Homogénéisation des systèmes
couplés: fluide/structure + transport réactif
Andro Mikeli ´c
Institut Camille Jordan, UMR 5208, UFR Math ´ematiques, Universit ´e Claude Bernard Lyon 1, Lyon, FRANCE
Introduction 1
L’argile est une roche sédimentaire, composée pour une large part de minéraux spécifiques, silicates en général d’aluminium plus ou moins hydratés, qui présentent une structure feuilletée (phyllosilicates) qui explique leur
plasticité, ou bien une structure fibreuse (sépiolite et palygorskite) qui explique leurs qualités d’absorption.
On les classe selon l’épaisseur des feuillets (0,7 ou 1 ou 1,4 nm).
Questions:
i) Peut-on modéliser la filtration à travers une telle
géométrie. Perméabilité (mesure de capacité d’un matériau à laisser passer un fluide)?
ii) Peut-on modéliser un écoulement ráctif, couplé avec la filtration/déformation, à travers une telle géométrie?
iii) Peut-on ajouter des effets électrocinétiques?
Asymptotic solution for an elastic channel 1
Figure 1: Schematic of a section of length L a long
elastic channel.
Asymptotic solution for an elastic channel 2
We consider the stationary laminar flow of incompressible Newtonian fluid through a 2-D channel with linearly elastic walls. The fluid is driven by the pressure gradient.
L is the length of the channel
ℓ is the half of the channel width in the undeformed state δ is the thickness of the walls in undeformed state
ℓ ∼ δ
ε = ℓ
L << 1. (1)
The outer walls of the channel are fixed. Due to symmetry, only the upper half of the fluid geometry will be considered, that is, the fluid and solid occupy
Asymptotic solution for an elastic channel 3
Ωf0 = {(x, y), 0 < x < L, 0 < y < ℓ} , (2) Ωs0 = {(x, y), 0 < x < L, l < y < ℓ + δ} , (3) respectively. Further, the height of the unknown boundary of the fluid-solid interface is denoted by γ(x). As a result, the fluid in the deformed configuration will occupy:
Ωf = {(x, y), 0 < x < L, 0 < y < γ(x)} . (4) To derive an asymptotic solution to the FSI problem first we need the model:
Find ΓI, v, p and u such that:
ΓI = n
y + u(y)|∀y ∈ ΓI0o
(the interface), (5)
Asymptotic solution for an elastic channel 4
In Ωf we have the Stokes system for the flow:
−µf∆v + ∇p = ~b in Ωf, (6)
∇ · v = 0 in Ωf, (7) In Ωs0 we have the equilibrium equations of the linear
elasticity
−∇ · (C : E(u)) = ~b0 in Ωs0, (8) S = C : E(u) = λstr (E(u)) I + 2µsE(u); 2Eij = ∂ui
∂xj + ∂uj
∂xi .
(9)
Asymptotic solution for an elastic channel 5
On the interface we have continuity of the contact forces, written in Lagrange’s configuration:
det(∇u + I)(−pI + 2µfD(v(y + u(y)))) (∇u + I)−T n0
= (C : E) n0 on ΓI0. (10) v satisfies the kinematic interface condition
v = 0 on ΓI. (11) and in addition v, p and u should also satisfy any boundary conditions that might be specified on ∂Ω\ΓI.
Observe that the position of the interface is part of the
boundary value problem, and the solid-fluid coupling term (10) makes it a nonlinear one.
Asymptotic solution for an elastic channel 6
STRATEGY:
The system (5)-(10) will be normalized for this particular geometry.
Then a formal expansion of the field variables
(pressure, velocity and displacements) with respect to the small parameter ε will be used to obtain an
asymptotic solution of (5)-(10).
This asymptotic solution will further be interpreted as a one-dimensional Darcy flow with a nonlinear
permeability.
Dimensionless form 1
x = Lx, y˜ = ℓy, p(x, y) =˜ IPp(˜˜ x,y˜), (12) v1(x, y) = ¯V1v˜1(˜x, y), v˜ 2(x, y) = ¯V2v˜2(˜x,y˜), (13) u1(x, y) = ¯U1u˜1(˜x, y), u˜ 2(x, y) = ¯U2u˜2(˜x,y˜). (14) In these notations the fluid (4) and solid (3) domain are
given by
Ωf = {(˜x,y˜) : 0 < x <˜ 1, 0 < y <˜ γ(˜˜ x)}, Ωs =
(˜x, y) : 0˜ < x <˜ 1, γ(˜˜ x) < y <˜ 1 + δ ℓ
,
Dimensionless form 2
˜
γ(˜x) = γ(x)
ℓ . (15)
∂p
∂x = IP L
∂p˜
∂x˜, ∂p
∂y = IP ℓ
∂p˜
∂y˜, (16)
∂vi
∂x = V¯i L
∂v˜i
∂x˜ , ∂vi
∂y = V¯i ℓ
∂v˜i
∂y˜ , (17)
∂ui
∂x = U¯i L
∂u˜i
∂x˜ , ∂ui
∂y = U¯i ℓ
∂u˜i
∂y˜ . (18)
It is clear that the scaling parameters can not be chosen independently. Below we will discuss the relations between different scaling parameters.
Dimensionless Stokes system
IP = µV¯1
Lε2 = µ4V1,maxℓ2L
ℓ2 = 4µV1,maxL. (19)
V¯2 = εV¯1 (20)
The Stokes system can now be rewritten as
−ε2∂2v˜1
∂x˜2 − ∂2v˜1
∂y˜2 + ∂p˜
∂x˜ = 0 (21)
−ε2 ∂2v˜2
∂x˜2 − ∂2v˜2
∂y˜2 + ε−2 ∂p˜
∂y˜ = 0
∂v˜1
∂x˜ + ∂v˜2
∂y˜ = 0.
Dimensionless Navier system
Ss =
(λs + 2µs)U¯L1 ∂∂u˜x˜1 + λsU¯l2 ∂∂u˜y˜2 µs ¯
U1
l ∂u˜1
∂y˜ + U¯L2 ∂∂u˜x˜2 µs ¯
U1
l ∂u˜1
∂y˜ + U¯L2 ∂∂u˜x˜2
(λs + 2µs)U¯l2 ∂∂u˜y˜2 + λs U¯L1 ∂∂u˜x˜1
Further assume that
U¯2 = δ, U¯1 = ε0U¯2 = δ.
Using this scaling for the displacements, the stress in the solid become
Ss = δ ℓ
"
(λs + 2µs)ε∂∂u˜x˜1 + λs ∂∂u˜y˜2 µs∂∂u˜y˜1 + µsε∂∂u˜x˜2 µs ∂∂u˜y˜1 + µsε∂∂u˜x˜2 (λs + 2µs)∂∂u˜y˜2 + λsε∂∂u˜x˜1
# .
(22)
Dimensionless Navier system 2
Further, it is also necessary to write the system of elasticity equations (8) in non-dimensional form. For an isotropic
solid, and in the absence of a body force, it is easy to see, the equation (8) reduces to
ε2δ(λs + 2µs)∂2u˜1
∂x˜2 + εδ(λs + µs) ∂2u˜2
∂x∂˜ y˜ + δµs∂2u˜1
∂y˜2 = 0, (23) ε2δµs ∂2u˜2
∂x˜2 + εδ(µs + λs) ∂2u˜1
∂x∂˜ y˜ + δ(λs + 2µs)∂2u˜2
∂y˜2 = 0. (24)
Asymptotic expansion
Consider now, an asymptotic expansions of the field variables with respect to the small parameter ε:
˜
vi = ˜vi0 + εv˜i1 + ε2v˜i2 + ... (25)
˜
p = ˜p0 + εp˜1 + ε2p˜2 + ...
˜
ui = ˜u0i + εu˜1i + ε2u˜2i + ... (26)
Asymptotic expansion for Stokes system
Substituting these expansions in Stokes system (21), and collecting terms corresponding to different powers of ε, we get:
at order ε−2 : ∂p˜0
∂y˜ = 0, at orderε−1 : ∂p˜1
∂y˜ = 0. (27)
Asymptotic expansion for Stokes system
Next, at order ε0 we obtain:
−∂2v˜10
∂y˜2 + ∂p˜0
∂x˜ = 0 (28a)
−∂2v˜20
∂y˜2 + ∂p˜2
∂y˜ = 0 (28b)
∂v˜10
∂x˜ + ∂v˜20
∂y˜ = 0. (28c)
⇒ p˜0 = ˜p0(x), p˜1 = ˜p1(x). (29) Integration of the equation (28a) with respect to y˜ yields:
∂v˜10
∂y˜ = ˜y∂p˜0
∂x˜ .
Asymptotic expansion for Stokes system 2
The last equation, after another integration with respect to
˜
y, from y˜ to γ˜(˜x), becomes:
˜
v10(˜x, γ˜(˜x)) − v10(˜x, y˜) = 1
2 γ˜2(˜x) − y˜2 ∂p˜0
∂x˜ . (30)
⇒ − ∂
∂x˜
1
3γ˜3(˜x)∂p˜0
∂x˜
= 0. (31)
Observe that the last equation can be interpreted as a Darcy flow in a 1-D porous medium. To do this, fix x˜ and define, the y˜−average operator h·iy˜:
hφ(˜˜ x, y)˜ iy˜ := 1 2
Z γ(˜˜ x)
−γ(˜˜ x)
φ(˜˜ x, y))dy˜ (32)
Asymptotic expansion for Stokes system 3
where φ(˜˜ x, y)˜ is a generic function. The average is with respect to the reference width of the fluid channel. Then:
hv˜1(˜x)i = −1
3γ˜3(˜x)∂p˜0
∂x˜ (33)
One would now recognize that equation (31) is the
conservation of mass for a flow with flux hv˜1(˜x)i, driven by a pressure gradient ∂∂p˜x˜0 . Furthermore, the quantity K˜:
K˜ := ˜K(˜γ3(˜x), x) =˜ −µf hv˜1(˜x)i
∂x˜p˜0(˜x) = µf
3 γ˜3(˜x) (34)
Asymptotic expansion for Stokes system 4
defined as the ratio of the mass flux and the pressure
gradient, can be interpreted as permeability of the channel.
The permeability is scaled, as usual, by the fluid viscosity, so it doesn’t depend on the fluid properties. In the rigid case (γ˜(˜x) = 1) it will coincide with the standard Darcy
permeability. When the channel is deformable however, it is not constant but will depend on the 3-rd power of the
channel opening. With the definition (34) in mind, equation (31) can be rewritten as
− ∂
∂x˜
K˜ (˜γ3(˜x))∂p˜0
∂x˜
= 0. (35)
We need an expression for γ˜ in terms of p˜0 which will close the last equation.
Asymptotic expansion for elasticity system
Here we get
˜
u01(˜x, y) =˜ − 1 + δℓ − y˜
µs c1(˜x) (36) and
˜
u02(˜x, y) =˜ − 1 + δℓ − y˜
λs + 2µs c2(˜x). (37)
Expansion of solid-fluid interface condition
First for the fluid stress Tf we get
Tf = IP
2ε2 ∂v˜1
∂x˜ − p˜ ε∂v˜1
∂y˜ + ε3 ∂v˜2
∂x˜ ε∂v˜1
∂y˜ + ε3 ∂v˜2
∂x˜ 2ε2∂v˜2
∂y˜ − p˜
(38)
Asymptotics for the solid-fluid interface
Next, the expansions for velocity and pressure from (25) are substituted into the last equation (38) to obtain
Tf = µfV¯1L ℓ2
"
−p˜0(˜x) 0 0 −p˜0(˜x)
#
+ O (ε) . (39) The normal component of the zeroth order term for the
stress tensor is calculated at the interface, at the reference configuration, for which the normal is simply n0 = e2:
Ss,0n0 = δ
ℓ (c1(˜x)e1 + c2(˜x)e2) . (40) Note that the last equation cannot be compared directly
with (39) as the first is in Lagrangian, while the former
Asymptotics for the solid-fluid interface 2
in Eulerian coordinates. However, the point (˜x, 1), at the interface in the reference configuration, corresponds to the interface point (˜xE, γ˜(˜xE)) in the Eulerian formulation, where
˜
xE = ˜x + ˜u01(˜x) and γ˜(˜xE) = 1 + ˜u02(˜x). (41) The fluid stress force at the interface in the reference
configuration is given by the expression at the left hand side in the formula (10). To use equation (10) one needs the
leading order of the deformation gradient:
F = ∇u+I = 1 + ∂u∂x1 ∂u∂y1
∂u2
∂x 1 + ∂u∂y2
!
= 1 + ε∂∂u˜x˜1 δl ∂∂u˜y˜1 ε∂∂u˜x˜2 1 + δl ∂∂u˜y˜2
!
(42)
Asymptotics for the solid-fluid interface 3
therefore, F = F0 + O (ε), where F0 = 1 δl ∂∂u˜y˜01
0 1 + δl ∂∂u˜y˜02
!
and F0−T =
1 + δ l
∂u˜02
∂y˜
−1 1 + δl ∂∂u˜y˜02
−δl ∂∂u˜y˜01
(43)
The insertion of (39), (40) and (43) into (10) now results in:
−IPp˜0(˜x)e2 = δ
ℓ (c1(˜x)e1 + c2(˜x)e2) + O(ε). (44)
⇒ u˜01 = c1(˜x) = 0,
˜
γ(˜x) = 1 + ˜u02(˜x, 1) = 1 + IP
λs + 2µs p˜0(˜x). (45)
Permeability of a long elastic channel
Equivalently, in dimensional variables, γ(x) = ℓ + ℓ
λs + 2µs p0(x). (46)
Permeability of a long elastic channel
K = K(p0(x), x) = 1 3ℓ
l + l
λs + 2µs p0(x) 3
= ℓ2 3
1 + p0(x) λs + 2µs
(47)
Long elastic channel: A typical solution
Comments:
For more details see: O. Iliev, A. Mikeli´c, P. Popov : On
upscaling certain flows in deformable porous media, Multiscale Model. Simul., Vol. 7 (2008), no. 1, p. 93-123.
For application to the blood flow see: A. Mikeli´c, G. Guidoboni, S.
Cani´c : Fluid-Structure Interaction in a Pre-Stressed Tube with Thickˇ Elastic Walls I: The Stationary Stokes Problem, Networks and
Heterogeneous Media, Vol. 2 (2007), p. 397 - 423.
In the above reference the following law was derived:
∆p(z) =
K (1−Ehσ2)R2 − PeR
u0s(R, z). (48)
where
K = 1 2
2 + h/R (1 + h/R)2,
Open questions:
u0s(R, z) is the radial displacement at the interface, and
∆p(z) is the transmural pressure. Notice that for K = 1 this is exactly the Law of Laplace.
Open questions:
Nonlinear elastic solid structure?
Electrokinetic effects?
Coupling with reactive flows?
Dynamic Permeability 1
Fluid-saturated porous media arise in contexts as
sedimentary rocks and soils and also in study of metallic foam structures, polymer gels and catalytic beds. One of basic properties of porous media is the fluid permeability .
This key macroscopic property of porous media is described by the well-known Darcy law
U~ (x) = −kdc
µ ∇p0(x), (49)
U~ is the average fluid velocity (= the filtration velocity)
∇p0 is the pressure gradient µ is the dynamic fluid viscosity kdc is the static fluid permeability.
Dynamic Permeability 2
Darcy’s law is valid for the slow flow of viscous fluids. It corresponds to the homogenization of the complex fluid motions taking place at the microscopic scale in the pore structure. Hence the permeability tensor kdc depends
nonlocally on the morphology of the pore space.
We note the considerable progress in understanding the Darcy law (49) by homogenizing the Stokes and the
Navier-Stokes fluid equations from the microscopic pore level to the macroscopic level, see e.g.
G. Allaire, in : One-phase newtonian flow, Homogenization and porous media, ed. U. Hornung, Springer 1997, pp. 45 - 68.
A. Mikeli´c, in : Homogenization theory and applications to filtration through porous media, Filtration in Porous Media and Industrial
Applications, Lecture Notes in Mathematics Vol. 1734, Springer, 2000, pp. 127-214.
Dynamic Permeability 3
Another important transport phenomena involve electrical conduction. Electrical formation factor F is given by
F = σ
σ∗ , (50)
where σ∗ is the effective electrical conductivity of a porous medium containing fluid of conductivity σ and insulating solid phase.
Relationship between the permeability and conductivity:
kdc = c1 ℓ2c
F . (51)
(see Katz and Thompson, Phys. Rev. B 1986. ) ℓc is a
characteristic length related to the threshold pressure in a mercury injection experiment.
Dynamic Permeability 4
c1 is a constant, estimated to be 4.4×10−3. ℓc is well defined only when the porous medium can be modeled as a
distribution of cylindrical pores on a lattice.
The formal analogy between Darcy’s law (49) and Ohm’s law for bulk electrical transport in porous media suggests a relationship between electricity and hydrodynamics at the macroscopic level. In this direction Johnson, Koplik and
Schwartz introduced a geometrical pore-size parameter Λ, which is always well defined for any porous medium and
which describes the effects of an internal boundary layer on a variety of processes such as electrical surface conduction etc. Let E be the local electrical field, then the parameter Λ is a weighted pore volume-to-surface area ratio
Λ = 2 R
|E(r)| dV
, (52)
Dynamic Permeability 5
where dV denotes an integration over the pore volume and dS denotes an integration over the pore-solid surface.
Based on plausible physical arguments, and backed by detailed computer simulations on random materials,
Johnson, Koplik and Schwartz found the formula kdc = c2 Λ2
8F , (53)
where c2 is a constant of order 1. We note that Λ/2 6= VP /S, where Vp is the pore volume and S is the surface area of the pore-grain interface. Λ is a length that is directly related to transport,; it is a measure of the dynamically connected part of the pore space.
Random Porous Media
Let us define a random porous medium after the article M. Avellaneda, S. Torquato : Flow in random porous media :
mathematical formulation, variational principles, and rigorous bounds , J. Fluid Mech. , Vol. 206 (1989) , p. 25-46.
The random porous medium is a domain of space V(ω) ∈ IR3 (where the realization ω is taken from probability space Ω) which consists of two regions : the pore space V1(ω),
through which fluid flows, of the porosity ϕ and a
solid-phase region V2(ω). Let ∂V(ω) be the surface between V1(ω) and V2(ω). The characteristic function of the pore
region is given by
I(x, ω) =
(1 x ∈ V1(ω),
0 x ∈ V2(ω). (54)
Random Porous Media 2
We assume that our porous medium is a realization of a
statistically homogeneous ergodic random porous medium, scaled by a small parameter a . This means that by means of Birkhoff’s ergodic theorem we equate ensemble
averages with infinite volume averages, i.e. we have
< f(x0) >= lim
|C|→∞
1
|C|
Z
C
f(x0 + x) dx (55) where < · · · > denotes ensemble averaging. For periodic porous media, this corresponds to the averaging over a unit cell.
We note that
ϕ =< I >= lim
|V1|,|V|→∞
|V1|
|V| ; S =< M >= lim
|∂V|,|V|→∞
|∂V|
|V|
Expos ´e `a l’Atelier ”ANALYSE MULTIECHELLE EN ELECTROCINETIQUE ET APPLICATIONS AUX MILIEUX POREUX”, CMAP, Ecole Polytechnique, Palaiseau, le 22 mars 2010 – p. 33/72(56)
Random Porous Media 3
Let us recall that the fluid permeability arising in Darcy’s law (49) can be expressed in terms of a certain scaled auxiliary velocity field for periodic media (see e.g. the minicours by G. Allaire) and random media (see e.g. papers by S.
Kozlov). Following Avellaneda and Torquato, we write it as (kdc)ij =< wji >, (57) where the auxiliary scaled velocity wi satisfies
∆wi = ∇πi − ei in V1, (58)
div wi = 0 in V1, (59)
wi = 0 on ∂V (60)
{wi, πi} are bounded functions . (61)
Random Porous Media 4
Now we consider a fluid-filled porous medium, with
connected pore space, under the excitation of an external harmonic source with frequency ω.
The pore structure at microlevel is characterized by a
typical length of the pore scale a. In order to be as close to the reality as possible, we suppose that the fluid has
dynamic viscosity µ, bulk modulus K and density ρf.
Intrinistic viscous relaxation time is t0 = ρfa2/µ the fluid sound velocity c0 = p
K/ρf.
We note that t0 is the time scale at which the inertial force density inside a pore, ρfa/t20 equals the viscous force
density µ/(t0a). The product of these two quantities defines the large scale L = t c .
Dynamic Permeability 6
The displacements and the Reynolds number are small and the linearized fluid/structure interaction is considered. A
dimensionless small parameter is ε = a/L. After making the change of variables y = x/a and τ = t/t0 and looking for
time harmonic motions with angular frequency ω, we find the following dimensionless form of the problem
−iωε~v = Div σ in V1 (62) σ = −pI + 2εD(~v) in V1 (63)
iωεp = div ~v in V1 (64)
εΘ = CD(~u) in V2 (65)
− ρs
ρf ω2ε~u = Div Θ in V2 (66)
~v = −iω~u and σ~n = Θ~n on ∂V. (67)
Dynamic Permeability 7
Here C is the fourth-order elastic tensor in the units of K and p, σ and Θ are in units of Lµ2/(ρfa3).
Using the modified expansion of Burridge and Keller, Jour.
Acoustic Soc. Amer. 1981 , the field quantities are expanded as a two-scale perturbation series in ε :
~v = v0(x, y) + εv1(x, y) + . . . , with x = r/L and y = r/a. Then the derivatives are expressed with respect to x and y
variables.
By equating terms with equal powers of ε, a hierarchy of equations for the quantities v0, v1, u0, u1, σ0, σ1, p0, p1 etc is obtained. The relevant lowest-order (ε−1) equations are :
Divyσ0 = 0 and σ0 = −p0I (68) divyv0 = 0 and Dy(u0) = 0. (69)
Dynamic Permeability 8
Now we conclude that u0 = u0(x) , σ0 = σ0(x) and, as far as the fluid is concerned, to the lowest order it may be regarded
as incompressible on the y scale.
Next following Biot’s articles v0 is expressed in terms of u0 and a relative fluid permeability w
v0 = −iω
u0(x) + w(x, y)~
. (70)
Now we introduce the corresponding auxiliary problem :
−iω∆yW ↔ = ∇yP↔ − ω2W↔ − I↔ in V1, (71)
div W ↔ = 0 in V1, (72)
W↔ = 0 on ∂V (73)
{W↔, P↔} are bounded functions . (74)
Dynamic Permeability 9
The average flow rate U~ is obtained from 70 by averaging w over the pore scale y :
U~ = −iω < ~w >y= −iω < W↔(x, y) >y
−∇xp0+ω2u0
(75)
Equation (75) may be viewed as a generalization of the
Darcy’s law to deformable elastic porous media. The elastic solid displacement actes as an additional source term of
the form ε2u0. Comparing (75) with the Darcy’s law implies the following definition of the dynamic permeability
κ↔(ω) = −iω < W↔(x, y) >y (76)
Dynamic Permeability 10
The real and imaginary parts of κ↔(ω) behave very differently. The real component is dominant at low
frequencies, where viscous stresses are important. The imaginary component dominates at high frequencies, because in this regime the fluid behaves as if it were
inviscid in almost all of the pore space and viscous effects are confined to a boundary layer of width p2µ/(ρfω) near the pore surface. More precisely,
ℜκ↔(ω) ∼ kdc and Im κ↔(ω) ∼ ωρf
µ < |W↔(y, 0)|2 >y, for ω ≪
(77)
where W↔(y, 0) is the solution for (71)-(73) for ω = 0.
The high frequencies limit was studied in number of papers.
Dynamic Permeability 11
The main result is the following : ℜκ↔(ω) ∼
√2 ΛF
ωρf µ
−3/2
and Im κ↔(ω) ∼ µ
ωF ρf , for ω ≫ 1,
(78)
where Λ is is a weighted pore volume-to-surface area ratio given by (52) and electrical formation factor F is given by (50). The limit is calculated using the observation
ω2W ↔ ∼ −I↔ + ∇yP↔ ∼ E, ω ≫ 1, (79) where E the electrical field satisfying
curl E = 0, div E = 0 (80)
in the pore geometry and the boundary condition E · n = 0
Dynamic Permeability 12
We note that the boundary value problem for the field E can be interpreted as the stationary potential flow of an ideal
fluid through the microstructure arising from an imposed pressure drop I↔.
Next, from dimensional analysis, the dynamic permeability can be written in the scaled form
κ↔(ω) = R2
8F f(ωρfR2
µ ), (81)
where R is the length scale defining the effective hydrodynamic pore radius of the microstructure.
As remarked by Avellaneda et al, the determination of
κ↔(ω) requires ideally an infinite set of measurements. At the other hand, empiric theory requires only the
measurements of Λ and F.
Dynamic Permeability 13
The approach is based on a hypothesis of universality of the dynamic response , supposed to be valid for a wide class of porous media.
Explicit formula of Johnson, Koplik and Dashen 1986:
κ(ω) = Λ2
F g(ωρfΛ2
µ ) with g(˜ω) = 1 8p
1 − iω/16˜ − iω˜ . (82) The permeability calculation for periodic porous media:
κ(ω) =
(κ0 + iC1ωρf/µ as ω → 0,
√2
ΛF (ωµ/ρf)−3/2 + iµ/(F ρfω) as ω → ∞. (83)
Geometry
Periodic 2-scale conv for porous media pb 1
Description of a periodic porous medium
We consider a periodic porous medium Ω =]0, L[3 in IR3 with a periodic arrangement of the pores. The formal description goes along the following lines: First we define the
geometrical structure inside the unit cell Y =]0, 1[3. Let Ys (the solid part) be a closed subset of Y¯ and Yf = Y\Ys (the fluid part). Now we make the periodic repetition of Ys all
over IR3 and set Ysk = Ys + k, k ∈ ZZ3. Obviously the set ES = S
k∈ZZ3 Ysk is a closed subset of IR3 and EF = IR3\ES is an open set in IR3. We make the following assumptions on Yf and EF :
(i) Yf is an open connected set of strictly positive measure, with a Lipschitz boundary and Ys has strictly positive
measure in Y¯ as well.
Periodic 2-scale conv for porous media pb 2
(ii) EF and the interior of ES are open sets with the
boundary of class C1,1, which are locally located on one side of their boundary. Moreover EF is connected and the solid part, ES, is supposed connected in IR3
Now we see that Ω =]0, L[3 is covered with a regular mesh of size ε, each cell being a cube Yiε = ε(Y + i), with
1 ≤ i ≤ N(ε) = |Ω|ε−3[1 + 0(1)]. We define Ysεi = (Πεi)−1(Ys) and Yfε
i = (Πεi)−1(Yf). For sufficiently small ε > 0 we
suppose L/ε ∈ IN and consider Tε = {k ∈ ZZ3|Ysεk ⊂ Ω} and define
Ωεs = [
k∈Tε
Ysεk, Γε = ∂Ωεs, Ωεf = Ω \ Ωεs.
Periodic 2-scale conv for porous media pb
Obviously, ∂Ωεf = ∂Ω ∪ Γε. The domains Ωεs and Ωεf
represent, respectively, the solid and fluid parts of a porous medium Ω.
0000000000000000000 0000000000000000000 1111111111111111111 1111111111111111111 0000000000000000000 0000000000000000000 0000000000000000000 1111111111111111111 1111111111111111111 1111111111111111111 0000000000000000000 0000000000000000000 0000000000000000000 1111111111111111111 1111111111111111111 1111111111111111111 0000000000000000000 0000000000000000000 0000000000000000000 1111111111111111111 1111111111111111111 1111111111111111111 0000000000000000000 0000000000000000000 0000000000000000000 1111111111111111111 1111111111111111111 1111111111111111111 0000000000000000000 0000000000000000000 0000000000000000000 1111111111111111111 1111111111111111111 1111111111111111111 0000000000000000000 1111111111111111111
00000000 00000000 00000000 11111111 11111111 11111111
00000000 00000000 00000000 11111111 11111111 11111111 00
11 00 11 00 11 00 11 00 11
000 111 000 111 000 111 000 111 000 111
00000000 11111111 00
00 00 00
11 11 11 11
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11 000000
000 111111 111
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11 000000
000 111111 111
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11 000000
000 111111 111
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11 000000
000 111111 111
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11 000000
000 111111 111 000000 000 111111 111 000000 000 111111 111 000000 000 111111 111 000000 000 111111 111
000000 000 111111 000111 000000 111111 111
000000 000 111111 111
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
000 111 000 111 000 111 000 111
00 11
00 11
00 11
00 11
00 11
00 11
000000 000 111111 111
000000 000 111111 111
000000 000 111111 111
000000 000 111111 111
0000 00 1111 11
0000 00 1111 11
0000 00 1111 11
0000 00 1111 000 11
111
z r
fluid
FLUID
0 < y < 1
0<r<1~ FLUID
FLUID
2ε
PERIODIC DOMAIN IN TUBE FLOW PERIODIC DOMAIN IN POROUS MEDIA FLOW
Biot’s law without dissipation 1
Obviously, ∂Ωεf = ∂Ω ∪ Γε. The domains Ωεs and Ωεf
represent, respectively, the solid and fluid parts of a porous medium Ω.We suppose small deformations and the
displacements are described by linearized equations in both media. More precisely, in Ωεf we have the linearized momentum equation for the time derivative of the fluid displacement uε, in Eulerian variables.
Ωεs is the reference configuration of a deformable elastic
body and the equation of linear elasticity for the deformation wε are in Lagrangean variables. In both domains all
quadratic terms are neglected. The interface between two media changes with the deformation of the structure.
Biot’s law without dissipation 2
If the pores contain an inviscid fluid then the kinematic interface condition is continuity of the normal velocities. It reads
∀t, ∀x ∈ Γε, ∂uε
∂t (x + wε(x, t), t) · ν = ∂wε
∂t (x, t) · ν.
After linearization and neglecting the term
∇x ∂uε
∂t (x, t) · wε(x, t), we get continuity of the normal displacements at Γε.
example 3: Biot’s law without dissipation 3
Let F ∈ C∞([0, T]; L2(Ω)3) and curl F ∈ C∞([0, T]; L2(Ω)3) . Then we consider the system
ρs ∂2wε
∂t2 − div(σ(wε)) = F ρs and σ(wε) = AD(wε) in Ωεs×]0, T[,
(84)
ρf ∂2uε
∂t2 + ∇pε = F ρf and div ∂u
ε
∂t = 0 in Ωεf×]0, T[, (85) (uε − wε) · ν = 0 and (pεI + σ(wε)) · ν = 0 on Γε×]0, T[ (86) {uε, wε, pε} are L − periodic. (87) uε(x, 0) = ∂uε
∂t (x, 0) = 0 in Ωεf; wε(x, 0) = ∂wε
∂t (x, 0) = 0 in Ωεs
(88)
Biot’s law without dissipation 4
The above system has a unique variational solution in V = {ϕ ∈ L2(Ω)3 ; ϕ ∈ H1(Ωεs)3, div ϕ = 0 in Ωεf,
div ϕ ∈ L2(Ω) and ϕ is L-periodic}. It is arbitrary smooth with respect to t.
After taking uε ∈ H3(0, T; L2(Ωεf)3) ∩ H2(0, T; H1per (Ωεf)3) , wε ∈ H3(0, T; L2(Ωεs)3) and pε ∈ H1(0, T; L2(Ωεf)) as the test function in (84)-(88) we obtain
∂uε
∂t
L∞(0,T;L2(Ωεf)3)+
∂wε
∂t
L∞(0,T;L2(Ωεs)3)+
∇wε
L∞(0,T;L2(Ωεs)9) ≤
(89)
Biot’s law without dissipation 5
The linearized incompressible Euler system doesn’t involve derivatives of the velocity field with respect to x and an
H1-estimate for the velocity doesn’t follow directly. One way to proceed is to use the H(div; Ωεf) estimate in the fluid part.
Nevertheless, after taking the curl of the linearized Euler system, we get
div uε = 0 and curl uε
L∞(0,T;L2(Ωεf)3) ≤ C (90) and it is natural to estimate the L2-norm of the ∇uε by the L2-norms of div uε and curl uε. Such estimate requires a boundary condition at the interface. In our model it is a given normal component uε · ν and then the H1-estimate uniform with respect to ε holds if and only if the first Betti number of Ωε is zero.
Biot’s law without dissipation 6
By assumptions, ∂Ωεf is a connected 2D manifold. Each
such manifold is homeomorphic to a sphere with ”handles"
and the first Betti number or the genus of ∂Ωεf is the number of handles. For a simply connected domain, the first Betti number is zero. Our domain is multiply connected and the estimate requires some effort. By supposing that
∂Yf ∈ C1,1 we get
∇uε
L∞(0,T;L2(Ωεf)9) ≤ C
ε . (91)
For the proof see [FeMi03] J. L. Ferr´ın , A. Mikeli ´c :
Homogenizing the Acoustic Properties of a Porous Matrix Containing an Incompressible Inviscid Fluid , Mathematical Methods in the Applied
Sciences, Vol. 26 (2003), p. 831-859.
Biot’s law without dissipation 7
We note that for isolated fluid parts the velocity gradient is uniformly bounded, leading to different results. For more details we refer to [FeMi] . The extension of the pressure field pε to Ωεs×]0, T[ is given by
˜
pε(x, t) =
pε(x, t) − |Ω1| R
Ωεf
pε(x, t) dx, x ∈ Ωεf,
−|Ω1| R
Ωεf
pε(x, t)dx, x ∈ Ωεs. (92)
Then R
Ω
˜
pεdiv ϕ = R
Ωεf
pεdiv ϕ, ∀ϕ ∈ H1per (Ω)3, and
p˜ε
H1(0,T;L20(Ω)) +
∇p˜ε
H1(0,T;H−per1 (Ω)3) ≤ C. (93)
Biot’s law without dissipation 8
If we have two different estimates for gradients in the solid and in the fluid part, then the classical way to proceed is by extending the deformation from Ωεf to Ω and then passing to the limit ε → 0. For precise recent results on
homogenization of Neumann problems in perforated domains with Lipschitz perforations we refer to
[Ace92] Acerbi, E., Chiadò Piat, V., Dal Maso, G., Percivale D. : An extension theorem from connected sets, and
homogenization in general periodic domains. Nonlinear Anal., TMA, 18 481–496 (1992) and to the book by Jikov, Kozlov and Oleinik [JKO].
Here we need the " triple field compactness " result from [All92], modified in [FeMi03]