Accepted Manuscript

On the well-posed coupling between free uid and porous viscous

ows
Philippe Angot
PII: S0893-9659(10)00252-1
DOI: 10.1016/j.aml.2010.07.008
Reference: AML 3282
To appear in: Applied Mathematics Letters
Received date: 31 October 2009
Revised date: 18 April 2010
Accepted date: 17 July 2010
Please cite this article as: P. Angot, On the well-posed coupling between free uid and porous
viscous ows, Appl. Math. Lett. (2010), doi:10.1016/j.aml.2010.07.008
This is a PDF le of an unedited manuscript that has been accepted for publication. As a
service to our customers we are providing this early version of the manuscript. The manuscript
will undergo copyediting, typesetting, and review of the resulting proof before it is published in
its nal form. Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal pertain.
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
Author manuscript, published in "Applied Mathematics Letters 24, 6 (2011) 803-810"
DOI : 10.1016/j.aml.2010.07.008
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
Arrirro Mxrnrmxrrcs Lrrrras, 2010 (ro xrrrxa)
On the well-posed coupling between free uid and porous viscous ows
Philippe Angot
Universit e de Provence & LATP - CMI, UMR CNRS 6632, 39 rue F. Joliot Curie, 13453 Marseille Cedex 13 - France.
Abstract
We present a well-posed model for the Stokes/Brinkman problem with a family of jump embedded boundary condi-
tions (J.E.B.C.) on an immersed interface with weak regularity assumptions. It is issued from a general framework
recently proposed for ctitious domain problems. Our model is based on algebraic transmission conditions combining
the stress and velocity jumps on the interface separating the uid and porous domains. These conditions are well
chosen to get the coercivity of the operator. Then, the general framework allows us to prove new results on the global
solvability of some models with physically relevant stress or velocity jump boundary conditions for the momentum
transport at a uid-porous interface. The Stokes/Brinkman problem with Ochoa-Tapia & Whitaker (1995) interface
conditions and the Stokes/Darcy problem with Beavers & Joseph (1967) conditions are both proved to be well-posed
by an asymptotic analysis. Up to now, only the Stokes/Darcy problem with Saman (1971) approximate interface
conditions with negligible tangential porous velocity was known to be well-posed.
Key words: Transmission problems, Jump embedded boundary conditions, Stokes/Brinkman problem, Stokes/Darcy
problem, Fluid/porous coupled ows, Well-posedness analysis, Asymptotic analysis, Vanishing viscosity, Singular
perturbation
2000 MSC: 34E15, 35J20, 35J25, 35J50, 35J55, 35Q30, 35Q35, 65J20, 76D03, 76D07, 76M45, 76S05, 86A60
1. Introduction
Notations. Let the domain R
d
(d =2 or 3 in practice) be an open bounded and Lipschitz continuous domain.
Let an interface R
d1
, Lipschitz continuous, separate into two disjoint connected subdomains: the uid domain

f
and the porous one
p
such that =
f

p
. The boundaries of the subdomains are respectively dened
by:
f
=
f
for
f
,
p
=
p
for
p
and =
f

p
for , see Fig. 1, assuming no cusp singularity
at . Let n be the unit normal vector on oriented from
p
to
f
and any unit tangential vector of a local
tangential basis (
1
, ,
d1
) on . For any quantity dened all over , the restrictions on
f
and
p
are denoted
by
f
and
p
respectively. For a function in H
1
(
f

p
), let

and
+
be the traces of

p
and

f
on each side
of respectively,

= (
+
+

)/2 the arithmetic mean of traces of , and [[]]

= (
+

) the jump of traces of

on oriented by n.
There exist in the literature dierent models with physically relevant stress or velocity jump boundary conditions
for the tangential momentum transport at the uid-porous interface , see e.g. [30, 20]. When the homogeneous
porous ow is to be governed by the Brinkman equation, cf. [13, 14, 1, 22, 3, 10], the interface condition below
linking the jump of shear stress with a continuous velocity was derived with volume averaging techniques by Ochoa-
Tapia and Whitaker [28] instead of the usual stress and velocity continuity boundary conditions at the interface [3]:
_
v
f
n

v
p
n
_

=

otw

K
v

and v
f
= v
p
= v

on , (1)
Email addresses: angot@cmi.univ-mrs.fr (Philippe Angot)
URL: http://www.latp.univ-mrs.fr/ angot (Philippe Angot)
Preprint submitted to Applied Mathematics Letters (to appear) July 30, 2010
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
Figure 1: Conguration for uid-porous ows inside the domain =
f

p
.
where the dimensionless parameter
otw
is of order of one; see [29, 20, 17, 33] for its characterization. We prove in
Section 3, as a by-product of our general framework recalled in Section 2, that stress jump boundary conditions of
this type yield a well-posed uid-porous Stokes/Brinkman problem whatever the dimensionless parameter
otw
0.
This was not already stated up to our knowledge.
When the porous ow is governed by the Darcy equation, see e.g. [22], the well-known Beavers and Joseph
interface condition [11] must be used. It links the shear stress at the interface with the jump of tangential velocity:
(v
f
n)

=

bj

K
_
v
f
v
p
_

and v
f
n = v
p
n = v n

on , (2)
where the dimensionless parameter
bj
= O(
1

) depends on the porosity and may vary between 0.1 and 4 [11, 2].
The approximate Saman interface condition [31], derived by homogenization techniques in [23], is also written
when the porous ltration tangential velocity can be neglected with respect to the uid velocity at the interface:
v
p

v
f

, i.e. for a permeability value K or Darcy number Da = K/H

2
suciently small. The global solvability
of the Stokes/Darcy problem with the Saman condition for v
p

0 is proved with a mixed hybrid formulation

in [24] whatever the dimensionless parameter
bj
0, and then by many others with various formulations, see
e.g. the recent review [18]. The only result of well-posedness for the full form of Beavers and Joseph condition
is recently established in [15] for
2
bj
suciently small. We prove in Section 4 by a singular perturbation in our
general framework with a vanishing viscosity that the above Beavers and Joseph interface conditions yield a well-
posed Stokes/Darcy problem whatever the parameter
bj
0. Here, the main diculty lies in how to give a sense to
the tangential trace of the porous velocity on the interface with minimal regularity assumptions. This is particularly
relevant for thin uid layers as for conducting fractures in porous media ows [8, 9, 15].
We rst begin in the next Section 2 by describing the general framework with jump embedded boundary conditions
studied in [6]. It is derived by a generalization to vector elliptic problems of a previous model stated for scalar
problems [4, 5]. A short version of the following results can be found in [7].
2. A well-posed Stokes/Brinkman problem with jump embedded boundary conditions
Let (v, p) p I+2 d(v) denote the Newtonian stress tensor dened with the eective viscosity in the porous
domain
p
, with = in the uid domain
f
and d(v)
1
2
(v + v
t
) being the strain rate tensor. We consider
the following Stokes/Brinkman problem including jump embedded boundary conditions (J.E.B.C.) on the interface
2
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
which link the trace jumps of both the stress vector (v, p) n and the velocity vector v through the interface :
(v, p) = f in
f
, (3)
(v, p) + K
1
v = f in
p
, (4)
v = 0 in
f

p
, (5)
v = 0 on
f

p
, (6)
[[(v, p) n]]

= Mv

on , (7)
(v, p) n

= S[[v]]

on . (8)
Here, the viscosity coecient and eective viscosity in the porous medium are bounded positive functions such
that
0
= min(, ) > 0, the symmetric permeability tensor K (K
i j
)
1i, jd
is uniformly positive denite, and the
transfer matrices S, M on are measurable, bounded and uniformly semi-positive matrices verifying ellipticity
assumptions:
K (L

())
dd
; K
0
> 0, R
d
, K(x)
1
K
0

2
a.e. in
p
. (A1)
M, S (L

())
dd
; M
0
, S
0
0, R
d
, M(x) M
0

2
, S(x) S
0

2
a.e. on . (A2)
With usual notations for Sobolev spaces, e.g. [27, 21], we now dene the Hilbert spaces:
H
1
0
f
(
f
)
d

_
w H
1
(
f
)
d
; w

f
= 0 on
f
_
, H
1
0
p
(
p
)
d

_
w H
1
(
p
)
d
; w

p
= 0 on
p
_
,
W
_
w L
2
()
d
, w

f
H
1
0
f
(
f
)
d
and w

p
H
1
0
p
(
p
)
d
; w = 0 in
f

p
_
equipped with the natural inner product and associated norm in H
1
(
f

p
)
d
.
Let us note that for v W satisfying (3) or (4) with f L
2
()
d
such that (v, p) L
2
()
d
, we can dene
(v, p) n

in H

1
2
()
d
, see [25, 12]. The model with the J.E.B.C. (7-8) also allows a possible pressure jump [[p]]

0
in H

1
2
() with additional regularity assumptions.
Then, as a consequence of the general framework stated in [6], the problem (3-8) satises in the nice weak
formulation below:
Find v W such that w W, a(v, w) = l(w) with
a(v, w) = 2
_

f
d(v) : d(w) dx + 2
_

p
d(v) : d(w) dx +
_

p
K
1
v wdx +
_

Mv

ds +
_

S[[v]]

[[w]]

ds
l(w) =
_

f wdx. (9)
Besides, the following well-posedness result is ensured by [6, Theorem 1.1].
Theorem 2.1 (Global solvability of Stokes/Brinkman model with J.E.B.C.). If the ellipticity assumptions (A1,A2)
hold, the problem (3-8) with f L
2
()
d
has a unique solution (v, p) W L
2
() satisfying the weak form (9) for all
w W and such that p
f
= p
f
0
+ C
0
+ C
1
/2 and p
p
= p
p
0
+ C
0
C
1
/2 where p
0
L
2
0
() = q L
2
(),
_

q dx = 0
and C
0
, C
1
are constants dened by:
C
0
=
1

_
(v, p
0
) n

S[[v]]

, n
_

1
2
,
and C
1
=
1

_
[[(v, p
0
) n]]

Mv

, n
_

1
2
,
.
Hence, to satisfy (7-8) in the sense of H

1
2
()
d
, the pressure eld p L
2
() must be adjusted from the zero-average
pressure p
0
L
2
0
() such that: (p p
0
)

= C
0
and [[p p
0
]]

= C
1
.
Moreover, there exists a constant
0
(
f
,
p
, K
0
,
0
) > 0 such that:
v
W
+ p
0

0,

c(
f
,
p
, , , K
1

0
f
0,
.
Remark 1 (Generalizations). For practical problems, the case of a nonhomogeneous Dirichlet boundary condition:
v = v
D
on
f

p
with v
D
H
1
2
(
f

p
)
d
and the compatibility condition
_

p
v
D
nds = 0, can be treated as well
by dening an ad-hoc divergence-free extension of v
D
, e.g. [32], and adding its contribution in the source term f of
the present problem (9). The generalization to unsteady Stokes/Brinkman problems is also straightforward.
3
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
3. The Stokes/Brinkman problem with Ochoa-Tapia & Whitaker interface conditions
We now consider that = /, where ]0, 1] is the porosity of the porous medium, and stress jump interface
conditions of Ochoa-Tapia & Whitakers type [28] like in (1), the original ones reading with

=
otw
and
n
= 0:
[[(v, p) n]]

= Mv with M
j j
=

, j = 1, , d 1, M
dd
=

n

K
n
and [[v]]

= 0 on , (10)
where M is a positive diagonal matrix with

,
n
0 a.e. on and K

, K
n
permeability coecients. Then, as a
consequence of the general framework stated in [6], the problem (3-6,10) satises in the weak formulation below:
Find v V = u H
1
0
()
d
; u = 0 such that,
2
_

f
d(v) : d(w) dx + 2
_

d(v) : d(w) dx +
_

p
K
1
v wdx +
_

Mv wds =
_

f wdx, w V. (11)
Besides, the following well-posedness result is ensured as a corollary of Theorem 2.1.
Corollary 3.1 (Global solvability of Stokes/Brinkman problem with OT-W). If the ellipticity assumptions (A1,A2)
hold, the problem (3-6,10) with f L
2
()
d
has a unique solution (v, p) V L
2
() satisfying the weak form (11) for
all w V and such that p
f
= p
f
0
+ C
1
/2 and p
p
= p
p
0
C
1
/2 with p
0
L
2
0
() and the constant C
1
dened by:
C
1
=
1

([[(v, p
0
) n]]

Mv , n)

1
2
,
.
Skrrcn or raoor. The existence and uniqueness of v V satisfying (11) is ensured by the Lax-Milgram Theorem. The
pressure eld p
0
L
2
0
() can be also recovered by the De Rham theorem [32, 12] which involves the inf-sup condition
between the velocity and pressure spaces [19]. Then, by constructing an ad-hoc divergence-free extension as for [6,
Theorem 1.1] (see also [12]), this allows to verify the stress jump condition (10) in H

1
2
()
d
with the pressure eld
p L
2
() tted such that we have formally [[p p
0
]]

= C
1
and (p p
0
)

= 0. 2
We can also interpret this solution as the limit solution of the problem (3-8) with penalized velocity jumps on
when the penalty parameter > 0 tends to zero and we have the following convergence result.
Theorem 3.2 (Convergence to Stokes/Brinkman problem with OT-W). For any > 0, the solution (v

, p

) of the
problem (3-8) from Theorem 2.1 with M dened in (10) and S =
1

I strongly converges to the solution (v, p) of

Corollary 3.1 in W L
2
() when 0. Moreover, there exists a constant C(
f
,
p
, , , K
0
, K
1
) > 0 such that
the following error estimate holds, being the weak limit of
1

[[v

]]

in L
2
()
d
:
v

v
W
+ p
0
p
0

0,
C jj
0,

and j[[v

]]

j
0,
jj
0,
.
With additional regularity assumptions such that H
1
2
()
d
, then the previous estimate becomes optimal in O().
Skrrcn or raoor. The solution v

W satises with (9) the weak form below:

2
_

f
d(v

) : d(w) dx + 2
_

d(v

) : d(w) dx +
_

p
K
1
v

wdx +
_

Mv

ds +
1

[[v

]]

[[w]]

ds
=
_

f wdx, w W. (12)
By choosing w = v

, we get using the Korn and Friedrichs-Poincar e inequalities in

f
,
p
together with the inequality:
a b (a
2
+ b
2
)/2, a, b R:

0
_

p
v

2
dx +
0
K
0
_

p
v

2
dx +
_

Mv

ds +
1

[[v

]]

2
ds
c(
f
,
p
)

0
jfj
2
0,
.
4
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
With this bound, there exists v W such that, up to a subsequence, v

tends to v in W or H
1
(
f

p
)
d
weakly when
0 and strongly in L
2
()
d
. Indeed, since the trace application is continuous, we have: v

p
= 0. Moreover we
have: j[[v

]]

j
0,
c(
f
,
p
,
0
, f)

and thus [[v]]

= 0, v

= v

and v belongs to the subspace V of W. Then p

0
dened by Theorem 2.1 is bounded in L
2
0
() since we have using the Ne cas theorem [32, 19]:
jp
0
j
0,
c(
f
,
p
)
_
jp
0
j
1,
f
+ jp
0
j
1,
p
_
C jv

j
W
+ jfj
0,
. (13)
Thus, there exists p
0
L
2
0
() such that, up to a subsequence, p
0
tends to p
0
weakly in L
2
(). Now taking the limit
of (12) when 0, there exists L
2
()
d
such that
1

[[v

]]

tends weakly to in L
2
()
d
and we get that v is the
unique solution in V (the uniqueness being proved directly with f = 0 and w = v V W) satisfying:
2
_

f
d(v) : d(w) dx + 2
_

d(v) : d(w) dx +
_

p
K
1
v wdx +
_

Mv

ds +
_

[[w]]

ds
=
_

f wdx, w W. (14)
Hence, v V also satises (11) for all w V. Besides, using test functions w = (

c
compactly supported either
in
f
or in
p
and such that div = 0 in
f
or in
p
respectively, and using the Stokes formula, we get with the De
Rham theorem [32, 12] the existence and uniqueness (
f
and
p
being connected) of the pressure restrictions p
0
f
and p
0
p
in L
2
0
(
f
) and L
2
0
(
p
) respectively. This denes the pressure eld p
0
= p
0
f
+ p
0
p
in L
2
0
() over the whole
domain such that (v, p
0
) veries the Stokes/Brinkman equations (3-5) a.e. in
f

p
.
Then, we can dene the pressure eld p L
2
() with p
0
and the constant C
1
as in Corollary 3.1 such that the
stress jump condition (10) is veried in H

1
2
()
d
. Moreover, the constant C
1

dened in Theorem 2.1 with (v

, p
0
)
satises: lim
0
C
1

= C
1
with the weak limits of (v

, p
0
) and the continuity of the trace applications. We can also give an
interpretation of . By writing the dierence between the weak form of problem (3-6,10) with test functions w W
using the Stokes formula and the limit weak form (14), it yields:
_
(v, p
0
) n

, [[w]]

1
2
,
= 0, w W. By
constructing an ad-hoc divergence-free extension in W of any function u in H
1
2
()
d
, as for [6, Theorem 1.1] (see also
[12, chap. III] for the Stokes/Neumann problem with a stress boundary condition), we dene the constant C
0
= lim
0
C
0

below, C
0

dened in Theorem 2.1 with (v

, p
0
), such that we have = (v, p
0
+ C
0
) n

in the sense of H

1
2
()
d
:
C
0
=
1

_
(v, p
0
) n

, n
_

1
2
,
, such that
_
(v, p
0
+ C
0
) n

, u
_

1
2
,
= 0, u H
1
2
()
d
.
To prove the strong convergence and the error estimate, we rst write the error equation being the dierence between
(9) satised by v

for all w W and (14) using the fact that: [[v]]

= 0 and v

= v

. Then, choosing w = v

v, we
get with the Cauchy-Schwarz inequality:
2
0
_

p
d(v

v)
2
dx +
0
K
0
_

p
v

v
2
dx + M
0
_

v
2
ds +
1

[[v

v]]

2
ds jj
0,
j[[v

v]]

j
0,
which simply gives using the Korn and Poincar e inequalities in
f
and
p
:
j[[v

]]

j
0,
= j[[v

v]]

j
0,
jj
0,
and jv

vj
W
C(
f
,
p
,
0
) jj
0,

. (15)
If belongs to H
1
2
()
d
, the last error estimate can be improved up to O() by constructing some adequate extensions
from in the subdomains
f
and
p
. Finally, the pressure estimate is obtained using the Ne cas theorem and we get:
jp
0
p
0
j
0,
c(
f
,
p
)
_
j(p
0
p
0
)j
1,
f
+ j(p
0
p
0
)j
1,
p
_
C jv

vj
W
,
which completes the proof. 2
5
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
4. The Stokes/Darcy problem with Beavers & Joseph interface conditions
We consider the problem (3-8) with the Dirichlet boundary condition (6) on
p
replaced by the stress boundary
condition of Neumann where is the outward unit normal vector on
p
and q H

1
2
(
p
)
d
given, e.g. q = p
e
:
v = 0 on
f
and (v
p
, p
p
) = p
p
+ v
p
= q on
p
. (16)
Let us dene the Hilbert space W
N
equipped with the natural inner product and norm in H
1
(
f

p
)
d
:
W
N
w L
2
()
d
, w

f
H
1
0
f
(
f
)
d
and w

p
H
1
(
p
)
d
; w = 0 in
f

p
.
Then, the following well-posedness result is ensured as a corollary of Theorem 2.1, see also [6, Theorem 2.1].
Corollary 4.1 (Global solvability of the Stokes/Brinkman model with J.E.B.C. and stress B.C.). With the assump-
tions of Theorem 2.1 and q H

1
2
(
p
)
d
, there exists a unique solution (v, p) W
N
L
2
() satisfying the weak form:
a(v, w) = l(w) +(q, w)

1
2
,
p
for all w W
N
with p
f
= p
f
0
+C
0
+C
1
/2 and p
p
= p
p
0
+C
0
C
1
/2 where p
0
L
2
0
() and
the constants C
0
, C
1
are dened as in Theorem 2.1 such that the equations (3-5) hold almost everywhere in
f

p
and (7-8) are satised in H

1
2
()
d
. Then, if the following compatibility condition holds:
C
0

1
2
C
1
= C
N
with C
N
=
1

((v, p
0
) q, )

1
2
,
p
,
the stress boundary condition (16) is also satised in H

1
2
(
p
)
d
and (v, p) W
N
L
2
() is the unique solution of the
problem (3-5,7-8,16).
For any > 0, let us now consider the solution (v

, p

) W
N
L
2
() of the problem (3-5,7-8,16) with a vanishing
viscosity = for the Brinkman problem in
p
. The condition (16) avoids the creation of a spurious boundary layer
along
p
for the Darcy problem when 0. The J.E.B.C. (7-8) are also calibrated as follows to obtain interface
conditions of Beavers & Josephs type [11] with a jump of tangential velocity (2) allowing a possible pressure jump:
[[(v, p) n]]

= Mv

with M
j j
= 0, j = 1, , d 1, M
dd
=

n

K
n
on , (17)
(v, p) n

= S[[v]]

with S
j j
=

, j = 1, , d 1, S
dd
=
1

on , (18)
where M, S are positive diagonal matrices with

=
bj
,
n
0 a.e. on and K

, K
n
permeability coecients.
Let us dene the Hilbert spaces
W
S/D

_
w L
2
()
d
, w

f
H
1
0
f
(
f
)
d
, w

p
L
2
(
p
)
d
; w = 0 in
f

p
_
equipped with the natural inner product and norm in H
1
(
f
)
d
L
2
(
p
)
d
and
W
SD

_
w W
S/D
; w L
2
(), [[w]]

L
2
()
d
, [[w n]]

= 0
_
equipped with the norm dened by: jwj
2
W
SD
= jwj
2
1,
f
+ jwj
2
0,
p
+ j wj
2
0,
+ j[[w]]

j
2
0,
.
We now prove the following convergence result which also ensures the well-posedness of the Stokes/Darcy pro-
blem with Beavers & Josephs type interface conditions (2,17) whatever the coecients

,
n
0 a.e. on .
Theorem 4.2 (Convergence to Stokes/Darcy problem with B-J). With the data f L
2
()
d
and q = 0, the solution
(v

, p

) in W
N
L
2
() for any > 0 from Corollary 4.1 of the problem (3-5,16,17,18) with a vanishing viscosity =
weakly converges to the solution (v, p) in W
S/D
L
2
() of the Stokes/Darcy problem with the interface conditions
(2,17) on when 0. Indeed, in the porous domain
p
, v
p
and p
p
satisfy the Darcy equation, i.e. Eq. (4) with
= 0, and p
p
belongs to H
1
(
p
) such that p
p
= 0 on
p
.
With additional regularity assumptions such that v
p
H
1
(
p
)
d
, then v W
SD
W
N
and we have the global error
estimate with C > 0 depending on the data, jvj
0,
p
, jj
0,
and dened as the weak limit of
1

[[v

n]]

in L
2
():
v

v
1,
f
+

v
1,
p
+v

v
0,
p
+ p
0
p
0

0,
C jj
0,

and j[[v

n]]

j
0,

_
2jvj
2
0,
p
+ jj
2
0,
_ 1
2
.
6
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
Skrrcn or raoor. The proof is here abridged without explaining most of the arguments already detailed in the proof
of Theorem 3.2. From (3-5,16,17,18) with the Stokes formula, the solution v

W
N
satises the weak form below:
2
_

f
d(v

) : d(w) dx + 2
_

p
d(v

) : d(w) dx +
_

p
K
1
v

wdx +
_

Mv

ds
+
d1

j=1
_

S
j j
[[v

j
]]

[[w
j
]]

ds +
1

[[v

n]]

[[w n]]

ds =
_

f wdx, w W
N
. (19)
By choosing w = v

, we get using the Korn inequality in

f
,
p
and Poincar e inequality in
f
:

0
_

f
v

2
dx + 2
_

p
v

2
dx +

0
K
0
2
_

p
v

2
dx + M
0
_

2
ds + S
0
d1

j=1
_

[[v

j
]]
2

ds
+
1

[[v

n]]
2

ds c(
f
,
p
,
0
, K
0
) jfj
2
0,
. (20)
With this bound, there exists v W
S/D
and v H
1
(
p
)
d
such that, up to a subsequence, v

tends to v in W
S/D
or
H
1
(
f
)
d
L
2
(
p
)
d
weakly when 0 (strongly in L
2
(
f
)
d
) and

v
p

tends to v in H
1
(
p
)
d
weakly. Indeed, since
the trace application is continuous, we have: v

f
= 0. Moreover we have: j[[v

n]]

j
0,
c(
f
,
0
, K
0
, f)

and thus
[[v n]]

= 0, v n

= v n

in L
2
(). Since [[v

]]

is bounded in L
2
() (for

> 0 and thus S

0
> 0) and because
v
f
H
1
(
f
)
d
has a trace in H
1
2
()
d
, there exists v

L
2
()
d
dened as the weak limit of the trace v
p

in L
2
()
d
.
Hence we dene the tangential velocity jump: [[v ]]

= (v
f

) L
2
() and we have v W
SD
.
Then p
0
dened by Corollary 4.1 is bounded in L
2
0
() because, using the Ne cas theorem as for (13), we have:
jp
0
j
0,
c(
f
,
p
)
_
jp
0
j
1,
f
+ jp
0
j
1,
p
_
C, since jv

j
1,
f
,

jv

j
1,
p
and jv

j
0,
p
are all bounded. Thus,
there exists p
0
L
2
0
() such that, up to a subsequence, p
0
weakly tends to p
0
in L
2
().
Now taking the limit of (19) when 0, there exists L
2
() such that
1

[[v

n]]

weakly tends to in L
2
()
and we get that v is the unique solution in W
SD
satisfying the weak form:
2
_

f
d(v) : d(w) dx +
_

p
K
1
v wdx +
_

Mv

ds +
d1

j=1
_

S
j j
[[v
j
]]

[[w
j
]]

ds +
_

[[w n]]

ds
=
_

f wdx, w W
N
. (21)
The existence and uniqueness of the solution v W
SD
to the above problem can be also a priori ensured by the
generalized Lax-Milgram theorem of Ne cas [27] with an inf-sup stability inequality. Besides, using test functions
w = (

c
compactly supported either in
f
or in
p
and such that div = 0 in
f
or in
p
respectively,
and using the Stokes formula, we get with the De Rham theorem the existence and uniqueness (
f
and
p
being
connected) of the pressure restrictions p
0
f
and p
0
p
in L
2
0
(
f
) and L
2
0
(
p
) respectively. This denes the pressure
eld p
0
= p
0
f
+ p
0
p
in L
2
0
() over the whole domain such that (v, p
0
) veries the Stokes/Darcy equations (3-5)
a.e. in
f

p
with = 0 in (4), i.e. the Darcy equation. Because of uniqueness, the whole sequence (v

, p
0
) weakly
converges to (v, p
0
) in W
S/D
L
2
0
().
Then, to satisfy the interface conditions (17,18) on , i.e. in H

1
2
()
d
, the pressure eld p L
2
() must be
adjusted from the zero-average pressure p
0
L
2
0
() such that: (p p
0
)

= C
0
and [[p p
0
]]

= C
1
, where the
constants C
0
, C
1
are calculated as in Theorem 2.1 with (v, p
0
) above dened. Since f
p
, v
p
L
2
(
p
)
d
, we have by the
Darcy equation that p
p
belongs to H
1
(
p
). The limit boundary condition (16) which reduces to: p
p

p
= 0 in H
1
2
(
p
)
can be also satised if the following compatibility condition holds:
C
0

1
2
C
1
= C
N
with C
N
=
1

p
p
0
ds, (22)
7
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
such that: p
f
= p
f
0
+ C
0
+ C
1
/2 and p
p
= p
p
0
+ C
N
dene the pressure solution p L
2
(
f
) H
1
0
p
(
p
). We can also
interpret in a similar way as in the proof of Theorem 3.2.
Now, if v
p
belongs to H
1
(
p
)
d
with sucient regularity assumption, then v W
SD
W
N
, v

= v
p

H
1
2
()
d
and we prove the strong convergence and a global error estimate in . The dierence equation between (19) and (21)
reads: for all w W
N
,
2
_

f
d(v

v) : d(w) dx + 2
_

p
d(v

v) : d(w) dx +
_

p
K
1
(v

v) wdx +
_

M(v

v)

ds
+
d1

j=1
_

S
j j
[[(v

v)
j
]]

[[w
j
]]

ds +
1

[[v

n]]

[[w n]]

ds = 2
_

p
d(v) : d(w) dx
_

[[w n]]

ds. (23)
Then, choosing w = (v

v) W
N
with [[v n]]

= 0, we get the error estimate for the velocity:

2
0
jd(v

v)j
2
0,
f
+ jd(v

v)j
2
0,
p
+
0
K
0
jv

vj
2
0,
p
+ M
0
j(v

v)

j
2
0,
+ S
0
d1

j=1
j[[(v

v)
j
]]

j
2
0,
+
1
2
j[[v

n]]

j
2
0,

1
2
_
2jvj
2
0,
p
+ jj
2
0,
_
(24)
which yields the result with the Korn and Poincar e inequalities in
f
or
p
. Finally, the pressure estimate is obtained
using the Ne cas theorem and we get with the Stokes and Darcy equations:
jp
0
p
0
j
0,
C
_
jv

vj
1,
f
+ jv

vj
0,
p
+ jv

j
0,
p
_
, (25)
which concludes the proof with (24) since

jv

j
0,
p
is bounded with (20). We thus obtain the given error estimate,
typical of the existence of a spurious boundary layer in this singular perturbation problem, see e.g. [26], as for the
L
2
-penalty method analysed in [3, 16]. 2
Acknowledgments
Special thanks to one of the reviewers for his valuable comments which have improved the preprint version.
References
[1] G. Aiixrar, Homogenization of the Navier-Stokes equations in open sets perforated with tiny holes. I: Abstract framework, a volume distri-
bution of holes. Arch. Ration. Mech. Anal. 113(3), 209-259, 1991.
[2] Pn. Axoor, Etude dune cellule mod elisant un bicouche uide-poreux - Probl` emes dinterface & an emom etrie laser-D oppler, M emoire de
DEA de M ecanique, Universit e Bordeaux I (www.tree.u-bordeaux1.fr), juin 1985.
[3] Pn. Axoor, Analysis of singular perturbations on the Brinkman problem for ctitious domain models of viscous ows, Math. Meth. in the
Appl. Sci. (M
2
AS ) 22(16), 1395-1412, 1999.
[4] Pn. Axoor, A model of fracture for elliptic problems with ux and solution jumps, C. R. Acad. Sci. Paris, Ser. I Math. 337(6), 425-430, 2003.
[5] Pn. Axoor, A unied ctitious domain model for general embedded boundary conditions, C. R. Acad. Sci. Paris, Ser. I Math. 341(11),
683-688, 2005.
[6] Pn. Axoor, A ctitious domain model for the Stokes/Brinkman problem with jump embedded boundary conditions, C. R. Math. Acad. Sci.
Paris, Ser. I 348(11-12), 697-702, 2010.
[7] Pn. Axoor, Well-posed Stokes/Brinkman and Stokes/Darcy problems for coupled uid-porous viscous ows, American Institute of Physics
(AIP) Conference Proceedings, 8th ICNAAM 2010, Rhodes (Greece), 19-25 sept. 2010.
[8] Pn. Axoor, F. Bo.ra, F. Htarar, Numerical modelling of ow in fractured porous media, in Finite Volumes for Complex Applications IV
Marrakech, F. Benkhaldoun, D. Ouazar, S. Raghay (Eds), pp. 249-260, Hermes Science, 2005.
[9] Pn. Axoor, F. Bo.ra, F. Htarar, Asymptotic and numerical modelling of ows in fractured porous media, Math. Model. Numer. Anal.
(M
2
AN) 43(2), 239-275, 2009.
[10] J.-L. Atarxtir, On the domain of validity of Brinkmans equation, Transport in Porous Media 79(2), 215-223, 2009.
[11] G.S. Brxvras, D.D. Josrrn, Boundary conditions at a naturally permeable wall, J. Fluid Mech. 30, 197-207, 1967.
[12] F. Bo.ra, P. Fxaarr,

El ements danalyse pour l etude de quelques mod` eles d ecoulements de uides visqueux incompressibles, Math ematiques
& Applications 52, Springer-Verlag, 2006.
[13] H.C. Barxkmxx, A calculation of the viscous force exerted by a owing uid on a dense swarm of particules, Appl. Sci. Res. A(1), 27-34,
1947.
8
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1
A
C
C
E
P
T
E
D
M
A
N
U
S
C
R
I
P
T
ACCEPTED MANUSCRIPT
[14] H.C. Barxkmxx, On the permeability of media consisting of closely packed porous particules, Appl. Sci. Res. A(1), 81-86, 1947.
[15] Y. Cxo, M. Gtxzataora, F. Htx, X. Wxxo, Coupled Stokes-Darcy model with Beavers-Joseph interface boundary condition, Comm. Math.
Sci. 8(1), 1-25, 2010.
[16] G. Cxaaot, P. Fxaarr, Boundary layer for a penalization method for viscous incompressible ow, Adv. Di. Eqs. 8(12), 1453-1480, 2003.
[17] M. Cnxxorsars, D. Jxmrr, Boundary conditions at a planar uid-porous interface for a Poiseuille ow, Int. J. Heat Mass Transfer 49(13-14),
2137-2150, 2006.
[18] M. Drscxccrxrr, A. Qtxarraoxr, Navier-Stokes/Darcy coupling: modeling, analysis and numerical approximation, Rev. Math. Complut. 22(2),
315-426, 2009.
[19] V. Graxtir xxo P.A. Rxvrxar, Finite Element Methods for the Navier-Stokes Equations, Springer Series in Comput. Math., 5, Springer-Verlag,
1986 (1rst edn. 1979).
[20] B. Go.rxt, D. Lntriirra, D. Goarx, M.G. Vrixaor, Momentum transport at a uid-porous interface, Int. J. Heat Mass Transfer 46, 4071-4081,
2003.
[21] P. Garsvxao, Elliptic problems in nonsmooth domains, Monographs and Studies in Mathematics 24, Adv. Publish. Prog., Pitman (Boston),
1985.
[22] U. Hoaxtxo (Eo.), Homogenization and porous media, Interdisciplinary Applied Mathematics 6, Springer-Verlag (New York), 1997.
[23] W. J xora, A. Mrkrir c, On the interface boundary condition of Beavers & Joseph and Saman, SIAM J. Appl. Math. 60(4), 1111-1127, 2000.
[24] W.L. Lx.rox, F. Scnrrwrck, I. Yorov, Coupling uid ow with porous media ow, SIAM J. Numer. Anal. 40, 2195-2218, 2003.
[25] J.-L. Lroxs, Probl` emes aux limites dans les equations aux d eriv ees partielles, Presses de lUniversit e de Montr eal, 1965.
[26] J.-L. Lroxs, Perturbations singuli` eres dans les probl` emes aux limites et en contr ole optimal, Lect. Notes in Math. 323, Springer (Berlin),
1973.
[27] J. Nr cxs, Les m ethodes directes en th eorie des equations elliptiques, Masson (Paris), 1967.
[28] J.A. Ocnox-Txrrx, S. Wnrrxkra, Momentum transfer at the boundary between a porous medium and a homogeneous uid - I. Theoretical
development, Int. J. Heat Mass Transfer 38, 2635-2646, 1995.
[29] J.A. Ocnox-Txrrx, S. Wnrrxkra, Momentum transfer at the boundary between a porous medium and a homogeneous uid - II. Comparison
with experiment, Int. J. Heat Mass Transfer 38, 2647-2655, 1995.
[30] L.E. Px.xr, B. Sraxtonxx, Analysis of the boundary condition at the interface between a viscous uid and a porous medium and related
modelling questions, J. Math. Pures Appl. 77, 317-354, 1998.
[31] P.G. Sxrrmxx, On the boundary condition at the surface of a porous medium, Stud. Appl. Math. L(2), 93-101, 1971.
[32] R. Trmxm, Navier-Stokes Equations; Theory and Numerical Analysis, North-Holland, 1986 (1rst edn. 1977).
[33] F.J. Vxio rs-Pxaxox, J. Aivxarz-Rxmarz, B. Go.rxt, J.A. Ocnox-Txrrx, Computation of jump coecients for momentum transfer between a
porous medium and a uid using a closed generalized transfer equation, Transp. in Porous Media 78, 439-457, 2009.
9
h
a
l
-
0
0
4
7
6
3
8
6
,

v
e
r
s
i
o
n

2

-

2
3

J
u
l

2
0
1
1