2014 NumerMath MortarMultiscaleMethod

Numer. Math.
(2014) 127:93–165 Numerische

DOI 10.1007/s00211-013-0583-z Mathematik
Mortar multiscale finite element methods

for Stokes–Darcy flows
Vivette Girault · Danail Vassilev · Ivan Yotov
Received: 25 January 2012 / Revised: 7 March 2013 / Published online: 29 September 2013
© Springer-Verlag Berlin Heidelberg 2013
Abstract We investigate mortar multiscale numerical methods for coupled Stokes

and Darcy flows with the Beavers–Joseph–Saffman interface condition. The domain
is decomposed into a series of subdomains (coarse grid) of either Stokes or Darcy
type. The subdomains are discretized by appropriate Stokes or Darcy finite elements.
The solution is resolved locally (in each coarse element) on a fine scale, allowing for
non-matching grids across subdomain interfaces. Coarse scale mortar finite elements
are introduced on the interfaces to approximate the normal stress and impose weakly
continuity of the normal velocity. Stability and a priori error estimates in terms of
the fine subdomain scale h and the coarse mortar scale H are established for fairly
general grid configurations, assuming that the mortar space satisfies a certain inf-sup
condition. Several examples of such spaces in two and three dimensions are given.
Numerical experiments are presented in confirmation of the theory.
D. Vassilev and I. Yotov were partially supported by the NSF grant DMS 1115856 and the DOE grant
DE-FG02-04ER25618.
V. Girault
UPMC-Paris 6 and CNRS, UMR 7598, 75230 Paris Cedex 05, France
e-mail: girault@ann.jussieu.fr
V. Girault
Department of Mathematics, TAMU, College Station, TX 77843, USA
D. Vassilev
Mathematics Research Institute, College of Engineering, Mathematics and Physical Sciences,
University of Exeter, Exeter EX4 4QF, UK
e-mail: D.Vassilev@exeter.ac.uk
I. Yotov (B)
Department of Mathematics, University of Pittsburgh, 301 Thackeray Hall, Pittsburgh, PA 15260, USA
e-mail: yotov@math.pitt.edu
123
94 V. Girault et al.
Mathematics Subject Classification 65N12 · 65N15 · 65N30 · 65N55 · 76D07 ·

76S05
1 Introduction
Mathematical and numerical modeling of coupled Stokes and Darcy flows has become
a topic of significant interest in recent years. Such coupling occurs in many appli-
cations, including surface water-groundwater interaction, flows through vuggy or
fractured porous media, industrial filters, fuel cells, and cardiovascular flows. The
most commonly used model is based on the experimentally derived Beavers–Joseph–
Saffman interface condition [10,55], a slip with friction condition for the Stokes flow
with a friction coefficient that depends on the permeability of the adjacent porous
media. Existence and uniqueness of a weak solution has been studied in [24,46].
Numerous stable and convergent numerical methods have been developed, see, e.g.,
[24,28,29,32,33,44,46,49,54] for methods based on different numerical discretiza-
tions suitable for each region, and [4,9,19,45,47,59] for approaches employing unified
finite elements. The full Beavers–Joseph condition was considered in [2,20]. A cou-
pling of Stokes–Darcy flows with transport was analyzed in [57]. The nonlinear system
of coupled Navier–Stokes and Darcy flows has been studied in [8,26,35].
In this paper we develop multiscale mortar methods for multi-domain non-matching
grid discretizations of Stokes–Darcy flows in two and three dimensions. Non-matching
grids provide flexibility in meshing complex geometries with relatively simple locally
constructed subdomain grids that are suitable for the choice of subdomain discretiza-
tion methods. Mortar finite elements play the role of Lagrange multipliers to impose
weakly interface conditions. In [46], a Lagrange multiplier approximating the normal
stress was introduced to impose continuity of the normal velocity for discretizations
involving mixed finite element methods for Darcy and conforming Stokes elements.
With a choice of the Lagrange multiplier space as the normal trace of the Darcy veloc-
ity space, the analysis in [46] applied to non-matching grids on the Stokes–Darcy
interface, although this was not explicitly noted. A similar choice was considered in
subsequent mortar discretizations for Stokes–Darcy flows [14,29,54]. Mortar methods
for mixed finite element discretizations for Darcy have been studied in [5,6,52,60].
The analysis in these papers allows for the mortar grid to be different from the traces
of the subdomain grids with the assumption that the mortar space satisfies a suitable
solvability condition that limits the number of mortar degrees of freedom. Mortar
discretizations for Stokes have been developed in [11,12]. There, the mortar grid was
chosen to be the trace of one of the neighboring subdomain grids, similar to the choice
in mortar methods for conforming Galerkin discretizations for second order elliptic
problems [13].
In this work we allow for non-matching grid interfaces of Stokes–Darcy, Stokes–
Stokes, and Darcy–Darcy types. We develop multiscale discretizations, where the
subdomains are discretized on a fine scale h and the mortar space is discretized on a
coarse scale H . Our method is based on saddle-point formulations in both regions and
employs inf-sup stable mixed finite elements for Darcy and conforming elements for
Stokes. The mortars approximate different physical variables and are used to impose
different matching conditions depending on the type of interface. On Stokes–Stokes
123
Mortar multiscale finite element methods for Stokes–Darcy flows 95
interfaces, the mortar functions represent the entire stress vector and impose weak con-
tinuity of the entire velocity vector. On Stokes–Darcy and Darcy–Darcy interfaces, the
mortars approximate the normal stress, which is just the pressure in the Darcy region,
and impose weak continuity of the normal velocity. The mortar spaces are assumed
to satisfy suitable inf-sup conditions, allowing for very general subdomain and mor-
tar grid configurations. We consider approximations of different polynomial degrees
on the three types of interfaces and the two types of subdomains. The mortar spaces
can be continuous or discontinuous, the latter providing localized mass conservation
across interfaces. Our method is more general than existing Stokes–Stokes mortar
methods [11,12] and Stokes–Darcy mortar methods [14,29,46,54]. On Darcy–Darcy
interfaces, our condition is closely related to the solvability condition considered in
[5,6,52,60].
The stability and convergence analysis relies on the construction of a bounded global
interpolant in the space of weakly continuous velocities that also preserves the velocity
divergence in the usual discrete sense. This is done in two steps, starting from suitable
local interpolants and correcting them to satisfy the interface matching conditions. The
correction step requires the existence of bounded mortar interpolants. This is a very
general condition that can be easily satisfied in practice. We present two examples in
2-D and one example in 3-D that satisfy this solvability condition. Our error analysis
shows that the global velocity and pressure errors are bounded by the fine scale local
approximation error and the coarse scale non-conforming error. Since the polynomial
degrees on subdomains and interfaces may differ, one can choose higher order mortar
polynomials to balance the fine scale and the coarse scale error terms and obtain
fine scale asymptotic convergence. The dependence of the stability and convergence
constants on the subdomain size is explicitly determined. In particular, the stability
and fine scale convergence constants do not depend on the size of subdomains, while
the coarse scale non-conforming error constants deteriorate when the subdomain size
goes to zero. This is to be expected, as the relative effect of the non-conforming error
becomes more significant in such regime. However, this dependence can be made
negligible by choosing higher order mortar polynomials, as mentioned above.
Our multiscale Stokes–Darcy formulation can be viewed as an extension of the
mortar multiscale mixed finite element (MMMFE) method for Darcy developed in
[6]. The MMMFE method provides an alternative to other multiscale methods in the
literature such as the variational multiscale method [3,41] and the multiscale finite
element method [22,40]. All three methods utilize a divide and conquer approach:
solve relatively small fine scale subdomain problems that are only coupled on the
coarse scale through a reduced number of degrees of freedom. The mortar multiscale
approach is more flexible as it allows for employment of a posteriori error estimation
to adaptively refine the mortar grids where necessary to improve the global accuracy
[6]. Following the non-overlapping domain decomposition approach from [37], it can
be shown that the global Stokes–Darcy problem can be reduced to a positive definite
coarse scale interface problem [58]. The latter can be solved using a preconditioned
Krylov space solver requiring Stokes or Darcy subdomain solves at each iteration. An
alternative more efficient implementation for MMMFE discretizations for Darcy was
developed in [31], where a multiscale flux basis giving the interface flux response for
each coarse scale mortar degree of freedom is precomputed. The multiscale flux basis
123
is used to replace the solution of subdomain problems by a simple linear combination.

The application of this methodology to the Stokes–Darcy interface problem will be
discussed in a forthcoming paper. We should mention that there have been a number of
papers in the literature studying domain decomposition methods for the Stokes–Darcy
problem, primarily in the two-subdomain case, see, e.g., [21,25,27,30,39].
1.1 Notation and preliminaries
Let be a bounded, connected Lipschitz domain of Rn , n = 2, 3, with boundary

∂ and exterior unit normal vector n, and let be a part of ∂ with positive n − 1
measure: || > 0. We do not assume that is connected, but if it is not connected, we
assume that it has a finite number of connected components. In the case when n = 3,
we also assume that is itself Lipschitz. Let
1
H0, () = {v ∈ H 1 (); v| = 0}.
1 () reads: There exists a constant P depending only

Poincaré’s inequality in H0,
on and such that
∀v ∈ H0,
1
(), v L 2 () ≤ P |v| H 1 () . (1.1)
The norms and spaces are made precise later on. The formula (1.1) is a particular case
of a more general result (cf. [15,50]):
Proposition 1.1 Let be a bounded, connected Lipschitz domain of Rn and let be
a seminorm on H 1 () satisfying:
1) There exists a constant P1 such that
∀v ∈ H 1 (), (v) ≤ P1 v H 1 () . (1.2)
2) The condition (c) = 0 for a constant function c holds if and only if c = 0. Then
there exists a constant P2 depending only on , such that
1
∀v ∈ H 1 (), v L 2 () ≤ P2 |v|2H 1 () + (v)2 2 . (1.3)
We recall Korn’s first inequality: There exists a constant C1 depending only on

and such that
∀v ∈ H0,
1
()n , |v| H 1 () ≤ C1 D(v) L 2 () , (1.4)
where D(v) is the deformation rate tensor, also called the symmetric gradient tensor:
1
D(v) = ∇ v + ∇ vT .
2
This is a particular case of the following more general result (see (1.6) in [16]):
123
Proposition 1.2 Let be a bounded, connected Lipschitz domain of Rn and let be

a seminorm on H 1 ()n satisfying:
1) There exists a constant C2 such that
∀v ∈ H 1 ()n , (v) ≤ C2 v H 1 () . (1.5)
2) The condition (m) = 0 for a rigid-body motion m holds if and only if m is a

constant vector. Then there exists a constant C3 depending only on , such that
1
∀v ∈ H 1 ()n , |v| H 1 () ≤ C3 D(v)2L 2 () + (v)2 2 . (1.6)
In particular, Proposition 1.2 implies that there exists a constant C , only depending
on such that (see (1.9) in [16]),
⎛ 2 ⎞ 21

⎜ ⎟
∀v ∈ H 1 ()n , |v| H 1 () ≤ C ⎝ D(v) L 2 () + curl v ⎠ ,
2 (1.7)

where | · | denotes the Euclidean vector norm.

For any non-negative integer m, recall the classical Sobolev space (cf. [1] or [50])

H m () = v ∈ L 2 (); ∂ k v ∈ L 2 () ∀ |k| ≤ m ,
equipped with the following seminorm and norm (for which it is a Hilbert space)
⎛ ⎞1 ⎛ ⎞1

2 2

|v| H m () = ⎝ |∂ k v|2 d x ⎠ , v H m () = ⎝ |v|2H k () ⎠ .
|k|=m 0≤|k|≤m
This definition is extended to any real number s = m + s for an integer m ≥ 0 and

0 < s < 1 by defining in dimension n the fractional semi-norm and norm
⎛ ⎞1
|∂ k v(x) − ∂ k v( y)|2 1
2
|v| H s () = ⎝ ⎠ 2
d x d y , v H s () = v2H m () + |v|2H s () .
|k|=m
|x − y| n+2 s

1 1
In particular, we shall frequently use the fractional Sobolev spaces H 2 () and H00
2
()
for a Lipschitz surface when n = 3 or curve when n = 2 with the following
seminorms and norms:
123
⎛ ⎞1
2 1
|v(x) − v( y)|2 2
|v| 1 =⎝ d xd y ⎠ v 1 = v2
2 () + |v| 2
1 ,
H 2 () |x − y| n H 2 () L
H 2 ()

(1.8)
⎛ ⎞1 1
2
|v(x)|2 2
|v| 1 = ⎝|v|2 1 + dx ⎠ v 1 2 2
= v L 2 () + |v| 1 ,
2 ()
H00 H 2 () d∂ (x) 2 ()
H00 2 ()
H00

(1.9)
where d∂ (x) denotes the distance from x to ∂. When is a subset of ∂ with
1
positive n − 1 measure, H00 2
() is the space of traces of all functions of H0,∂\
1 ().
The above norms (1.8) and (1.9) are not equivalent except when is a closed surface
1 1
or curve. The dual space of H 2 () is denoted by H − 2 ().
In addition to these spaces, we shall use the Hilbert space

H (div; ) = v ∈ L 2 ()n ; div v ∈ L 2 () ,
equipped with the graph norm

1
2
v H (div;) = v2L 2 () + div v2L 2 () .
1
The normal trace v · n of a function v of H (div; ) on ∂ belongs to H − 2 (∂)
(cf. [34]). The same result holds when is a part of ∂ and is a closed surface. But
1
if is not a closed surface, then v · n belongs to the dual of H00
2
(). When v · n = 0
on ∂, we use the space
H0 (div; ) = {v ∈ H (div; ); v · n = 0 on ∂} .
2 Problem statement
2.1 Coupled Stokes and Darcy systems
Let be partitioned into two non-overlapping regions: the region of the Darcy flow,
d , and the region of the Stokes flow, s , each one possibly non-connected, but with
a finite number of connected components, and with Lipschitz-continuous boundaries
∂d and ∂s :
= d ∪ s .
Let d = ∂d ∩ ∂, s = ∂s ∩ ∂, sd = ∂d ∩ ∂s . The unit normal vector on
sd exterior to d , respectively s , is denoted by nd , respectively ns . In dimension
123
three, we assume that d , s , and sd also have Lipschitz-continuous boundaries. Let
f d be the gravity force in d , f s a given body force in s , let νd > 0, respectively
νs > 0, be the constant viscosity coefficient of the Darcy, respectively Stokes flow, let
K be the rock permeability tensor in d , let qd be an external source or sink term in
d , and let α > 0 be the slip coefficient in the Beavers–Joseph–Saffman law [10,55],
determined by experiment. As far as the data are concerned, we assume on one hand
that K is bounded, symmetric and uniformly positive definite in d : there exist two
constants, λmin > 0 and λmax > 0 such that
∀x ∈ d , ∀χ ∈ Rn , λmin |χ|2 ≤ K (x)χ · χ ≤ λmax |χ|2 , (2.1)
and we assume on the other hand, that the source qd satisfies the solvability condition

qd d x = 0. (2.2)
d
The fluid velocity and pressure in d , respectively s , are denoted by ud and pd ,

respectively by us and ps . The stress tensor of the Stokes flow is denoted by σ (us , ps ),
σ (us , ps ) = − ps I + 2νs D(us ).
In the Darcy region d , the pair (ud , pd ) satisfies
νd K −1 ud + ∇ pd = f d in d , (2.3)
div ud = qd in d , (2.4)
ud · n = 0 on d . (2.5)
In the Stokes region s , the pair (us , ps ) satisfies
− div σ (us , ps ) ≡ −2νs div D(us ) + ∇ ps = f s in s , (2.6)

div us = 0 in s , (2.7)
us = 0 on s . (2.8)
These operators suggest to look for (ud , pd ) in H (div; d ) × H 1 (d ) and (us , ps )
in H 1 (s )n × L 2 (s ). The Darcy and Stokes flows are coupled on sd through the
following interface conditions
us · ns + ud · nd = 0 on sd , (2.9)

− σ (us , ps )ns · ns ≡ ps − 2νs D(us )ns · ns = pd on sd , (2.10)
√ √
Kl Kl
− σ (us , ps )ns · τ l ≡ − 2 D(us )ns · τ l = us · τ l , on sd , 1 ≤l ≤ n −1,
νs α α
(2.11)
123
whereτ l , 1≤ l ≤ n − 1 is an orthogonal system of unit tangent vectors on sd and

K l = K τ l ·τ l . Conditions (2.9) and (2.10) incorporate continuity of flux and normal
stress, respectively. Condition (2.11) is known√ as the Beavers–Joseph–Saffman law
[10,42,55] describing slip with friction, where K l /α is a friction coefficient.
2.2 First variational formulation
For any functions ϕd defined in d and ϕs defined in s , it is convenient to define ϕ

in by ϕ|d = ϕd and ϕ|s = ϕs . With this notation, regarding the data, we assume
that f ∈ L 2 ()n , we extend qd by zero in s , i.e. we set qs = 0 and owing to (2.2),
the extended function q belongs to L 20 (). Before setting problem (2.3)–(2.11) into
an equivalent variational formulation, it is useful to interpret the interface conditions
(2.10)–(2.11). First we observe from the regularity of f s that each row of σ (us , ps )
1
belongs to H (div; s ); hence σ (us , ps )ns belongs to H − 2 (∂s )n , and in particular
1
is well-defined as an element of the dual of H00 2
(sd )n , which is a distribution space
on sd . But without further information, it cannot be multiplied directly by the normal
or tangent vectors, since the boundary is only Lipschitz-continuous. To bypass this
difficulty, following [35] we define the function on sd

n−1
νs α
g = − pd n s − √ (us · τ l )τ l , (2.12)
l=1
Kl
and replace (2.10)–(2.11) by the condition
σ (us , ps )ns = g on sd . (2.13)

1
As the traces of pd and of all components of us on sd belong to H 2 (sd ), Sobolev’s
imbeddings [1] imply that g belongs to L r (sd )n for any finite r when n = 2 and
r = 4 when n = 3. Hence condition (2.13) makes sense. Let us check that it implies
(2.10)–(2.11). First, prescribing (2.13) guarantees that σ (us , ps )ns belongs at least to
L 4 (sd )n and thus can be multiplied by the normal or tangent vectors. Then by virtue
of this regularity, (2.12), (2.13) are equivalent to:

n−1

n−1
νs α
σ (us , ps )ns · ns ns + σ (us , ps )ns · τ l τ l = − pd ns − √ (us · τ l )τ l ,
l=1 l=1
Kl
and therefore, by identifying on both sides the components of the normal and tangent
vectors (that forms an orthonormal set), we derive (2.10)–(2.11). Hence (2.13) is the
interpretation of (2.10)–(2.11).
Now, let (ud , pd ) ∈ H (div; d ) × H 1 (d ) and (us , ps ) ∈ H 1 (s )n × L 2 (s ) be
a solution of (2.3)–(2.11). In order to set (2.3)–(2.11) in variational form, we take the
scalar product of (2.3) and (2.6) respectively with any test function v d in H 1 (d )n
satisfying v d · n = 0 on d , and any v s in H 1 (s )n satisfying v s = 0 on s . Then we
apply Green’s formula in d and s . This yields
123

νd K −1 ud · v d − pd div v d + pd v d · n d = f d · vd , (2.14)
d d sd d

2 νs D(us ) : D(v s )− ps div v s − σ (us , ps )ns , v s sd = f s · vs , (2.15)
s s s
1
where ·, ·sd denotes the duality pairing between H00 2
(sd )n and its dual space. The
validity of (2.14) and (2.15) follows from the above considerations. By summing
(2.14) and (2.15), by using the fact that nd = −ns , and by applying (2.13), the term
on the interface, say I , reads

I = − σ (us , ps )ns , v s sd − pd v d · n s = − g · vs − pd v d · n s .
sd sd sd
Then the expression (2.12) for g yields

n−1

νs α
I = √ (us · τ l )(v s · τ l ) + pd [v · n], (2.16)
l=1
Kl
sd sd
where the jump [v · n] is defined by
[v · n] = v s · ns + v d · nd .
Finally, we eliminate this jump by enforcing strongly the transmission condition (2.9)
on the test function v. In view of the interior terms in (2.14) and (2.15) and what
remains in (2.16), we see that we can reduce the regularity of our functions and work
in the space
X̃ = {v ∈ H (div; ); v s ∈ H 1 (s )n , v|s = 0, (v · n)|d = 0}, (2.17)
which is a Hilbert space equipped with the norm

1
∀v ∈ X̃ , v X̃ = v2H (div;) + |v s |2H 1 ( ) 2 . (2.18)
s
Note that the restriction of v · n on sd belongs at least to L 4 (sd ). Let us denote
W = L 20 () with the norm wW = w L 2 () . Then we propose the following
variational formulation: Find (u, p) ∈ X̃ × W such that

∀v ∈ X̃ , νd K −1 ud · v d + 2 νs D(us ) : D(v s ) − p div v
d s

n−1
νs α
+ √ (us · τ l )(v s · τ l ) = f · v, (2.19)
l=1
Kl
sd
123

∀w ∈ W, w div u = w qd . (2.20)
d
Lemma 2.1 For any data f in L 2 ()n and qd in L 20 (d ), any solution (u, p) ∈ X̃ ×W
of problem (2.19)–(2.20) satisfies pd ∈ H 1 (d ) and solves (2.3)–(2.11). Conversely,
any solution of (2.3)–(2.11), (u, p) ∈ X̃ ×W with pd ∈ H 1 (d ), solves (2.19)–(2.20).
Proof It stems from the above considerations that any solution (ud , pd ) ∈ H (div; d )
×H 1 (d ) and (us , ps ) ∈ H 1 (s )n × L 2 (s ) of (2.3)–(2.11) is such that (u, p)
belongs to X̃ × L 2 () and solves (2.19)–(2.20). Moreover as is connected, (2.19)
only defines p up to an additive constant and this constant can be chosen so that p
belongs to W .
Conversely, let (u, p) ∈ X̃ × W solve (2.19)–(2.20), and denote its restriction to
d and s as above. By choosing smooth test functions with compact support first in
d and next in s , we immediately derive that (ud , pd ) is a solution of (2.3)–(2.5)
and (us , ps ) is a solution of (2.6)–(2.8). Furthermore, since both f d and νd K −1 ud
belong to L 2 (d )n , (2.3) implies that pd ∈ H 1 (d ).
It remains to recover the transmission conditions (2.9)–(2.11). First, (2.9) is a con-
sequence of the definition (2.17) of X̃ . Next, we recover (2.14) and (2.15) by taking
the scalar product of (2.3) and (2.6) with a function v ∈ X̃ that is smooth in d and
in s , and by applying Green’s formula in both regions. By comparing with (2.19),
this gives
n−1
νs α
pd v d · nd − σ (us , ps )ns , v s sd = √ (us · τ l )(v s · τ l ).
l=1
Kl
sd sd
By taking into account the orientation of the normal, the regularity of v, and the
definition (2.12) of g, this is equivalent to:
n−1

νs α
σ (us , ps )ns , v s sd = − pd v s · n s − √ (us · τ l )(v s · τ l ) = g · vs .
l=1
Kl
sd sd sd
As the trace space of v s on sd is large enough, this implies (2.13).
2.3 Existence and uniqueness of the solution

For any functions ud , v d in L 2 (d )n and us , v s in H 1 (s )n , we define the bilinear
form

n−1
νs α
ã(u, v) = νd K −1 ud · v d + 2 νs D(us ) : D(v s ) + √ (us · τ l )(v s · τ l ),
Kl
d s l=1
sd
(2.21)
123
and for any functions v d ∈ H (div; d ), v s ∈ H (div; s ) and w ∈ L 2 (), we define

the bilinear form

b̃(v, w) = − w div v d − w div v s . (2.22)
d s
Note that ã(·, ·) is continuous on X̃ × X̃ and b̃(·, ·) is continuous on X̃ × L 2 ():

νd
∀(u, v) ∈ X̃ × X̃ , |ã(u, v)| ≤ ud L 2 (d ) v d L 2 (d )
λmin
+2 νs ∇ us L 2 (s ) ∇ v s L 2 (s )

n−1
νs α
+ √ us · τ l L 2 (sd ) v s · τ l L 2 (sd ) ,
l=1
λmin
∀(v, w) ∈ X̃ × L 2 (), |b̃(v, w)| ≤ v X̃ w L 2 () . (2.23)
Then (2.19)–(2.20) has the familiar form : Find (u, p) ∈ X̃ × W such that

∀v ∈ X̃ , ã(u, v) + b̃(v, p) = f · v, (2.24)

∀w ∈ W, b̃(u, w) = − w qd . (2.25)
d
Next, we set
X̃ 0 = {v ∈ X̃ ; div v = 0}, (2.26)
and more generally, for a given function g ∈ W , we define the affine variety
X̃ g = {v ∈ X̃ ; div v = g}. (2.27)
Then we consider the reduced problem : Find u ∈ X̃ q such that

∀v ∈ X̃ 0 , ã(u, v) = f · v, (2.28)

recall that q is qd extended by zero on s . It is well known [18,34] that showing equiv-
alence between problems (2.28) and (2.24)–(2.25) amounts to proving the following
inf-sup condition.
Lemma 2.2 There exists a constant β > 0 such that
b̃(v, w)
∀w ∈ W, sup ≥ βwW . (2.29)
v∈ X̃
v X̃
123
Proof Let w ∈ W . The inf-sup condition between H01 ()n and L 20 () implies that
there exists a function v ∈ H01 ()n such that
1
div v = w in and |v| H 1 () ≤ w L 2 () ,
κ
where κ depends only on ; see for example [34] or [18]. Then v belongs to X̃ and a
2 1
straightforward computation yields (2.29) with β ≥ κ/ P∂ + 2 2 , where P∂ is the
Poincaré constant in (1.1).
Lemma 2.2 implies that (2.28) and (2.24)–(2.25) are equivalent in the following sense.
Proposition 2.1 Let ( f , qd ) be given in L 2 ()n × L 20 (d ). Let (u, p) ∈ X̃ × W be a

solution of (2.24)–(2.25). Then u solves (2.28). Conversely, let u ∈ X̃ qd be a solution
of (2.28). Then there exists a unique p in W such that (u, p) satisfies (2.24)–(2.25).
Now, let us prove that (2.28), and hence (2.24)–(2.25), is well-posed. This relies on
the ellipticity of ã(·, ·) on X̃ 0 . If s is connected and |s | > 0, a partial ellipticity
result for ã(·, ·) follows directly from Korn’s inequality (1.4):
νs νd
∀v ∈ X̃ , ã(v, v) ≥ 2 |v s |2H 1 ( ) + v d 2L 2 ( ) , (2.30)
C12 s λmax d
with the constant C1 of (1.4). If s is connected and |s | = 0, then sd = ∂s up to
a set of zero measure, and proving the analogue of (2.30) makes use of (1.7) and the
tangential components on sd . Indeed, we have formally
n−1
a.e. on ∂s , (v s × ns )(x) ≤ |(v s · τ l )(x)|, (2.31)
l=1
and therefore

n−1

curl v s = − v s × ns ⇒ curl v s ≤ |v s · τ l | .
s sd s sd l=1
Hence (1.7) and a straightforward manipulation yield for all v s in H 1 (s )n :

n−1
νs α
2νs D(v s )2L 2 ( ) +
√ v s · τ l 2L 2 ( )
s
l=1
Kl sd

νs α
≥ 2 min 2, √ |v s |2H 1 ( ) . (2.32)
C λmax |sd | s
123
As a consequence (2.30) is replaced by

νd νs α
∀v ∈ X̃ , ã(v, v) ≥ v d 2L 2 ( ) + 2 min 2, √ |v s |2H 1 ( ) .
λmax d C λmax |sd | s
(2.33)
Finally, if s is not connected, the analogue of (2.30) holds on all connected compo-
nents of s that are adjacent to s and the analogue of (2.33) holds on all connected
components of s that are not adjacent to s .
It remains to establish that the mapping:
1
v → |v| X̃ 0 = |v s |2H 1 ( ) + v d 2L 2 ( ) 2 (2.34)
s d
is a norm on X̃ 0 equivalent to v X̃ . This is the object of the next lemma.

Lemma 2.3 There exists a constant C4 such that
1
∀v ∈ X̃ 0 , v s L 2 (s ) ≤ C4 |v s |2H 1 ( ) + v d 2L 2 ( ) 2 . (2.35)
s d
Proof Let us assume that s is connected; the case when s is not connected is treated
as above. If |s | > 0, (2.35) follows from Poincaré’s inequality (1.1) applied in s
and does not require the norm of v d in the right-hand side. When |s | = 0, the proof
of (2.35) is a variant of the proof of Peetre–Tartar’s Lemma [51]. Let us recall its
argument. Assume that (2.35) is not true. Then there exists a sequence (v m ) in X̃ 0
such that
lim v m
d L 2 (d ) = lim |v s | H 1 (s ) = 0 and ∀m, v s L 2 (s ) = 1.
m m
m→∞ m→∞
As X̃ 0 is reflexive, this implies that there exists a function v ∈ X̃ 0 such that
lim v m = v weakly in X̃ .
m→∞
Moreover v s = c, a constant vector, and v d = 0. Then the fact that v belongs to X̃

implies that c · ns = 0 on sd , and since sd coincides a.e. with ∂s , this implies that
c = 0. Thus
lim v m = 0 weakly in H 1 (s )n ,

m→∞ s
hence
lim v m = 0 strongly in L 2 (s )n .

m→∞ s
s L 2 (s ) = 1.
This contradicts the fact that for all m v m
Therefore (2.35) combined with either (2.30) or (2.33) yields the ellipticity of ã(·, ·).
123
Fig. 1 Domain decomposition

with multiple connected
components
Lemma 2.4 There exists a constant C5 > 0 such that
∀v ∈ X̃ 0 , ã(v, v) ≥ C5 v2X̃ . (2.36)
With the continuity of ã(·, ·) and b̃(·, ·), the ellipticity of ã(·, ·) on X̃ 0 , and the inf-
sup condition (2.29), the Babuška–Brezzi’s theory [7,17] implies immediately that
(2.24)–(2.25) is well-posed.
Theorem 2.1 Problem (2.24)–(2.25) has a unique solution (u, p) ∈ X̃ × W and there
exists a constant C that depends only on d , s , λmin , λmax , α, νd , and νs , such that

u X̃ + p L 2 () ≤ C f L 2 () + qd L 2 (d ) . (2.37)
In turn, the well-posedness of problem (2.3)–(2.11) stems from Lemma 2.1.
2.4 Domain decomposition of the Darcy and Stokes regions
Let s , respectively d , be decomposed into Ms , respectively Md , non-overlapping,

open Lipschitz subdomains:
Ms Md
s = ∪i=1 s,i , d = ∪i=1 d,i .
Set M = Md + Ms ; according to convenience we can also number the subdomains

with a single index i, 1 ≤ i ≤ M, the Darcy subdomains running from Ms +1 to M. An
example of a domain decomposition with multiple connected components is depicted
in Fig. 1. Let ni denote the outward unit normal vector on ∂i . For 1 ≤ i ≤ M, let
the boundary interfaces be denoted by i , with possibly zero measure:
i = ∂i ∩ ∂,
and for 1 ≤ i < j ≤ M, let the interfaces between subdomains be denoted by i j ,

again with possibly zero measure:
i j = ∂i ∩ ∂ j .
123
In addition to sd , let dd , respectively ss , denote the set of interfaces between subdo-
mains of d , respectively s . Then, assuming that the solution (u, p) of (2.3)–(2.11)
is slightly smoother, we can obtain an equivalent formulation by writing individually
(2.3)–(2.11) in each subdomain i , 1 ≤ i ≤ M, and complementing these systems
with the following interface conditions
[uu d · n] = 0, [ pd ] = 0 on dd , (2.38)

[us ] = 0, [σ (us , ps )n] = 0 on ss , (2.39)
where the jumps on an interface i j , 1 ≤ i < j ≤ M, are defined as
[v · n] = v i · ni + v j · n j , [σ n] = σ i ni + σ j n j , [v] = (vi − v j )|i j ,
using the notation vi = v|i . The smoothness requirement on the solution is meant
to ensure that the jumps [ud · n], respectively [σ (us , ps )n], are well-defined on each
interface of dd , respectively ss .
Finally, let us prescribe weakly the interface conditions (2.38), (2.39), and (2.9)
by means of Lagrange multipliers, usually called mortars. For this, it is convenient to
attribute a unit normal vector ni j to each interface i j of positive measure, directed
from i to j (recall that i < j). The basic velocity spaces are:
X d = {v ∈ L 2 (d )n ; v d,i := v|d,i ∈ H (div; d,i ), 1 ≤ i ≤ Md ,

1
v · ni j ∈ H − 2 (i j ), i j ∈ dd ∪ sd , v · n = 0 on d }, (2.40)
X s = {v ∈ L (s ) ; v s,i := v|s,i ∈ H (s,i ) , 1 ≤ i ≤ Ms , v = 0 on s },
2 n 1 n
and the mortar spaces are:
1 1
∀i j ∈ ss , i j = H − 2 (i j )n , ∀i j ∈ sd ∪ dd , i j = H 2 (i j ).
(2.41)
Then we replace X̃ (see (2.17)) by
X = {v ∈ L 2 ()n ; v d := v|d ∈ X d , v s := v|s ∈ X s }, (2.42)
we keep W = L 20 () for the pressure, and we define the mortar spaces
n 1
s = {λ ∈ D (ss ) ; λ|i j ∈ H − 2 (i j )n for all i j ∈ ss },
1
sd = {λ ∈ L 2 (sd ); λ|i j ∈ H 2 (i j ) for all i j ∈ sd }, (2.43)
1
d = {λ ∈ L (dd ); λ|i j ∈ H (i j ) for all i j ∈ dd }.
2 2
123
We equip these spaces with broken norms:

⎛ ⎞1 M 21
2

Md s
|||v||| X d = ⎝ v2H (div;d,i ) ⎠ , |||v||| X s = v2H 1 ( ,
s,i )
i=1 i=1
⎛ ⎞1
1
2
|||v||| X = |||v|||2X d + |||v|||2X s , |||λ|||s = ⎝ ⎠ ,

2
λ2 1
H − 2 (i j )
i j ∈ss
⎛ ⎞1 ⎛ ⎞1
2 2

|||λ|||sd =⎝ λ 2
1
⎠ , |||λ||| = ⎝
d
λ 2
1
⎠ .
H 2 (i j ) H 2 (i j )
i j ∈sd i j ∈dd
Note that in most geometrical situations, X d (and hence X ) is not complete for the
above norm, but none of the subsequent proofs require its completeness.
The matching condition between subdomains is weakly enforced through the fol-
lowing bilinear forms:

∀v ∈ X s , ∀μ ∈ s , bs (v, μ) = [v], μi j ,
i j ∈ss

∀v ∈ X d , ∀μ ∈ d , bd (v, μ) = [v · n], μi j , (2.44)
i j ∈dd

∀v ∈ X, ∀μ ∈ sd , bsd (v, μ) = [v · n], μi j .
i j ∈sd
For the velocity and pressure in the Darcy and Stokes regions, we use the following
bilinear forms:

∀(u, v) ∈ X s × X s , as,i (u, v) = 2 νs D(us,i ) : D(v s,i )
s,i

n−1
νs α
+ √ (us ·τ l )(v s ·τ l ), 1 ≤ i ≤ Ms ,
l=1 ∂ ∩
Kl
s,i sd

∀(u, v) ∈ X d × X d , ad,i (u, v) = νd K −1 ud,i · v d,i , 1 ≤ i ≤ Md , (2.45)
d,i

∀v ∈ X, ∀w ∈ L 2 (), bi (v, w) = − wdiv v i , 1 ≤ i ≤ M.
i
Then we set

Ms
Md
∀(u, v) ∈ X × X, a(u, v) = as,i (u, v) + ad,i (u, v),
i=1 i=1
123

M
∀(v, w) ∈ X × L 2 (), b(v, w) = bi (v, w).
i=1
The second variational formulation reads: Find (u, p, λsd , λd , λs ) ∈ X × W × sd ×

d × s such that

∀v ∈ X, a(u, v) + b(v, p) + bsd (v, λsd ) + bd (v, λd ) + bs (v, λs ) = f · v,

∀w ∈ W, b(u, w) = − w qd , (2.46)
d
∀μ ∈ sd , bsd (u, μ) = 0, ∀μ ∈ d , bd (u, μ) = 0, ∀μ ∈ s , bs (u, μ) = 0.
It remains to prove that (2.46) is equivalent to (2.3)–(2.11) when the solution is

sufficiently smooth. Since we know from Theorem 2.1 that (2.3)–(2.11) has a unique
solution, equivalence will also establish that (2.46) is uniquely solvable.
Theorem 2.2 Assume that the solution (u, p) ∈ X̃ × W of (2.3)–(2.11) with pd ∈

H 1 (d ) satisfies
1
∀i j ∈ dd ∪ sd , (ud · nd )|i j ∈ H − 2 (i j ), ∀i j ∈ ss ,
1
(σ (us , ps )ns )|i j ∈ H − 2 (i j )n .
Then (2.46) is equivalent to (2.3)–(2.11).
Proof The argument is similar to that used in proving Lemma 2.1. Let (u, p) be a
solution of (2.3)–(2.11) satisfying the above regularity. Take the scalar product in each
i of (2.3) and (2.6) with a test function v in X , apply Green’s formula and add the
corresponding equations. In view of (2.16) and the regularity of (u, p), this gives:

a(u, v) + b(v, p) − σ (us , ps )ni j , [v]i j + pd , [v · n]i j = f · v.
i j ∈ss i j ∈dd ∪sd
(2.47)
We recover the first equation in (2.46) by defining
∀i j ∈ ss , λs |i j = −σ (us , ps )|i j ni j ,

(2.48)
∀i j ∈ dd , λd |i j = pd |i j , ∀i j ∈ sd , λsd |i j = pd |i j ,
and the remaining equations follow from the regularity of (u, p).
Conversely, let (u, p, λsd , λd , λs ) in X × W × sd × d × s be a solution of
(2.46). By choosing smooth test functions with compact support in each i we recover
the interior equations (2.3), (2.4), (2.6), (2.7) in each subdomain. On one hand, we
easily derive from the last equation of (2.46) that u has no jump through the interfaces
123
of ss . Hence u ∈ H 1 (s )n . On the other hand, we pick an index i with 1 ≤ i ≤ Ms

and an interface i j in ss , we take a function v in X , smooth in i , zero outside i ,
and zero on ∂i \i j . By taking the scalar product of (2.6) in i with v, applying
Green’s formula, comparing with (2.46), and using the same process in j , we find
λs |i j = −σ (us,i , ps,i )|i j ni j = −σ (us, j , ps, j )|i j ni j .
As λs is single-valued, this implies that σ (us , ps )|i j ni j has no jump through i j .

This is true for all interfaces in ss . Therefore (2.6) is satisfied in s . By applying
a similar process to the interfaces of dd , we derive first that (ud , pd ) belongs to
H (div; d ) × H 1 (d ) and next that (2.3) is satisfied in d . Finally, the third equation
in (2.46) implies that u belongs to H (div; ), therefore u is in X and the pair (u, p)
solves (2.19)–(2.20); by virtue of Lemma 2.1, it also solves (2.3)–(2.11).
3 Discretization
3.1 Meshes and discrete spaces
In view of discretization, we assume from now on that and all its subdomains i ,
1 ≤ i ≤ M, have polygonal or polyhedral boundaries. Since none of the subdomains
overlap, they form a mesh, Td of d and Ts of s , and the union of these meshes
constitutes a mesh T of . Furthermore, we suppose that this mesh satisfies the
following assumptions:
Hypothesis 3.1 1. T is conforming, i.e. it has no hanging nodes.

2. The subdomains of T can take at most L different geometrical shapes, where L
is a fixed integer independent of M.
3. T is shape-regular in the sense that there exists a real number σ , independent of
M such that
diam(i )
∀i, 1 ≤ i ≤ M, ≤ σ, (3.1)
diam(Bi )
where diam(i ) is the diameter of i and diam(Bi ) is the diameter of the largest
ball contained in i . Without loss of generality, we can assume that diam(i ) ≤ 1.
As each subdomain i is polygonal or polyhedral, it can be entirely triangulated.

Let h > 0 denote a discretization parameter, and for each h, let Ti h be a regular family
of partitions of i made of triangles or tetrahedra T in the Stokes region and triangles,
tetrahedra, parallelograms, or parallelepipeds in the Darcy region, with no matching
requirement at the subdomains interfaces. Thus the meshes are independent and the
parameter h < 1 is allowed to vary with i, but to reduce the notation, unless necessary,
we do not indicate its dependence on i. By regular, we mean that there exists a real
number σ0 , independent of i and h such that
123
hT
∀i, 1 ≤ i ≤ M, ∀T ∈ Ti h , ≤ σ0 , (3.2)
ρT
where h T is the diameter of T and ρT is the diameter of the largest ball contained
in T .
In addition we assume that each element of Ti h has at least one vertex in i . For
the interfaces, let H > 0 be another discretization parameter and for each H and each
j denote a regular family of partitions of i j into segments, triangles or
i < j, let Ti H
parallelograms of diameter bounded by H . These partitions may not match the traces
of the subdomain grids.
On these meshes, we define the following finite element spaces. In the Stokes
region, for each s,i , let (X s,i
h , W h ) ⊂ H 1 ( )n × L 2 ( ) be a pair of finite
s,i s,i s,i
element spaces satisfying a local uniform inf-sup condition for the divergence. More
h
precisely, setting X 0,s,i = X s,i
h ∩ H 1 ( )n and W h
0 s,i 0,s,i = Ws,i ∩ L 0 (s,i ), we assume
h 2

that there exists a constant βs > 0, independent of h and the diameter of s,i , such
that

s,i w div v
h h
∀i, 1 ≤ i ≤ Ms , inf sup ≥ βs . (3.3)
w h ∈W0,s,i
h
v h ∈X h |v h | H 1 (s,i ) w h L 2 (s,i )
0,s,i
h
In addition, since X 0,s,i ⊂ H01 (s,i )n , it satisfies a Korn inequality: There exists a

constant α > 0, independent of h and the diameter of s,i , such that
∀i, 1 ≤ i ≤ Ms , ∀v h ∈ X 0,s,i
h
, D(v h ) L 2 (s,i ) ≥ α |v h | H 1 (s,i ) . (3.4)
There are well-known examples of pairs satisfying (3.3) (cf. [34]), such as the mini-
element, the Bernardi–Raugel element, or the Taylor-Hood element. Similarly, in the
Darcy region, for each d,i , let (X d,i
h , W h ) ⊂ H (div; ) × L 2 ( ) be a pair of
d,i d,i d,i
mixed finite element spaces satisfying a uniform inf-sup condition for the divergence.
h
More precisely, setting X 0,d,i = X d,i
h ∩ H (div; ) and W h
0 d,i 0,d,i = Wd,i ∩ L 0 (d,i ),
h 2

we assume that there exists a constant βd > 0 independent of h and the diameter of
d,i , such that

d,i w h div v h
∀i, 1 ≤ i ≤ Md , inf sup ≥ βd . (3.5)
w h ∈W0,d,i
h
v h ∈X 0,d,i
h v h H (div;d,i ) w h L 2 (d,i )
Furthermore, we assume that
∀i, 1 ≤ i ≤ Md , ∀v h ∈ X d,i
h
, div v h ∈ Wd,i
h
. (3.6)
Again, there are well-known examples of pairs satisfying (3.5) and (3.6) (cf. [18] or
[34]), such as the Raviart–Thomas elements, the Brezzi–Douglas–Marini elements,
the Brezzi–Douglas–Fortin–Marini elements, the Brezzi–Douglas–Durán–Fortin ele-
ments, or the Chen–Douglas elements. Then we discretize straightforwardly X d and
X s by
123
X dh = {v ∈ L 2 (d )n ; v|d,i ∈ X d,i

h
, 1 ≤ i ≤ Md , v · n = 0 on d },
X sh = {v ∈ L 2 (s )n ; v|s,i ∈ X s,i
h
, 1 ≤ i ≤ Ms , v = 0 on s },
and we set
Wdh = {w ∈ L 2 (d ); w|d,i ∈ Wd,i

h
}, Wsh = {w ∈ L 2 (s ); w|s,i ∈ Ws,i
h
},
X h = {v ∈ L 2 ()n ; v|d ∈ X dh , v|s ∈ X sh },
W h = {w ∈ L 20 (); w|d ∈ Wdh , w|s ∈ Wsh }.
The finite elements regularity implies that X dh ⊂ X d , X sh ⊂ X s and X h ⊂ X . Of

course, W h ⊂ W .
In the mortar region, we take finite element spaces sH , dH , and sd
H . These spaces
consist of continuous or discontinuous piecewise polynomials. We will allow for vary-

ing polynomial degrees on different types of interfaces. Although the mortar meshes
and the subdomain meshes so far are unrelated, we need compatibility conditions
between sH , sd H and H on one hand, and X h and X h on the other hand. The
d d s
following set of conditions is fairly crude and will be sharpened further on.
1. For all i j ∈ ss ∪ sd , i < j, and for all v ∈ X̃ , there exists v h ∈ X s,i
h , vh = 0
on ∂s,i \i j satisfying

v · ni j =
h
v · ni j . (3.7)
i j i j
2. For all i j ∈ ss , i < j, and for all v ∈ X̃ , there exists v h ∈ X s,

h , v h = 0 on
j
∂s, j \i j satisfying

∀μ H ∈ sH , μ H · vh = μ H · v. (3.8)
i j i j
3. For all i j ∈ dd ∪ sd , i < j, and for all v ∈ X̃ , there exists v h ∈ X d,
h ,
j
v · n j = 0 on ∂d, j \i j satisfying
h

∀μ H
∈ dH , ∀μ H
∈ sd
H
, μ v · ni j =
H h
μ H v · ni j . (3.9)
i j i j
Condition (3.7) is very easy to satisfy in practice and it trivially holds true for all
examples of Stokes spaces considered in this paper. Conditions (3.8) and (3.9) state
that the mortar space is controlled by the traces of the subdomain velocity spaces.
Both conditions are easier to satisfy for coarser mortar grids. Condition (3.8) is more
general than previously considered in the literature for mortar discretizations of the
Stokes equations [11,12]. Examples for which (3.8) holds are given in the Appendix:
123
Sect. 7.1 in 2-D and Sect. 7.2 in 3-D. The condition (3.9) on dd is closely related
to the mortar condition for Darcy flow in [5,6,52,60] and on sd , it is more general
than existing mortar discretizations for Stokes–Darcy flows on [14,29,46,54]. It is
discussed in more detail in Sect. 4.4.
H H H
Lemma 3.1 Under assumptions (3.8) and (3.9), the only solution λsd , λd , λs in
sd × d × s to the system
H H H
∀v h ∈ X h , bs (v h , λsH ) + bd (v h , λdH ) + bsd (v h , λsd

H
)=0 (3.10)
is the zero solution.
Proof Consider any i j ∈ ss with i < j; the proof for the other interfaces being the
same. Take an arbitrary v in H01 ()n and v h associated with v by (3.8), extended by
zero outside s, j . Then on one hand,

λsH · v = λsH · v h = bs (v h , λsH ),
i j i j
and on the other hand,
bd (v h , λdH ) = bsd (v h , λsd

H
) = 0.
Therefore

∀v ∈ H01 ()n , λsH · v = 0,
i j
thus implying that λsH = 0.
Finally, we define
Vdh = {v ∈ X dh ; ∀μ ∈ dH , bd (v, μ) = 0}, Vsh = {v ∈ X sh ; ∀μ ∈ sH , bs (v, μ) = 0},

V h = {v ∈ X h ; v|d ∈ Vdh , v|s ∈ Vsh , ∀μ ∈ sd
H
, bsd (v, μ) = 0}, (3.11)
Z = {v ∈ V ; ∀w ∈ W , b(v, w) = 0}.
h h h
3.2 Variational formulations and uniform stability of the discrete problem
The discrete version of the second variational formulation (2.46) is: Find (uh , p h , λsd
H,
λd , λs ) ∈ X × W × sd × d × s such that
H H h h H H H

∀v h ∈ X h , a(uh , v h )+b(v h , p h )+bsd (v h , λsd
H ) + b (v h , λ H ) + b (v h , λ H ) =
d d s s f · vh ,

123

∀w h ∈ W h , b(uh , w h ) = − w h qd , (3.12)
d
∀μ H ∈ sd
H , b (uh , μ H ) = 0, ∀μ H ∈ H , b (uh , μ H ) = 0,
sd d d
∀μ H ∈ sH , bs (uh , μ H ) = 0.
The last three equations of (3.12) state that uh ∈ V h . Therefore, we can extract from
(3.12) the following reduced formulation: Find uh ∈ V h , p h ∈ W h such that

∀v ∈ V , a(u , v ) + b(v , p ) =
h h h h h h
f · vh ,

(3.13)
∀w h ∈ W h , b(uh , w h ) = − w h qd .
d
Lemma 3.2 Problems (3.12) and (3.13) are equivalent.

Proof Clearly, (3.12) implies (3.13). Conversely, if the pair (uh , p h ) solves (3.13),
H , λ H , λ H such that all these variables satisfy (3.12) is an easy conse-
existence of λsd d s
quence of Lemma 3.1 and an algebraic argument.
In view of this equivalence, it suffices to analyze problem (3.13). From the Babuška–
Brezzi’s theory, uniform stability of the solution of (3.13) stems from an ellipticity
property of the bilinear form a in Z h and an inf-sup condition of the bilinear form b.
Let us prove an ellipticity property of the bilinear form a, valid when n = 2, 3. For
this, we make the following assumptions on the mortar spaces:
Hypothesis 3.2 1. On each i j ∈ dd ∪ sd , dH |i j and sd H|
i j contain at least
P0 .
2. On each i j ∈ ss , on each hyperplane F ⊂ i j , sH | F contains at least Pn0 .
3. On each i j ∈ ss , sH |i j contains at least Pn1 .
The second assumption guarantees that ni j ∈ sH |i j ; it follows from the third assump-
tion when i j is flat. The third assumption is solely used for deriving a discrete Korn
inequality; it can be relaxed, as we shall see in Sect. 7.2. The first two assumptions
imply that all functions v h in V h satisfy
M
M

div v h = v h · ni = [v h · n] = 0.
i=1 i=1∂ i< j
i i ij
Therefore, the zero mean-value restriction on the functions of W h can be relaxed.

Thus the condition v h ∈ Z h and (3.6) imply in particular that div v dh = 0 in d,i ,
1 ≤ i ≤ Md . Hence
∀v h ∈ Z h , |||v dh ||| X d = v dh L 2 (d ) . (3.14)
First, we treat the simpler case when |s | > 0 and s is connected.
123
Lemma 3.3 Let |s | > 0 and s be connected. Then under Hypothesis 3.2, we have
νd νs
∀v h ∈ Z h , a(v h , v h ) ≥ |||v h |||2 + 2 2 |||v sh |||2X s , (3.15)
λmax d X d C
where the constant C only depends on the shape regularity of Ts .
Proof Since v sh ∈ Vsh and Pn1 ∈ ss

H|
i j for each i j ∈ ss , then P1 [v s ] = 0, where
h
P1 is the orthogonal projection on Pn1 for the L 2 norm on each i j . Thus, as s is

connected and |s | > 0, inequality (1.12) in [16] gives

Ms
Ms
∀v sh ∈ Vsh , |v sh |2H 1 ( ≤ C2 D(v sh )2L 2 ( , (3.16)
s,i ) s,i )
i=1 i=1
where the constant C only depends on the shape regularity of Ts . Hence we have the
analogue of (2.30):
νs
Ms
νd
∀v h ∈ Z h , a(v h , v h ) ≥ |||v dh |||2X d + 2 2 |v sh |2H 1 ( ) . (3.17)
λmax C s,i
i=1
Finally the above argument permits to apply formula (1.3) in [15] in order to recover the
full norm of X s in the right-hand side of (3.17). In fact, it is enough that Pn0 ∈ ss
H|
i j
for each i j ∈ ss .
Now we turn to the case when s is connected and |s | = 0, consequently sd =
∂s , up to a set of zero measure.
Lemma 3.4 Let |s | = 0 and s be connected, i.e. sd = ∂s . Then under Hypoth-
esis 3.2, we have

νd νs α
∀v h ∈ Z h , a(v h , v h ) ≥ |||v dh |||2X d + 2 min 2, √ |||v sh |||2X s ,
λmax C λmax |sd |
(3.18)
where the constant C only depends on the shape regularity of Ts .
Proof All constants in this proof only depend on the shape regularity of Ts . Since
(3.14) holds, it suffices to derive a lower bound for as (v sh , v sh ). To begin with, as
v sh ∈ Vsh , we apply Theorem 4.2 if n = 2 or 5.2 if n = 3 in [16]:
M

Ms s

h 2
∀v sh ∈ Vsh , |v sh |2H 1 ( ) ≤C 2
D(v sh )2L 2 ( ) + (v s ) ,
s,i s,i
i=1 i=1
123
where the functional is a suitable seminorm. Let us choose

∀v ∈ X s , (v) = v × ns . (3.19)

sd
Clearly, is a seminorm on X s . Next, considering that

∀v ∈ H (s ) , (v) = curl v ,
1 n

s
it is easy to check that if m is a rigid body motion, then (m) = 0 if and only if m
is a constant vector. Finally, observing that behaves exactly like the functional 2
of Example 2.4 in [16], we see that satisfies all assumptions of Theorem 4.2 or 5.3
in [16]. Thus
⎛ 2 ⎞

M s M s
2⎜ ⎟
∀v s ∈ Vs ,
h h
|v s | H 1 ( ) ≤ C ⎝
h 2
D(v s ) L 2 ( ) +
h 2
v s × ns ⎠ .
h
s,i s,i
i=1 i=1 sd
(3.20)
In order to recover the full norm of X s in the left-hand side of (3.20), we apply Theorem
5.1 in [15] with the functional
n−1

(v) = |v s · τ l |.
l=1
sd
Again, is a seminorm on X s and since sd is a closed curve or surface, the condition
(c) = 0 for a constant vector c implies that c = 0. The remaining assumptions of
this theorem easily follow by observing that has the same behavior as the functional
1 of Example 4.2 in [15]. This yields
⎛ ⎛ ⎞2 ⎞
Ms n−1

⎜ ⎜ ⎟ ⎟
∀v sh ∈ Vsh , v sh 2L 2 ( ) ≤ C 2 ⎝ |v sh |2H 1 ( ) + ⎝ |v sh · τ l |⎠ ⎠ . (3.21)
s s,i
i=1 l=1
sd
By combining (3.20) and (3.21), we obtain

⎛ 2 ⎛ ⎞2 ⎞

⎜
Ms n−1
⎜ ⎟ ⎟
∀v sh ∈ Vsh , |||v sh |||2X s ≤ C 2 ⎜
⎝ D(v sh )2L 2 ( ) + v sh × ns + ⎝ |v sh · τ l |⎠ ⎟⎠.
s,i
i=1 sd l=1
sd
(3.22)
Then (3.18) follows from (3.22) by arguing as in deriving (2.31).
123
The case when s is not connected follows from Lemmas 3.3 or 3.4 applied to each
connected component of s according that it is or is not adjacent to s .
To control the bilinear form b in s , we make the following assumption: There exists
a linear approximation operator sh : H01 ()n → Vsh satisfying for all v ∈ H01 ()n :
–

∀i, 1 ≤ i ≤ Ms , div sh (v) − v = 0. (3.23)
s,i
– For any i j in sd ,

h
s (v) − v · ni j = 0. (3.24)
i j
– There exists a constant C independent of v, h, H , and the diameter of s,i , 1 ≤

i ≤ Ms , such that
|||sh (v)||| X s ≤ C|v| H 1 () . (3.25)
A general construction strategy discussed in Sect. 4.1 gives an operator sh that satisfies
(3.23) and (3.24). The stability bound (3.25) is shown to hold for the specific examples
presented in Sects. 7.1 and 7.2.
Lemma 3.5 Assume that an operator sh satisfying (3.23)–(3.25) exists; then there
exists a linear operator sh : H01 ()n → Vsh such that for all v ∈ H01 ()n ,
Ms

∀w h ∈ Wsh , w h div(sh (v) − v) = 0, (3.26)
i=1
s,i

h
∀i j ∈ sd , s (v) − v · ni j = 0, (3.27)
i j
and there exists a constant C independent of v, h, H , and the diameter of s,i ,

1 ≤ i ≤ Ms , such that
|||sh (v)||| X s ≤ C|v| H 1 () . (3.28)
Proof The operator sh is constructed by correcting sh :
sh (v) = sh (v) + csh (v)
123
where csh (v)|s,i ∈ X 0,s,i

h and

∀w h
∈ W0,s,i
h
,1 ≤ i ≤ Ms , w div
h
csh (v) = w h div v − sh (v) . (3.29)
s,i s,i
Existence of csh (v) follows directly from (3.3) and with the same constant
1
|csh (v)| H 1 (s,i ) ≤ div(v − sh (v)) L 2 (s,i ) . (3.30)
βs
The restriction w h ∈ W0,s,i

h is relaxed by applying (3.23) and using the fact that csh (v)
h . Finally, (3.28) follows from the above bound and (3.25).
belongs to X 0,s,i
The idea of constructing the operator sh via the interior inf-sup condition (3.3)
and the simplified operator sh satisfying (3.23) and (3.25) is not new. It can be found
for instance in [36] and [12].
To control the bilinear form b in d , we make the following assumption: There
exists a linear operator dh : H01 ()n → Vdh satisfying for all v ∈ H01 ()n :
Md

∀w h ∈ Wdh , w h div dh (v) − v = 0. (3.31)
i=1
d,i
– For any i j in sd ,

∀μ H ∈ sd
H
, μ H dh (v) − sh (v) · ni j = 0. (3.32)
i j
– There exists a constant C independent of v, h, H , and the diameter of d,i ,

1 ≤ i ≤ Md , such that
|||dh (v)||| X d ≤ C|v| H 1 () . (3.33)
The construction of the operator dh is presented in Sect. 4. In particular, the general
construction strategy discussed in Sect. 4.1 gives an operator that satisfies (3.31) and
(3.32). The stability bound (3.33) is shown to hold for various cases in Sect. 4.4.
The next lemma follows readily from the properties of sh and dh .
123
Lemma 3.6 Under the above assumptions, there exists a linear operator h ∈
L(H01 ()n ; V h ) such that for all v ∈ H01 ()n
M

∀w h ∈ W h , w h div h (v) − v = 0, (3.34)
i=1
i
||| (v)||| X ≤ C|v| H 1 () ,

h
(3.35)
with a constant C independent of v, h, H , and the diameter of i , 1 ≤ i ≤ M.
Proof Take h (v)|s = sh (v) and h (v)|d = dh (v). Then (3.34) follows from
(3.26) and (3.31). The matching condition of the functions of V h at the interfaces of
sd holds by virtue of (3.32). Finally, the stability bound (3.35) stems from (3.28) and
(3.33).
The following inf-sup condition between W h and V h is an immediate consequence of

a simple variant of Fortin’s Lemma [18,34] and Lemma 3.6.
Theorem 3.1 Under the above assumptions, there exists a constant β > 0, indepen-
dent of h, H , and the diameter of i j for all i < j, such that
b(v h , w h )
∀w h ∈ W h , sup ≥ β w h L 2 () . (3.36)
v h ∈V h |||v h ||| X
Finally, well-posedness of the discrete scheme (3.13) follows from Lemma 3.3 or 3.4
and Theorem 3.1.
Corollary 3.1 Under the above assumptions, problem (3.13) has a unique solution
(uh , p h ) ∈ V h × W h and

|||uh ||| X + p h L 2 () ≤ C f L 2 () + qd L 2 (d ) , (3.37)
with a constant C independent of h, H , and the diameter of i j for all i < j.
Proof Since (3.13) is set into finite dimension, it is sufficient to prove uniqueness,
and as uniqueness follows from (3.37), it suffices to prove this stability estimate. Thus
let (uh , p h ) solve (3.13), which is a typical linear problem with a non-homogeneous
constraint. By virtue of the discrete inf-sup condition (3.36), there exists a function
uqh ∈ V h such that

∀w ∈ W ,
h h
b(uqh , w h ) =− w h qd ,
d
1
|||uqh ||| X ≤ qd L 2 (d ) . (3.38)
β
123
Then u0h = uh − uqh solves (3.13) with qd = 0 and the coercivity condition (3.15) or
(3.18) and the discrete inf-sup condition (3.36) imply that
|||u0h ||| X + p h L 2 () ≤ C( f L 2 () + qd L 2 (d ) ).
With (3.38), this gives (3.37).
4 Elements of construction of sh and hd
Constructing an operator sh with values in Vsh , satisfying (3.23)–(3.25), uniformly

stable with respect to the diameter of the subdomains and interfaces, is not straight-
forward, particularly in 3-D. On the other hand, a general construction of dh in d
can be found in [5], and we shall adapt it so that it matches suitably sh on sd . Let
us describe our strategy in each region.
4.1 General construction strategy
Let v ∈ H01 ()n . In s , we propose the following three-step construction.
1. Starting step. In each s,i , 1 ≤ i ≤ Ms , take sh (v) = S h (v), where S h (v) is
a Scott and Zhang [56] approximation operator, constructed so that S h (v)|∂s,i
only uses values of v restricted to ∂s,i , and in particular, S h (v)|s = 0. Thus
S h (v)|s,i ∈ X s,i
h and S h (v) ∈ X h .
s
2. First correction step. For each i j ∈ ss ∪ sd with i < j, correct sh (v) in s,i
by setting
sh (v)|s,i := sh (v)|s,i + ci,

h
ij
(v),
h
where ci,ij
(v) ∈ X s,i i,i j (v) = 0 on ∂s,i \i j ,
h , ch

h
ci,ij
(v) · ni j = v − S h (v)|s,i · ni j , (4.1)
i j i j
and satisfies suitable bounds. The existence of ci, h

ij
(v) (without bounds) is guar-
anteed by assumption (3.7). In particular, one can define it on i j to satisfy (4.1),
extend it by zero on ∂s,i \i j , and extend it arbitrarily inside s,i to a function in
i,i j (v) only affects values at points located in the
h . Note that the correction ch
X s,i
interior of s,i and in the interior of i j . It has no influence on values at points
located on the other interfaces. For this reason, the discrete term that appears in
the right-hand side of (4.1) is the trace on i j of S h (v) coming from s,i ; this
term has not been yet corrected. Once this correction step is performed on all
i j ∈ ss ∪ sd with i < j, the resulting function sh (v) satisfies on all these
interfaces
123

h
s (v)|s,i − v · ni j = 0. (4.2)
i j
Note that this step does not modify the trace on i j of S h (v) coming from s, j .
This trace will be modified in the next step.
3. Second correction step. For each i j ∈ ss with i < j, correct sh (v) in s, j by
setting
sh (v)|s, j := sh (v)|s, j + chj,i j (v),
where chj,i j (v) ∈ X s, j,i j (v) = 0 on ∂s, j \i j ,

h , ch
j

∀μ H
∈ sH , μ ·
H
chj,i j (v) = μ H · sh (v)|s,i − S h (v)|s, j , (4.3)
i j i j
and satisfies suitable bounds. The existence of chj,i j (v) (without bounds) is guar-
anteed by assumption (3.8). In particular, one can define it on i j to satisfy (4.3),
extend it by zero on ∂s, j \i j , and extend it arbitrarily inside s, j to a func-
j,i j (v) has no influence on values at
h . Here also, the correction ch
tion in X s, j
points located on the other interfaces. Once this correction step is performed on all
i j ∈ ss with i < j, the resulting function sh (v) satisfies on all these interfaces
bs (sh (v), μ H ) = 0 for all μ H ∈ sH , i.e.

∀μ H ∈ sH , [sh (v)] · μ H = 0. (4.4)
i j
Finally, if the second assumption in Hypothesis 3.2 holds, then n ∈ sH and (4.2)
and (4.4) imply that on all these interfaces,

h
s (v)|s, j − v · n = 0. (4.5)
i j
Therefore sh (v) satisfies (3.23) and (3.24). Specific constructions of the cor-
h
rections ci,ij
(v) and chj,i j (v) that guarantee that sh (v) also satisfies (3.25) are
presented in the forthcoming subsections.
Next, we propose the following two-step construction algorithm in d .
1. Starting step. Set Pdh (v) = R h (v) ∈ X dh , where R h (v) is a standard mixed approx-
imation operator associated with Wdh . It preserves the normal component on the
boundary:
123

∀i j ⊂ ∂d,k , 1 ≤ k ≤ Md , ∀v h ∈ X dh , v h · ni j R h (v)|d,k − v · ni j = 0,
i j
(4.6)
and satisfies

∀1 ≤ i ≤ Md , ∀w h ∈ Wdh , w h div R h (v) − v = 0. (4.7)
d,i
2. Correction step. It remains to prescribe the jump condition. For each i j ∈ dd ∪
sd with i < j, correct Pdh (v) in d, j by setting:
Pdh (v)|d, j := Pdh (v)|d, j + chj,i j (v),
where chj,i j (v) ∈ X d, j,i j (v) · n j = 0 on ∂d, j \i j , div c j,i j (v) = 0 in
h , ch
j
h
d, j ,

∀μ H
∈ dH , μ H chj,i j (v) · ni j = μ H R h (v)|d,i − R h (v)|d, j · ni j ,
i j i j
(4.8)

∀μ H ∈ sd
H
, μ H chj,i j (v) · ni j = μ H sh (v)|s,i − R h (v)|d, j · ni j ,
i j i j
and chj,i j (v) satisfies adequate bounds. Existence of a non necessarily divergence-
free chj,i j (v) (without bounds) follows from (3.9); it suffices to extend suitably
R h (v)|d, j and R h (v)|d,i or sh (v)|s,i . The zero divergence will be prescribed
in the examples. Note that chj,i j (v) has no effect on interfaces other than i j and
no effect on the restriction of Pdh (v) in d,i or on that of sh (v) in s,i . Therefore
these corrections can be superimposed.
When step 2 is done on all i j ∈ dd ∪ sd with i < j, the resulting function Pdh (v)
has zero normal trace on d , it belongs to Vdh since, due to the first equation in (4.8),
it satisfies for all i j ∈ dd with i < j

∀μ H ∈ dH , μ H [Pdh (v) · n] = 0, (4.9)
i j
and, as the corrections are assumed to be divergence-free in each subdomain,

∀w h ∈ Wdh , ∀1 ≤ i ≤ Md , w h div Pdh (v) − v = 0. (4.10)
d,i
123
Furthermore, due to the second equation in (4.8), it satisfies for all i j ∈ sd ,

∀μ H
∈ sd
H
, μ H sh (v)|s,i − Pdh (v)|d, j · ni j = 0. (4.11)
i j
Therefore, taking dh (v) = Pdh (v) in d , it satisfies (3.31) and (3.32).
h
The remainder of this section is devoted to corrections ci, ij
(v) and chj,i j (v) where
the constants in the stability bounds (3.25) and (3.33) are shown to be independent
of the discretization parameters and the diameter of the subdomains. In particular,
the bounds for the corrections will stem from the fact that each involves differences
between v and good approximations of v. Beforehand, we need to refine the assump-
tions on the meshes at the interfaces and refine Hypothesis 3.1 on the mesh of subdo-
mains.
Hypothesis 4.1 For i < j, let T be any element of Ti h that is adjacent to i j , and let
{T } denote the set of elements of T jh that intersect T . The number of elements in this
set is bounded by a fixed integer and there exists a constant C such that
|T | |T |−1 ≤ C.
The same is true if the indices i and j of the triangulations are interchanged. These
constants are independent of i, j, h, and the diameters of the interfaces and subdo-
mains.
Hypothesis 4.2 1. Each i , 1 ≤ i ≤ M, is the image of a “reference” polygonal or

polyhedral domain by an homothety and a rigid body motion:
ˆ i ), x = Fi ( x̂) = Ai Ri x̂ + bi ,
i = Fi ( (4.12)
where Ai = diam(i ), Ri is an orthogonal matrix with constant coefficients and

bi a constant vector.
2. There exists a constant σ1 independent of M such that for any pair of adjacent
subdomains i and j , 1 ≤ i, j ≤ M, we have
Ai A−1
j ≤ σ1 . (4.13)
By (4.12) diam( ˆ i ) = 1. In addition, it follows from Hypothesis 3.1 that on one

hand the reference domains ˆ i can take at most L geometrical shapes and on the other
hand,
1
∀i, 1 ≤ i ≤ M, diam( B̂i ) ≥ , (4.14)
σ
ˆ i and σ is the constant of (3.1).

where B̂i is the largest ball contained in
123
h
4.2 A construction of ci,ij
(v)
h
In this section we construct a correction ci,ij
(v) satisfying (4.1) and suitable bounds
needed to establish the stability estimate (3.25). Recall the approximation properties
of the Scott and Zhang operator of degree r ≥ 1. Let v ∈ H t (s,i ), for some real
number t ≥ 1 and let T be an element of Ti h , 1 ≤ i ≤ Ms . Let T denote the macro-
element that is used for defining the values of S h (v) in T . Then (cf. [56]) there exists
a constant C that depends only on r , t, and the shape regularity of Ti h , such that the
following local approximation error holds
min(r,t)
v − S h (v) L 2 (T ) + h T |v − S h (v)| H 1 (T ) ≤ Ch T |v| H min(r,t) (T ) . (4.15)
Owing to the regularity of Ti h , when (4.15) is summed over all T in Ti h , it gives
v − S h (v) L 2 (s,i ) + h|v − S h (v)| H 1 (s,i ) ≤ Ch min(r,t) |v| H min(r,t) (s,i ) . (4.16)
Let v ∈ H01 ()n . Consider the case when n = 3, the case n = 2 being simpler, and
also consider the case when i j is a polyhedral surface, not necessarily a flat plane. Let
h on , and T h
i,i j the trace of Ti on i j , 1 ≤ i ≤ Ms ,
h
X i, denote the trace of X s,i h
ij ij
i < j; Ti,
h
ij
is a triangular mesh of each flat face in i j . In all cases considered, the
restriction to each T of functions of X s,ih contains at least Pn . Now, choose one of
1
the faces, say F, of i j . For each interior vertex ak of Ti,
h , 1 ≤ k ≤ N , on F, let
ij F
Ok denote the macro-element of all triangles of Ti,
h
ij
(i.e. faces on F) that share the
vertex ak . Thus
NF
F = ∪k=1 Ok .
The set {Ok } is not a partition of F, but it can be transformed into a partition by setting
1 = O1 , and recursively k = Ok \ ∪k−1

=1 .
Note that some k may be empty. By construction, we have
NF
F = ∪k=1 k , k ∩ = ∅, k = , k ⊂ Ok .
This can be done for all flat faces of i j . For each k, let bk be the piecewise P1 “bubble”
function such that
bk (a ) = δk, ,
123
extended by zero to all vertices inside s,i , and define ci,

h
ij
(v) by

NF
1
h
ci,ij
(v) = ck bk , where ck = v− S h (v)|s,i , ck = 0, if k = ∅.
Ok bk
F⊂i j k=1 k
(4.17)
h
Lemma 4.1 The correction ci,ij
(v) defined by (4.17) satisfies (4.1), more precisely

∀F ⊂ i j , h
ci,ij
(v) = v − S h (v)|s,i , (4.18)
F F
and there exists a constant C independent of h and the diameter of s,i and i j such
that
∀v ∈ H01 ()n , |ci,

h
ij
h
(v)| H 1 (s,i ) ≤ C|v| H 1 (s,i ) , ci,ij
(v) L 2 (s,i ) ≤ Ch|v| H 1 (s,i ) .
(4.19)
Proof For each k, the support of the trace of bk on i j is contained in a single flat
plane, say F, therefore,

NF
NF
NF NF

h
ci,ij
(v) = ck bk = ck bk = ck bk = v − S h (v)|s,i
F F k=1 k=1 F k=1 Ok k=1
k

= v − S h (v)|s,i ,
F
since the set (k ) is a partition of .

To derive the estimate (4.19), we consider the faces T in k , pass to the reference
element T̂ , where T is the element of Ti h adjacent to T , apply a trace theorem in T̂ ,
and revert to T . This gives

v− S (v)|s,i ≤
h
v − S h (v)|s,i L 1 (T )

k T ⊂k
1

≤ Ĉ |T | |T |− 2 v− S h (v) L 2 (T ) +h T |v− S h (v)| H 1 (T ) .
T ⊂k
In this proof, Ĉ denotes constants that depend only on the reference element and the
shape regularity of the triangulation. Then considering the regularity of Ti h , the local
approximation formula (4.15) with r = t = 1 implies that
123

1 −1
v − S (v)|s,i ≤ Ĉ
h
h T |T | |T |− 2 |v| H 1 (T ) ≤ Ĉ|k | ρk 2 |v| H 1 (Dk ) ,

k T ⊂k
(4.20)
where Dk is the set of elements of Ti h where the values of v are taken for computing
S h (v), and
ρk = min T ∈Dk ρT .
Therefore, considering that |k | ≤ |Ok |, ck defined by (4.17) satisfies
−1
|ck | ≤ Ĉ ρk 2 |v| H 1 (Dk ) . (4.21)
Now,
2 2
NF NF

|ci,
h
(v)|2H 1 ( =
ck ∇ bk = ck ∇ bk .
ij s,i )
F⊂i j k=1 T ∈T h
F⊂i j k=1
s,i i T
But given an element T in Ti h there is at most a fixed (and small) number of indices
k such that bk |T = 0. Therefore
2
NF

|ci,
h
ij
(v)|2H 1 (T ) = ck ∇ bk ≤ Ĉ |ck |2 |bk |2H 1 (T ) ,
F⊂i j k=1 k
T
where the sum runs over all indices k such that bk |T = 0. By substituting (4.21) into
this inequality and estimating the norm of bk , we easily derive that
1 1
|ci,
h
(v)|2H 1 (T ) ≤ Ĉ |T ||v|2H 1 (D ) . (4.22)
ij ρk ρT2 k
k
Since ρk ρT2 has the same order as |T |, then by summing (4.22) over all T in Ti h and
considering that there is at most a fixed (and small) number of repetitions in the Dk ,
we obtain the first inequality in (4.19). The second inequality in (4.19) follows in a
similar manner from the representation (4.17) and bound (4.21).
4.3 A construction of chj,i j (v) in s . The 2-D case
Constructing a suitable correction solving (4.3) is fairly complex because it amounts

to satisfying many conditions per interface. In 2-D, it can be done by following the
sharp approach of Crouzeix and Thomée [23]. This is a general procedure that can
123
also be used in 3-D, but as observed in Remark 4.1, it restricts the meshes. To bypass
this restriction, we present a more ad hoc construction in 3-D, which is postponed to
the Appendix: Sect. 7.2.
We slightly modify the underlying Scott and Zhang operator by asking that the
modified operator S̃ h be continuous at the end points of all interfaces i j in ss ∪ sd ,
and coincide elsewhere with S h . This only concerns the set of points, say ak , 1 ≤ k ≤
P, that do not lie on s , since S h (v) is necessarily zero on s . Note that these are all
interior cross points of the triangulation of subdomains T . To achieve continuity at
any such point a, we pick an element Ta in s with vertex a and set

1
S̃ h (v)(a) = v.
|Ta |
Ta
h
Both corrections ci,ij
(v) and chj,i j (v) are defined to satisfy respectively (4.1)
and (4.3) with S h (v) replaced by S̃ h (v). For v ∈ H 1 (), this does not change the
approximation properties (4.16) of S̃ h and hence (4.19), except that the domain of
definition of v is possibly a little larger than s,i or s, j near the end points of i j .
Notation 4.1 From now on, we denote by ˜ s,k the possibly slightly enlarged domain
of definition of v in case S̃ h (v) is used, with the convention that it coincides with s,k
when S̃ h (v) is not used.
Owing to the continuity of S̃ h (v) at the end points of i j and the fact that all previously
made corrections vanish at the end points of the interfaces i j in ss ∪sd , the integrand
in the right-hand side of (4.3) also vanishes at the end points of i j . Thus the trace of
1
sh (v)|s,i − S̃ h (v)|s, j on i j belongs to H00
2
(i j )2 , see Sect. 1.1. We shall see that
this property is crucial for estimating uniformly the gradient of chj,i j (v).
We propose to split the construction of chj,i j (v) into two parts:
1 1
1. Construct an operator πsh ∈ L(H00
2
(i j )2 , X hj,i j ) such that for all ∈ H00
2
(i j )2 ,

∀μ H ∈ sH , (πsh () − ) · μ H = 0,
i j
πsh () = 0, at the end points of i j , (4.23)

|πsh ()| 1 ≤ κs || 1 ,
2 ( )
H00 2 ( )
H00
ij ij
with a constant κs that is independent of h, H , and the measure of i j .

2. Define chj,i j (v) to be a suitable extension of πsh (sh (v)|s,i − S̃ h (v)|s, j ) inside
s, j so that it satisfies all uniform properties required for the stability bound (3.25).
Note that chj,i j (v) satisfies (4.3) by construction. Also, the existence of the operator
πsh implies that condition (3.8) holds.
123
We now discuss the construction of an operator satisfying (4.23). Uniformity with

respect to the diameters of i j and s, j is obtained by reverting to the reference
subdomains. Let ˆ j be associated with s, j by Hypothesis 4.2 and let ˆ = F −1 (i j ).
j
Let the image by F j−1 of T j,
h
ij
be denoted by T̂ j . Similarly, set
X̂ j = {v̂ = v h ◦ F j−1 ; v h ∈ X hj,i j , v h = 0, at the end points of i j },

ˆ = {μ̂ = μ H ◦ F −1 ; μ H ∈ sH }.
j
1
Then, given ˆ in H00
2 ˆ 2 ˆ in X̂ j unique solution of
() , if we construct π̂ j ()

ˆ
∀μ̂ ∈ , ˆ · μ̂ =
π̂ j () ˆ · μ̂, (4.24)
ˆ ˆ
and satisfying
ˆ
|π̂ j ()| 1 ˆ
< Ĉ|| 1 , (4.25)
H00 ˆ
2 () ˆ
2 ()
H00
ˆ then by reverting to s, j and defining π h ()

with a constant Ĉ independent of j and , j
by
ˆ
π hj () ◦ F j = π̂ j ( ◦ F j ) = π̂ j (), (4.26)
we shall obtain an operator satisfying (4.23). Indeed, uniqueness of the solution of

(4.24) will guarantee that the mapping π̂ j is linear, and the inequality in (4.23) with the
1
2
same constant will follow from the invariance of the H00 seminorm by a homothety
and a rigid body motion. But since explicit constructions of π̂ j are fairly technical,
we postpone them to the Appendix: Sect. 7.1 and discuss the extension part (2) now.
First we construct a suitable function v in H 1 (s, j ).
1
Lemma 4.2 For any s, j ⊂ s , any interface ⊂ ∂s, j and any ∈ H00
2
() there
exists a function v = E() ∈ H (s, j ) satisfying
1
v| = , v|∂s, j \ = 0,
the mapping E is linear and there exists a constant C independent of , v, and the
diameter of and s, j such that
|v| H 1 (s, j ) ≤ C|| 1 , v L 2 (s, j ) ≤ C A j || 1 , (4.27)

2 ()
H00 2 ()
H00
where A j is the diameter of s, j .
123
1
Proof With the notation of Hypothesis 4.2, let ˆ = ◦ F j . Then ˆ belongs to H00
2 ˆ
();
ˆ j and the extended function, say ˆ0 belongs to
hence it can be extended by zero to ∂
1
ˆ j ) with
H 2 (∂
|ˆ0 | 1
ˆ j)
ˆ
≤ Ĉ|| 1 . (4.28)
H 2 (∂ ˆ
2 ()
H00
1
ˆ j ); H 1 (
There exists an extension operator Ê ∈ L(H 2 (∂ ˆ j )) such that
Ê 1
ˆ j );H 1 (
ˆ j ))
≤ Ĉ. (4.29)
L(H 2 (∂
Take v̂ = Ê(ˆ0 ) and v = v̂◦F j−1 . The mapping → v is linear, v belongs to H 1 (s, j ),
its trace satisfies the statement of the lemma, and it remains to check (4.27). The
following inequalities stem from (4.29), (4.28) and a straightforward scaling argument:
|v| H 1 (s, j ) ≤ |v̂| H 1 (ˆ j ) ≤ Ĉ|ˆ0 | 1

ˆ j)
ˆ
≤ Ĉ|| 1 ≤ Ĉ|| 1 ,
H 2 (∂ H00 ˆ
2 () 2 ()
H00
(4.30)
v L 2 (s, j ) ≤ Ĉ A j v̂ L 2 (ˆ j ) ≤ Ĉ A j || 1 .
2 ()
H00
Next, we take = i j and define
chj,i j (v) = S h (w), w = E(πsh (sh (v)|s,i − S̃ h (v)|s, j )), (4.31)
where the Scott and Zhang interpolant S h is constructed so that S h (w) vanishes on
∂s, j \i j . The uniform approximation properties (4.16) of S h , (4.25), and (4.27)
imply, with constants C independent of h and the diameters of i j , s,i and s, j :

chj,i j (v) H 1 (s, j ) ≤ C|w| H 1 (s, j ) ≤ C {sh (v)|s,i − S̃ h (v)|s, j } 1 , (4.32)
2 ( )
H00 ij
and it remains to derive a uniform bound for the last norm in terms of v. This is the
object of the next lemma.
Lemma 4.3 Let i j ∈ ss , i < j. There exists a constant C independent of h and the
diameters of i j , s,i and s, j , such that

∀v ∈ H01 ()2 , {sh (v)|s,i − S̃ h (v)|s, j } 1 ≤ C|v| H 1 (˜ s,i ∪˜ s, j ) , (4.33)
2 ( )
H00 ij
˜ s,k is defined in Notation 4.1.

where
123
Proof Recall that the trace on i j of sh (v)|s,i is S̃ h (v)|s,i + ci,

h
ij
(v). As ci,
h
ij
(v)
vanishes on ∂s,i \i j , then by passing to the reference subdomain ˆ i , we obtain
|ci,
h
(v)| 1 = |ci,
h
(v) ◦ Fi | 1
ij 2 ( )
H00 ij
ij ˆ
2 ()
H00
≤ Ĉ|ci,
h
ij
(v) ◦ Fi | H 1 (ˆ i ) ≤ Ĉ|ci,
h
ij
(v)| H 1 (s,i ) .
Therefore it is bounded owing to (4.19), with ˜ s,i instead of s,i . Thus it suffices to
consider the jump [ S̃ h (v)] through i j .
First, by writing [ S̃ h (v)] = [ S̃ h (v) − v], by passing to the reference subdomains
ˆ
k , k = i, j, and by using the approximation properties (4.16) of S h , we derive
h
S̃ (v) − v | 1 = {( S̃ h (v)−v) ◦ Fk |ˆ } 1 ≤ Ĉ( S̃ h (v)−v) ◦ Fk H 1 (ˆ )
s,k
H 2 (i j ) k ˆ
H 2 () k
1
≤ Ĉ | S̃ h (v) − v|2H 1 ( ) + A−2 S̃ h (v) − v2L 2 ( )
2
s,k k s,k
1
≤ Ĉ |v|2 1 ˜ + (h A−1 2 |v|2 2
) ˜ . (4.34)
H (s,k ) k 1 H (s,k )
Since h < Ak , this bounds the first part of the norm and it remains to estimate

1 h 2
[ S̃ (v)] ,
d(x)
i j
where d(x) denotes the distance between x and the end points of i j . This part does
not have the same immediate bound as the above correction because the jump does
not vanish on the other boundaries of the subdomains. Let us first consider the case
when i j is a straight line. There is no loss of generality in assuming that i j lies on
the axis y = 0 with end point at the origin: i j =]0, L[. Let 0 = x0 < x1 < · · · <
x N < x N +1 = L denote the nodes of Ti, h
ij
and exceptionally set h n = xn+1 − xn .
Then d(x) = Min(x, L − x) and since
L L
1 h 2 1 h 2 1 h 2
[ S̃ (v)] ≤ [ S̃ (v)] + [ S̃ (v)] ,
d(x) x L−x
i j 0 0
it suffices to examine the first term:
L xn+1
1 h 2 N
1 h 2
[ S̃ (v)] = [ S̃ (v)] .
x x
0 n=0 xn

For any n ≥ 1, we have xn ≥ h n−1 , and since [ S̃ h (v)] belongs to H 1 (0, L) ∩
2
C 0 (0, L) , we can write
123
xn+1 xn+1⎛ x
⎞2
1 h 2 1 d h
[ S̃ (v)] ≤ ⎝[ S̃ h (v)(xn )] + [ S̃ (v)(t)]dt ⎠ d x
x xn dt
xn xn xn
2
2h n h d
≤ [ S̃ h (v)(xn )]2 +
n [ S̃ h (v)] . (4.35)
h n−1 2 dx 2
L (xn ,xn+1 )
An equivalence of norms yields
2h n h 2 2h n 1
[ S̃ (v)(xn )] ≤ [ S̃ h (v)]2L ∞ (xn ,xn+1 ) ≤ Ĉ [ S̃ h (v)]2L 2 (x ,x )
h n−1 h n−1 h n−1 n n+1
1
= Ĉ ( S̃ h (v)|s,i − v) − ( S̃ h (v)|s, j − v)2L 2 (x ,x ) .
h n−1 n n+1
By arguing as in the proof of Lemma 4.1, we obtain

1
S̃ h (v)|s,i − v L 2 (xn ,xn+1 ) ≤ Ĉh T2 |v| H 1 (˜ n,i ) , (4.36)
where T is the element of Ti h adjacent to [xn , xn+1 ] and ˜ n,k denotes the set of
elements in s,k used in defining S̃ (v)|s,k on [xn , xn+1 ].
h
The estimation of the second term is more technical because now S̃ h (v) is con-
structed on T jh whereas [xn , xn+1 ] is the side of an element of Ti h . Let {T } denote the
set of elements of T jh that intersect [xn , xn+1 ]. The argument of Lemma 4.1 gives:
1
S̃ h (v)|s, j − v L 2 (xn ,xn+1 ) ≤ Ĉ h T2 |v| H 1 (˜ n, ) . (4.37)

Then combining (4.36) and (4.37), using the regularity of the triangulation and Hypoth-
esis 4.1, we obtain
2h n h 2
[ S̃ (v)(xn )] ≤ Ĉ |v|2H 1 (˜ ) + |v|2H 1 (˜ ) . (4.38)
h n−1 n,i n, j
Similarly,
h 2n
d h
2
1
[ S̃ (v)] 2 ≤ Ĉ [ S̃ h (v)]2L 2 (x ,x )
h n−1 d x L (xn ,xn+1 ) h n−1 n n+1
1
= Ĉ [ S̃ h (v) − v]2L 2 (x ,x )
h n−1 n n+1
2
≤ Ĉ |v| H 1 (˜ ) + |v| H 1 (˜ ) .
2
(4.39)
n,i n, j
When n = 0, as [ S̃ h (v)(0)] = 0, there only remains the second term in the first line
of (4.35), and we immediately derive
123
x1 x1 x
1 h 2 1 d h 2
[ S̃ (v)] = [ S̃ (v)] dt d x
x x dt
0 0 0
x1
d h
2
d

2
≤ [ S̃ (v)] 2 ≤ h1 [ S̃ h (v)] 2
dx L (0,x) dx L (0,x1 )
0
2
≤ Ĉ |v| H 1 (˜ ) + |v|2H 1 (˜ ) .
0,i 0, j
Finally, when i j is a polygonal line, the above argument is valid on the segments that
share an end point with i j and is simpler on the other segments as d(x) is not small
compared to h.
Inequalities (4.32) and (4.33) imply the following.
Corollary 4.1 Let i j ∈ ss , i < j. There exists a constant C independent of h and
the diameters of i j , s,i and s, j , such that
∀v ∈ H01 ()2 , chj,i j (v) H 1 (s, j ) ≤ C|v| H 1 (˜ s,i ∪˜ s, j ) . (4.40)
Corollary 4.2 The approximation operator sh constructed in Sect. 4.1 with the above
h
corrections ci,ij
(v) and chj,i j (v) satisfies assumption (3.25).
Proof Since sh (v) is constructed by correcting the Scott-Zhang interpolant S̃ h with
h
ci,ij
(v) and chj,i j (v), assumption (3.25) follows from (4.16), (4.19), (4.40), and the
Poincaré inequality (1.1).
Remark 4.1 The above correction chj,i j (v) has a straightforward extension in 3-D,
but its use is limited because now it requires continuity of S̃h (v) on the edges of
subdomains; thus the meshes must match on these edges. This is not required by the
correction defined in Sect. 7.2.
4.4 A construction of chj,i j (v) in d
Here we construct a correction chj,i j (v) in d satisfying (4.8) and suitable continuity
bounds that are needed to establish the stability estimate (3.33). Recall that the exis-
tence of chj,i j (v) relies on (3.9). In the construction below we directly show that (3.9)
holds for a wide range of mesh configurations.
Let v be given in H01 ()n . Recall that the mixed approximation operator R h defined
in each d,i takes its values in X dh and satisfies (4.6) on each i j ⊂ ∂d,k , 1 ≤ k ≤
Md , and (4.7) in each d,i , 1 ≤ i ≤ Md :
123

∀v h ∈ X dh , v h · ni j R h (v)|d,k − v · ni j = 0,
i j

∀w h ∈ Wdh , w h div R h (v) − v = 0.
d,i
Furthermore there exists a constant C independent of h and the geometry of d,i , such
that
∀v ∈ H01 ()n , R h (v) H (div;d,i ) ≤ Cv H 1 (d,i ) , 1 ≤ i ≤ Md . (4.41)
This is easily established by observing that the moments defining the degrees of free-
dom of R h (v) are invariant by homothety and rigid-body motion; in particular the
normal vector is preserved. In addition, it satisfies (3.6):
∀i, 1 ≤ i ≤ Md , ∀v h ∈ X d,i
h
, div v h ∈ Wd,i
h
.
The above properties also imply (3.5): for all i, 1 ≤ i ≤ Md ,

d,i w h div v h
inf sup ≥ βd ,
w h ∈W0,d,i
h
v h ∈X 0,d,i
h v h H (div;d,i ) w h L 2 (d,i )
with a constant βd > 0 independent of h and Ai .

Now, let i j ∈ dd ∪ sd ; by analogy with the situation in the Stokes region, we
j,i j the trace space of X d, j on i j . Following [5], we define the space
h
denote by X d, h
of normal traces
X nj,i j = {w · ni j ; w ∈ X d,
h
j,i j },
and the orthogonal projection Q hj,i j from L 2 (i j ) into X nj,i j . Then we make the
following assumption: There exists a constant C, independent of H , h, i, j, and the
diameters of i j and d, j , such that
∀μ H ∈ dH , ∀μ H ∈ sd
H
, μ H L 2 (i j ) ≤ CQ hj,i j (μ H ) L 2 (i j ) . (4.42)
It is shown in [60] that (4.42) holds for both continuous and discontinuous mortar
spaces, if the mortar grid Ti Hj is a coarsening by two of T j,i j . A similar inequality
h
for more general grid configurations is shown in [52]. Formula (4.42) implies that the
projection Q hj,i j is an isomorphism from the restriction of sd
H , respectively H , to
d
i j , say sd,i
H
j respectively d,i j , onto its image in X j,i j , and the norm of its inverse
H n
is bounded by C. Then a standard algebraic argument shows that its dual operator,
namely the orthogonal projection from X nj,i j into sd,i H , respectively H , is also
j d,i j
an isomorphism from the orthogonal complement in X nj,i j of the projection’s kernel
123
onto sd,i
H , respectively H , and the norm of its inverse is also bounded by C. As
j d,i j
a consequence, for each f ∈ L 2 (i j ), there exists v h · ni j ∈ X nj,i j such that

∀μ H ∈ dH , ∀μ H ∈ sd
H
, μ v · ni j =
H h
f μH ,
i j i j
and there exists a constant C independent of h, and the diameter of i j , such that
v h · ni j L 2 (i j ) ≤ C f L 2 (i j ) .
This implies that (3.9) holds. Furthermore, the solution v h · ni j is unique in the orthog-
onal complement of the projection’s kernel and by virtue of this uniqueness, v h · ni j
depends linearly on f . This result permits to partially solve (4.8).
Lemma 4.4 Let v ∈ H01 ()n . Under assumption (4.42), for each i j ∈ dd ∪ sd ,
there exists w h · ni j ∈ X nj,i j such that

∀μ H
∈ dH , μ w · ni j =
H h
μ H R h (v) · n ,
i j i j (4.43)

w h · ni j L 2 (i j ) ≤ C R h (v) · n L 2 (i j ) ,

∀μ H
∈ sd
H
, μ w · ni j =
H h
μ H sh (v)|s,i − R h (v)|d, j · ni j ,
i j i j (4.44)

w h · ni j L 2 (i j ) ≤ C sh (v)|s,i − R h (v)|d, j · ni j L 2 (i j ) ,
with the constant C of (4.42). The mapping v → w h · ni j is linear.
Lemma 4.4 constructs a normal trace w h · ni j on i j and we must extend it inside

d, j . To this end, let h ∈ L 2 (∂d, j ) be the extension of w h · ni j by zero on ∂d, j .
Next, we solve the problem: Find q ∈ H 1 (d, j ) ∩ L 20 (d, j ) such that
∂q
q = 0 in d, j , = h on ∂d, j . (4.45)
∂n j
3
Lemma 4.5 Problem (4.45) has one and only one solution q ∈ H 2 (d, j )∩ L 20 (d, j )
and

|q| H 1 (d, j ) ≤ C A j [R h (v) · n] L 2 (i j ) ,
|q| 3 ≤ C[R h (v) · n] L 2 (i j ) , i j ∈ dd , (4.46)
H 2 (d, j )

|q| H 1 (d, j ) ≤ C A j sh (v)|s,i − R h (v)|d, j · ni j L 2 (i j ) ,
123

|q| 3 ≤ C sh (v)|s,i − R h (v)|d, j · ni j L 2 (i j ) , i j ∈ sd , (4.47)
H 2 (d, j )
with constants C independent of h, H , q, i, j, and the diameters of i j and d, j .
Proof Since μ H contains the constant functions, h satisfies on i j ∈ dd , using

(4.43) and (4.6),

h
h = w h · ni j = (R (v) − v) · n] = 0,
∂d, j i j i j
and this implies the unique solvability of (4.45). A similar argument holds on i j ∈ sd
using (4.44) and (3.24). In order to check (4.46), we pass to the reference subdomain
ˆ j associated with d, j , let n̂ j be its exterior unit normal vector, ˆ = F −1 (i j ),
j
q̂ = q ◦ F j and ˆ = h ◦ F j = (w h ◦ F j ) · n̂ j ; then q̂ ∈ H 1 (
ˆ j ) ∩ L 2 (
0
ˆ j ) is the
unique solution of
∂ˆ q̂
ˆ q̂ = 0 in
ˆ j, = A j ˆ on ∂
ˆ j.
∂ˆ n̂ j
As ˆ is in L 2 (∂
ˆ j ), it follows from [43] that q̂ belongs to H 2 (
ˆ j ) and 3
q̂ 3 ˆ 2 ˆ ,
≤ Ĉ A j
ˆ j)
H 2 ( L (∂ j )
ˆ j . Then (4.46) is a direct

with a constant Ĉ that only depends on the geometry of
consequence of (4.43) and the following bounds:
n−1 n
−1
ˆ 2 ˆ , |q| H 1 ( ) = A 2
h L 2 (i j ) = A j 2 |q̂| H 1 (ˆ j ) ,
L (∂ j ) d, j j
n−3
|q| 3 = A j 2 |q̂| 3
ˆ j)
.
H 2 (d, j ) H 2 (
The argument for (4.47) is similar, using (4.44).

1
Now define c = ∇ q in d, j . Then c belongs to H (div; d, j ) ∩ H 2 (d, j )n and
div c = 0. Therefore R h (c) is well defined [18] and satisfies the approximation prop-
erty for divergence-free functions [48]
1
c − R h (c) L 2 (d, j ) ≤ Ch r |c| H r (d, j ) , 0 < r ≤ , (4.48)
2
with a constant C independent of h, j, and the diameter of d, j . We are now ready to
define the correction chj,i j (v). In particular, take chj,i j (v) = R h (c) applied in d, j .
Note that chj,i j (v) belongs to X d, j,i j (v) = 0
h , and (4.7) and (4.6) imply that div ch
j
123
in d, j and chj,i j (v) · n j = h = w h · ni j on i j . Therefore (4.43) and (4.44) imply

that chj,i j (v) satisfies (4.8). Furthermore, (4.48) yields
chj,i j (v) L 2 (d, j ) ≤ R h (c) − c L 2 (d, j ) + c L 2 (d, j )

1
≤ Ch 2j |c| 1 +c L 2 (d, j ) ,
H 2 (d, j )
with a constant C independent of the geometry of d, j . Considering that h j ≤ A j ,

Lemma 4.5 gives

chj,i j (v) L 2 (d, j ) ≤ C A j [R h (v) · n] L 2 (i j ) , i j ∈ dd , (4.49)

chj,i j (v) L 2 (d, j ) ≤ C A j (sh (v)|s,i − R h (v)|d, j ) · ni j L 2 (i j ) , i j ∈ sd ,
(4.50)
with a constant C independent of h, H , v, i, j, and the diameters of i j and d, j .
Corollary 4.3 The approximation operator dh constructed in Sect. 4.1 with correc-
tions chj,i j (v) described above satisfies assumption (3.33).
Proof Since dh is a correction of the mixed interpolant R h , which is stable in the sense
of (4.41), it remains to bound chj,i j (v) H (div;d, j ) . By construction div chj,i j (v) = 0.
In the following we will make use of the trace inequality [38]
−1 1
∀ϕ ∈ H 1 (d, j ), ϕ L 2 (i j ) ≤ C A j 2 ϕ L 2 (d, j ) + A 2j |ϕ| H 1 (d, j ) . (4.51)
For i j ∈ dd , using (4.49), (4.6), and (4.51), we have

chj,i j (v) L 2 (d, j ) ≤ C A j (v · ni L 2 (i j ) + v · n j L 2 (i j ) ) ≤ Cv H 1 (d,i ∪d, j ) ,
using also that A j ≤ 1. For i j ∈ sd , we employ (4.50) and (4.1) to obtain

chj,i j (v) L 2 (d, j ) ≤ C A j (v − R h (v)|d, j ) · n j L 2 (i j )

+ (v − (S h (v) + ci,
h
ij
(v))|s,i ) · ni L 2 (i j ) ,
h
with ci,ij
(v) defined in (4.1) and constructed in Sect. 4.2, and if necessary S h replaced
by S̃ h . For the first term in the right-hand side, using (4.6), we have
(v − R h (v)|d, j ) · n j L 2 (i j ) ≤ v · n j L 2 (i j ) ≤ Cv H 1 (d, j ) .
For the second term on the right, using (4.51) and the approximation property (4.16)
of S h or S̃ h we have
123
−1 1
(v − S h (v)|s,i ) · ni L 2 (i j ) ≤ C(Ai 2 v− S h (v) L 2 (s,i ) + Ai2 |v − S h (v)| H 1 (s,i ) )
−1 1
≤ C(Ai 2 h i |v| H 1 (˜ s,i ) + Ai2 |v| H 1 (˜ s,i ) )
≤ C|v| H 1 (˜ s,i ) ,
using that h i < Ai ≤ 1, with a similar bound for ci,

h
ij
(v) L 2 (˜ s,i ) , in view of (4.19).
The proof is completed by combining all bounds and using the Poincaré inequality
(1.1).
5 Error analysis
In this section we establish a priori error estimates for our method. Let us assume
that the finite element spaces X sh and Wsh in s contain at least polynomials of degree
rs and rs − 1, respectively. Let X dh and Wdh in d contain at least polynomials of
degree rd and ld , respectively. In all cases under consideration, either ld = rd or
ld = rd − 1. Let sd H , H , and H contain at least polynomials of degree r , r ,
d s sd dd
and rss , respectively. In the analysis we will make use of the following well known
approximation properties of the mixed interpolant R h [18,34]: for all v ∈ H t ()n ,
t ≥ 1, there exists a constant C that depends only on rd , ld , t, and the shape regularity
of Ti h , such that for all T in Ti h
v − R h (v) L 2 (T ) ≤ Ch r |v| H r (T ) , 1 ≤ r ≤ min(rd + 1, t), (5.1)

div(v − R (v)) L 2 (T ) ≤ Ch |div v| H r (T ) , 0 ≤ r ≤ min(ld + 1, t − 1),
h r
(5.2)

1
(v − R h (v)|d,i ) · n L 2 (T ) ≤ Ch r |v · n| H r (T ) , 0 ≤ r ≤ min rd + 1, t − ,
2
(5.3)
where T denotes an arbitrary element of the trace Ti,

h
ij
on i j . We begin with the
following approximation result for the operator h defined in Lemma 3.6.
Lemma 5.1 Under the assumptions of Lemma 3.6, the operator h ∈ L(H01 ()n ; V h )
n
satisfies for all v ∈ H t () ∩ H01 () , t ≥ 1,
|||v − h (v)||| X s ≤ Ch r |v| H r +1 () , 0 ≤ r ≤ min(rs , t − 1), (5.4)

|||v − h (v)||| X d ≤ C h r v r + 1 + h q div v H q () + h s v H s+1 () ,
H 2 ()

1 1
≤r ≤ min rd + 1, t − , 0 ≤ q ≤ min(ld + 1, t −1), 0 ≤ s ≤ min(rs , t −1).
2 2
(5.5)
Proof To show (5.4), recall that h (v)|s = sh (v), sh (v) = sh (v) + csh (v), where
csh (v) is defined in (3.29), and sh is a correction of the Scott-Zhang operator con-
structed in Sect. 4. The triangle inequality and (3.30) imply that
123
v − sh (v) H 1 (s,i ) ≤ Cv − sh (v) H 1 (s,i ) .
Recall that sh (v) is constructed by correcting the Scott-Zhang interpolant S h or

h
S̃ h with ci,ij
(v) and chj,i j (v). Since v − S h (v) H 1 (s,i ) is bounded optimally in
(4.16), it remains to bound ci,
h
ij
(v) H 1 (s,i ) and chj,i j (v) H 1 (s, j ) . The bound on
ci,
h
ij
(v) H 1 (s,i ) follows from a simple modification of the argument in the proof of
Lemma 4.1. More precisely, by using (4.15) with 0 ≤ r ≤ min(rs , t − 1), (4.20)
can be modified as

−1
v − S (v)|s,i ≤ Ch rT |k | ρk 2 |v| H r +1 (Dk ) .
h

k
Propagating the above bound through the proof gives
ci,
h
ij
(v) L 2 (s,i ) +h|ci,
h
ij
(v)| H 1 (s,i ) ≤ Ch r +1 |v| H r +1 (˜ s,i ) , 0 ≤r ≤ min(rs , t − 1).
We proceed with the bound on chj,i j (v) H 1 (s, j ) . Considering first the 2-D case
presented in Sect. 4.3, inequality (4.34) in the proof of Lemma 4.3 can be modified,
using the approximation property (4.16) of S̃ h , as

| S̃ h (v) − v |s,k | 1 ≤ Ch r |v| H r +1 (˜ s,k ) , 0 ≤ r ≤ min(rs , t − 1),
H 2 (i j )
and (4.38) and (4.39) can be modified, using the approximation property (4.15) of S̃ h ,
as
xn+1
1 h 2

[ S̃ (v)] ≤ Ĉh 2r |v|2H r +1 (˜ ∪˜ ) , 0 ≤ r ≤ min(rs , t − 1),
x n,i n, j
xn
implying
chj,i j (v) H 1 (s, j ) ≤ Ch r |v| H r +1 (˜ s,i ∪˜ s, j ) , 0 ≤ r ≤ min(rs , t − 1).
We next consider the 3-D case presented in Sect. 7.2. Modifying the proof of
Lemma 7.4 in a similar way gives
chj,i j (v) L 2 (s, j ) + h|chj,i j (v)| H 1 (s, j ) ≤ Ch r +1 |v| H r +1 (s,i ∪s, j ) , 0 ≤ r ≤ min(rs , t − 1).
A combination of the above inequalities gives (5.4).

We continue with the proof of (5.5). Recall that h (v)|d = dh (v), where
dh (v) = R h (v) + chj,i j (v) is a corrected mixed interpolant with a correction sat-
isfying (4.8). Since v − R h (v) H (div;d,i ) is bounded optimally in (5.1)–(5.2), it
123
remains to bound chj,i j (v) constructed in Sect. 4.4. By construction div chj,i j (v) = 0,
so we only need to control chj,i j (v) L 2 (d, j ) . Consider first i j ∈ dd . Using (5.3),
bound (4.49) gives
1
chj,i j (v) L 2 (d, j ) ≤ C A 2j h r |v| H r (i j ) ≤ Ch r v r + 1 ,
H 2 (d,i ∪d, j )

1
0 < r ≤ min rd +1, t − ,
2
where we have used the trace inequality [38]
−1
|ϕ| H r (i j ) ≤ Ai 2 ϕ 1 , r > 0. (5.6)
H r + 2 (d,i )
For i j ∈ sd , we modify the argument in Corollary 4.3 to use the full approximation
properties (4.16) of S h or S̃ h to obtain

chj,i j (v) L 2 (d, j ) ≤ C h r v r + 1 ˜ ˜ d, j )
+ h s v H s+1 (˜ s,i ∪˜ d, j ) ,
H 2 (s,i ∪

1
0 < r ≤ min rd + 1, t − , 0 ≤ s ≤ min(rs , t − 1).
2
A combination of the above inequalities completes the proof of (5.5).
Next, we need to approximate the functions for the pressure. For any q ∈ L 2 (i ), let
P h q be its L 2 (i )-projection onto Wih := W h |i ,
(q − P h q, w h ) = 0, ∀w h ∈ Wih ,
satisfying the approximation property
q − P h q0,i ≤ C h r |q| H r (i ) , 0 ≤ r ≤ ri + 1, (5.7)
where ri is the polynomial degree in the space Wih : ri = rs − 1 in s and ri = ld

in d .
5.1 Error estimates
From the error equation:
∀v h ∈ Z h , ∀qh ∈ Wh , a(u − uh , v h ) = −(b(v h , p − q h ) + bsd (v h , λsd )

+bd (v h , λd ) + bs (v h , λs )), (5.8)
123
and Lemmas 3.3 and 3.4, we immediately derive

|||u − uh ||| X ≤ C inf |||u − v h ||| X + inf p − w h W + CRuh , (5.9)
v h ∈V h (qd ) w h ∈W h

where V h (qd ) = {v ∈ V h ; ∀w ∈ W h , b(v, w) = d w qd }, and
1

Ruh = sup bsd (v h
, λ sd ) + bd (v h
, λ d ) + bs (v h
, λ s ) (5.10)
v h ∈Z h |||v h ||| X
is the consistency error at the interfaces. Similarly a simple variant of (5.8) with
v h ∈ V h and Theorem 3.1 yield

p − p h W ≤ C |||u − uh ||| X + inf p − w h W + CRhp , (5.11)
w h ∈W h
where Rhp is given by (5.10) with Z h replaced by V h . As the bound on the approxi-
mation error of W h follows from (5.7) and Lemma 5.1 estimates the approximation
error of V h (qd ), it suffices to bound Rhp . Note that by virtue of (3.11), we have for all
μsH ∈ sH , μdH ∈ dH , μsd H ∈ H ,
sd
1 H

Rhp = sup bsd (v h
, λ sd −μ H
sd )+bd (v h
, λ d −μd
H
)+bs (v h
, λ s −μ s .
)
v h ∈V h |||v h ||| X
(5.12)
The bound for bs (v h , λs − μsH ) is straightforward because v h is measured in H 1 in

each s,i , and therefore its trace can be considered on each i, j . But deriving an
optimal bound for the other terms is more intricate because v h is measured in H (div)
1
in each d,i and this only controls the trace of the normal component in H − 2 (∂d,i ).
Consequently, λsd − μsd H and λ − μ H must belong to H 21 (∂ ). In 2-D, this is
d d d,i
easily achieved by interpolating λsd and λd with a variant of the Scott and Zhang
interpolant that is continuous at the vertices of the subdomains; a similar idea was
used in Sect. 4.3. But in 3-D, this construction requires an additional assumption on
the mortar grids in the Darcy region.
Hypothesis 5.1 For each d,i in 3-D, the mortar grids Ti H j on all i j ⊂ ∂d,i \∂
are chosen such that their traces on the boundaries of i j coincide.
This assumption implies conformity of mortar grids along interface boundaries in
each connected region in d and it allows us to consider the space dH,c , which is the
subset of continuous functions in dH ∪ sd H . On ∪ , let I H be the Scott–Zhang
dd sd
H,c
interpolant in d . We assume that its definition depends only on function values on
dd ∪ sd . On ss , let I H be the L 2 (i j )-orthogonal projection operator in sH . This
operator has the following approximation properties, assuming sufficient smoothness
of μ and μ (the index stands for dd or sd):
123
∀i j ∈ , ∀τ ∈ i j , μ− I H (μ) H t (τ ) ≤ C H r −t |μ| H r (τ ) , 0 ≤ t ≤ 1, t ≤r ≤r +1,

(5.13)
∀i j ∈ ss , ∀τ ∈ i j , μ − I H (μ) L 2 (τ ) ≤ C H r |μ| H r (τ ) , 0 ≤ r ≤ rss + 1,
(5.14)
where τ is the macroelement used in defining I H (μ) restricted to τ . The union of

all τ for τ in ∂i may slightly overlap interfaces that are not part of ∂i ; we denote
the overlap by O.
Lemma 5.2 There exists a constant C independent of h, H , and the diameters of the
subdomains such that, for all v h ∈ X h :

Ms
−1
|bs (v h , λs − I H (λs ))| ≤ C H s Ai 2 v h H 1 (s,i ) |λs | H s (∂s,i ∩ss ∪O) ,
i=1
0 ≤ s ≤ rss + 1, (5.15)
provided λs is sufficiently smooth.

Proof We have

bs (v , λs − I (λs )) =
h H
[v h ] · (λs − I H (λs ))
i j ∈ss
ij

Ms
≤ v h L 2 (∂s,i ∩ss ) λs − I H (λs ) L 2 (∂s,i ∩ss ) .
i=1
Then (5.15) follows readily from (5.14) and the trace inequality (4.51).
Lemma 5.3 Under Hypothesis 5.1, there exists a constant C independent of h, H ,
and the diameters of the subdomains such that, for all v h ∈ X h :
|bsd (v h , λsd − I H (λsd )) + bd (v h , λd − I H (λd ))|

⎛
Md
≤C ⎝
1 1
v h H (div;d,i ) H q− 2 |λ| H q (∂d,i ∩dd ∪O) + H r − 2 |λ| H r (∂d,i ∩sd ∪O)
i=1
⎞

Md
−1
+ A j 2 v h H 1 (s, j ) H r |λ| H r (∂d,i ∩sd ∪O) ⎠ ,
i=1 j
1 1
≤ q ≤ rdd +1, ≤r ≤rsd + 1, (5.16)
2 2
provided λsd and λd are sufficiently smooth, and where the sum over j runs over all
s, j adjacent to d,i .
Proof In the following λ denotes either λd or λsd depending on the type of interface.
We can write
123
bsd (v h , λ − I H (λ)) + bd (v h , λ − I H (λ))

= [v h · n](λ − I H (λ))
i j ∈sd ∪dd
ij

Md
= v h · ni (λ − I H (λ)) + v h · ns (λ − I H (λ)),
i=1∂ i j ∈sd
d,i ij
where we have used that v h · n = 0 on ∂d,i ∩ ∂ and λ − I H (λ) has been extended
1 1
continuously from H 2 (∂d,i \∂) to H 2 (∂d,i ). The argument of Lemma 5.2 can
be used to bound the second sum above by the last sum in (5.16). Thus it is left to
bound the first sum.
Consider first one subdomain d,i that is not adjacent to sd . Let us switch to the
reference domain ˆ i and let Ê denote the extension operator defined in (4.29) relative
ˆ 1
to i . For any f in H 2 (∂d,i ), define E( f ) by: E( f ) = Ê( f ◦ Fi ); therefore E
1
maps H 2 (∂d,i ) into H 1 (d,i ). Thus, by Green’s formula

v h · ni (λ − I H (λ)) = v h · ni E(λ − I H (λ))
∂d,i ∂d,i

= div v h E(λ − I H (λ)) + v h · ∇ E(λ − I H (λ)).
d,i d,i
By reverting to the reference domain and observing that I H is invariant under the rigid
body motion Fi , this reads

v h · ni (λ − I H (λ)) ≤ Ain−1 v̂ H (div;
ˆ ˆ i ) Ê(λ̂ − Î(λ̂)) H 1 (
ˆ i ).

∂d,i
On one hand, the bound in (4.29) gives
Ê(λ̂ − Î(λ̂)) H 1 (ˆ i ) ≤ Ĉλ̂ − Î(λ̂)) 1

ˆ i)
≤ Ĉλ̂ − Î(λ̂)) 1
ˆ i \∂ )
ˆ
H 2 (∂ H 2 (∂
1
2
≤ Ĉ λ̂ − Î(λ̂))2 1 ,
H ˆ
2 ()
ˆ
⊂∂ ˆ
ˆ i \∂
with constants here and below independent of h, H , and the diameters of d,i and
ˆ and H 1 (),
i j . By space interpolation between L 2 () ˆ the estimates (5.13) summed
ˆ
on with t = 0 and t = 1 readily imply that
r − 12 1
λ̂ − Î(λ̂)) 1
ˆ
≤ Ĉ Ĥi |λ̂| H r (ˆ ) , ≤ r ≤ rdd + 1, (5.17)
H 2 () 2
123
ˆ i is related to the mesh size Hi on ∂i by

where the mesh size Ĥi on ∂
Hi = Ai Ĥi .
ˆ
In the case when r is not an integer, since (5.17) is stated in the reference region ,
we choose the fractional seminorm in the right-hand side to be defined by (1.8) and
incorporate the equivalence constant into Ĉ. The motivation for this choice is that the
intrinsic norm is easily transformed by Fi . Therefore
r − 12
λ̂ − Î(λ̂)) 1
ˆ i)
≤ Ĉ Ĥi |λ̂| H r (∂ ˆ \∂ ˆ ∪Ô)
H 2 (∂ i
r− 1 r − n−1
≤ Ĉ Ĥi 2 Ai 2 |λ| H r (∂d,i \∂ ∪O) .
On the other hand,
− n2
1
2
v̂ H (div;
ˆ ˆ i ) = Ai Ai2 div v h 2L 2 ( )
+ v h 2L 2 ( .
d,i d,i )
By collecting these inequalities, we derive

r − 1 h

v · ni (λ − I (λ)) ≤ Ĉ Ai Ĥi
h H 2 v
H (div;d,i ) |λ| H r (∂d,i \∂ ∪O ) ,

∂d,i
1
≤ r ≤ rdd + 1.
2
By applying (5.17) to a portion of sd , using (5.13) the same argument handles the
case when d,i is adjacent to sd . Then (5.16) follows easily from these estimates.
Remark 5.1 We can skip Hypothesis 5.1, work locally on each i j , and use the L 2
norm for both factors v h · ni and λ − I H (λ). But this gives rise to a factor of the form
1
Hi / h 2 in (5.16), that is clearly not optimal when h is very small compared to Hi .
Let Rh denote either Ruh or Rhp . The next lemma estimates Rh .
Lemma 5.4 Under Hypothesis 5.1, the consistency error Rh satisfies

1 1 1 1
R h ≤ C A − 2 H r − 2 pd 1 + A− 2 H q− 2 pd q+ 1
H r + 2 (d ) H 2 (d )

+A−1 H s (us 3 + ps s+ 1 ) ,
H s+ 2 (s ) H 2 (s )
1 1
≤ q ≤ rdd + 1, ≤ r ≤ rsd + 1, 0 < s ≤ rss + 1, (5.18)
2 2
where A = min1≤i≤M Ai .
123
Proof By combining (5.15) and (5.16), we easily derive

⎛ ⎛ ⎞1 ⎛ ⎞1
2 2
⎜ Md Md
Rh ≤ C ⎝ H q− 21 ⎝ |λ|2 ⎠ +H r − 12 ⎝ |λ|2 r ⎠
H q (∂ d,i ∩dd ∪O ) H (∂d,i ∩sd ∪O )
i=1 i=1
M 21
s
+H s Ai−1 |λs |2H s (∂s,i ∩ss ∪O)
i=1
⎞
M 21
s
⎟
+H r Ai−1 |λ|2H r (∂s,i ∩sd ∪O) ⎠,
i=1
1 1
≤ q ≤ rdd + 1, ≤ r ≤ rsd + 1, 0 ≤ s ≤ rss + 1.
2 2
In view of the representation (2.47), local trace theorems yield:
−1
|λsd | H r (∂d,i ∩sd ) ≤ C Ai 2 pd 1 ,
H r + 2 (d,i )
−1
|λd | H q (∂d,i ∩dd ) ≤ C Ai 2 pd 1 ,
H q+ 2 (d,i )
− 12
|λs | H s (∂s,i ∩ss ) ≤ C Ai us s+ 23 + ps s+ 21 .
H (s,i ) H (s,i )
Then (5.18) follows from these bounds, that are easily extended to include O, and the
fact that Ai−1 < H −1 .
Now we use the abstract error bounds (5.9) and (5.11), the approximation results
(5.4), (5.5), and (5.7), and the consistency error bound (5.18) to obtain the following
convergence result.
Theorem 5.1 We have
|||u − uh ||| X + p − p h W

≤ C h r1 (u H r1 +1 () + p H r1 () ) + h r2 u 1
H r2 + 2 ()
+ h r3 (div u H r3 () + p H r3 () )
+ A−1 H r4 (us 3 + ps 1 )
H r4 + 2 (s ) H r4 + 2 (s )
− 21 r5 − 21 − 21 r6 − 21

+A H pd r5 + 21 +A H pd r6 + 21 ,
H (d ) H (d )
1
0 ≤ r1 ≤ rs , ≤ r2 ≤ rd + 1, 0 ≤ r3 ≤ ld + 1,
2
1 1
0 < r4 ≤ rss + 1, ≤ r5 ≤ rdd + 1, ≤ r6 ≤ rsd + 1.
2 2
123
Remark 5.2 In the above estimate, the fine scale subdomain approximation error terms
are of optimal order with constants independent of the size of the subdomains A. The
constants of the coarse scale mortar consistency error terms deteriorate with decrease
in A, since in that case the number of interfaces grows. Nevertheless, higher order
mortar polynomials can be employed to balance the error terms, giving optimal fine
scale convergence.
6 Numerical tests
In this section we validate our analysis by carrying out some numerical experiments.
In all tests the computational domain is taken to be = s ∪ d , where s =
(0, 1) × ( 21 , 1) and d = (0, 1) × (0, 21 ). For simplicity we set
σ (us , ps ) = − ps I + νs ∇us
in the Stokes equation in s , and
K = KI
in the Darcy equation in d , where K is a positive constant.

To test for convergence we construct the following analytical solution satisfying
the flow equations in s and d along with the conditions on the interface sd :

(2 − x)(1.5 − y)(y − ξ )
us = 3 y2 ,
− y3 + 2 (ξ
+ 1.5) − 1.5ξ y − 0.5 + sin(ωx)

ω cos(ωx)y
ud = ,
χ (y + 0.5) + sin(ωx)
sin(ωx) + χ
ps = − + νs (0.5 − ξ ) + cos(π y),
2K
χ (y + 0.5)2 sin(ωx)y
pd = − − ,
K 2 K
where
√
νs K 1−G −30ξ −17
νs = 0.1, K = 1, α = 0.5, G = ,ξ = , χ= , and ω = 6.0.
α 2(1 + G) 48
The right-hand sides f s , f d , and qd for the Stokes–Darcy flow system are obtained
by substituting the analytical solution into (2.6), (2.3), and (2.4), respectively. The
boundary conditions are as follows: for the Stokes region, the velocity us is specified
on the left and top boundaries, and the normal and tangential stresses (σ ns ) · ns
and (σ ns ) · τ s are specified on the right boundary; for the Darcy region, the normal
velocity ud · nd is specified on the left boundary and the pressure pd is specified
on the bottom and right boundaries. Each region s and d is divided into two
123
Fig. 2 Non-matching 1
subdomain meshes
8 × 12–12 × 16
0.8
0.6
Y
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
X
subdomains, giving a total of four subdomains. The subdomain grids do not match
across the interfaces. The Stokes subdomains are discretized by the Taylor–Hood
triangular finite elements with quadratic velocities and linear pressures (rs = 2). The
Darcy subdomains are discretized by the lowest order Raviart–Thomas rectangular
finite elements (rd = ld = 0). We use discontinuous piecewise linear mortars on all
interfaces (rss = rdd = rsd = 1). To test convergence, we solve the problem on a
sequence of grid refinements. On the coarsest level, the subdomain grids are 3 × 4 in
the lower left and upper right subdomains √ and 2 × 3 in the other two subdomains. We
test two cases, H = 2h and H = h. In both cases the coarsest mortar grids have a
single element per interface. In the first case the mortar grids are refined by two each
time the subdomain grids are refined by two. In the second case the mortar grids are
refined by two each time the subdomain grids are refined by four. The non-matching
grids on the middle level of refinement are shown in Fig. 2. The computed solution
and the numerical error on this grid level with mortar meshes H = 2h are displayed in
Fig. 3. The numerical errors and convergence rates on all refinement levels are reported
in Tables 1, 2, 3, 4, 5, and 6. We report separately the errors in s and d in their
respective norms. In the Darcy region we also report the discrete L 2 errors · M,d
for the pressure and the velocity, computed via the use of the midpoint quadrature rule
on each element.
In the case H = 2h we observe second order of convergence for us − ush H 1 (s )
and ps − psh L 2 (s ) , as well as first order convergence for ud − udh H (div;d ) and
pd − pdh L 2 (d ) . These rates are consistent with the error terms O(h rs ) = O(h 2 ) and
O(h rd +1 ) = O(h ld +1 ) = O(h) that appear in Theorem 5.1. Note that in the case H =
1
2h the interface consistency error terms are O(h rss +1 ) = O(h 2 ) and O(h rdd + 2 ) =
1 3
O(h rsd + 2 ) = O(h 2 ). While the error terms in Theorem 5.1 depend on the global
solution norms, the results indicate that the lower convergence rates in d and on sd
do not affect the second order convergence in s . This suggests that it may be possible
123
1 1
3 0.1
P-error
P
0.8 0.6 0.8 0.014
0.4 0.012
0.2 0.01
0 0.008
-0.2 0.006
0.6 -0.4 0.6 0.004
-0.6 0.002
0
Y
Y
-0.8
-1 -0.002
-1.2 -0.004
0.4 0.4 -0.006
-0.008
-0.01
-0.012
-0.014
0.2 0.2 -0.016
-0.018
-0.02
-0.022
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
X X
Fig. 3 Test 1: computed solution (left) and error (right) on subdomain meshes 8 × 12–12 × 16 and mortar
meshes H = 2h
Table 1 Test 1: H = 2h
Mesh us − ush H 1 ( ) Rate ps − psh L 2 ( ) Rate

s s
2×3 3×4 3.93e−01 2.64e−02

4×6 6×8 8.95e−02 2.13 6.37e−03 2.05
8×12 12×16 2.10e−02 2.09 1.53e−03 2.06
16×24 24×32 5.08e−03 2.05 3.75e−04 2.03
32×48 48×64 1.25e−03 2.02 9.29e−05 2.01
Numerical errors and convergence rates in s
Mesh ud − udh H (div;d ) Rate pd − pdh L 2 ( ) Rate

d
2×3 3×4 3.51e+00 1.01e−01

4×6 6×8 1.79e+00 0.97 5.05e−02 1.00
8×12 12×16 9.00e−01 0.99 2.52e−02 1.00
16×24 24×32 4.50e−01 1.00 1.26e−02 1.00
32×48 48×64 2.20e−01 1.00 6.28e−03 1.00
Numerical errors and convergence rates in d
to establish interior convergence error estimates. Such estimates were derived in [53]
for mortar discretizations for Darcy flow. We also observe superconvergence at the cell
centers in the Darcy region, as indicated by the O(h 2 ) convergence of pd − pdh M,d
3
and the O(h 2 ) convergence of ud − udh M,d . Superconvergence for the pressure on
unstructured grids and for the velocity on logically rectangular grids in mixed finite
element discretizations for Darcy flow has been extensively studied in the literature,
see, e.g. [5] and the references therein. The reduced superconvergence order for the
3
Darcy velocity is due to the O(h 2 ) mortar error term, as shown in [5].
123
Mesh ud − udh M,d Rate pd − pdh M,d Rate
2×3 3×4 2.60e−01 1.65e−02

4×6 6×8 6.71e−02 1.95 4.13e−03 2.00
8×12 12×16 2.09e−02 1.68 1.01e−03 2.03
16×24 24×32 6.85e−03 1.61 2.50e−04 2.01
32×48 48×64 2.32e−03 1.56 6.22e−05 2.01
Numerical errors and convergence rates in d using the discrete · M,d norm
√
Table 4 Test 2: H = h
Mesh us − ush H 1 ( ) Rate ps − psh L 2 ( ) Rate

s s
2×3 3×4 3.93e−01 2.64e−02

8×12 12×16 5.59e−02 1.41 3.51e−03 1.46
32×48 48×64 1.03e−02 1.22 6.02e−04 1.27
Numerical errors and convergence rates in s
√
Mesh ud − udh H (div;d ) Rate pd − pdh L 2 ( ) Rate

d
2×3 3×4 3.51e+00 1.01e−01

8×12 12×16 9.10e−01 0.97 2.53e−02 1.00
32×48 48×64 2.30e−01 0.99 6.32e−03 1.00
Numerical errors and convergence rates in d
√
Mesh ud − udh M,d Rate pd − pdh M,d Rate
2×3 3×4 2.60e−01 1.65e−02

8×12 12×16 1.10e−01 0.62 3.65e−03 1.09
32×48 48×64 5.01e−02 0.57 7.57e−04 1.13
Numerical errors and convergence rates in d using the discrete · M,d norm
√
In the case H = h, we observe approximately O(h) convergence for all error
norms. Note that in this case the interface consistency error terms are O(h (rss +1)/2 ) =
1 1 3
O(h) and O(h (rdd + 2 )/2 ) = O(h (rsd + 2 )/2 ) = O(h 4 ), so their effect on the conver-
gence in the Stokes and Darcy regions is more significant. In this multiscale case, one
may utilize higher order mortars to recover optimal fine scale subdomain convergence,
see [6] for the Darcy case.
123
7 Appendix
This Appendix is devoted to specific examples of construction of chj,i j (v) in s .
7.1 The 2-D case
Recall that the construction of the correction chj,i j (v) in s in 2-D required an operator
1
πsh satisfying (4.23). In particular, given ˆ in H00
2 ˆ 2 ˆ
() , one needs to construct π̂ j ()
in X̂ j , unique solution of (4.24):

ˆ
∀μ̂ ∈ , ˆ · μ̂ =
π̂ j () ˆ · μ̂,
ˆ ˆ
and satisfying (4.25).

Here we present two simple examples of how to construct such an operator when
i j is a straight line segment and the traces of the discrete spaces on i j are piecewise
P1 finite elements, Without loss of generality, we can suppose that ˆ is the segment
[0, 1], that the nodes of T̂ j are 0 = x̂0 < x̂1 < · · · < x̂ N < x̂ N +1 = 1, and we set
ĥ n = x̂n+1 − x̂n , 0 ≤ n ≤ N . We choose each component of the functions of X̂ j
piecewise P1 in the subintervals [x̂n , x̂n+1 ], with degrees of freedom at the nodes x̂n ,
1 ≤ n ≤ N.
7.1.1 First example
As a first example, we choose each component of the mortar functions of ˆ also

piecewise P1 in the subintervals [x̂n , x̂n+1 ], with degrees of freedom at the nodes x̂n ,
n = 0, 2 ≤ n ≤ N − 1, and n = N + 1. The nodes x̂1 and x̂ N are deleted so that
ˆ have the same dimension 2N ; thus the matrix of the linear system (4.24)
X̂ j and
is square. In order to express these functions in terms of a basis, it is convenient to
modify slightly the standard Lagrange basis functions for the velocity. More precisely,
we define
1 1
ϕ̂0 (x̂)|[x̂0 ,x̂2 ] = (x̂2 − x̂), ϕ̂ N +1 (x̂)|[x̂ N −1 ,x̂ N +1 ] = (x̂ − x̂ N −1 ),
ĥ 0 + ĥ 1 ĥ N + ĥ N −1
1 1
ϕ̂2 (x̂)|[x̂0 ,x̂2 ] = (x̂ − x̂0 ), ϕ̂2 (x̂)|[x̂2 ,x̂3 ] = (x̂3 − x̂),
ĥ 0 + ĥ 1 ĥ 2
1
ϕ̂ N −1 (x̂)|[x̂ N −2 ,x̂ N −1 ] = (x̂ − x̂ N −2 ), (7.1)
ĥ N −1
1
ϕ̂ N −1 (x̂)|[x̂ N −1 ,x̂ N +1 ] = (x̂ N +1 − x̂),
ĥ N −1 + ĥ N
1 1
ϕ̂n (x̂)|[x̂n−1 ,x̂n ] = (x̂ − x̂n−1 ), ϕ̂n (x̂)|[x̂n ,x̂n+1 ] = (x̂n+1 − x̂),
ĥ n−1 ĥ n
n = 1, 3 ≤ n ≤ N − 2, n = N ,
123
extended by zero elsewhere. We take

N
X̂ j = v̂ ∈ H01 (0, 1)2 ; v̂(x̂) = v̂ n ϕ̂n (x̂) , (7.2)
n=1
−1

N
ˆ = μ̂ ∈ H (0, 1) ; μ̂(x̂) = μ̂(x̂0 )ϕ̂0 (x̂)+
1 2
μ̂(x̂n )ϕ̂n (x̂)+ μ̂(x̂ N +1 )ϕ̂ N +1 (x̂) .
n=2
(7.3)
On one hand, the set {ϕ̂0 , ϕ̂n , 2 ≤ n ≤ N − 1, ϕ̂ N +1 } is indeed a Lagrange basis for
ˆ On the other hand, the set {ϕ̂n , 1 ≤ n ≤ N } is a basis but not a Lagrange basis for
.
X̂ j ; however, it is easy to check that
v̂ 1 = v̂(x̂1 ) − v̂(x̂2 )ϕ̂2 (x̂1 ), v̂ n = v̂(x̂n ), 2 ≤ n ≤ N − 1,

v̂ N = v̂(x̂ N ) − v̂(x̂ N −1 )ϕ̂ N −1 (x̂ N ). (7.4)
The system (4.24) can be decoupled into two independent systems in R N , one for each
component, and each system has the form Mα = b, where α ∈ R N is the unknown,
the non zero coefficients of M per row are
1 1
M1,1 = (ĥ 0 + 2ĥ 1 ), M1,2 = (ĥ 0 + ĥ 1 ),
6 6
1 1 ĥ 2
M2,1 = (2ĥ 0 + ĥ 1 ), (ĥ 0 + ĥ 1 + ĥ 2 ), M2,3 =
M2,2 = ,
6 3 6
ĥ n−1 1 ĥ n
Mn,n−1 = , Mn,n = (ĥ n−1 + ĥ n ), Mn,n+1 = , 3 ≤ n ≤ N − 2,
6 3 6
ĥ N −2 1
M N −1,N −2 = , M N −1,N −1 = (ĥ N −2 + ĥ N −1 + ĥ N ),
6 3
1 1
M N −1,N = (ĥ N −1 + 2ĥ N ), M N −1,N = (2ĥ N −1 + ĥ N ),
6 6
1
M N ,N = (ĥ N −1 + ĥ N ).
6
ˆ the components of b ∈ R N are
Denoting by ˆ a generic component of ,
x̂2 x̂3 x̂n+1
b1 = ˆϕ̂0 , b2 = ˆϕ̂2 , bn = ˆϕ̂n , 3 ≤ n ≤ N − 2,

x̂0 x̂0 x̂n−1
(7.5)
x̂N +1 x̂N +1
b N −1 = ˆϕ̂ N −1 , b N = ˆϕ̂ N +1 .
x̂ N −2 x̂ N −1
The matrix M is tridiagonal but not symmetric, the coefficients of its three diagonals
are all strictly positive, and it is strictly diagonally dominant, hence invertible, but
123
this is not sufficient to obtain a sharp estimate for its inverse. To this end, we use
the approach of Crouzeix and Thomée [23]. More precisely, let D be the principal
diagonal of M and factor M into M = D(I + K ) where K is a tridiagonal matrix
with principal diagonal zero. Denote the diagonal terms of D by dn > 0 and note that
1
ˆ and α. Recall that | · | denotes the
D 2 is well-defined. The next lemma relates π̂ j ()
Euclidean vector norm.
Lemma 7.1 For each component ˆ of ˆ we have
ˆ 2 ˆ ≤ 2| D 2 α|.
1
π̂ j () L () (7.6)
Proof Considering the support of the basis functions ϕ̂n , we have
x̂2
N −2 n+1
x̂
ˆ 22
π̂ j () = (α1 ϕ̂1 + α2 ϕ̂2 ) +
2
(αn ϕ̂n + αn+1 ϕ̂n+1 )2
ˆ
L ()
x̂0 n=2 x̂
n
x̂N +1
+ (α N −1 ϕ̂ N −1 + α N +1 ϕ̂ N +1 )2 .
x̂ N −1
By substituting the expression of these integrals and rearranging terms, we derive

N −2
ˆ 2 2 ≤ 2 α12 (ĥ 0 + ĥ 1 ) + α22 (ĥ 0 + ĥ 1 + ĥ 2 ) +
π̂ j () αn2 (ĥ n−1 + ĥ n )
ˆ
L () 3
n=3

+α N −1 (ĥ N −2 + ĥ N −1 + ĥ N ) + α N (ĥ N −1 + ĥ N ) .
2 2
In view of the diagonal terms dn of D, this can be written

N −1

ˆ 22 ĥ 0 + ĥ 1 2 ĥ N −1 + ĥ N
π̂ j () ˆ
L ()
≤2 α12 + αn dn + α 2N .
3 3
n=2
But
ĥ 0 + ĥ 1 ĥ 0 + ĥ 1 ĥ N −1 + ĥ N
= 2d1 < 2d1 , and similarly < 2d N .
3 ĥ 0 + 2ĥ 1 3
Hence
N 21

ˆ 2 ˆ ≤2
π̂ j () αn2 dn , (7.7)
L ()
n=1
whence (7.6).
123
ˆ relies on a bound for | D 2 α|. But D 2 α

This lemma shows that an L 2 bound for π̂ j ()
1 1
has also the following expression
1 1 1 −1 1
D 2 α = I + D 2 K D− 2 D− 2 b .
Therefore Lemma 7.1 implies that
1
ˆ 2 ˆ ≤ 2 I + D 2 K D− 2 −1 2 | D− 2 b|,
1 1
π̂ j () L () (7.8)
where · 2 denotes the matrix norm subordinated by the Euclidean norm, and more
generally · p is the matrix norm subordinated by the l p vector norm. The next lemma
1
gives a bound for | D− 2 b|.
Lemma 7.2 We have

√
ˆ
1
| D− 2 b| ≤ 3 ˆ .
L 2 () (7.9)
Proof By integrating the basis functions ϕ̂n , we derive
1
N
| D− 2 b|2 = dn−1 bn2
n=1

1 ĥ 0 + ĥ 1 ˆ 2 ĥ 0 + ĥ 1 + ĥ 2 ˆ 2
≤ L 2 (x̂ ,x̂ ) + L 2 (x̂ ,x̂ )
3 d1 0 2 d2 0 3

N −2
ĥ n−1 + ĥ n ˆ 2 ĥ N −2 + ĥ N −1 + ĥ N ˆ 2
+ L 2 (x̂ ,x̂ ) + L 2 (x̂
dn n−1 n+1 d N −1 N −2 , x̂ N +1 )
n=3

ĥ N −1 + ĥ N ˆ 2
+ L 2 (x̂ . (7.10)
dN N −1 , x̂ N +1 )
Then proceeding as in Lemma 7.1, we obtain

N
ˆ 22 ˆ 22 ˆ 22 ˆ 22
1
| D− 2 b|2 ≤ L (x̂
+ + +
0 , x̂ 1 ) L (x̂ 0 , x̂ 2 ) L (x̂ n−1 , x̂ n+1 ) L (x̂ N , x̂ N +1 )
n=1
ˆ 22
+ ,
L (x̂ N −1 , x̂ N +1 )
whence (7.9).
1 1 −1
It remains to evaluate I + D 2 K D− 2 2 . Following Crouzeix and Thomée, we
introduce the following constraint on the mesh length ĥ n :
123
Hypothesis 7.1 There exist two constants c0 > 0 and γ > 0 independent of N such
that
ĥ n
∀n, m, 0 ≤ n, m ≤ N , ≤ c0 γ |n−m| . (7.11)
ĥ m
This condition holds for quasi-uniform meshes, but it is also satisfied by much more
general meshes.
Proposition 7.1 If Hypothesis 7.1 holds with suitable constants c0 and γ , then there
exists a constant C independent of N and j such that
1 1 −1
I + D 2 K D− 2 2 ≤ C. (7.12)
Hence with the same constant C,

√
∀ˆ ∈ L 2 () ˆ 2 ˆ ≤ 2 3C
ˆ 2 , π̂ j ()
L ()
ˆ 2 ˆ .
L () (7.13)
Proof The proof follows the ideas of [23] but is more complex because the functions of
X̂ j vanish at the end points of ˆ whereas those of
ˆ do not. Let us prove convergence
of the series
∞
1 1
D 2 K r D− 2 2 ;
r =1
this will yield
∞

1 1 −1 1 1
I + D 2 K D− 2 2 ≤ 1 + D 2 K r D− 2 2 . (7.14)
r =1
As K has three diagonals, then K r has 2r + 1 diagonals. In addition, since the coef-
ficients of these diagonals are non negative, this implies that
1 1 1 1
D 2 K r D− 2 1 ≤ (2r + 1) D 2 K r D− 2 ∞ ,
and by interpolation,
1 1 1 1 1
D 2 K r D− 2 2 ≤ (2r + 1) 2 D 2 K r D− 2 ∞ .
Therefore
1 1
d 1 1
D 2 K r D− 2 2 ≤
n 2
sup (2r + 1) 2 K r∞ . (7.15)
|n−m|≤2r dm
123
Thus a sufficient condition for the series convergence is:

d 1 δ r
n 2
sup ≤c , (7.16)
|n−m|≤2r dm K ∞
where c > 0 and 0 < δ < 1 are two constants independent of r . First, assuming
Hypothesis 7.1, a simple but tedious computation gives for 0 < γ < 2
dn
sup ≤ 2c0 (1 + γ + γ 2 )γ 2r . (7.17)
|n−m|≤2r dm
Hence the series converges if

1
γ < Min 2, .
K ∞
It is easy to check that

ĥ 0 + ĥ 1 ĥ N + ĥ N −1
1 ĥ 0
K ∞ = Max , , 1+ ,
ĥ 0 + 2ĥ 1 ĥ N + 2ĥ N −1 2 ĥ 0 + ĥ 1 + ĥ 2

1 ĥ N
1+ .
2 ĥ N −2 + ĥ N −1 + ĥ N
To simplify the notation, set θ = c0 γ . Owing to Hypothesis 7.1, for θ ≥ 1, we have

1 θ θ2 1 θ2
K ∞ ≤ 1 + Max , = 1 + . (7.18)
2 2 + θ 1 + θ + θ2 2 1 + θ + θ2
Take for instance θ = 23 ; then (7.18) yields K ∞ < 43 . Therefore the series converges
for γ = 43 and in turn c0 = 98 . Then (7.13) is an immediate consequence of Lemmas
7.1 and 7.2.
ˆ in H 1 ()
Now we estimate π̂ j () ˆ 2 for ˆ ∈ H 1 ()
ˆ 2 . First, we have the analogue of
0
Lemma 7.1.
Lemma 7.3 We keep the notation of Lemma 7.1. Under Hypothesis 7.1 with γ > 1,
we have
1
ˆ 1 ˆ ≤ 2 c0 γ 2 2 − 1
|π̂ j ()| H () c0 γ (1 + γ ) + 2 + | D 2 α|. (7.19)
3 1+γ
Proof Let us denote the derivation on ˆ with a prime. Then

N
ˆ=
π̂ j () αn ϕ̂n ,
n=1
123
and a straightforward computation gives

−2

1 1 1 1
N
1 1
ˆ 2 2
π̂ j () ≤2 α12 + +α22 + + αn2 +
ˆ
L () ĥ 0 ĥ 1 ĥ 0 + ĥ 1 ĥ 2 n=3 ĥ n−1 ĥ n

1 1 1 1
+α 2N −1 + + α 2N + .
ĥ N −2 ĥ N −1 + ĥ N ĥ N −1 ĥ N
Hypothesis 7.1 yields

1 1 1 1 1 1 c0 γ 2
+ ≤ (1 + c0 γ ), + ≤ c0 γ (1 + γ ) + 2 + ,
ĥ 0 ĥ 1 2d1 ĥ 0 + ĥ 1 ĥ 2 3d2 1+γ
1 1 2
+ ≤ (1 + c0 γ ),
ĥ n−1 ĥ n 3dn

1 1 1 c0 γ 2
+ ≤ c0 γ (1 + γ ) + 2 + ,
ĥ N −2 ĥ N −1 + ĥ N 3d N −1 1+γ
1 1 1
+ ≤ (1 + c0 γ ).
ĥ N −1 ĥ N 2d N
Then (7.19) follows readily from the fact that for γ > 1,
c0 γ 2
c0 γ (1 + γ ) + 2 + > 2(1 + c0 γ ).
1+γ
Comparing with (7.8) and the proof of Lemma 7.2, we see that a direct estimate of
ˆ relies on a bound for | D− 23 b|, which we can hardly expect to be sharp. This
π̂ j ()
difficulty can be bypassed by using the fact that ˆ belongs to H01 () ˆ 2 and arguing
ˆ = Iˆ()
as in [23]. Uniqueness of the solution of (4.24) in X̂ j implies that π̂ j ( Iˆ()) ˆ
ˆ
where I denotes the standard Lagrange interpolant in X̂ j , that is well-defined because
ˆ is a line segment. This permits to write

ˆ = π̂ j (ˆ − Iˆ())
π̂ j () ˆ + ( Iˆ()
ˆ − )
ˆ + ˆ . (7.20)
First, we easily derive that
ˆ − |
| Iˆ() ˆ 1 ˆ ≤ 2||
ˆ 1 ˆ .
H () H ()
Therefore
ˆ 1 ˆ ≤ 3||
|π̂ j ()| ˆ 1 ˆ + |π̂ j (ˆ − Iˆ())|
ˆ ˆ , (7.21)
H () H () H 1 ()
and it suffices to derive a bound for this last term. Let Mα = b be the system (4.24)
for a generic component of ˆ − Iˆ().
ˆ Applying (7.12) and the analogue of (7.8), we
obtain
123
1 −1
|π̂ j (ˆ − Iˆ())|
ˆ −1 3 3
ˆ ≤ I + D 2 K D2
H 1 () 2 | D− 2 b| ≤ C| D− 2 b|, (7.22)
with the constant C of (7.12), provided c0 and γ are chosen as in the proof of Propo-
1 1
sition 7.1. Here we use the fact that (7.15) is also valid for D− 2 K r D 2 2 . It remains
3
to estimate | D− 2 b|.
Proposition 7.2 Let ˆ belong to H01 ()

ˆ 2 . If Hypothesis 7.1 holds with the constants
c0 and γ chosen as in the proof of Proposition 7.1, then each component ˆ of ˆ satisfies
√
ˆ 1 ˆ ,
3
| D− 2 b| < 30c|| H () (7.23)
where c depends only on the interpolant Iˆ. Therefore,

√
∀ˆ ∈ H01 () ˆ 1 ˆ < 3 + 30cC ||
ˆ 2 , |π̂ j ()|
H ()
ˆ 1 ˆ ,
H () (7.24)
with the constants c and C of (7.23) and (7.12).
Proof We sketch the proof. To simplify, set w = ˆ − Iˆ(). ˆ Arguing as in Lemma 7.2,
we recover a bound for | D b| by replacing dn by dn and ˆ by w in (7.10). The key
− 23 3
point here is that the properties of the interpolant Iˆ imply

w2L 2 (x̂ ˆ21
≤ c ĥ 20 || ˆ21
+ ĥ 21 || ,
0 , x̂ 1 ) H (x̂ 0 , x̂ 1 ) H (x̂ 1 , x̂ 2 )
for a constant c independent of N , with analogous expressions for the norms in the
3
other subintervals. The factors involving ĥ n in | D− 2 b| can be bounded by applying
(7.11) with c0 γ = 23 , as in the proof of Proposition 7.1. This gives

N
|D − 23 ˆ21
b|2 < 10c || ˆ21
+ || + ˆ21
|| ˆ21
+ ||
H (x̂ 0 , x̂ 1 ) H (x̂ 0 , x̂ 2 ) H (x̂ n−1 , x̂ n+1 ) H (x̂ N , x̂ N +1 )
n=1

ˆ21
+|| ,
H (x̂ N −1 , x̂ N +1 )
and (7.23) follows.
Finally, space interpolation between (7.13) and (7.24) gives
1
∀ˆ ∈ H00
2 ˆ 2 ˆ
() , |π̂ j ()| 1 ˆ
< Ĉ|| 1 , (7.25)
H00 ˆ
2 () ˆ
2 ()
H00
with a constant Ĉ independent of N .
123
7.1.2 Second example
In the first example, ˆ and X̂ j have the same dimension, but of course this is not
necessary. For the unique solvability of (3.8), it is sufficient that the set of degrees
of freedom of ˆ be an adequate subset of those of X̂ j , sufficiently rich to guarantee
a good approximation property of the Lagrange interpolant Iˆ. Let us briefly give
another simple choice of ˆ with roughly half as many degrees of freedom as in the
first example. Let N be an odd integer, keep the same ϕ̂0 and ϕ̂ N +1 as in (7.1), and
choose
1 1
ϕ̂1 (x̂)|[x̂0 ,x̂1 ] = (x̂ − x̂0 ), ϕ̂1 (x̂)|[x̂1 ,x̂2 ] = (x̂2 − x̂),
ĥ 0 ĥ 1
1 1
ϕ̂ N (x̂)|[x̂ N −1 ,x̂ N ] = , (x̂ − x̂ N −1 ) ϕ̂ N (x̂)|[x̂ N ,x̂ N +1 ] = (x̂ N +1 − x̂),
ĥ N −1 ĥ N
(7.26)
1
ϕ̂2n (x̂)|[x̂2n−2 ,x̂2n ] = (x̂ − x̂2n−2 ),
ĥ 2n−2 + ĥ 2n−1
1 N
ϕ̂2n (x̂)|[x̂2n ,x̂2n+2 ] = (x̂2n+2 − x̂), 1 ≤ n ≤ ,
ĥ 2n + ĥ 2n+1 2
extended by zero elsewhere. We take

⎧ ⎫
⎨ N /2
⎬
X̂ j = v̂ ∈ H01 (0, 1)2 ; v̂(x̂) = v̂ 1 ϕ̂1 (x̂) + v̂(x̂2n )ϕ̂2n (x̂) + v̂ N ϕ̂ N (x̂) ,
⎩ ⎭
n=1
(7.27)
⎧ ⎫
⎨ (N
+1)/2 ⎬
ˆ =
μ̂ ∈ H 1 (0, 1)2 ; μ̂(x̂) = μ̂(x̂2n )ϕ̂2n (x̂) . (7.28)
⎩ ⎭
n=0
With this choice, (4.24) is uniquely solvable and it can be checked that all the results
established for the first example carry over to this second example, with different
constants.
7.2 The 3-D case
In addition to restricting the mesh in 3-D (see Remark 4.1), the above proofs are
complex because they apply to a situation where the matrix of the system is irreducible.
The proofs are much easier, when (3.8) represents local and independent conditions,
but in general, this can only be achieved by suitably modifying the discrete velocity
spaces. In this section we present an example for which (3.8) holds by explicitly
constructing the solution of (4.3). The correction satisfies suitable continuity bounds
that are needed to establish the stability estimate (3.25). We follow the approach of
BenBelgacem [12] who presents a local construction in 3-D by adding to the space
h in
s, j a stabilizing bubble function in each face T of the trace T j,i j of the
X s, h
j
123
triangulation T jh on i j and choosing for H constant vectors on each face T . More

precisely, for each j, 1 ≤ j ≤ Ms , and for each T in T jh , let P(T ) denote a polynomial
space such as the mini-element or Bernardi–Raugel element, used in approximating
the velocity of the Stokes problem with order one. For each face T of T j, h , let T be
ij
the element of T jh with face T , and define
b T | T = λ1 λ2 λ3 ,
where λk , k = 1, 2, 3, denote the barycentric coordinates of the three vertices of T ;

it is extended by zero outside T . Then set
j = {v ∈ C (s, j ) ; ∀T ∈ T j , T not adjacent to i j , v|T ∈ P(T ),

h 0 3 h
X s,
∀T ∈ T jh , T adjacent to i j , v|T ∈ P(T ) + bT R3 },
(7.29)
and choose
H = {μ H ∈ L 2 (i j )3 ; ∀T ∈ T j,
h
ij
, μ H |T ∈ P30 }. (7.30)
As bT vanishes at all vertices of T , it does not change the approximating properties

of P(T ). Note that H does not satisfy Hypothesis 3.2 (3). Nevertheless a discrete
Korn inequality holds in Vsh , see Propositions 7.3 and 7.4 below. With this choice, we
can show that (3.8) holds. Indeed, for v ∈ X̃ , it is easy to see that

1
vh = v T bT , where v T = v
T bT

T ∈T j,
h
T
ij
satisfies (3.8). We now define

1 h
chj,i j (v) = cT bT , where cT = s (v)|s,i − S h (v)|s, j ,
T bT

T ∈T j,
h
T
ij
(7.31)
where sh (v)|s,i is computed in the preceding subsection by correcting S h (v)|s,i

h
with ci,ij
(v) defined in (4.17).
Lemma 7.4 The correction chj,i j (v) defined by (7.31) satisfies (4.3) and there exists
a constant C independent of h and the diameter of s,i , s, j , and i j such that for
all v in H01 ()n
|chj,i j (v)| H 1 (s, j ) ≤ C|v| H 1 (s,i ∪s, j ) , chj,i j (v) L 2 (s, j ) ≤ Ch|v| H 1 (s,i ∪s, j ) .
(7.32)
123
Proof For any face T of T j,

h , we can write
ij

1 h
cT = s (v)|s,i − v − (S h (v)|s, j − v) .
T bT

T
But on i j ,
sh (v)|s,i = S h (v)|s,i + ci,

h
ij
(v),
owing to the support of each correction. Therefore

1
c =
T
h
ci,ij
(v) + (S h (v)|s,i − v) − (S h (v)|s, j − v) = A1 + A2 + A3 .
T bT

T
Let T be the element of T jh adjacent to T . By arguing as in the proof of Lemma 4.1,

we easily derive
| A3 bT | H 1 (T ) ≤ Ĉ|v| H 1 (T ) , (7.33)
where T is the macro-element of T jh required to define S h (v) in T . The estimation

of A2 is similar, but more technical because T belongs to T jh whereas S h (v)|s,i is
constructed on Ti h . Following the argument used in the proof of Lemma 4.3, we derive

| A2 bT | H 1 (T ) ≤ Ĉ |v| H 1 (T ) , (7.34)

where {T } denote the set of elements of Ti h that intersect T .

The bound for A1 follows the same lines. With the notation of Lemma 4.1,

1
A1 = ck bk ,
T bT

T k
where the sum runs over the indices k for which Ok ∩T = ∅. According to Hypothesis
4.1, the number of terms in this sum is bounded by a fixed integer. Thus
−1
| A1 | ≤ Ĉ ρk 2 |v| H 1 (Dk ) ,
k
and Hypothesis 4.1 implies that
| A1 bT | H 1 (T ) ≤ Ĉ|v| H 1 (˜ T ) , (7.35)

123
where ˜ T is a macro-element in s,i and again the number of elements in ˜ T is

bounded by a fixed integer. All constants are independent of h and the diameter of
s,i , s, j , and i j . Finally, the first inequality in (7.32) follows readily by summing
(7.33)–(7.35) and applying Hypothesis 4.1. The argument for the second inequality in
(7.32) is similar.
7.2.1 A discrete Korn inequality in Vsh
Since all assumptions except Hypothesis 3.2 (3) hold, it suffices to examine the jump
of functions of Vsh through the interfaces i j ∈ ss . The next two lemmas show that
the projection of this jump on polynomials of P1 is small.
Lemma 7.5 Let i j ∈ ss , i < j. There exists a constant C, independent of h and
the diameters of i j , s,i , and s, j such that

1
3
∀v h ∈ Vsh , ∀ p ∈ P31 , [v h ] · p ≤ Ch 2 |i j | 2 |v h | H 1 (Ds,i ) + |v h | H 1 (Ds, j ) ,

i j
(7.36)
where Ds,k denotes the layer of elements in s,k adjacent to i j .

Proof All constants in this proof are independent of h and the diameters of i j , s,i ,
and s, j . Let T be an element (i.e. a face) of T j,
h , which is the mesh of H . There is
ij
no loss of generality in assuming that T lies on the x1 − x2 plane. Consider a generic
component v h of v h . Since the jump already satisfies

[v h ] = 0,
T
it suffices to bound the product of the jump by xk , k = 1, 2, say x1 . Then for any
constant c,
⎛ ⎞

1
[v h ]x1 = [v h ] − c ⎝x1 − x1 ⎠ .
|T |
T T T
A scaling argument gives

1
x1 − x 1
≤ Ĉ|T | 2 h T .
|T |
1

T L 2 (T )
Similarly,

inf [v h ] − c L 2 (T ) ≤ Ĉh T |[v h ]| H 1 (T ) ≤ Ĉh T |v h |s,i | H 1 (T ) + |v h |s, j | H 1 (T ) .
c∈R
123
On one hand, equivalence of norms gives
−1
|v h |s, j | H 1 (T ) ≤ Ĉ ρT 2 |v h |s, j | H 1 (T ) ,
where T is the element of T jh adjacent to T . On the other hand, arguing as in the

proof of Lemma 7.4, we prove that
−1
|v h |s,i | H 1 (T ) ≤ Ĉ ρT 2 |v h |s,i | H 1 (T ) ,
where T is the union of all elements of Ti h that intersect T . Then taking the product
and summing over all faces T of T j,
h , we obtain
ij
⎛ ⎞1
2
⎜ ⎟ h
⎜ |T |⎟
3
[v ]x1 ≤ Ĉh 2
h
⎝ ⎠ |v | H 1 (Ds,i ) + |v | H 1 (Ds, j ) ,
h

i j T ∈T j,
h
ij
whence (7.36).
Lemma 7.6 Let i j ∈ ss , i < j, and let ([v h ]) ∈ P31 denote the projection of [v h ]
onto P31 on i j . There exists a constant C, independent of h and the diameters of i j ,
s,i , and s, j such that
3
∀v h ∈ Vsh , ([v h ]) L 2 (i j ) ≤ Ch 2 |v h | H 1 (Ds,i ) + |v h | H 1 (Ds, j ) . (7.37)
Proof By definition ([v h ]) ∈ P31 solves

∀ p ∈ P31 , ([v h ]) · p = [v h ] · p.
i j i j
ˆ j , this reads
By passing to the reference subdomain

∀ p̂ ∈ P31 , ˆ |i j |−1
([v h ]) ◦ F j · p̂ = || [v h ] · p.
ˆ i j
The coefficients of the polynomial of degree one ([v h ]) ◦ F j solve a linear system
whose matrix only depends on .ˆ Therefore, in view of (7.36),
1 3
ˆ ≤ Ĉh 2 |v | H 1 (Ds,i ) + |v | H 1 (Ds, j ) .
([v h ]) L 2 (i j ) ≤ |i j | 2 ([v h ]) ◦ F j L ∞ () h h
This last result enables us to prove Korn’s inequality in Vsh .
123
Proposition 7.3 Let |s | > 0 and s be connected. Then there exists h 0 > 0 such
that for all h ≤ h 0 ,

Ms
Ms
∀v h ∈ Vsh , |v h |2H 1 ( ≤C D(v h )2L 2 ( ; (7.38)
s,i ) s,i )
i=1 i=1
the constants C and h 0 are independent of h and the diameters of the interfaces i j
and the subdomains s,k .
Proof Let v h ∈ Vsh . Formula (1.12) in [16] gives, with a constant C1 that depends
only on the shape regularity of the mesh of subdomains T ,

Ms
|v h |2H 1 (
s,i )
i=1
⎛ ⎞
Ms 1
≤ C1 ⎝ D(v h )2 2 L (s,i )
+ ([v h ])2L 2 ( ) ⎠
diam(i j ) ij
i=1 i< j
⎛ ⎞
Ms 1
≤ C1 ⎝ D(v h )2 2 L (s,i )
+ Ĉ h 3 |v h |2H 1 ( + |v h |2H 1 ( ⎠,
diam(i j ) s,i ) s, j )
i=1 i< j
owing to (7.37). But h < diam(i j ) and there exists an h 0 > 0 such that
1
∀h ≤ h 0 , h 2 (C1 Ĉ)−1 ≤ .
2
Since C1 and Ĉ are independent of h and the diameters of the interfaces i j and the
subdomains, then so is h 0 . Hence for all h ≤ h 0 ,

Ms
Ms
|v h |2H 1 ( ) ≤ 2C D(v h )2L 2 ( .
s,i s,i )
i=1 i=1
By arguing as in the proof of Lemma 3.4, we easily derive Korn’s inequality when
|s | = 0 and s is connected.
Proposition 7.4 Let |s | = 0 and s be connected, i.e. sd = ∂s . Then there exists
h 0 > 0 such that for all h ≤ h 0 ,
∀v h ∈ Vsh ,
⎛ ⎛ ⎞2 ⎞

Ms
Ms
2
⎜ ⎜ ⎟ ⎟
|v h |2H 1 ( ≤C ⎝ D(v ) L 2 (s,i ) + ⎝
h 2
|v sh · τ l |⎠ ⎠ ; (7.39)
s,i )
i=1 i=1 l=1
sd
123
the constants C and h 0 are independent of h and the diameters of the interfaces i j
and the subdomains s,k .
The case when s is not connected follows from these two propositions applied to
each connected component of s according that it is or not adjacent to s .
References
1. Adams, R.A.: Sobolev Spaces. Academic Press, New York (1975)

2. Angot, P.: On the well-posed coupling between free fluid and porous viscous flows. Appl. Math. Lett.
24, 803–810 (2011)
3. Arbogast, T.: Analysis of a two-scale, locally conservative subgrid upscaling for elliptic problems.
SIAM J. Numer. Anal. 42, 576–598 (2004)
4. Arbogast, T., Brunson, D.S.: A computational method for approximating a Darcy–Stokes system
governing a vuggy porous medium. Comput. Geosci. 11, 207–218 (2007)
5. Arbogast, T., Cowsar, L.C., Wheeler, M.F., Yotov, I.: Mixed finite element methods on non-matching
multiblock grids. SIAM J. Numer. Anal. 37, 1295–1315 (2000)
6. Arbogast, T., Pencheva, G., Wheeler, M.F., Yotov, I.: A multiscale mortar mixed finite element method.
Multiscale Model. Simul. 6, 319–346 (2007)
7. Babuška, I.: The finite element method with Lagrangian multipliers. Numer. Math. 20, 179–192 (1973)
8. Badea, L., Discacciati, M., Quarteroni, A.: Numerical analysis of the Navier–Stokes/Darcy coupling.
Numer. Math. 115, 195–227 (2010)
9. Badia, S., Codina, R.: Unified stabilized finite element formulations for the Stokes and the Darcy
problems. SIAM J. Numer. Anal. 47, 1971–2000 (2009)
10. Beavers, G.S., Joseph, D.D.: Boundary conditions at a naturally impermeable wall. J. Fluid. Mech 30,
197–207 (1967)
11. Ben Belgacem, F.: The mixed mortar finite element method for the incompressible Stokes problem:
convergence analysis. SIAM J. Numer. Anal. 37, 1085–1100 (2000) (electronic)
12. Ben Belgacem, F.: A stabilized domain decomposition method with nonmatching grids for the Stokes
problem in three dimensions. SIAM J. Numer. Anal. 42, 667–685 (2004) (electronic)
13. Bernardi, C., Maday, Y., Patera, A.T.: A new nonconforming approach to domain decomposition: the
mortar element method. In Nonlinear Partial Differential Equations and Their Applications. Collège
de France Seminar, Vol. XI (Paris, 1989–1991). Pitman Res. Notes Math. Ser., vol. 299, pp. 13–51.
Longman Sci. Tech., Harlow (1994)
14. Bernardi, C., Rebollo, T.C., Hecht, F., Mghazli, Z.: Mortar finite element discretization of a model
coupling Darcy and Stokes equations. M2AN Math. Model. Numer. Anal. 42, 375–410 (2008)
15. Brenner, S.C.: Poincaré-Friedrichs inequalities for piecewise H 1 functions. SIAM J. Numer. Anal. 41,
306–324 (2003)
16. Brenner, S.C.: Korn’s inequalities for piecewise H 1 vector fields. Math. Comp. 73, 1067–1087 (2004)
17. Brezzi, F.: On the existence, uniqueness and approximation of saddle-point problems arising from
Lagrange multipliers. RAIRO. Anal. Num. 2, 129–151 (1974)
18. Brezzi, F., Fortin, M.: Mixed and Hybrid Finite Element Methods. Springer, New York (1991)
19. Burman, E., Hansbo, P.: A unified stabilized method for Stokes’ and Darcy’s equations. J. Comput.
Appl. Math. 198, 35–51 (2007)
20. Cao, Y., Gunzburger, M., Hu, X., Hua, F., Wang, X., Zhao, W.: Finite element approximations for
Stokes-Darcy flow with Beavers–Joseph interface conditions. SIAM J. Numer. Anal. 47, 4239–4256
(2010)
21. Chen, W., Gunzburger, M., Hua, F., Wang, X.: A parallel Robin–Robin domain decomposition method
for the Stokes–Darcy system. SIAM J. Numer. Anal. 49, 1064–1084 (2011)
22. Chen, Z., Hou, T.Y.: A mixed multiscale finite element method for elliptic problems with oscillating
coefficients. Math. Comp. 72, 541–576 (2003)
23. Crouzeix, M., Thomee, V.: The stability in L p and W p1 of the l2 projection onto finite element function
spaces. Math. Comput. 48, 521–532 (1987)
24. Discacciati, M., Miglio, E., Quarteroni, A.: Mathematical and numerical models for coupling surface
and groundwater flows. Appl. Numer. Math. 43, 57–74 (2002) (19th Dundee Biennial Conference on
Numerical, Analysis (2001))
123
25. Discacciati, M., Quarteroni, A.: Convergence analysis of a subdomain iterative method for the finite
element approximation of the coupling of Stokes and Darcy equations. Comput. Vis. Sci. 6, 93–103
(2004)
26. Discacciati, M., Quarteroni, A.: Navier–Stokes/Darcy coupling: modeling, analysis, and numerical
approximation. Rev. Mat. Complut. 22, 315–426 (2009)
27. Discacciati, M., Quarteroni, A., Valli, A.: Robin-Robin domain decomposition methods for the Stokes-
Darcy coupling. SIAM J. Numer. Anal. 45, 1246–1268 (2007) (electronic)
28. Ervin, V.J., Jenkins, E.W., Sun, S.: Coupled generalized nonlinear Stokes flow with flow through a
porous medium. SIAM J. Numer. Anal. 47, 929–952 (2009)
29. Galvis, J., Sarkis, M.: Non-matching mortar discretization analysis for the coupling Stokes–Darcy
equations. Electron. Trans. Numer. Anal. 26, 350–384 (2007)
30. Galvis, J., Sarkis, M.: FETI and BDD preconditioners for Stokes–Mortar–Darcy systems. Commun.
Appl. Math. Comput. Sci. 5, 1–30 (2010)
31. Ganis, B., Yotov, I.: Implementation of a mortar mixed finite element method using a multiscale flux
basis. Comput. Methods Appl. Mech. Eng. 198, 3989–3998 (2009)
32. Gatica, G.N., Meddahi, S., Oyarzúa, R.: A conforming mixed finite-element method for the coupling
of fluid flow with porous media flow. IMA J. Numer. Anal. 29, 86–108 (2009)
33. Gatica, G.N., Oyarzúa, R., Sayas, F.-J.: Analysis of fully-mixed finite element methods for the Stokes–
Darcy coupled problem. Math. Comp. 80, 1911–1948 (2011)
34. Girault, V., Raviart, P.-A.: Finite Element Methods for Navier–Stokes Equations. Springer series in
Computational Mathematics, vol. 5. Springer, Berlin (1986)
35. Girault, V., Rivière, B.: DG approximation of coupled Navier–Stokes and Darcy equations by Beaver–
Joseph–Saffman interface condition. SIAM J. Numer. Anal. 47, 2052–2089 (2009)
36. Girault, V., Scott, L.R.: A quasi-local interpolation operator preserving the discrete divergence, Calcolo,
pp. 1–19 (2003)
37. Glowinski, R., Wheeler, M.F.: Domain decomposition and mixed finite element methods for elliptic
problems. In: Glowinski, R., Golub, G.H., Meurant, G.A., Periaux, J. (eds.) First International Sym-
posium on Domain Decomposition Methods for Partial Differential Equations, pp. 144–172. SIAM,
Philadelphia (1988)
38. Grisvard, P.: Elliptic problems in nonsmooth domains. Pitman, Boston (1985)
39. Hoppe, R.H.W., Porta, P., Vassilevski, Y.: Computational issues related to iterative coupling of subsur-
face and channel flows. Calcolo 44, 1–20 (2007)
40. Hou, T.Y., Wu, X.H.: A multiscale finite element method for elliptic problems in composite materials
and porous media. J. Comput. Phys. 134, 169–189 (1997)
41. Hughes, T.J.R.: Multiscale phenomena: Green’s functions, the Dirichlet-to-Neumann formulation,
subgrid scale models, bubbles and the origins of stabilized methods. Comput. Methods Appl. Mech.
Eng. 127, 387–401 (1995)
42. Jäger, W., Mikelić, A.: On the interface boundary condition of Beavers, Joseph, and Saffman. SIAM
J. Appl. Math. 60, 1111–1127 (2000)
43. Jerrison, D.S., Kenig, C.: The Neumann problem on Lipschitz domains. Bull. AMS 4, 203–207 (1981)
44. Kanschat, G., Rivière, B.: A strongly conservative finite element method for the coupling of Stokes
and Darcy flow. J. Comput. Phys. 229, 5933–5943 (2010)
45. Karper, T., Mardal, K.-A., Winther, R.: Unified finite element discretizations of coupled Darcy–Stokes
flow. Numer. Methods Partial Differ. Equ. 25, 311–326 (2009)
46. Layton, W.J., Schieweck, F., Yotov, I.: Coupling fluid flow with porous media flow. SIAM J. Numer.
Anal. 40, 2195–2218 (2003)
47. Mardal, K.A., Tai, X.-C., Winther, R.: A robust finite element method for Darcy-Stokes flow. SIAM J.
Numer. Anal. 40, 1605–1631 (2002) (electronic)
48. Mathew, T.P.: Domain decomposition and iterative refinement methods for mixed finite element dis-
cretizations of elliptic problems, PhD thesis, Courant Institute of Mathematical Sciences, New York
University, Tech. Rep. 463 (1989)
49. Mu, M., Xu, J.: A two-grid method of a mixed Stokes–Darcy model for coupling fluid flow with porous
media flow. SIAM J. Numer. Anal. 45, 1801–1813 (2007)
50. Nečas, J.: Les Méthodes directes en théorie des équations elliptiques. Masson, Paris (1967)
51. Peetre, J.: Espaces d’interpolation et théorème de soboleff. Ann. Inst. Fourier 16, 279–317 (1966)
52. Pencheva, G., Yotov, I.: Balancing domain decomposition for mortar mixed finite element methods.
Numer. Linear Algebra Appl 10, 159–180 (2003)
123
53. Pencheva, G., Yotov, I.: Interior superconvergence in mortar and non-mortar mixed finite element
methods on non-matching grids. Comput. Methods Appl. Mech. Eng. 197, 4307–4318 (2008)
54. Rivière, B., Yotov, I.: Locally conservative coupling of Stokes and Darcy flows. SIAM J. Numer. Anal.
42, 1959–1977 (2005)
55. Saffman, P.G.: On the boundary condition at the surface of a porous media. Stud. Appl. Math., L,
pp. 93–101 (1971)
56. Scott, L.R., Zhang, S.: Finite element interpolation of nonsmooth functions satisfying boundary con-
ditions. Math. Comp. 54, 483–493 (1990)
57. Vassilev, D., Yotov, I.: Coupling Stokes–Darcy flow with transport. SIAM J. Sci. Comput. 31, 3661–
3684 (2009)
58. Vassilev, D., Wang, C., Yotov, I.: Domain decomposition for coupled Stokes and Darcy flows. Comput.
Methods Appl. Mech. Eng. (to appear)
59. Xie, X., Xu, J., Xue, G.: Uniformly-stable finite element methods for Darcy–Stokes–Brinkman models.
J. Comput. Math. 26, 437–455 (2008)
60. Yotov, I.: Mixed finite element methods for flow in porous media, PhD thesis, Rice University, Hous-
ton, Texas (1996). TR96-09, Dept. Comp. Appl. Math., Rice University and TICAM report 96–23,
University of Texas at Austin
123

2014 NumerMath MortarMultiscaleMethod

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2014 NumerMath MortarMultiscaleMethod

Uploaded by

Copyright:

Available Formats

Numer. Math.

(2014) 127:93–165 Numerische

Mortar multiscale finite element methods

Vivette Girault · Danail Vassilev · Ivan Yotov

Abstract We investigate mortar multiscale numerical methods for coupled Stokes

Mathematics Subject Classification 65N12 · 65N15 · 65N30 · 65N55 · 76D07 ·

is used to replace the solution of subdomain problems by a simple linear combination.

1.1 Notation and preliminaries

Let be a bounded, connected Lipschitz domain of Rn , n = 2, 3, with boundary

1 () reads: There exists a constant P depending only

∀v ∈ H 1 (), (v) ≤ P1 v H 1 () . (1.2)

We recall Korn’s first inequality: There exists a constant C1 depending only on

Proposition 1.2 Let be a bounded, connected Lipschitz domain of Rn and let  be

1) There exists a constant C2 such that

∀v ∈ H 1 ()n , (v) ≤ C2 v H 1 () . (1.5)

2) The condition (m) = 0 for a rigid-body motion m holds if and only if m is a

where | · | denotes the Euclidean vector norm.

This definition is extended to any real number s = m + s  for an integer m ≥ 0 and

equipped with the graph norm

H0 (div; ) = {v ∈ H (div; ); v · n = 0 on ∂} .

2.1 Coupled Stokes and Darcy systems

∀x ∈ d , ∀χ ∈ Rn , λmin |χ|2 ≤ K (x)χ · χ ≤ λmax |χ|2 , (2.1)

The fluid velocity and pressure in d , respectively s , are denoted by ud and pd ,

σ (us , ps ) = − ps I + 2νs D(us ).

In the Darcy region d , the pair (ud , pd ) satisfies

In the Stokes region s , the pair (us , ps ) satisfies

− div σ (us , ps ) ≡ −2νs div D(us ) + ∇ ps = f s in s , (2.6)

whereτ l , 1≤ l ≤ n − 1 is an orthogonal system of unit tangent vectors on sd and

2.2 First variational formulation

For any functions ϕd defined in d and ϕs defined in s , it is convenient to define ϕ

and replace (2.10)–(2.11) by the condition

σ (us , ps )ns = g on sd . (2.13)

Then the expression (2.12) for g yields

where the jump [v · n] is defined by

X̃ = {v ∈ H (div; ); v s ∈ H 1 (s )n , v|s = 0, (v · n)|d = 0}, (2.17)

which is a Hilbert space equipped with the norm

As the trace space of v s on sd is large enough, this implies (2.13).

2.3 Existence and uniqueness of the solution

and for any functions v d ∈ H (div; d ), v s ∈ H (div; s ) and w ∈ L 2 (), we define

Note that ã(·, ·) is continuous on X̃ × X̃ and b̃(·, ·) is continuous on X̃ × L 2 ():

X̃ 0 = {v ∈ X̃ ; div v = 0}, (2.26)

X̃ g = {v ∈ X̃ ; div v = g}. (2.27)

Then we consider the reduced problem : Find u ∈ X̃ q such that

Proposition 2.1 Let ( f , qd ) be given in L 2 ()n × L 20 (d ). Let (u, p) ∈ X̃ × W be a

Hence (1.7) and a straightforward manipulation yield for all v s in H 1 (s )n :

As a consequence (2.30) is replaced by

is a norm on X̃ 0 equivalent to v X̃ . This is the object of the next lemma.

As X̃ 0 is reflexive, this implies that there exists a function v ∈ X̃ 0 such that

Moreover v s = c, a constant vector, and v d = 0. Then the fact that v belongs to X̃

lim v m = 0 weakly in H 1 (s )n ,

lim v m = 0 strongly in L 2 (s )n .

Fig. 1 Domain decomposition

Lemma 2.4 There exists a constant C5 > 0 such that

∀v ∈ X̃ 0 , ã(v, v) ≥ C5 v2X̃ . (2.36)

In turn, the well-posedness of problem (2.3)–(2.11) stems from Lemma 2.1.

2.4 Domain decomposition of the Darcy and Stokes regions

Let s , respectively d , be decomposed into Ms , respectively Md , non-overlapping,

Set M = Md + Ms ; according to convenience we can also number the subdomains

and for 1 ≤ i < j ≤ M, let the interfaces between subdomains be denoted by i j ,

[uu d · n] = 0, [ pd ] = 0 on dd , (2.38)

where the jumps on an interface i j , 1 ≤ i < j ≤ M, are defined as

[v · n] = v i · ni + v j · n j , [σ n] = σ i ni + σ j n j , [v] = (vi − v j )|i j ,

X d = {v ∈ L 2 (d )n ; v d,i := v|d,i ∈ H (div; d,i ), 1 ≤ i ≤ Md ,

∀v ∈ H 1 (), (v) ≤ P1 v H 1 () . (1.2)

Proposition 1.2 Let be a bounded, connected Lipschitz domain of Rn and let be

∀v ∈ H 1 ()n , (v) ≤ C2 v H 1 () . (1.5)

2) The condition (m) = 0 for a rigid-body motion m holds if and only if m is a

This definition is extended to any real number s = m + s for an integer m ≥ 0 and

|||v||| X = |||v|||2X d + |||v|||2X s , |||λ|||s = ⎝ ⎠ ,

where the functional is a suitable seminorm. Let us choose

Clearly, is a seminorm on X s . Next, considering that

|||sh (v)||| X s ≤ C|v| H 1 () . (3.25)

|||sh (v)||| X s ≤ C|v| H 1 () . (3.28)

Proof The operator sh is constructed by correcting sh :

sh (v) = sh (v) + csh (v)

|||dh (v)||| X d ≤ C|v| H 1 () . (3.33)

||| (v)||| X ≤ C|v| H 1 () ,

4 Elements of construction of sh and hd

Constructing an operator sh with values in Vsh , satisfying (3.23)–(3.25), uniformly

sh (v)|s,i := sh (v)|s,i + ci,

sh (v)|s, j := sh (v)|s, j + chj,i j (v),