MOXReport No.

97 dicembre 2006

Mathematical analysis of the

Navier-Stokes/Darcy coupling

Lori Badea, Marco Discacciati, Alfio Quarteroni

MOX, Dipartimento di Matematica F. Brioschi

Politecnico di Milano, Via Bonardi 9 - 20133 Milano (Italy)
mox@mate.polimi.it

http://mox.polimi.it

Mathematical analysis of the

Navier-Stokes/Darcy coupling
Lori Badea1 , Marco Discacciati2 , Alfio Quarteroni3
1. Institute of Mathematics of the Romanian Academy,
P.O. Box 1-764, 014700 Bucharest, Romania. Lori.Badea@imar.ro
2. RICAM - J. Radon Institute for Computational and Applied Mathematics,
Altengerberstrae 69, A-4040 Linz, Austria. marco.discacciati@oeaw.ac.at
3. IACS - Chair of Modeling and Scientific Computing, EPFL, CH-1015 Lausanne,
Switzerland, and MOX, Politecnico di Milano, P.zza Leonardo da Vinci 32, I-20133
Milano, Italy. alfio.quarteroni@epfl.ch

Abstract
We consider a differential system based on the coupling of the Navier-Stokes
and Darcy equations for modeling the interaction between surface and subsurface flows. We formulate the problem as an interface equation, we analyze
the associated (nonlinear) Steklov-Poincare operators, and we prove its wellposedness.

Introduction and setting of the problem

Let Rd (d = 2, 3) be a bounded domain, decomposed as the union of

two non intersecting subdomains f and p separated by an interface , i.e.
=
f
p , f p = and
f
p = . We suppose the boundaries f

and p to be Lipschitz continuous. From the physical point of view, is a

surface separating an upper domain f filled by a fluid, from a lower domain
p formed by a porous medium. We assume that the fluid contained in f has
a fixed upper surface (i.e., we do not consider the case of free-surface fluid) and
can filtrate through the underlying porous medium.
In order to describe the motion of the fluid in f , we introduce the Navier-Stokes
equations: t > 0,
t uf T(uf , pf ) + (uf )uf
uf

=
=

f
0

in f ,
in f ,

(1)

where T(uf , pf ) = (uf + T uf ) pf I is the Cauchy stress tensor, > 0

is the kinematic viscosity of the fluid, while uf and pf are the fluid velocity
and pressure, respectively; is the gradient operator with respect to the space
coordinates.
In the lower domain p , we define the piezometric head = z + pp /(f g), where
z is the elevation from a reference level, pp is the pressure of the fluid in p , f
its density and g is the gravity acceleration.
1

The fluid motion in p is described by the equations:

nup
up

=
=

K
0

in p ,
in p ,

(2)

where up is the fluid velocity, n is the volumetric porosity and K is the hydraulic
conductivity tensor K = diag(K1 , . . . , Kd ) with Ki L (p ), i = 1, . . . , d.
The first equation is Darcys law. In the following, we shall denote K = K/n =
diag(Ki /n) (i = 1, . . . , d).
For the sake of clarity, in our analysis we shall adopt homogenous boundary conditions. The treatment of non-homogeneous conditions involves some additional
technicalities, but neither the guidelines of the theory nor the final results are
affected. We refer the reader to [7]. In particular, for the Navier-Stokes problem
we impose the no-slip condition uf = 0 on f \, while for the Darcy problem,
we set the piezometric head = 0 on bp and we require the normal velocity to
be null on p : up np = 0 on p , where p = bp p (see Fig. 1). np
and nf denote the unit outward normal vectors to the surfaces p and f ,
respectively, and we have nf = np on . We suppose nf and np to be regular
enough, and we indicate n = nf for simplicity of notation.
f
f
np

nf

bp
Figure 1: Schematic representation of a 2D vertical section of the computational
domain

We supplement the Navier-Stokes and Darcy problems with the following conditions on :
up n = uf n ,
i (T(uf , pf ) n) = uf i ,
n (T(uf , pf ) n)

= g ,

i = 1, . . . , d 1 ,

(3)
(4)
(5)

where i (i = 1, . . . , d 1) are linear independent unit tangential vectors to

the boundary , and is the characteristic length of the pores of the porous
medium.
Conditions (3)-(5) impose the continuity of the normal velocity on , as well as
that of the normal component of the normal stress, but they allow pressure to
be discontinuous across the interface.
A rigorous mathematical justification of these interface conditions, usually denoted as Beavers-Joseph-Saffman conditions, can be found in [12, 13, 14].
2

The same interface conditions have been considered in [8, 9, 10, 15] for the
coupling of Stokes and Darcy equations.
From Sect. 1.1 on, we focus on the steady problem obtained by dropping the
time derivative in the momentum equation (1).
Then, after writing the weak form of the global problem (Sect. 1.1), we reformulate it as a nonlinear problem in one sole interface unknown. In Sect. 3,
we introduce and analyze some nonlinear extension operator that will be used
in Sect. 4 to write the nonlinear interface problem S = 0, whose unknown
solution is the common value of the normal velocity uf n = up n across .
S is a nonlinear Steklov-Poincare operator. Finally, in Sect. 4.1, we prove the
well-posedness of this interface problem.

1.1

Weak form of the global problem

From now on, we shall consider the steady case for the Navier-Stokes equations
and, instead of (2), we will use the following formulation for Darcy problem:
(K) = 0

find :

in p .

(6)

We define the functional spaces:

Hf = {v (H 1 (f ))d : v = 0 on f \ },
Hp = { H 1 (p ) : = 0 on bp },

(7)
(8)

and let Q = L2 (f ). We denote by | |1 the H 1 -seminorm, and by k k0 the

L2 -norm; it will always be clear form the context whether we are referring to
the norms and seminorms in f or p .
Then, we introduce the bilinear forms
Z

af (v, w) =
(v + T v) (w + T w)
v, w (H 1 (f ))d , (9)
2
f
Z
qv
v (H 1 (f ))d , q Q ,
(10)
bf (v, q) =
f

ap (, ) =

, H 1 (p ) ,

(11)

and, for all v, w, z (H 1 (f ))d , the trilinear form

cf (w; z, v) =

[(w )z] v =

d Z
X

i,j=1

wj

zi
vi .
xj

(12)

The coupling conditions (3)-(5) can be incorporated in the weak formulation of

the Navier-Stokes/Darcy problem as natural conditions on . In fact, the latter
reads:
find uf Hf , pf Q, Hp such that for all v Hf , q Q, Hp ,
af (uf , v) + cf (uf ; uf , v) + bf (v, pf ) + g ap (, )
Z X
Z
Z
Z
d1

+ g (v n) g (uf n) +
f v , (13)
(uf j )(v j ) =
j=1

bf (uf , q) = 0 .

(14)

In the next sections, we will consider the issue of the well-posedness of (13)-(14).
In our analysis, we shall use some classical existence and uniqueness results for
nonlinear problems, that we anticipate, for the sake of clarity, in the following
section.

General existence and uniqueness results

In this section we recall some existence and uniqueness results for nonlinear
saddle-point problems, referring the reader to, e.g., [3, 4, 5, 6, 11] for a rigorous
study.
Let (X, k kX ) and (Y, k kY ) be two real Hilbert spaces. Consider a bilinear
continuous form b(, ) : X Y R, (v, q) b(v, q), and a trilinear form
a(; , ) : X X X R, (w, u, v) a(w; u, v), where, for w X the mapping
(u, v) a(w; u, v) is a bilinear continuous form on X X.
Then, we consider the following problem: given l X , find a pair (u, p) X Y
satisfying
a(u; u, v) + b(v, p) = hl, vi
v X,
(15)
b(u, q) = 0
q Y.
Introducing the linear operators A(w) L(X; X ) for w X, and B L(X; Y ):
hA(w)u, vi = a(w; u, v) u, v X ,
hBv, qi = b(v, q)

v X, q Y ,

problem (15) becomes: find (u, p) X Y such that

A(u)u + B T p
Bu

in X ,
in Y .

= l
= 0

(16)

Taking V = Ker(B), we associate (15) with the problem:

find u V :

a(u; u, v) = hl, vi

v V ,

(17)

or, equivalently: find u V s.t. A(u)u = l in V , where the linear operator

L(X ; V ) is defined by h l, vi = hl, vi, v V .
If (u, p) is a solution of problem (15), then u solves (17). The converse may be
proven provided an inf-sup condition holds. Indeed, the following results can be
proved.
Theorem 2.1 (Existence and uniqueness) Suppose that:
1. the bilinear form a(w; , ) is uniformly elliptic in the Hilbert space V with
respect to w, i.e. there exists a constant > 0 such that
a(w; v, v) kvk2X

v, w V ;

2. the mapping w A(w) is locally Lipschitz-continuous in V , i.e. there

exists a continuous and monotonically increasing function L : R+ R+
such that for all m > 0
|a(w1 ; u, v) a(w2 ; u, v)| L(m)kukX kvkX kw1 w2 kX
u, v V , w1 , w2 Sm with Sm = {w V : kwkX m};
4

(18)

3. it holds

k lkV
L
2

k lkV

< 1.

(19)

Then, problem (17) has a unique solution u V .

We consider now problem (15).
Theorem 2.2 Assume that the bilinear form b(, ) satisfies the inf-sup condition: > 0
b(v, q)
.
(20)
inf sup
qY vX kvkX kqkY
Then, for each solution u of (17) there exists a unique p Y such that the pair
(u, p) is a solution of (15).

Some nonlinear extension operators: definition and analysis

In this section we apply domain decomposition methods at the differential level

to study the Navier-Stokes/Darcy problem. We identify the subdomains with
f and p , then we introduce and analyze some nonlinear extension operators
that will be used in Sect. 4 to write the Steklov-Poincare interface equation
associated to the coupled problem.
In our analysis we adopt the interface condition
uf j = 0 on ,

j = 1, . . . , d 1,

(21)

instead of (4). This simplification is acceptable from the physical viewpoint

since the term in (4) involving the normal derivatives of uf is multiplied by ,
and the velocity uf can be supposed of order O() in the neighborhood of
(see [13]), so that the left hand side can be approximated to zero. We point
out that this simplification does not alter the coupling structure, since (4) is a
boundary condition for the fluid problem in f and not a coupling condition.
1/2
We consider the trace space = H00 () (see [16]) and the spaces
Hf = {v Hf : v j = 0 on , j = 1, . . . , d 1} ,

Hp0

= { Hp : = 0 on } .

(22)
(23)

Moreover, we consider two linear continuous extension operators:

R1 : Hf
R2 : H

1/2

such that (R1 ) n = on

() Hp

such that R2 = on ,

(24)
1/2

() . (25)

We can prove the following result (see [7]).

Proposition 3.1 The coupled Navier-Stokes/Darcy problem (13)-(14) can be
reformulated in the equivalent multidomain form:

find uf Hf , pf Q, Hp such that

af (uf , w) + cf (uf ; uf , w) + bf (w, pf )=
bf (uf , q) = 0
q Q ,
ap (, ) = 0
Hp0 ,
Z
(uf n) = ap (, R2 )
,
Z
Z
f (R1 ) af (uf , R1 )
g =

f w

w (H01 (f ))d , (26)

(27)
(28)
(29)

cf (uf ; uf , R1 ) bf (R1 , pf )

(30)

Our aim is now to reformulate the coupled problem (26)-(30) as an interface

equation in a scalar unknown defined on corresponding to the trace of the
fluid normal velocity uf n on . First of all, we need to introduce and analyze
some further extension operators.
Let us consider the (unknown) interface variable = (uf n)| .
Due to the incompressibility constraint in f and to the boundary conditions
imposed on f \ , it must be 0 with


Z
=0 .
(31)
0 = :

Then, let us define the linear extension operator:

Rf = (R1f , Rf2 ),

Rf : 0 Hf Q0 ,

(32)

satisfying (R1f ) n = on , and

w (H01 (f ))d ,
af (R1f , w) + bf (w, Rf2 ) = 0
(33)
1
q Q0 ,
bf (Rf , q) = 0
R
where Q0 = {q Q : f q = 0}. Moreover, we consider the linear extension
operator
Rp : 0 Hp ,
Rp
(34)
such that
ap (Rp , ) =

Hp .

(35)

Finally, let us introduce the following nonlinear extension operator:

Rf : 0 Hf Q0 ,

Rf = (R1f , R2f )

such that (R1f ) n = on , and, for all v (H01 (f ))d , q Q0 ,

af (R1f , v) + cf (u + R1f ; u + R1f , v) + bf (v, R2f ) =
bf (R1f , q) =

0,
0,

where u (H01 (f ))d satisfies the linear auxiliary Stokes problem:

6

(36)

find u (H01 (f ))d , Q0 :

af (u , v) + bf (v, ) =

f v

v (H01 (f ))d ,

(37)

q Q0 .

bf (u , q) = 0

Let us finally recall that the following estimate holds:

2
kf k0 .
C

|u |1

3.1

(38)

Existence and uniqueness of the operator Rf

We face now the issue of the existence and uniqueness of the extension operator Rf . With this purpose, we define the auxiliary (homogeneous) nonlinear
operator
R0 = (R10 , R20 ),

R0 : 0 (H01 (f ))d Q0 ,

with Ri0 = Rif Rfi ,

(39)

i = 1, 2.

Clearly, R10 n = 0 on , and it satisfies:

af (R10 , v) + cf (u + R1f + R10 ; u + R1f + R10 , v)
+bf (v, R20 ) =
bf (R10 , q)

0,

(40)

0,

for all v (H01 (f ))d , q Q0 . Remark that problem (40) is analogous to (36),
but here R10 (H01 (f ))d , while R1f Hf .
We consider the functional space
Vf0 = {v (H01 (f ))d : v = 0

in f } ,

(41)

and, given 0 , we define the form:

a(w; z, v) = af (z, v) + cf (w; z, v) + cf (u + R1f ; z, v)
+ cf (z; u + R1f , v)

w, z, v (H 1 (f ))d , (42)

and the functional

h, vi = cf (u + R1f ; u + R1f , v)

v (H 1 (f ))d .

(43)

Thus, we can rewrite (40) as: given 0 ,

a(R10 ; R10 , v) = h, vi

find R10 Vf0 :

v Vf0 .

(44)

Finally, let us recall some useful inequalities: the Poincare inequality (see, e.g.,
[17] p. 11)
Cf > 0 : kvk0 Cf |v|1
v Hf ,
(45)
the Korn inequality (see, e.g., [18] p. 149): v = (v1 , . . . , vd ) Hf
C > 0 :

2
d 
X
vl
vj
+
C kvk21 ,
xl
xj

f j,l=1

(46)

and

C > 0 :

1 |z|1 |v|1
|c(w; z, v)| C|w|

w, z, v Hf ,

(47)
1

which follows from the Poincare inequality (45) and the inclusion (H (f ))d
(L4 (f ))d (for d = 2, 3) due to the Sobolev embedding theorem (see [1]).
We can now state the following result.
Proposition 3.2 Let f L2 (f ) such that

kf k0
1 C2
<

,
2
8 C

(48)

where C and C are the constants introduced in (46) and (47), respectively. If


C
2
1
(49)
kf k0 ,
0 : |Rf |1 <

C
4C
then, there exists a unique nonlinear extension Rf = (R1f , R2f ) Hf Q0 .

Remark 3.1 Notice that (49) imposes a constraint on . In particular, since

the norms |R1f |1 and kk are equivalent (see [9], Lemma 4.1), this condition
implies that a unique extension Rf exists, provided the norm of is small
enough. In our specific case, this means that we would be able to consider an
extension Rf only if the normal velocity across the interface is sufficiently
small.
Finally, remark that (48) guarantees that the radius of the ball in (49) is positive.
Proof. The proof is made of several steps and it is based on Theorems 2.1-2.2.
1. Let v, w Vf0 and 0 . Then, we have
a(w; v, v) = af (v, v) + cf (w; v, v)
+ cf (u + R1f ; v, v) + cf (v; u + R1f , v).

(50)

Integrating by parts and recalling that w Vf0 , then

Z
Z
1
1
2
cf (w; v, v) =
w n|v|
w|v|2 = 0 ,
2 f
2 f
where |v| is the Euclidian norm of the vector v. Moreover, denoting by nj the
components of the unit outward normal vector nf to f , we have
Z X
d
(u + R1f )i
1
vj
vi
cf (v; u + Rf , v) =
xj
f i,j=1
=
=
+

d Z
X

i,j=1

d Z
X

i,j=1

d Z
X

i,j=1

d Z
X

(vi vj )(u + R1f )i nj

(vi vj )(u + R1f )i +
xj

f
i,j=1
d Z
X
vj
vi
vj (u + R1f )i
vi (u + R1f )i
xj
x
j

f
i,j=1

vj nj (u + R1f )i vi

= cf (v; v, u + R1f ).
8

Finally, (u + R1f ) = 0 by construction, so that cf (u + R1f ; v, v) = 0.

Then (50) becomes:
a(w; v, v) = af (v, v) cf (v; v, u + R1f ) ,

(51)

and using the inequalities (46) and (47) we obtain:

a(w; v, v)

C 2
2 |u + R1 |1
|v| C|v|
1
f
2  1

C
1
2

C(|u |1 + |Rf |1 )
|v|1
2



C
2kf k0
1
2

|v|1
,
+ |Rf |1
C
2
C

the last inequality following from (38). Then, thanks to (49), the bilinear form
a(w; , ) is uniformly elliptic on Vf0 with respect to w, with constant a (independent of w):


C
2kf k0
1

+ |Rf |1 .
C
a =
2
C
2. Still using (47), we easily obtain:
1 w2 |1 |v|1 |z|1 .
|a(w1 ; z, v) a(w2 ; z, v)| = |cf (w1 w2 ; z, v)| C|w
3. We have
k k(Vf0 )

| cf (u + R1f ; u + R1f , v)|

|v|1
vVf0 ,v6=0
sup

+ R1 |2 |v|1
C|u
f 1
|v|1
vVf0 ,v6=0
2

2
1

kf k0 + |Rf |1 ,
C
C

sup

so that
C

k k(Vf0 )
2a

<1

owing to (49).
4. Thanks to (49) and 13, a(; , ) and satisfy the hypotheses of Theorem 2.1,
which allows us to conclude that there exists a unique solution R10 Vf0 to
(44).
5. Since the inf-sup condition is satisfied, Theorem 2.2 guarantees that there
exists a unique solution (R10 , R20 ) to (40). The thesis follows from (39). 2

The interface equation associated to the coupled problem

In this section we reformulate the global coupled problem (26)-(30) as an interface equation depending solely on = (uf n)| .
9

We formally define the nonlinear pseudo-differential operator S : 0 0 ,

hS, i

af (R1f + u , R1 ) + cf (R1f + u ; R1f + u , R1 )

Z
+bf (R1 , R2f + )
f (R1 )
f

g(Rp )

0 , .

(52)

We have the following equivalence result, whose proof follows the guidelines of
Theorem 4.1 in [9].
Theorem 4.1 The solution to (26)-(30) can be characterized as follows:
uf = R1f + u , pf = R2f + + pf , = Rp ,
(53)
R
where pf = (meas(f ))1 f pf , and 0 is the solution of the nonlinear
interface problem:
hS, i = 0
0 .
(54)
Moreover, pf can be obtained from by solving the algebraic equation
pf = (meas())1 hS, i,
where is a fixed function such that
Z
1
=1.
meas()

(55)

Notice that a more useful characterization of the operator S can be provided.

Indeed, with the special choice R1 = R1f in (52), thanks to (33), we obtain
bf (R1f , R2f + ) = 0

, 0 .

Moreover, owing to (39), for , 0 , we have

hS, i

af (R10 + R1f + u , R1f )

+cf (R10 + R1f + u ; R10 + R1f + u , R1f )

Z
Z
1
f (Rf ) + g(Rp ) .

By taking R10 ( (H01 (f ))d ) as test function in (33), we obtain:

af (R1f , R10 ) + bf (R10 , Rf2 ) = 0 .
Finally, since Rf2 Q0 , owing to (40) it follows that af (R1f , R10 ) = 0, so
that, for all , 0 , the operator S can be characterized as
hS, i

af (R1f + u , R1f )

+cf (R10 + R1f + u ; R10 + R1f + u , R1f )

Z
Z
f (R1f ) .
+ g(Rp )

10

(56)

4.1

Existence and uniqueness result

We study the existence and uniqueness of the solution of the nonlinear interface
problem (54).
Note that in view of (56), S is defined in terms of the operator R10 , which,
thanks to (40), satisfies in its turn the following problem for all v Vf0 :
af (R10 , v) + cf (R10 + R1f + u ; R10 + R1f + u , v) = 0 .

(57)

Therefore, in order to prove the existence and uniqueness of the solution of

the interface problem, we have to consider (54), with the operator S as in
(56), coupled with (57), i.e., we have to guarantee at once the existence and
uniqueness of 0 and R10 Vf0 . To this aim we apply Theorem 2.1
replacing the space V by the product space W = 0 Vf0 endowed with the
norm:

v = (, v) W .
(58)
k
v kW = (|R1f |21 + |v|21 )1/2
We introduce the trilinear form and the linear functional associated with our
problem in the space W . For any fixed w
= (, w) W , we define the following
operator depending on w:

0 (Vf0 ) ,
(A0 (, w)(, u), Af (, w)(, u))

A(, w) :
W
A(, w) : (, u)

where, for every test function 0 ,

hA0 (, w)(, u), i

af (R1f , R1f ) + cf (w + R1f ; u + R1f , R1f )

+cf (u + R1f ; u , R1f )
+cf (u ; u + R1f , R1f ) +

whereas for any test function v Vf0 ,

hAf (, w)(, u), vi

g(Rp ) ,

= af (u, v) + cf (w + R1f ; u + R1f , v)

+cf (u ; u + R1f , v) + cf (u + R1f ; u , v) .

We indicate by a
the form associated to the operator A:
a
(w;
u
, v) = hA0 (, w)(, u), i + hAf (, w)(, u), vi

(59)

for all w
= (, w), u = (, u), v = (, v) W .
Next, we define two functionals 0 : 0 R and f : Vf0 R as:
Z
f (R1f ) af (u , R1f ) cf (u ; u , R1f )
0 ,
h0 , i =
f

hf , vi

and denote

v Vf0 ,

cf (u ; u , v)

vi = h0 , i + hf , vi
h,

v = (, v) W .

(60)

Thus, the problem defined by (54) and (57) can be reformulated as:
find u = (, R10 ) W :

vi
a
(
u; u
, v) = h,

v = (, v) W .

(61)

We shall prove the existence and uniqueness of the solution only in a closed
convex subset of W .
11

Lemma 4.1 Let f L2 (f ) be such that

kf k0
C2
< Cmin ,
2

(62)

where 0 < Cmin < 1/16 is a suitable positive constant, depending only on Cf
and C , to guarantee that
!

3 2
2
C1
+ 1 C2 > 0 ,
(63)
4
where
4kf k0
C

C1 =
,

C
4C

2
C2 =

4C
4
kf k0 + 2 2 kf k20
Cf kf k0 +
C
C

(64)

Moreover, let
0 rm < rM C1
be two constants defined as
p
C1 C12 2C2
rm =
2
If we consider

and

(65)

rM = C1

C2 / 2 .

(66)

B r = {w
= (, w) W : |R1f |1 r} ,

(67)

rm < r < rM ,

(68)

with
then, there exists a unique solution u
=

(, R10 )

B r to (61).

Remark 4.1 Since the constants C1 and C2 in (64) depend on kf k0 , the condition (63) can be viewed as an inequality in kf k0 . By simple algebraic calculations
we can prove that there exists a constant,
Cmin =

2+( 32 + 2)(Cf + C4 ) [2+( 32 + 2)(Cf + C4 )]2 41 (104 2)

2(104 2)

(69)

such that any kf k0 satisfying (62) is a solution of this inequality. Moreover, we

get that 0 < Cmin < 1/16, so that C1 > 0, and rm and rM in (66) satisfy (65).
Proof. The proof is composed of several parts.
1. For each w
= (, w) B r the bilinear form a
(w;
, ) is uniformly coercive on
W.
By definition, for all v = (, v) we have
Z
a
(w;
v, v) = af (R1f , R1f ) + af (v, v) + g(Rp )

+cf (w + R1f ; v + R1f , v +

+cf (v + R1f ; u , v + R1f )

R1f )

+cf (u ; v + R1f , v + R1f ) .

12

R
Thanks to (35), we have g(Rp ) 0. Using the inequalities (46) and (47),
the estimate (38) and the fact that w Vf0 , we obtain
a
(w;
v, v)

Ck
1f |1 (|R1f |21 + |v|21 )
(|R1f |21 + |v|21 ) 2C|R
2
|1 (|R1 |2 + |v|2 )
4C|u
f 1
1
Ck
1 2
2
1 |1 (|R1 |2 + |v|2 )

(|Rf |1 + |v|1 ) 2C|R

f
f 1
1
2
2
4C
kf k0 (|R1f |21 + |v|21 ) .
Ck

Thus,

a
(w;
v, v) a (|R1f |21 + |v|21 ) ,

having set

a =

Ck
1 |1 8C kf k0 .
2C|R
f
2
Ck

(70)

(71)

Condition a > 0 is equivalent to |R1f |1 < C1 , which is satisfied in view of (65)

and (68). Thus, the bilinear form a
(w;
, ) is uniformly coercive with respect to
any w
Br .
Thanks to the Lax-Milgram Lemma (see, e.g., [17] p. 133) the operator A(w)

L(W ; W ) is invertible for each w
B r . Moreover, the inverse T (w)
= (A(w))
1
belongs to L(W ; W ) and it satisfies
kT (w)k
L(W ;W )

1
.
a

i.e., (61)
Now, we prove that there exists a unique u
B r such that u
= T (
u),
has a unique solution in Br .
2. v T (
v ) maps B r into B r and is a strict contraction in B r .
For all v = (, v) B r we have

W kT (
W ||||W .
kT (
v )k
v )kL(W ;W ) ||||
a
Moreover,

W
||||

sup
v
W,
v 6=0

Z





1
1
1
f (Rf ) af (u , Rf ) cf (u ; u , v + Rf )

 f

k
v kW

|21 |v + R1f |1
Cf kf k0 |R1f |1 + 2|u |1 |R1f |1 + C|u
k
v kW
v
W,
v 6=0


|2 (|R1 |1 + |v|1 )
Cf kf k0 + 2|u |1 + C|u
f
1
sup
k
v
k
W
v
W,
v 6=0


2

2 Cf kf k0 + 2|u |1 + C|u |1
!

4C
4
2
2 Cf kf k0 +
kf k0 + 2 2 kf k0 ,
(72)
C
C
sup

13

the last inequality following from (38). From (72) and (71), corresponding to
some w
= (, w) B r , condition
W
||||
r
a
is equivalent to
2r2 2C1 r + C2 0 ,
that is rmin r rmax with
p
C1 C12 2C2
rmin =
2

and

rmax =

C1 +

p
C12 2C2
.
2

Since rm = rmin < rM rmax , it follows that for any v B r with r satisfying
(68), T (
v ) belongs to B r .
Finally, to prove that the map v T (
v ) is a strict contraction in B r , we
should guarantee (see [11] p. 282) that for any w
1 , w
2 B r

W ||||W L(r)kw
1 w
2 kW < kw
1 w
2 kW ,
k(T (w1 ) T (w2 ))k
2a

(73)

L(r) being the Lipschitz continuity constant associated to A. However,

|h(A(w1 ) A(w2 ))(
u), vi| = |
a(w
1 ; u
, v) a
(w
2 ; u
, v)|
= |cf (w1 + R1f 1 (w2 + R1f 2 ); u + R1f , v + R1f )|
1 + R1f 1 w2 R1f 2 |1 |u + R1f |1 |v + R1f |1
C|w

w
1 w
2 kW k
ukW k
v kW ,
2 2Ck

Thus, condition
so that L(r) = 2 2C.

W
||||
L(r) < 1
2a
is equivalent to

C2
r2 2C1 r + C12 > 0
2

i.e.,
r < rMIN

= C1 C2 / 2

or

r > rMAX

= C1 + C2 / 2 .

Condition r > rMAX does not fit with the previous restrictions on r. However,
since rM = rMIN , (73) is satisfied for any r in the interval (68).
3. The existence and uniqueness of the solution u = (, R10 ) B r to (61) is
now a simple consequence of the Banach contraction theorem (see, e.g., [19]).
2
The following theorem is a direct consequence of the previous lemma.
Theorem 4.2 If (62) holds with Cmin given in (69), and rm and rM defined
in (66), then problem (61) has a unique solution u
= (, R10 ) in the set
= (, w) W : |R1f |1 < rM } ,
BrM = {w
14

and it satisfies |R1f |1 rm . In particular, it follows that (54) has a unique

solution in the set SrM 0 ,
SrM = { 0 : |R1f |1 < rM } ,
and it satisfies |R1f |1 rm .
Proof. Since problem (54) has a solution if and only if u
= (, R10 ) is a
solution of problem (61), we prove only the first part of theorem.
From the previous Lemma 4.1, if (62) holds, with Cmin given in (69), (61) has
at least a solution in BrM as it has a solution in B r BrM , for any rm <
r < rM . To prove the uniqueness, let us assume that (61) has two solutions
u
1 = (1 , (R10 )1 ) 6= u
2 = (2 , (R10 )2 ) in BrM . Then, r1 = |R1f 1 |1 < rM and
r2 = |R1f 2 |1 < rM . Therefore, any set B r with max{rm , r1 , r2 } < r < rM
contains two different solutions of problem (61). This contradicts the result of
Lemma 4.1. Now, let u
= (, R10 ) be the unique solution of problem (61) in
BrM . According to Lemma 4.1, it belongs to each B r BrM with rm < r < rM ,
and consequently |R1f |1 rm .
2
Remark 4.2 Notice that condition (62) is analogous to what is usually required
to prove existence and uniqueness of the solution of the Navier-Stokes equations.
Moreover, we have proved that the solution is unique in SrM . Thus, in view of
Remark 3.1, Theorem 4.2 states that the solution is unique only for sufficiently
small normal velocities across the interface . Finally, notice that (62) implies (48) and that SrM is included in the set (49), so that the existence and
uniqueness of the nonlinear extension R10 is ensured as well.
The reformulation of the coupled problem as an interface equation is interesting
in view of setting up iterative substructuring methods to compute the solution
of the global problem. In particular, we can split the operator S into a nonlinear
part, say Sf , associated to the Navier-Stokes problem in f , and a linear part
Sp related to the Darcy equation in p , and exploit them to set up iterative
methods inspired to the Dirichlet-Neumann and Neumann-Neumann schemes
in domain decomposition (see, e.g., [18]). This study will be the object of a
forthcoming work [2].

References
[1] R.A. Adams. Sobolev Spaces. Academic Press, New York, 1975.
[2] L. Badea, M. Discacciati, and A. Quarteroni. Iterative substructuring
methods for the Navier-Stokes/Darcy coupling. Technical report, 2007.
In preparation.
[3] F. Brezzi, J. Rappaz, and P.A. Raviart. Finite dimensional approximation
of nonlinear problems. Part I: Branches of nonlinear solutions. Numer.
Math., 36:125, 1980.
[4] F. Brezzi, J. Rappaz, and P.A. Raviart. Finite dimensional approximation
of nonlinear problems. Part II: Limit points. Numer. Math., 37:128, 1981.
15

[5] F. Brezzi, J. Rappaz, and P.A. Raviart. Finite dimensional approximation

of nonlinear problems. Part III: Simple bifurcation points. Numer. Math.,
38:130, 1981.
[6] G. Caloz and J. Rappaz. Numerical Analysis for Nonlinear and Bifurcation
Problems. In Handbook of Numerical Analysis, Vol. V, pages 487637.
North-Holland, Amsterdam, 1997.
[7] M. Discacciati. Domain Decomposition Methods for the Coupling of Surface
and Groundwater Flows. PhD thesis, Ecole Polytechnique Federale de
Lausanne, Switzerland, 2004.
[8] M. Discacciati, E. Miglio, and A. Quarteroni. Mathematical and numerical
models for coupling surface and groundwater flows. Appl. Numer. Math.,
43:5774, 2002.
[9] M. Discacciati and A. Quarteroni. Analysis of a domain decomposition
method for the coupling of Stokes and Darcy equations. In F. Brezzi,
A. Buffa, S. Corsaro, and A. Murli, editors, Numerical Mathematics and
Advanced Applications, ENUMATH 2001, pages 320. Springer. Milan,
2003.
[10] M. Discacciati and A. Quarteroni. Convergence analysis of a subdomain
iterative method for the finite element approximation of the coupling of
Stokes and Darcy equations. Comput. Visual. Sci., 6:93103, 2004.
[11] V. Girault and P.A. Raviart. Finite Element Methods for Navier-Stokes
Equations. Theory and Algorithms. Springer, Berlin, 1986.
[12] W. J
ager and A. Mikelic. On the boundary conditions at the contact
interface between a porous medium and a free fluid. Ann. Scuola Norm.
Sup. Pisa Cl. Sci., 23:403465, 1996.
[13] W. J
ager and A. Mikelic. On the interface boundary condition of Beavers,
Joseph and Saffman. SIAM J. Appl. Math., 60(4):11111127, 2000.
[14] W. J
ager, A. Mikelic, and N. Neuss. Asymptotic analysis of the laminar
viscous flow over a porous bed. SIAM J. Sci. Comput., 22(6):20062028,
2001.
[15] W.L. Layton, F. Schieweck, and I. Yotov. Coupling fluid flow with porous
media flow. SIAM J. Num. Anal., 40:21952218, 2003.
[16] J.L. Lions and E. Magenes. Probl`emes aux Limites Non Homog`enes et
Applications, volume 1. Dunod, Paris, 1968.
[17] A. Quarteroni and A. Valli. Numerical Approximation of Partial Differential Equations. Springer, Berlin, 1994.
[18] A. Quarteroni and A. Valli. Domain Decomposition Methods for Partial
Differential Equations. Oxford Science Publications, Oxford, 1999.
[19] K. Yosida. Functional Analysis. Springer, Berlin, 1974.

16