(D−2r)/2d (29.1)
2d
(−1)i r2i

J0 (r) = (29.2)
22i (i!)2
i=0
d
εb = 0.365 + 0.22 (29.3)
⎧ D
⎨ 7.45 − 3.15 d D ∈ [2.02 − 13.0]
a= D d (29.4)
⎩ 7.45 − 11.25 d D ≥ 13.0
D d
d
b = 0.315 − 0.725 (29.5)
r D

External magnetic field H 0= − αt


0 αz + β (30)
2  π μ m̄  
6 kT
Volume-average Langevin magnetization ||ml || = φm m̄ coth 3 0
dm ||H0 + he || − ||H0 + he ||−1 (31.1)
 6 kT πdm μ0 m̄
3

H0r + her H0z + hez


ml = t
||ml ||0 ||ml || (31.2)
||H0 + he || ||H0 + he ||
 D/2  D/2 −1

Wall bypass fraction BPw = rε(r)vz (r) dr( rU0 dr) (32)
D/2−d/2 D/2−d/2
F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60 55

Table 3
Complete volume-average ferrohydrodynamic model
Volume conservation ε␥ + ε␬ = 1 (33)
 
Continuity ∇ · ε␥ v␥ =0 (34)

∂        
Linear momentum ρ ε␥ v␥ + ρε␥ ∇ v␥ • v␥ + ρ∇ • ṽ␥ ⊗ ṽ␥ =
∂t ␥ ␥ ␥ 
    ε␥     ∇ε␥   1
−ε␥ ∇ p␥ + ε␥ ρg + (η + ζ)ε␥ Δ v␥ + v␥ + ∇ v␥ • + 2ζε␥ ∇ × ␻␥ + 2ζ n ×
␥ ␥ ε␥ ␥ ␥ ε␥ ␥ |V | Γ ␥␬
   γκ

     
1

˜ ␥ dσ + μ0 ∇ H̃ 0 + h̃␥ • M̃ ␥ + μ0 ε␥ ∇ H 0 + h␥ + (H̃ 0 + h̃␥ ⊗ n␥␬ dσ) •
␥ |V |
 Γγκ
  1
M␥ (−p̃␥ I + (η + ζ)∇ ṽ␥ ) • n␥␬ dσ (35)
␥ |V |
Γγκ
∂        
Internal angular momentum ρI ε␥ ␻␥ + ρI∇ ␻␥ • ε␥ v␥ + ρI∇ • ␻ ˜ ␥ ⊗ ṽ␥ =
∂t  ␥ ␥ ␥
  ∇ε␥           
+2ζε␥ ∇ × v␥ + × v␥ − 2 ␻␥ + ε␥ μ0 M ␥ × (H 0 + h␥ ) + μ0 M̃ ␥ × H̃ 0 + h̃␥ +
␥ ε␥ ␥ ␥ ␥ ␥
   

   
    1   
 1
η ε␥ Δ ω␥ + (η + λ )ε␥ ∇∇ • ␻␥ + (η + λ )∇ ˜ ␥ • n␥␬ dσ
␻ + η +λ ˜ ␥ n␥␬ dσ + η ∇ •
∇ •␻
␥ ␥ |V | |V |
   
Γγκ Γγκ

1 1
˜ ␥ ⊗ n␥␬ dσ
␻ + η ∇␻
˜ ␥ • n␥␬ dσ (36)
|V | Γγκ
|V | Γγκ
∂            
Magnetization relaxation ε␥ M ␥ + ∇ε␥ M ␥ • v␥ + ε␥ M ␥ ∇ • v␥ + ∇ • M̃ ␥ ⊗ ṽ␥ =
∂t   ␥ ␥ ␥ ␥ ␥
    ε␥    
+ε␥ ␻␥ × M ␥ + ω̃␥ × M̃ ␥ − ( M ␥ − M l ) (37)
␥ ␥ τ ␥ ␥

     
Maxwell-flux law ∇ • (ε␥ h␥ + ε␥ M ␥ + ε␬ h␬ )=0 (38)
␥ ␥ ␬

  1
ε␬ ∇ • h␬ − h̃␬ • n␥␬ dσ = 0 (39)
␬ |V | Γ␥␬
∇ • H0 = 0 (40)

   
Ampère-Maxwell law ∇ × (ε␥ h␥ + ε␬ h␬ )=0 (41)
␥ ␬

  1
ε␬ ∇ × h␬ − n␥␬ × h̃␬ dσ = 0 (42)
␬ |V | Γ␥␬
∇ × H0 = 0 (43)

   
LMM approximation h␬ = (1 + Ξ) h␥ (44)
␬ ␥

in (36). In addition, the integrand of the last closure term in virtually opens onto a deadlock when attempting to infer a
the linear momentum balance equation (35) contains the sum mean-field effective magnetic permeability of the composite
of the ferrofluid dynamic and vortex viscosities multiplied by ferrofluid-granular system. When fluid and solid magnetic per-
the gradient of the fluctuating velocity on Γ ␥␬ . This suggests meabilities do not differ much and a linear magnetic materials
that unlike classical drag formulation, the drag function here (LMM) behavior is valid (case of low magnetic field linear-
must account for the occurrence in the laminar Darcy or Ergun Langevin limit), the effect of contacts between non-magnetic
equation term of the total viscosity rather than only the dynamic grains is not very important and the Maxwell-Garnett theory
viscosity. can be used (Khuzir, Bossis, Bashtovoi, & Volkova, 2003).
The closures (39) and (42) hide in the granular induced mag- This theory assumes random distribution of grains in a car-
netic field vector a monumental complexity associated with rier medium subject to a local Lorentz field. For spheres in
the hydrodynamic perturbation of the ferrofluid dynamic mag- densely packed beds with LMMs, the previous closures can
netization vector as well as the non-linear Langevin equation be replaced by (44) which relate the ferrofluid demagnetiz-
between equilibrium magnetization, external and equilibrium ing field and the induced magnetic field in the granular phase.
demagnetizing fields ((31.1)–(31.2)). In the general case, this Due to lack of better alternatives, we extend in this work the
56 F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60

LMM approximation even to the non-linear region. This arti- ing the modified Ergun drag function in (45) and (46). Solution
fice allows ignoring the closures associated with (39) and (42). of the coupled partial differential and integral equations is per-
In addition, the solid magnetic susceptibility, χ␬ , is neglected formed using a finite difference scheme on an Aspen Custom
with respect to the initial ferrofluid magnetic susceptibility, Modeler platform. Discretization meshes along r and z direc-
χ0 , yielding the LMM approximation correction (28). In addi- tions are, respectively, 0.2 and 4 mm. To ensure convergence of
tion, since ferrofluid magnetization is the chief determinant the numerical scheme, both fast-Newton method for non-linear
of both solid induced magnetic and ferrofluid demagnetizing solver with convergence criterion on residuals, and a MA48 lin-
fields, these should remain small with respect to the external ear solver are used. Typical simulation duration is 30 min on a
field to attempt further simplifications to be discussed next. 3 GHz CPU and 1.5 Giga octet RAM computer with a 10−5 for
Closure of the eight remaining terms entails mathematical com- the absolute equation tolerance.
plications and physical efforts beyond the scope of this study. It is noteworthy that the ferrofluid velocity and nanopar-
Rather, we will restrict to solve the zero-order formulation and ticle spin density fields are invariant to switching H0 into
leave the complete formulation as an open problem for future −H0 because the induced magnetic field in the solid and the
research. demagnetizing field in the ferrofluid have been neglected in the
ferrohydrodynamic bloc of equations. For consistency purposes,
the H0 z-projection is always positive. Furthermore, parameter
2.4. Zero-order axisymmetric volume-average model
β of the magnetic field (30) is chosen such that H0z = 0 at the
entrance (positive gradient) or at the exit (negative gradient).
Assuming that all the variables in the physical problem are
Fig. 3 illustrates the evolution of the fractional bypass nor-
independent of the azimuthal coordinate (axisymmetry), the pro-
malized with respect to the bypass when H0 = 0 as a function of
jections of Table 3 model equations along the three (r, θ, z)
the z-component external magnetic field positive and negative
coordinate axes are given in Table 4 together with the boundary
gradients. By convention, the external magnetic field is always
conditions chosen at the four peripheral boundaries of the porous
bulk-flow oriented. The volume-average asynchrony, defined
medium: z = 0, z = L, r = 0, r = R. We will assume without proof,
earlier, corresponds, regardless of magnetic gradient sign, to
uniqueness of solution for the system of equations describing
Fig. 1a case. Therefore, there is confirmation of a magnetovis-
this ferrohydrodynamic model.
cothickening effect, which tends to reduce the preferential flow
The pressure field is assumed to depend only on the axial
alongside the wall. The maximum reduction could be as low as
coordinate. The external magnetic field is purposely chosen to
18% at the highest negative gradient. The non-symmetry of the
exhibit only radial and axial dependences while fulfilling the
bypass with respect to the vertical axis, which is more severe for
divergenceless condition, thus H0θ = 0 ((40), (57.1)). This entails
negative gradient than for positive gradient, is due to a global
that azimuthal component of the equilibrium magnetization vec-
outcome of magnetoviscosity and Kelvin body force density.
tor is zero. In addition, the boundary conditions ((69)–(80),
Because of the gross approximations made in Section 2.4 regard-
(93)–(96)) as well as ((48)–(50), (54)) are verified by the trivial
ing the demagnetizing field and its gradient with respect to the
set ωr = ωz = 0, mθ = 0 and vθ = 0. These are taken as solutions for
external magnetic field and its gradient, the curve in the vicinity
our model. As for the equilibrium magnetization on eθ direction,
of dH0z /dz = 0 has to be considered with caution because of the
the last equality mθ = 0 is also coherent with H0θ = 0. Therefore,
approximation of the neglect of the demagnetizing field under
according to ((31.1)–(31.2)), it is not unrealistic to assume also
low external magnetic fields.
that hθ = 0 (57.2).
Preliminary simulations reveal that typically
||h|| < 10%||H0 || suggesting a further simplifica-
tion of the model towards decoupling magnetostatics
((55)–((58.1)–(58.3))) from the ferrohydrodynamic equa-
tions (45)–(54). Strictly speaking, this requires that the norm
1/2 1/2
condition (t ∇ h- : ∇ h- ) << (t ∇ H : ∇ H ) , due to the
- - - -0 - -0
Kelvin force terms in (46), (47) must also be verified; which
is always the case except near the wall, the entrance (positive
gradient) and the exit (negative gradient) regions at low
magnetic fields.

3. Implementation and discussion

Table 5 contains the simulation parameters of the simplified


zero-order axisymmetric volume-average ferrohydrodynamic
model of Table 4. Note that a fluid superficial velocity of 0.2 m/s
is chosen such that ferrofluid inertia cannot be ignored (non-
Darcy flow). The consensual values for the laminar and inertial Fig. 3. Normalized bypass fraction for various z-component external magnetic
Ergun constants are used, i.e., respectively, 150 and 1.75; yield- field gradients.
F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60 57

Table 4
Simplified zero-order axisymmetric volume-average ferrohydrodynamic model
 D/2
U0 D2
Continuity ε vz r dr =
8

0
(45)
∂vr v2 ∂vr
Linear momentum (radial) ρ vr − θ + vz =
∂r r ∂z
 
∂2 vr 1 ∂vr vr ∂2 vr vr d2 ε 1 dε 1 dε ∂vr ∂ωθ
(η + ζ) + − 2 + 2 + + + − 2ζ +
∂r2 r ∂r r ∂z ε dr2 r dr ε dr ∂r ∂z
μ0
dH ∂hr

∂hr

0r
mr + + mz
ε
vr 1 − ε
dr ∂r
η+ζ 1−ε
∂z
ρ 2

− 150 2 + 1.75 vr + v2z + v2θ


d ε d ε d
∂v
  (46)
z ∂vz ∂2 vz 1 ∂vz ∂2 vz vz d2 ε 1 dε 1 dε ∂vz
Linear momentum (axial) ρ vr + vz = (η + ζ) + + + + + +
∂r ∂z ∂r2 r ∂r ∂z2 ε dr2 r dr ε dr ∂r
∂ω ωθ

μ0
dH0z ∂hz

∂hz

θ
2ζ + + mz + + mr −
∂r r ε dz ∂z
∂r dp
vz 1 − ε η+ζ 1−ε ρ 2
150 2 + 1.75 vr + v2z + v2θ − − ρg
d ε d ε d dz
∂v vθ

∂vθ

(47)
θ
Linear momentum (azimuthal) ρ vr + + vz = (η +
 ∂r r ∂z  ∂ω

2
∂ vθ 1 ∂vθ vθ 2
∂ vθ vθ d2 ε 1 dε 1 dε ∂vθ r ∂ωz
ζ) + − + + + + + 2ζ − +
∂r2 r ∂r r2 ∂z2 ε dr2 r dr ε dr ∂r ∂z ∂r
μ0
H0r hr

vθ 1 − ε
η+ζ 1−ε ρ 

mθ + − 150 2 + 1.75 v2r + v2z + v2θ


ε r r d ε d ε d
∂ω ωθ ∂ωr

∂vθ

μ0
(48)
r
Internal angular momentum (radial) ρI vr − vθ + vz = −2ζ + 2ωr + mθ (H0z + hz )
∂r r ∂z ∂z ε
∂ω ∂ωz

vθ ∂vθ vθ dε

μ0
(49)
z
Internal angular momentum (axial) ρI vr + vz = 2ζ + + − 2ωz − mθ (H0r + hr )
∂r ∂z ε dr
∂ω ωr ∂ωθ

r ∂r ε (50)
θ
Internal angular momentum (azimuthal) ρI vr + vθ + vz =
∂v ∂r ∂v r v dε ∂z
μ
r z z 0
2ζ − − − 2ωθ + (mz (H0r + hr ) − mr (H0z + hz ))
∂z ∂r ε dr ε
∂mr mθ ∂mr
vr ∂vr ∂vz

(51)
Magnetization relaxation (radial) vr − vθ + vz + mr + + = ωθ mz − ωz mθ − τ −1 (mr − εmr ) (52)
∂r
∂mz
r
∂mz
v
∂z
∂vr
r
∂vz
∂r
∂z
= −ωθ mr + ωr mθ − τ −1 (mz − εmlz )
r
Magnetization relaxation (axial) vr + vz + mz + + (53)
∂r
∂m m∂zr ∂mθ r
v ∂z∂v
∂r ∂vz

= ωz mr − ωr mz − τ −1 mθ
θ r r
Magnetization relaxation (azimuthal) vr + vθ + vz + mθ + +
∂r r ∂z r ∂r
mr ∂mr ∂mz
h ∂hr ∂hz

∂z dε
(54)
r
Maxwell-flux law (demagnetizing field) + + = −ε + + + Ξ(H0r + hr ) − (1 −
r ∂r ∂z r
∂r ∂z

dr
H0r ∂H0r ∂H0z hr ∂hr ∂hz
ε) Ξ + + + (Ξ + 1) + +
r ∂r ∂z r ∂r ∂z
∂h ∂hz


(55)
r
Ampère-Maxwell law (azimuthal) (1 + Ξ(1 − ε)) − + Ξ(H0z + hz ) =0
∂z ∂r dr (56)
Azimuthal external magnetic field H0θ = 0 (57.1)

Azimuthal demagnetizing field hθ = 0 (57.2)

Induced magnetic field in granular phase hr = (1 + Ξ)hr (58.1)

hz = (1 + Ξ)hz (58.2)

hθ = 0 (58.3)

d2 dp
Boundary conditions on packed bed (0) = 0
dz2 dz
domain peripheral boundaries (59)
d dp
(L) = 0
dz dz
  (60)
vr (r, 0) = 0 r ∈ 0; D/2
(61)
∂vr
(r, L) = 0 r ∈ [0; D/2]
∂z (62)
58 F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60

Table 4 (Continued )
vr (0, z) = 0 z ∈ ]0; L[ (63)
vr (D/2, z) = 0 z ∈ ]0; L[ (64)
vz (r, 0) = Uo r ∈ [0; D/2] (65)
∂vz
(r, L) = 0 r ∈ [0; D/2] (66)
∂z
∂vz
(0, z) = 0 z ∈ ]0; L[ (67)
∂r
vz (D/2, z) = 0 z ∈ ]0; L[ (68)
vθ (r, 0) = 0 r ∈ [0; D/2] (69)
∂vθ
(r, L) = 0 r ∈ [0; D/2] (70)
∂z
vθ (0, z) = 0 z ∈ ]0; L[ (71)
vθ (D/2, z) = 0 z ∈ ]0; L[ (72)
ωr (r, 0) = 0 r ∈ [0; D/2] (73)
ωr (r, L) : Extension by continuity (49) r ∈ [0; D/2] (74)
ωr (0, z) = 0 z ∈ ]0; L[ (75)
ωr (D/2, z) = 0 z ∈ ]0; L[ (76)
ωz (r, 0) = 0 r ∈ [0; D/2] (77)
ωz (r, L) : Extension by continuity (50) r ∈ [0; D/2] (78)
∂ωz
(0, z) = 0 z ∈ ]0; L[ (79)
∂r
ωz (D/2, z) = 0 z ∈ ]0; L[ (80)
ωθ (r, 0) = 0 r ∈ [0; D/2] (81)
∂2 ωθ
(r, L) = 0 r ∈ [0; D/2] (82)
∂z2
ωθ (0, z) = 0 z ∈ ]0; L[ (83)
ωθ (D/2, z) : Extension by continuity (51) z ∈ ]0; L[ (84)
mr (r, 0) = 0 r ∈ [0; D/2] (85)
mr (r, L) : Extension by continuity (52) r ∈ ]0; D/2] (86)
mr (0, z) = 0 z ∈ ]0; L[ (87)
mr (D/2, z) : Extension by continuity (52) z ∈ ]0; L[ (88)
mz (r, 0) = 0 r ∈ [0; D/2] (89)
mz (r, L) = Extension by continuity (53) z ∈ ]0; D12] (90)
∂mz
(0, z) = 0 z ∈ ]0; L[ (91)
∂r
mz (D/2, z) : Extension by continuity (53) z ∈ ]0; L[ (92)
mθ (r, 0) = 0 r ∈ [0; D/2] (93)
mθ (r, L) : Extension by continuity (54) r ∈ ]0; D/2] (94)
mθ (0, z) = 0 z ∈ ]0; L] (95)
mθ (D/2, z) : Extension by continuity (54) z ∈ ]0; L] (96)
hr (r, 0) = 0 r ∈ [0; D/2] (97)
hr (r, L) : Extension by continuity (55) r ∈ ]0; D/2] (98)
hr (0, z) = 0 z ∈ ]0; L] (99)
hr (D/2, z) : Extension by continuity (55) z ∈ ]0; L[ (100)
hz (r, 0) = 0 r ∈ [0; D/2] (101)
hz (r, L) : Extension by continuity (56) r ∈ ]0; D/2] (102)
∂hz
(0, z) = 0 z ∈ ]0; L] (103)
∂r
hz (D/2, z) : Extension by continuity (56) z ∈ ]0; L[ (104)
F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60 59

Table 5
Selected numerical property values used in ferrohydrodynamic model
Anisotropy constant K (kJ/m3 ): 23 Ferrofluid viscosity η (mPa s): 0.2 Relaxation, Néelian τ N (␮s): 18.4
Bed height L (cm): 10 Ferrofluid vortex viscosity ζ (mPa s): 0.07 Second coefficient of viscosity λ (Pa s): 0
Boltzmann constant k (yJ/K): 13.8 Initial magnetic susceptibility χ0 : 1.28 Shear/bulk spin viscosities η , λ (kg m/s/rad): 0
Demagnetization LMM correction Ξ: 0.23 Larmor frequency f0 (GHz): 1 Superficial ferrofluid velocity U0 (m/s): 0.2
Diameter of column D (cm): 2 Magnetic particle density ρp (kg/m3 ): 5300 Surfactant density ρS (kg/m3 ): 500
Diameter, hydrodynamic coated dh (nm): 28 Liquid carrier viscosity η0 (mPa s): 0.1 Temperature T (K): 300
Diameter, magnetic (uncoated) dm (nm): 15 Magnetic dipole moment m̄ (MA/m): 0.45 Volume fraction, hydrodynamic φh (%): 23.4
Diameter of packing d (mm): 2 Moment of inertia density I (am2 ): 45 Volume fraction, magnetic φm (%): 3.6
Ferrofluid density ρ (kg/m3 ): 1056 Relaxation, Brownian τ B (␮s): 0.83 Vacuum permeability μ0 (␮N/A2 ): 26

Fig. 4 shows the outlet pressure gradient normalized with


respect to the pressure gradient at H0 = 0 for the conditions sim-
ulated for Fig. 3. Although the negative gradient region coincides
with the expected magneto-viscothickening, the positive gradi-
ent region seems to contradict Fig. 3 findings since pressure
drops in the positive gradient branch are predicted to be lower
than when H0 = 0. This result is easily explained by the fact that
spatially inhomogeneous magnetic fields give rise to a Kelvin
body force, which produces work. In the positive gradient case,
the Kelvin body force z-component is oriented streamwise thus
creating conditions of hypogravity with a tendency to chase the
ferrofluid more quickly out of the bed. As a result of this mag-
netic levitation, pressure gradient diminishes. On the contrary,
when the magnetic field gradient is negative, the Kelvin body
force acts counterstreamwise to the flow like gravitation and
opposes to the ferrofluid from exiting the bed. This artificial
hypergravity state gives rise to a resistive work the ferrofluid
must fight against to be able to get out of the bed. As a follow
Fig. 5. Typical (r,z) contour plot of radial magnetization component for a neg-
up to this study, additional work on the topic could focus on rig- ative z-component external magnetic field gradient.
orous calculations of the demagnetizing field effects at very low
external magnetic fields as well as on the proper formulation
of the closure terms involving the fluctuations of spin den-

sity, magnetization and demagnetizing and induced magnetic


fields.
Fig. 5 exhibits a typical contour plot of the radial compo-
nent of dynamic magnetization for a negative gradient external
magnetic field of −2000 Oe/m.

Acknowledgements

The authors gratefully acknowledge Laval University and


Total S.A. for their fellowship supports.

References

Bacri, J. C., Perzynski, R., Shliomis, M. I., & Burde, G. I. (1995). Nega-
tive viscosity effect in a magnetic fluid. Physical Review Letters, 75(11),
2128–2131.
Fig. 4. Algebraic reduced outlet pressure gradient for various z-component Felderhof, B. U. (2001). Flow of a ferrofluid down a tube in an oscillating
external magnetic field gradients. magnetic field. Physical Review E, 64(021508), 1–7.
60 F. Larachi, D. Desvigne / China Particuology 5 (2007) 50–60

Khuzir, P., Bossis, G., Bashtovoi, V., & Volkova, O. (2003). Flow of magne- Schumacher, K. R., Sellien, I., Knoke, G. S., Cader, T., & Finlayson, B.
torheological fluid through porous media. European Journal of Mechanics A. (2003). Experiment and simulation of laminar and turbulent ferrofluid
B/Fluids, 22, 331–343. pipe flow in an oscillating magnetic field. Physical Review E, 67(026308),
McTague, J. P. (1969). Magnetoviscosity of magnetic colloids. Journal of Chem- 1–11.
ical Physics, 51, 133–136. Shliomis, M. I. (1972). Effective viscosity of magnetic suspensions. Soviet
Odenbach, S. (2003). Magnetic fluids—Suspensions of magnetic dipoles Physics JETP, 34, 1291–1294.
and their magnetic control. Journal of Physics: Condensed Matter, 15, Whitaker, S. (1999). Theory and applications of transport in porous media.
S1497–S1508. Dordrecht: Kluwer Academic Press.
Rinaldi, C., & Zahn, M. (2002). Effects of spin viscosity on ferrofluid flow Zeuner, A., Richter, R., & Rehberg, I. (1998). Experiments on negative and
profiles in alternating and rotating magnetic fields. Physics of Fluids, 14, positive magnetoviscosity in an alternating magnetic field. Physical Review
2847–2870. E, 58, 6287–6293.
Rosensweig, R. E. (1997). Ferrohydrodynamics. New York: Dover Publica-
tions.

You might also like