Finite Volume Methods for Solving PDEs

Finite Volume Methods
Schemes and Analysis

course at the University of Wroclaw
Robert Eymard1 , Thierry Gallouët2 and Raphaèle Herbin3
April, 2008
1
Université Paris-Est Marne-la-Vallée
2
Université de Provence
3
Université de Provence
Contents
Introduction 3
1 Finite volume discretization

of balance laws 4
1.1 A generic example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Finite volume meshes and space-time discretizations . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Linear diffusion problems 9

2.1 Actual problems and schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2 Numerical results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Further analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3.1 The convergence proof in the (harmonic averaging) isotropic case . . . . . . . . . . . . . 25
2.3.2 The convergence proof in the case of the gradient schemes . . . . . . . . . . . . . . . . . 27
2.3.3 The convergence proof in the case of the mixed-type schemes . . . . . . . . . . . . . . . 33
3 Nonlinear convection - degenerate diffusion problems 34

3.1.1 Finite volume approximation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.1.2 A numerical example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.2 Further analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.1 Continuous definitions and main convergence result . . . . . . . . . . . . . . . . . . . . . 37
3.2.2 Existence, uniqueness and discrete properties . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.3 Compactness of a family of approximate solutions . . . . . . . . . . . . . . . . . . . . . 45
3.2.4 Convergence towards an entropy process solution . . . . . . . . . . . . . . . . . . . . . . 46
3.2.5 Uniqueness of the entropy process solution. . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4 Navier-Stokes equations 58
4.2 Finite volume scheme with staggered variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.1 The MAC scheme on regular grids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.2 An example, built on the pressure grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3 Further analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.3 Finite volume scheme with collocated variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.1 Discrete scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.2 Further analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.3 Numerical example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
1
2
5 Discrete functional analysis 70

5.1 The topological degree argument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2 Discrete Sobolev embedding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
?
5.2.1 Discrete embedding of W 1,1 (Ω) in L1 (Ω) . . . . . . . . . . . . . . . . . . . . . . . . . 70
1,p p?
5.2.2 Discrete embedding of W (Ω) in L (Ω), 1 < p < d . . . . . . . . . . . . . . . . . . . 72
5.2.3 Discrete embedding of W 1,p (Ω) in Lq (Ω), for some q > p . . . . . . . . . . . . . . . . . 73
5.3 Compactness results for bounded families in discrete W 1,p (Ω) norm . . . . . . . . . . . . . . . . 73
5.3.1 Compactness in Lp (Ω) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3.2 Regularity of the limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
5.4 Properties in the case of the ∆−adapted discretizations . . . . . . . . . . . . . . . . . . . . . . . 76
5.5 Lemma used for time translates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Bibliography 83
Introduction
Our aim is to describe and analyse numerical schemes, issued from cell centred finite volume methods, applied
to various physical problems. For each problem, we present a mathematical model and some relevant finite vol-
ume schemes. The efficiency of the schemes is then illustrated by some numerical results. In a second step, a
mathematical analysis of the schemes is carried out.
The first chapter introduces the notion of mesh, suitable for finite volume discretizations. In the second and third
chapters, we focus on problems which are modeled by a scalar partial differential equation: we shall successively
deal with pure diffusion operators, isotropic or not, and with convection-diffusion operators including the pure
hyperbolic case.
In the fourth chapter, we focus on the incompressible viscous flow (Navier-Stokes equations).
We finally provide, in the fifth chapter, some mathematical results, used in the convergence and error analysis of
the schemes.
3
Chapter 1
Finite volume discretization

of balance laws
1.1 A generic example

Let us take a generic example, dedicated to show the need of complete definitions for the space and time discretiza-
tions. Let Ω be a polygonal open subset of Rd , T ∈ R, and let us consider a scalar balance law written under the
general form:
ut + div(F (u, ∇u)) + s(u) = 0 on Ω × (0, T ), (1.1)
where F ∈ C 1 (R × Rd , Rd ) and s ∈ C(R, R), and u is a real function of (x, t) ∈ Ω × (0, T ). In order to give the
main ideas of the finite volume schemes, we need to introduce the set M containing the control volumes, which are
the subsets of Ω on which balance equations are written (see Definition 1.1). Such a balance equation is obtained
from the above conservation law by integrating it over a control volume K ∈ M and applying the Stokes formula:
Z Z Z
ut dx + F (u, ∇u) · nK dγ(x) + s(u) dx = 0,
K ∂K K
where nK stands for the unit normal vector to the boundary ∂K outward to K and γ denotes the integration with
respect to the (d − 1)–dimensional Lebesgue measure. Let us denote by E the set of edges (faces in 3D) of the
mesh, and EK the set of edges which form the boundary ∂K of the control volume K. With these notations, the
above equation reads:
Z X Z Z
ut dx + F (u, ∇u) · nK dγ(x) + s(u) dx = 0.
K σ∈EK σ K
Let δt = T /M , where M ∈ N, M ≥ 1, and let us perform an explicit Euler discretization of the above equation
(an implicit or semi-implicit discretization could also be performed, and is sometimes preferable, depending on
the type of equation). We then get:
Z X Z Z
u(n+1) − u(n) (n) (n)
dx + F (u , ∇u ) · nK dγ(x) + s(u(n) ) dx = 0,
K δt σ K
σ∈EK
where u(n) denotes an approximation of u(·, t(n) ), with t(n) = nδt (see Definition 1.6 for a general definition
of a time discretization). Let us then introduce the discrete unknowns (one per control volume and time step)
(n)
(uK )K∈M, n∈N ; assuming the existence of such a set of real values, we may define a piecewise constant function
by: X (n)
(n) (n)
uM ∈ HM (Ω) : uM = uK 1K ,
K∈M
4
5
where HM (Ω) denotes the space of functionsR from Ω to R which are constant on each control volume of the mesh
M. In order to define the scheme, the fluxes σ F (u(n) , ∇u(n) ) · nK dγ(x) need to be approximated as a function
(n)
of the discrete unknowns. We denote by FK,σ (uM ) the resulting numerical flux, the expression of which depends
on the type of flux to be approximated. Let us now give this expression for various simple examples.
First we consider the case of a linear convection equation, that is equation (1.1) where the flux F (u, ∇u) reduces
to F (u, ∇u) = vu, v ∈ Rd , and s(u) = 0:
ut + div(vu) = 0 on Ω. (1.2)
In order to approximate the flux vu · n on the edges of the mesh, one needs to approximate the value of u on
these edges, as a function of the discrete unknowns uK associated to each control volume K. This may be done
in several ways. A straightforward choice is to approximate the value of u on the edge σ = K|L separating the
control volumes K and L by the mean value 21 (uK + uL ). This yields the following numerical flux:
(cv,c) uK + uL
FK,σ (uM ) = vK,σ
2
R
where vK,σ = σ v · nK,σ , and nK,σ denotes the unit normal vector to the edge σ outward to K. This centred
choice is known to lead to stability problems, and is therefore often replaced by the so–called upstream choice,
which is given by:
(cv,u) + −
FK,σ (uM ) = vK,σ uK − vK,σ uL , (1.3)
where x+ = max(x, 0) and x− = − min(x, 0). Hence we see that for purely convective problems, we do not
require many geometrical properties on the mesh. Such meshes are formally defined in definition 1.1, where we
only need to give sufficient properties in order to compute the normal to the boundary.
If we now consider a linear convection diffusion reaction equation, that is equation (1.1) with F (u, ∇u) =
−Λ∇u + vu, v ∈ Rd , and s(u) = bu, b ∈ R:
ut + div(−Λ∇u + vu) + bu = 0 on Ω, (1.4)
the flux through a given edge then reads:

Z Z
F (u) · nK,σ = (−Λ∇u + v u) · nK,σ ,
σ σ
R
so that we now need to discretize the additional term σ Λ∇u · nK,σ ; this diffusion flux involves the knowledge of
the derivatives at the boundary. In order to give approximations of this term, we need to define finite differences,
consistent with ∇u, which implies that u is evaluated at some particular points of the mesh. This is the purpose
of Definition of ”pointed meshes” (see Definition 1.2 and the following ones, each of them suited with particular
meshes frameworks).
1.2 Finite volume meshes and space-time discretizations

As announced in the first section of this chapter, we now present the precise definitions of space and time dis-
cretizations, used all along this book for defining the different schemes. General finite volume meshes are used for
partial differential equations including first order, second order (scarcely more). In the case of convection prob-
lems, we only need of a partition of the domain in control volumes, whereas in the case of diffusion problems, we
also need to define points into the control volumes (see below).
We shall consider a series of properties, and a discretisation will be defined by reference to some of them.
Let us consider that
d ∈ N? ,
Ω ⊂ Rd is open, polygonal, bounded, connected, (1.5)
∂Ω = Ω \ Ω is the boundary of Ω, assumed to be Lipschitz-continuous.
6
In the above definition, polygonal means that ∂Ω is a finite union of subsets of hyperplanes of Rd .
Let us give a definition for a polygonal finite volume discretisation of Ω.
Definition 1.1 (Polygonal finite volume space discretization of Ω)

Under hypothesis (1.5) a polygonal finite volume discretization of Ω, is given by the pair (M, E), where:
1. M is a finite family of non empty connected open disjoint subsets of Ω (the “control volumes”) such that
Ω = ∪K∈M K. For any K ∈ M, let ∂K = K \ K be the boundary of K, m(K) > 0 denote the measure
of K and hK denote the diameter of K.
2. E is a finite family of disjoint subsets of Ω (the “edges” of the mesh), such that, for all σ ∈ E, σ is a non
empty open subset of a hyperplane of Rd , whose (d-1)-dimensional measure m(σ) is stricly positive. We
assume that, for all K ∈ M, there exists a subset EK of E such that ∂K = ∪σ∈EK σ. We then denote by
Mσ = {K ∈ M, σ ∈ EK }. We then assume that, for all σ ∈ E, either Mσ has exactly one element and
then σ ⊂ ∂Ω (the set of these edges, called boundary edges, is denoted by Eext ) or Mσ has exactly two
elements (the set of these edges, called interior edges, is denoted by Eint ). For all K ∈ M and σ ∈ EK , we
denote for a.e. x ∈ σ by nK,σ the unit vector normal to σ outward to K.
For all K ∈ M, we denote by NK the set of the neighbours of K:
NK = {L ∈ M \ {K}, ∃σ ∈ Eint , Mσ = {K, L}}. (1.6)
We then denote by HM (Ω) the set of functions from Ω to R, constant on each element of M. The size of the
discretization is defined by:
h(M) = sup{diam(K), K ∈ M}. (1.7)
We then define the space XD = RM×E , containing all the families of reals ((uK )K∈M , (uσ )σ∈E ), and we
consider the subspace XD,0 of all families ((uK )K∈M , (uσ )σ∈E ) such that uσ = 0 if σ ∈ Eext . We then define,
for u ∈ XD , ΠM u ∈ HM (Ω) as the element of HM (Ω) equal a.e. in K to the value uK , for all K ∈ M.
In order to study the schemes introduced in this paper for diffusion operators, we also define the notion of pointed
finite volume discretization.
Definition 1.2 (Pointed polygonal finite volume space discretization of Ω)

Under hypothesis (1.5), a pointed polygonal finite volume discretization of Ω, is given by the triplet (M, E, P),
where:
1. (M, E) is a polygonal finite volume discretization of Ω in the sense of definition 1.1.

2. P is a family of points of Ω indexed by M and E, denoted by P = ((xK )K∈M , (xσ )σ∈E ), such that for all
K ∈ M, xK ∈ K and for all σ ∈ E, xσ ∈ σ. We then denote by dK,σ the abscissae of xK on the straight
line with origine xσ and oriented by nK,σ .
We then define PM : C(Ω) → HM (Ω) by
PM ϕ(x) = ϕ(xK ) for a.e. x ∈ K, ∀K ∈ M, ∀ϕ ∈ C(Ω). (1.8)
and PD : C(Ω) → XD by
PD ϕK = ϕ(xK ), ∀K ∈ M, PD ϕσ = ϕ(xσ ), ∀σ ∈ E, ∀ϕ ∈ C(Ω). (1.9)
Remark 1.1 The conditions xK ∈ K and xσ ∈ σ can easily be relaxed.

7
xK
dKσ
xσ
Figure 1.1: Pointed mesh
Thanks to the algebraic definition of dK,σ , we have the relation

X
m(σ)dK,σ = d m(K), ∀K ∈ M. (1.10)
σ∈EK
Definition 1.3 (Pointed star-shaped polygonal finite volume space discretization of Ω)

Under hypothesis (1.5), a pointed polygonal finite volume discretization of Ω, denoted by (M, E, P), is said to be
star-shaped, if all K ∈ M is xK -star-shaped, which means that for all x ∈ K, the property [xK , x] ⊂ K holds
(this is equivalent to dK,σ ≥ 0 for all σ ∈ EK ). A pointed polygonal finite volume discretization of Ω, denoted by
(M, E, P), is said to be strictly star-shaped if, for all K ∈ M and σ ∈ EK , dK,σ > 0.
For all K ∈ M and σ ∈ EK , we denote by DK,σ the cone with vertex xK and basis σ
DK,σ = {txK + (1 − t)y, t ∈ (0, 1), y ∈ σ}, (1.11)
and we then define the set LD (Ω) of all functions g which are constant in each cone DK,σ (we
S then denote by gK,σ
the corresponding value). Note that HM (Ω) ⊂ LD (Ω). We denote, for all σ ∈ E, Dσ = K∈Mσ DK,σ (this set
is sometimes called the “diamond” associated to the edge σ). We then define dσ , for all σ ∈ E, by
dσ = dL,σ + dK,σ , ∀σ ∈ Eint , Mσ = {K, L},

(1.12)
dσ = dK,σ , ∀σ ∈ Eext , Mσ = {K}.
Remark 1.2 The above definition applies to a large variety of meshes. Note that no hypothesis is made on the
convexity of the control volumes, which enables that “hexahedra” with non planar faces can be used (in fact, such
sets have then 12 faces if each non planar face is shared in two triangles, but only 6 neighbouring control volumes).
Remark 1.3 The common boundary of two neighbouring control volumes can include more than one edge.
Definition 1.4 (∆−adapted pointed polygonal finite volume space discretization of Ω)

Under hypothesis (1.5), a strictly star-shaped pointed polygonal finite volume discretization of Ω, denoted by
(M, E, P), is said to be a ∆−adapted pointed polygonal finite volume space discretization of Ω, if, for all σ ∈ Eint ,
denoting by K, L ∈ M the two control volumes such that Mσ = {K, L}, then the straight line (xK , xL ) is
orthogonal to σ. In this case, we denote by xσ the intersection between σ and the line (xK , xL ) (assumed to
belong to σ for the sake of simplicity). Then, for K, L ∈ M, such that there exists σ ∈ Eint with Mσ = {K, L},
σ is unique and one denotes σ = K|L. For all K ∈ M and σ ∈ EK ∩ Eext , we denote by xσ the orthogonal
projection of xK on σ.
An example of two neighbouring control volumes K and L of M is depicted in Figure 1.2.

8
m(σ)
xK
xL
dKL
K
L K|L dK,σ
Figure 1.2: Notations for a ∆-adapted mesh
Definition 1.5 (∆−super-adapted pointed polygonal finite volume space discretization of Ω) Under hypothe-
sis (1.5), a ∆−adapted pointed polygonal finite volume discretization of Ω, denoted by (M, E, P), is said to be
∆−super adapted, if, for all σ ∈ Eint , denoting by K, L ∈ M the two control volumes such that Mσ = {K, L},
then the straight line (xK , xL ) is orthogonal to σ and intersects σ at the point xσ , equal to the center of gravity
of σ.
Definition 1.6 (Time discretization of (0, T )) A time discretization of (0, T ) is given by an integer value N and
by an increasing sequence of real values (tn )n∈[[0,N +1]] with t0 = 0 and tN +1 = T . The time steps are then
defined by δtn = tn+1 − tn , for n ∈ [[0, N ]].
Definition 1.7 (Space-time discretization of Ω × (0, T )) A finite volume discretization D of Ω × (0, T ) is the
family D = (M, E, (xK )K∈M , N, (tn )n∈[[0,N ]] ), where M, E, (xK )K∈M is a finite volume mesh of Ω in the
sense of one of the Definitions 1.1-1.5 and N , (tn )n∈[[0,N +1]] is a time discretization of (0, T ) in the sense of
Definition 1.6. For a given mesh D, one defines:
|D| = max(h(M), (δtn )n∈[[0,N ]] )

Chapter 2
Linear diffusion problems
2.1 Actual problems and schemes

We wish to develop and study some schemes for the approximation of the weak solution ū of the following diffusion
problem with full anisotropic tensor:
−div(Λ∇ū) = f in Ω,
(2.1)
ū = 0 on ∂Ω,
under the following assumptions:
Ω is an open bounded connected polygonal subset of Rd , d ∈ N? , (2.2)
Λ is a measurable function from Ω to Md (R), where Md (R) denotes the set of

d × d matrices, such that for a.e. x ∈ Ω, Λ(x) is symmetric,
the lowest and the largest eigenvalues of Λ(x), denoted by λ(x) and λ(x), (2.3)
are such that λ, λ ∈ L∞ (Ω) and there exists λ0 ∈ R with
0 < λ0 ≤ λ(x) ≤ λ(x) for a.e. x ∈ Ω,
and
f ∈ L2 (Ω). (2.4)
We give the classical weak formulation in the following definition.
Definition 2.1 (Weak solution) Under hypotheses (2.2)-(2.4), we say that ū is a weak solution of (2.1) if

 Zū ∈ H01 (Ω), Z
(2.5)
 Λ(x)∇ū(x) · ∇v(x)dx = f (x)v(x)dx, ∀v ∈ H01 (Ω).
Ω Ω
Remark 2.1 For the sake of clarity, we restrict ourselves here to the numerical analysis of Problem (2.1), however,
the present analysis readily extends to convection-diffusion-reaction problems and coupled problems. Indeed,
we emphasize that proofs of convergence or error estimate can easily be adapted to such situations, since the
discretization methods of all these terms are independent of one another, and the treatment of convection and
reaction term is well-known exact (see [26] or [20]).
We consider in this chapter a family of schemes based on the definition of a numerical flux, using an unknown
located at the edges of the mesh. A variety of situations with respect to a possible elimination of this unknown are
then considered, within always ensuring the following properties:
1. The matrices of the generated linear systems are expected to be sparse, symmetric, positive and definite.
9
10
2. We wish to be able to prove the convergence of the discrete solution and an associate gradient to the solution
of the continuous problem and its gradient, and to show error estimates.
Under hypothesis (1.5), let D = (M, E, P) be a discretization of Ω in the sense of definition 1.2 (we shall also
consider below the particular case of ∆−adapted discretizations in the sense of definition 1.4).
Common features of the schemes

The idea of the schemes described in this chapter is to find an approximation of the solution of (2.1) by setting
up a system of discrete equations for a family of values ((uK )K∈M , (uσ )σ∈E ) in the control volumes and on the
interfaces. Without a suitable elimination procedure, the number of unknowns is therefore card(M) + card(E).
Following the idea of the finite volume framework, equation (2.1) is integrated over each control volume K ∈ M,
which formally gives (assuming sufficient regularity on u and Λ) the following balance equation on the control
volume K:
X µ Z ¶ Z
− Λ(x)∇u(x) · nK,σ dγ(x) = f (x)dx.
σ∈EK σ K
R
The flux − σ Λ(x)∇u(x)·nK,σ dγ(x) is approximated by a function FK,σ (u) of the values ((uK )K∈M , (uσ )σ∈E )
at the “centers” and at the interfaces of the control volumes (in all the cases presented below, FK,σ (u) only depends
on uK and all (uσ0 )σ0 ∈EK ). We give the définition of the sets XD and XD,0 (also provided in Definition (1.1))
where the discrete unknowns lie, that is to say:
XD = {v = ((vK )K∈M , (vσ )σ∈E ), vK ∈ R, vσ ∈ R}, (2.6)
XD,0 = {v ∈ XD such that vσ = 0 ∀σ ∈ Eext }. (2.7)
A discrete equation corresponding to (2.1) is then: find u ∈ XD,0 such that
X Z
FK,σ (u) = f (x)dx, ∀K ∈ M. (2.8)
σ∈EK K
The values uσ on the interfaces are then introduced so as to allow for a consistent approximation of the normal
fluxes in the case of an anisotropic operator and a general, possibly nonconforming mesh. Thanks to the defi-
nition of XD,0 , the values uσ for the boundary faces or edges are prescribed by the discrete counterpart of the
homogeneous Dirichlet boundary condition:
uσ = 0, ∀σ ∈ Eext . (2.9)
We then need card(Eint ) equations to ensure that the problem is well posed. Following the finite volume ideas, we
may write the continuity of the discrete flux for all interior edges, that is to say:
FK,σ (u) + FL,σ (u) = 0, for σ ∈ Eint such that Mσ = {K, L}. (2.10)
We now have card(M) + card(Eint ) unknowns and equations. Note that u ∈ XD,0 is solution of (2.8)-(2.9)-(2.10)
if and only if it satisfies 
 u ∈ XD,0 , Z
X
hu, vi = v f (x)dx, for all v ∈ XD,0 , (2.11)
 F K
K∈M K
where we set X X
hu, viF = FK,σ (u)(vK − vσ ). (2.12)
K∈M σ∈EK
Indeed, taking vK = 1, vL = 0 for all L ∈ M with L 6= K and vσ = 0 for all σ ∈ E leads to (2.8). Taking vσ = 1
for σ ∈ Eint , vσ0 = 0 for all σ 0 ∈ E with σ 0 6= σ and vK = 0 for all K ∈ M provides (2.10). Reciprocally, for all
v ∈ XD,0 , multiplying (2.8) by vK and using (2.10) gives (2.11).
In the following, we review different choices for the meshes and for the expression of the numerical fluxes, leading
to different properties for the resulting scheme. Let us first turn to the simple case of isotropic diffusion, using
∆−adapted pointed meshes.
11
The case of the Laplace operator on ∆−adapted pointed meshes

We assume in this section the hypothesis (2.2) and that
Λ(x) = Id where Id is the identity application of Rd for a.e. x ∈ Ω. (2.13)
Let D be an admissible discretization of Ω in the sense of Definition 1.4. The finite volume approximation to
Problem (2.1) is given as the solution u = ((uK )K∈M , (uσ )σ∈E ) ∈ XD,0 of (2.8)-(2.9)-(2.10), in which we define
the numerical flux by:
uσ − uK
FK,σ (u) = −m(σ) , ∀K ∈ M, ∀σ ∈ EK . (2.14)
dK,σ
The mathematical analysis of this scheme is given in section 2.3.1. Nevertheless, let us rewrite the scheme, re-
marking that we can immediately eliminate uσ for Mσ = {K, L} from the relation (2.10). Indeed we get
dL,σ uK + dK,σ uL
uσ = . (2.15)
dL,σ + dK,σ
Hence we can identify the set of all elements of XD,0 , such that the condition (2.15) is ensured, with the set
HM (Ω), defined in Definition 1.1 as the set of the piecewise constant functions in each control volume. Reporting
the above expression of uσ given by (2.15) in (2.14), we get
δK,σ u
FK,σ (u) = −m(σ) , ∀K ∈ M, ∀σ ∈ EK , (2.16)
dσ
where we define, for all u ∈ HM (Ω)
δK,σ u = uL − uK , ∀σ ∈ Eint , Mσ = {K, L},
(2.17)
δK,σ u = −uK , ∀σ ∈ Eext , Mσ = {K},
and where we recall that
dσ = dL,σ + dK,σ , ∀σ ∈ Eint , Mσ = {K, L},
dσ = dK,σ , ∀σ ∈ Eext , Mσ = {K}.
Then the solution of (2.8) is also solution of the classical 2-point flux finite volume scheme studied for example in
[20]: find u ∈ HM (Ω), such that
X Z
δK,σ u
− m(σ) = f (x)dx, ∀K ∈ M.
dσ K
σ∈EK
Recall that the above scheme also writes

u ∈ HM (Ω), Z
X X δK,σ u δK,σ v
m(σ)dK,σ = f (x)v(x)dx, ∀v ∈ HM (Ω). (2.18)
dσ dσ Ω
K∈M σ∈EK
The case of harmonic averaging of an isotropic tensor

We assume in this section the hypothesis (2.2) and that
Λ(x) = λ(x)Id where Id is the identity application of Rd
with λ ∈ L∞ (Ω) such that there exists λ0 ∈ (0, +∞) with (2.19)
λ0 ≤ λ(x) for a.e. x ∈ Ω.
We consider in this section the so-called “harmonic averaging” case. Let D be an admissible ∆−adapted dis-
cretization of Ω in the sense of Definition 1.4. The finite volume approximation to Problem (2.1) is given as the
solution u = ((uK )K∈M , (uσ )σ∈E ) ∈ XD,0 of (2.8)-(2.9)-(2.10), in which we define the numerical flux by:
uσ − uK
FK,σ (u) = −m(σ)λK , ∀K ∈ M, ∀σ ∈ EK . (2.20)
dK,σ
12
In (2.20), we set Z
1
λK = λ(x)dx, ∀K ∈ M.
m(K) K
Note that, eliminating uσ for all σ ∈ Eint leads to find u ∈ HM (Ω) such that
X Z
δK,σ u
− λσ m(σ) = f (x)dx, ∀K ∈ M, (2.21)
dσ K
σ∈EK
denoting by λσ , for all σ ∈ E, the harmonic averaged value defined by:

λK λL
dK,σ dL,σ
λσ = d σ , ∀σ ∈ Eint , Mσ = {K, L}
λK
+ dλL,σ
L (2.22)
dK,σ
λσ = λK , ∀σ ∈ Eext , Mσ = {K}.
Then the solution of (2.21) is also solution of
uD ∈ HM (Ω), Z
X X δK,σ u δK,σ v
λσ m(σ)dK,σ = f (x)v(x)dx, ∀v ∈ HM (Ω). (2.23)
dσ dσ Ω
K∈M σ∈EK
Elimination of uσ in the general case

In the case where the relation (2.19) does not hold, we are then led to look for an expression of FK,σ (u) depending
on uK and all (uσ0 )σ0 ∈EK . In this case, it is no longer possible to locally eliminate uσ for all σ ∈ Eint , only using
(2.10).
Remark 2.2 Note that in the case of regular simplicial conforming meshes (triangles in 2D, tetrahedra in 3D),
there is an algebraic possibility to express the unknowns (uσ )σ∈E as local affine combinations of the values
(uK )K∈M [35] and therefore to eliminate them. The idea is to remark that the linear system constituted by
the equations (2.8) for all K ∈ MS , where MS is the set of all simplices sharing the same interior vertex S,
and (2.10) for all the interior edges such that Mσ ⊂ MS , presents as many equations as unknowns uσ , for
σ ∈ ∪K∈MS EK . Indeed, the number of edges in ∪K∈MS EK such that Mσ 6⊂ MS is equal to the number of
control volumes in MS . Unfortunately, there is at this time no general result on the invertibility nor the symmetry
of the matrix of this system, and this method does not apply to other types of meshes than the simplicial ones.
We may then choose to use the weak discrete form (2.12) as an approximation of the bilinear form a(·, ·), but with
a space of dimension smaller than that of XD,0 . This can be achieved by expressing the value of u on any interior
interface σ ∈ Eint as a consistent barycentric combination of the values uK :
X
uσ = βσK uK , (2.24)
K∈M
where (βσK ) K∈M is a family of real numbers, with βσK 6= 0 only for some control volumes K close to σ, and such
σ∈Eint
that X X
βσK = 1 and xσ = βσK xK , ∀σ ∈ Eint . (2.25)
K∈M K∈M
We recall that the values uσ , σ ∈ EP
ext are set to 0 in order to respect the boundary conditions. This ensures that if
ϕ is a regular function, then ϕσ = K∈M βσK ϕ(xK ) is a consistent approximation of ϕ(xσ ) for σ ∈ Eint . Hence
the new scheme reads:
 X

 Find u ∈ X D,0 such that u σ = βσK uK , ∀σ ∈ Eint , and

X Z K∈M
X (2.26)

 βσK vK , ∀σ ∈ Eint .
 hu, viF = vK f (x)dx, for all v ∈ XD,0 with vσ =
K∈M K K∈M
13
This method has been shown in [25] to be efficient in the case of a problem where Λ = Id (for the approximation of
the viscous terms in the Navier-Stokes problem). Unfortunately, because of a poor approximation of the local flux
at strongly hetereogenous interfaces, this approach is not sufficient to provide accurate results for some types of
flows in heterogeneous media, as we shall show in section 2.2. This is especially true when using coarse meshes, as
is often the case in industrial problems. Therefore it is possible to take advantage of both techniques: we shall use
equation (2.12) and keep the unknowns uσ on the edges which require them, for instance those where the matrix
Λ is discontinuous: hence (2.10) will hold for all edges associated to these unknowns; for all other interfaces, we
shall impose the values of u using (2.24), and therefore eliminate these unknowns. Let us decompose the set Eint
of interfaces into two non intersecting subsets, that is: Eint = A ∪ Ehyb , Ehyb = Eint \ A. The interface unknowns
associated with A will be computed by using the barycentric formula (2.24).
Remark 2.3 Note that, although the accuracy of the S scheme is increased in practice when the points where the
matrix Λ is discontinuous are located within the set σ∈Ehyb σ, such a property is not needed in the mathematical
study of the scheme.
Let us introduce the space XD,A ⊂ XD,0 defined by:
XD,A = {v ∈ XD such that vσ = 0 for all σ ∈ Eext and vσ satisfying (2.24) for all σ ∈ A}. (2.27)
The composite scheme which we consider in this work reads:

½
Find u ∈ XPD,A such that:
R (2.28)
hu, viF = K∈M vK K f (x)dx, for all v ∈ XD,A .
We therefore obtain a scheme with card(M) + card(Ehyb ) equations and unknowns. It is thus less expensive while
it remains precise (for the choice of numerical flux given below) even in the case of strong heterogeneities (see
section 2.2).
Note that with the present scheme, (2.10) holds for all σ ∈ Ehyb , but not generally for any σ ∈ A. However, fluxes
between pairs of control volumes can nevertheless be identified. These pairs are no longer necessarily connected
by a common boundary, but determined by the stencil used in relation (2.24).
Remark 2.4 (Other boundary conditions) In the case of Neumann or Robin boundary conditions, the discrete
space XD,A is modified to include the unknowns associated to the corresponding edges, and the resulting discrete
weak formulation is then straightforward.
Remark 2.5 (Extension of the scheme) There is no additional difficulty to replace (2.24) in the definition of
(2.27) by X X 0
uσ = βσK uK + βσσ uσ0 , ∀σ ∈ A,
K∈M σ 0 ∈Ehyb
with X X 0 X X 0
βσK + βσσ = 1 and xσ = βσK xK + βσσ xσ0 , ∀σ ∈ A.
K∈M σ 0 ∈Ehyb K∈M σ 0 ∈Ehyb
Then all the mathematical properties shown below still hold.
Let us apply these principles in two situations. The first one is the case of an anisotropic tensor and of a ∆−adapted
mesh, for which we shall be able to get back a standard finite volume scheme. The second one is that of an
anisotropic tensor and of a general pointed mesh.
The case of an anisotropic tensor and of a ∆−adapted mesh

Under hypotheses (2.2)-(2.4), let D be a ∆−adapted finite volume discretization of Ω in the sense of Definition
1.4. For the sake of simplicity, we set A = Eint , consider the space XD,Eint ⊂ XD,0 defined by:
dK,σ vL + dL,σ vK
XD,Eint = {v ∈ XD such that vσ = 0 for all σ ∈ Eext and vσ = for all σ ∈ Eint }. (2.29)
dK,σ + dL,σ
14
Indeed, it is possible, in the particular case of a ∆−adapted finite volume discretization of Ω, to use the property
that xσ is located, for Mσ = {K, L}, on the line (xK , xL ), and therefore, express (2.24) using only the two
values uK and uL .
We then introduce a discrete gradient ∇D : XD,Eint → HM (Ω)d , given by constant values in each control
volume: X
1 uσ − uK
∇K u = m(σ)(xσ − xK ) , ∀K ∈ M, ∀u ∈ XD,Eint . (2.30)
m(K) dK,σ
σ∈EK
A first natural scheme, inspired by the finite element framework, would be

u ∈ XD,Eint ,
Z Z
Λ(x)∇D u(x) · ∇D v(x)dx = f (x)ΠM v(x)dx, ∀v ∈ XD,Eint .
Ω Ω
(recall that ΠM v is defined in Definition 1.1). But this scheme is not convenient, since it is possible to get ∇D u = 0
with u 6= 0. Hence a stabilization is necessary. To this purpose, we define an element RD u ∈ LD (Ω), given by
the expression
p µ ¶
uσ − uK
RK,σ u = αK d − ∇K u · nK,σ , ∀K ∈ M, ∀σ ∈ EK , ∀u ∈ HM (Ω). (2.31)
dK,σ
where (αK )K∈M is a family of given positive real values.
The modified finite volume approximation to Problem (2.1) is given, for a given family of positive reals (αK )K∈M ,
as the solution of the following equation:

 u Z ∈ XD,Eint , Z
(2.32)
 (Λ(x)∇D u(x) · ∇D v(x) + RD u(x)RD v(x)) dx = f (x)ΠM v(x)dx, ∀v ∈ XD,Eint .
Ω Ω
Remark 2.6 In the case Λ = Id, if D is a ∆−superadapted discretization in the sense of definition 1.5 (recall that
it means xσ − xK = dK,σ nK,σ , for all K ∈ M and σ ∈ EK ) and αK = 1 for all K ∈ M, then the scheme (2.32)
gives back the scheme provided above for the Laplace case. Indeed, we get that
X X m(σ)dK,σ µ uσ − uK ¶µ
vσ − vK
¶
m(DK,σ )RK,σ uRK,σ v = d − ∇K u · nK,σ − ∇K v · nK,σ ,
d dK,σ dK,σ
σ∈EK σ∈EK
which leads to
X X uσ − uK vσ − vK
m(DK,σ )RK,σ uRK,σ v = m(σ)dK,σ − m(K)∇K u · ∇K v,
dK,σ dK,σ
σ∈EK σ∈EK
P P
using σ∈EK m(σ)dK,σ uσdK,σ
−uK
nK,σ = m(K)∇K u and σ∈EK m(σ)dK,σ nK,σ ntK,σ = m(K)Id.
In the sequel, we shall refer to the scheme (2.32) as the gradient scheme on adapted meshes, since it is obtained
from a discretization formula on the gradient, the consistence of which results from the orthogonality property of
the mesh.
We may note that scheme (2.32) leads to
X³
m(K)ΛK ∇K u · ∇K v
K∈M Z
X ¡ uσ − uK ¢¡ vσ − vK ¢´ X
+αK m(σ)dK,σ − ∇K u · nK,σ − ∇K v · nK,σ = vK f (x)dx,
dK,σ dK,σ K
σ∈EK K∈M
setting Z
1
ΛK = Λ(x)dx.
m(K) K
15
This expression of the scheme can then be put under the form
X X X X Z
0
AσK σ (uσ − uK )(vσ − vK ) = vK f (x)dx,
K∈M σ∈EK σ 0 ∈EK K∈M K
0
where the matrix (AσK σ )σ0 ,σ∈EK is symmetric. Let us define
X 0
FK,σ (u) = − AσK σ (uσ0 − uK ), ∀σ ∈ EK , ∀K ∈ M. (2.33)
σ 0 ∈EK
Taking for v ∈ XD,Eint , vK = 1 and vL = 0 for all L ∈ M \ {K}, we obtain that the above equation is equivalent
to finding the values (uK )K∈M , solution of the following system of equations:
X Z
b
FK,σ (u) = f (x)dx, ∀K ∈ M, (2.34)
σ∈EK K
where
dK,σ dL,σ
FbK,σ (u) = FK,σ (u) − FL,σ (u), ∀σ ∈ Eint , Mσ = {K, L} (2.35)
dK,σ + dL,σ dK,σ + dL,σ
and
FbK,σ (u) = FK,σ (u), ∀σ ∈ Eext , Mσ = {K}. (2.36)
This is indeed a finite volume scheme (in the sense of the respect of the local balance), since
FbK,σ (u) = −FbL,σ (u), ∀σ ∈ Eint , Mσ = {K, L}.
Gradient schemes on general non matching grids

We now finally relax the assumption that it was possible to mesh the domain using a ∆−adapted mesh: indeed,
actual problems involve more and more often general grids, which do not satisfy this orthogonality condition. This
is for instance the case in subsurface flow modelling, where parallelepipedic (or quadrangular in 2D) meshes are
widely used to mesh the underground layers. In the case of geologigal faults for instance, the independent meshing
of geological layers results in non conforming meshes (see e.g. Figure 2.1) which are clearly not ∆−adapted
meshes.
Figure 2.1: Non conforming subsurface flow modelling
Hence we consider that D = (M, E, P) is only a discretization of Ω in the sense of definition 1.2, and we then
denote by xσ the center of gravity of σ ∈ E. As we did in the case of an anisotropic diffusion, on ∆−adapted
meshes, we again identify the numerical fluxes FK,σ (u) through the mesh dependent bilinear form h·, ·iF defined
in (2.12), using the expression of a discrete gradient. Indeed let us assume that, for all u ∈ XD , we have constructed
a discrete gradient ∇D u, we then seek a family (FK,σ (u)) K∈M such that
σ∈EK
X X Z
FK,σ (u)(vK − vσ ) = ∇D u(x) · Λ(x)∇D v(x)dx, ∀u, v ∈ XD . (2.37)
K∈M σ∈EK Ω
16
Remark 2.7 It is then always possible to deduce an expression for FK,σ (u) satisfying (2.37), under the suffi-
cient condition that, for all K ∈ M and a.e. x ∈ K, ∇D u(x) may be expressed as a linear combination of
(uσ − uK )σ∈EK , the coefficients of which are measurable bounded functions of x. This property is ensured in the
construction of ∇D u(x) given below.
Then, in order to ensure the desired properties 2 and 3, we shall see in Section 2.3.2 that it suffices that the discrete
gradient satisfies the following properties.
1. For a sequence of space discretisations of Ω with mesh size tending to 0, if the sequence of associated grid
functions is bounded in some sense, then their discrete gradient is expected to converge, at least weakly in
L2 (Ω)d , to the gradient of an element of H01 (Ω);
2. If ϕ is a regular function from Ω to R, the discrete gradient of the piecewise function defined by taking the
value ϕ(xK ) on each control volume K and ϕ(xσ ) on each edge σ is a consistent approximation of the
gradient of ϕ.
Le us first define: X
1
∇K u = m(σ)(uσ − uK )nK,σ , ∀K ∈ M, ∀u ∈ XD , (2.38)
m(K)
σ∈EK
where nK,σ is the outward to K normal unit vector, m(K) and m(σ) are the usual measures (volumes, areas, or
lengths) of K and σ. The consistency of formula (2.38) stems from the following geometrical relation:
X
m(σ)nK,σ (xσ − xK )t = m(K)Id, ∀K ∈ M, (2.39)
σ∈EK
where (xσ − xK )t is the transpose of xσ − xK ∈ Rd , and Id is the d × d identity matrix. Indeed, for any linear
function defined on Ω by ψ(x) = G · x with G ∈ Rd , assuming that uσ = ψ(xσ ) and uK = ψ(xK ), we get
uσ − uK = (xσ − xK )t G = (xσ − xK )t ∇ψ, hence (2.38) leads to ∇K u = ∇ψ.
Since the coefficient of uK in (2.38) is in fact equal to zero, a reconstruction of the discrete gradient ∇D u solely
based on (2.38) cannot lead to a definite discrete bilinear form in the general case. Hence, we now introduce:
∇K,σ u = ∇K u + RK,σ u nK,σ , (2.40)
with √
d
RK,σ u = (uσ − uK − ∇K u · (xσ − xK )) , (2.41)
dK,σ
(recall that d is the space dimension and dK,σ is the Euclidean distance between xK and σ). We may then define
∇D u as the piecewise constant function equal to ∇K,σ u a.e. in the cone DK,σ with vertex xK and basis σ:
∇D u(x) = ∇K,σ u for a.e. x ∈ DK,σ . (2.42)
We can then prove that the discrete gradient defined by (2.38)-(2.42) meets the required properties (see Lemmas
2.6 and 2.7). In order to identify the numerical fluxes FK,σ (u) using the relation (2.37), we put the discrete gradient
under the form X 0
∇K,σ u = (uσ0 − uK )y σσ ,
σ 0 ∈EK
with  √ µ ¶

 m(σ) d m(σ)
 nK,σ + 1− nK,σ · (xσ − xK ) nK,σ if σ = σ 0
σσ 0 m(K) dK,σ√ m(K)
y = (2.43)

 m(σ 0 ) d
 nK,σ0 − m(σ 0 )nK,σ0 · (xσ − xK )nK,σ otherwise .
m(K) dK,σ m(K)
Thus:
Z X X X 0
∇D u(x) · Λ(x)∇D v(x)dx = Aσσ
K (uσ − uK )(vσ 0 − vK ), ∀u, v ∈ XD , (2.44)
Ω K∈M σ∈EK σ 0 ∈EK
17
with:
X Z
0 00 00
σ0
Aσσ
K = yσ σ
· ΛK,σ00 y σ and ΛK,σ00 = Λ(x)dx. (2.45)
σ 00 ∈EK DK,σ00
0
Then we get that the local matrices (AσσK )σσ 0 ∈EK are symmetric and positive, and the identification of the numer-
ical fluxes using (2.37) leads to the expression:
X 0
FK,σ (u) = Aσσ
K (uK − uσ 0 ). (2.46)
σ 0 ∈EK
The properties provided by this definition (which could not have been obtained using natural expansions of regular
functions) are shown in Lemma 2.8. Then Theorem 2.1 shows that these properties are sufficient to provide the
convergence of the scheme. Note that the proof of this property holds for general heterogeneous, anisotropic and
possibly discontinuous fields Λ, for which the solution u of (2.5) is not in general more regular than u ∈ H01 (Ω).
The local consistency property provided by definition (2.46) is only detailed in the error estimate theorem 2.2, in
the case where Λ and u are regular enough.
√
Remark 2.8 The choice of the coefficient d in (2.41) is not compulsory, and any fixed positive value could be
substituted; it is motivated by the fact that it provides a diagonal matrix AK in the case of isotropic diffision
and of meshes which satisfy nK,σ = xσdK,σ −xK
(triangular, rectangular, orthogonal parallelepipedic meshes but
unfortunately not general tetrahedric meshes), and yields the usual two point scheme. In this case, the formula
(2.38) leads to the discrete gradient which was introduced in [22].
Mixed-type schemes on general non matching grids

We again consider that D = (M, E, P) is only a discretization of Ω in the sense of definition 1.2, and we then
denote by xσ the center of gravity of σ ∈ E. Our purpose is again to define a relation between the numerical
fluxes FK,σ (u) and the values uK , (uσ0 )σ0 ∈EK . But the idea is now, defining the space of the numerical fluxes YD
YD = {((FK,σ )σ∈EK )K∈M , FK,σ ∈ R}. (2.47)
to connect an element F ∈ YD with an element u ∈ XD,0 by

X X
hF, GiY = (uK − uσ )GK,σ , ∀G ∈ YD , (2.48)
K∈M σ∈EK
through a bilinar form h·, ·iY on YD . This bilinear form allows to consider the matrix MK defined by
0
σσ
MK = h1K,σ , 1K,σ0 iY ,
and then the bilinear form can be rewritten in the equivalent form
X X X
σσ 0
hF, GiY = MK FK,σ GK,σ0 .
K∈M σ∈EK σ 0 ∈EK
The scheme is then expressed by the relations (2.8)-(2.9)-(2.10) in addition to (2.48). Hence we must now define
a suitable bilinear form on YD (this problem is explored by the mixed finite element method or the mimetic finite
difference method [6]). Instead of defining general orthogonality properties in the space YD , we here consider an
idea similar to (2.40). Let us define, in the spirit of (2.39),
X
m(K)ΛK vK (F ) = − FK,σ (xσ − xK ), (2.49)
σ∈EK
with Z
1
ΛK = Λ(x)dx.
m(K) K
18
Note that the definition X

hF, GiY = m(K)ΛK vK (F ) · vK (G)
K∈M
leads to a non-invertible matrix. We then introduce the linear form

FK,σ
HK,σ (F ) = − − ΛK vK (F ) · nK,σ , (2.50)
m(σ)
and √ xσ − xK
vK,σ (F ) = vK (F ) + dHK,σ (F )Λ−1
K . (2.51)
dK,σ
We may then define vD F as the piecewise constant function equal to vK,σ (F ) a.e. in the cone DK,σ with vertex
xK and basis σ:
vD F (x) = vK,σ (F ) for a.e. x ∈ DK,σ , (2.52)
and we set, following the same idea as (2.37)
Z
hF, GiY = Λ(x)vD F (x) · vD G(x)dx. (2.53)
Ω
√
Remark 2.9 As we noticed for the gradient scheme, the choice of the coefficient d in (2.51) is not compulsory,
and any fixed positive value could be substituted; it is again motivated by the fact that it provides a diagonal
matrix MK in the case of isotropic diffision and of meshes which satisfy nK,σ = xσdK,σ−xK
(triangular, rectangular,
orthogonal parallelepipedic meshes but unfortunately not general tetrahedric meshes), and yields the usual two
point scheme. In this case, the formula (2.38) leads to the discrete gradient which was introduced in [22].
Remark 2.10 Let us consider the case where Λ is constant in each control volume K ∈ M. Note that, for a
slightly different definition of vD , we may find exactly a mixed finite element method. Indeed, let us define
x − xK
bK,σ (F )(x) = vK (F ) + HK,σ (F )Λ−1
v K , ∀x ∈ DK,σ , (2.54)
dK,σ
and
bD F (x) = v
v bK,σ (F )(x) for a.e. x ∈ DK,σ , ∀K ∈ M, ∀σ ∈ EK . (2.55)
This definition ensures the property
FK,σ
bK,σ (F )(x) · nK,σ = −
ΛK v , ∀x ∈ σ, ∀σ ∈ EK , ∀K ∈ M.
m(σ)
Therefore, under the condition (2.10), we get that ΛK v bD F ∈ Hdiv(Ω) (it is easy to see the continuity of the normal
component on the boundary of each cone DK,σ ). Defining VD = {ΛK v bD F, F ∈ YD such that (2.10) holds },
the scheme (2.8)-(2.9)-(2.10), in addition to (2.53) with (2.54) resumes to a classical mixed finite element method,
with an explicit basis for the fluxes, different from the Raviart-Thomas one on triangles or rectangles, since it can
be written find g ∈ VD and u ∈ HM (Ω) such that
Z Z
−1
g(x) · Λ (x)b g (x)dx = − u(x)divb
g (x)dx, ∀b g ∈ VD , (2.56)
Ω Ω
and Z Z
u
b(x)divg(x)dx = u
b(x)f (x)dx, ∀b
u ∈ HM (Ω). (2.57)
Ω Ω
19
2.2 Numerical results

We present some numerical results obtained with various choices of A in the scheme (2.28),(2.12) with the flux
(2.46), which we synthetize here for the sake of clarity:

 Find u ∈ XD,A (that is (uK )K∈M , (uσ )σ∈EZhyb ), such that:

 X X X

 FK,σ (u)(vK − vσ ) = vK f (x)dx, for all v ∈ XD,A ,
K∈M σ∈E K∈M K (2.58)


K X

 Aσσ
0
 with FK,σ (u) = K (uσ 0 − uK ), ∀K ∈ M, ∀σ ∈ EK .

σ 0 ∈EK
Order of convergence We consider here the numerical resolution of Equation (2.1) supplemented by a homoge-
neous Dirichlet boundary condition; the right hand side is chosen so as to obtain an exact solution to the problem,
so as to easily compute the error between the exact and approximate solutions. We consider Problem (2.1) with a
constant matrix Λ: µ ¶
1.5 .5
Λ= , (2.59)
.5 1.5
and choose f : Ω → R and f such that the exact solution to Problem (2.1) is ū : Ω → R defined by ū(x, y) =
16x(1 − x)y(1 − y) for any (x, y) ∈ Ω. Note that in this case, the composite scheme is in fact the cell centred
scheme, there are no edge unknowns. (2.24).
Let us first consider conforming meshes, such as the triangular meshes which are depicted on Figure 2.2 and
uniform square meshes.
Figure 2.2: Regular conforming coarse and fine triangular grids
For both A = ∅ (hybrid scheme) and A = Eint (cell centred scheme), the order of convergence is close to 2 for
the unknown u and 1 for its gradient. Of course, the hybrid scheme is almost three times more costly in terms of
number of unknowns than the cell centred scheme for a given precision. However, the number of nonzero terms in
the matrix is, again for a given precision on the approximate solution, larger for the cell centred scheme than for
the hybrid scheme. Hence the number of unknowns is probably not a sufficient criterion for assessing the cost of
the scheme.
Results were also obtained in the case of uniform square or rectangular meshes. They show a better rate of
convergence of the gradient (order 2 in the case of the hybrid scheme and 1.5 in the case of the centred scheme),
even though the rate of convergence of the approximate solution remains unchanged and close to 2.
We then use a rectangular nonconforming mesh, obtained by cutting the domain in two vertical sides and using a
rectangular grid of 3n × 2n (resp. 5n × 2n) on the first (resp. second side), where n is the number of the mesh,
n = 1, . . . 7. . Again, the order of convergence which we obtain is 2 for u and around 1.8 for the gradient. We
give in Table 2.1 below the errors obtained in the discrete L2 norm for u and ∇u for a nonconforming mesh and
(in terms of number of unknowns) and for the rectangular 4 × 6 and 4 × 10 conforming rectangular meshes, for
both the hybrid and cell centred schemes. We show on Figure 2.3 the solutions for the corresponding grids (which
looks very much the same for both schemes).
20
NU NM ²(u) ²(∇u)
n Hyb Cent Hyb Cent Hyb Cent Hyb Cent
C1 130 48 874 488 1.28E-01 1.20E-01 1.64E-02 3.57E-02
NC 182 64 1334 724 1.03E-01 9.43E-02 1.66E-02 3.69E-02
C2 222 80 1542 864 7.61E-02 7.09E-02 9.18E-03 2.44E-02
Table 2.1: Error for the non conforming rectangular mesh, hybrid scheme (Hyb) and centred (Cent) schemes. For
both schemes: NU is the number of unknowns in the resulting linear system, NM the number of non zero terms in
the matrix, ²(u) the discrete L2 norm of the error on the solution and ²(∇u) the discrete L2 norm of the error on
the gradient. C1 and C2 are the two conforming meshes represented on the left and the right in Figure 2.3, and NC
the non conforming one represented in the middle.
Figure 2.3: The approximate solution for conforming and nonconforming meshes. Left: conforming 8 × 6 mesh,
center: non conforming 4 × 6, 4 × 10 mesh, right: conforming 10 × 10.
Further detailed results on several problems and conforming, non conforming and distorted meshes may be found
in [23].
The case of a highly heterogeneous tilted barrier
We now turn to a heterogeneous case. The domain Ω =]0, 1[×]0, 1[ is composed of 3 subdomains, which are
depicted in Figure 2.4: Ω1 = {(x, y) ∈ Ω; ϕ1 (x, y) < 0}, with ϕ1 (x, y) = y − δ(x − .5) − .475, Ω2 = {(x, y) ∈
Ω; ϕ1 (x, y) > 0, ϕ2 (x, y) < 0}, with ϕ2 (x, y) = ϕ1 (x, y) − 0.05, Ω3 = {(x, y) ∈ Ω; ϕ2 (x, y) > 0}, and δ = 0.2
is the slope of the drain (see Figure 2.4). Dirichlet boundary conditions are imposed by setting the boundary values
to those of the analytical solution given by u(x, y) = −ϕ1 (x, y) on Ω1 ∪ Ω3 and u(x, y) = −ϕ1 (x, y)/10−2 on
Ω2 .
The permeability tensor Λ is heterogeneous and isotropic, given by Λ(x) = λ(x)Id, with λ(x) = 1 for a.e.
x ∈ Ω1 ∪ Ω3 and λ(x) = 10−2 for a.e. x ∈ Ω2 . Note that the isolines of the exact solution are parallel to the
boundaries of the subdomain, and that the tangential component of the gradient is 0. We use the meshes depicted
in figure 2.4. Mesh 3 (containing 10 × 25 control volumes) is obtained from Mesh 1 by the addition of two layers
of very thin control volumes around each of the two lines of discontinuity of Λ: because of the very low thickness
of these layers, equal to 1/10000, the picture representing Mesh 3 is not different from that of Mesh 1.
We get the following results for the approximations of the four fluxes at the boundary.
nb. unknowns matrix size x=0 x=1 y=0 y=1

analytical -0.2 0.2 1. -1.
mesh 1 210 2424 -1.17 1.17 3.51 -3.51
centred mesh 2 1000 11904 -0.237 0.237 1.104 -1.104
mesh 3 250 2904 -0.208 0.208 1.02 -1.02
mesh 1 239 2583 -0.2 0.2 1. -1.
composite
mesh 2 1020 12036 -0.2 0.2 1. -1.
mesh 1 599 4311 -0.2 0.2 1. -1.
hybrid
mesh 2 2890 21138 -0.2 0.2 1. -1.
21
Ω1
Ω2
Ω3
Figure 2.4: Domain and meshes used for the tilted barrier test: mesh 1 (10 × 21 center), mesh 2 (10 × 100 right)
Note that the values of the numerical solution given by the hybrid and composite schemes are equal to those of the
analytical one (this holds under the only condition that the interfaces located on the lines ϕi (x, y) = 0, i = 1, 2
are not included in A, and that, for all σ ∈ A, all K ∈ M with βσK 6= 0 are included in the same subdomain Ωi ).
Note that Mesh 3, which leads to acceptable results for the computation of the fluxes, is not well designed for such
a coupled problem, because of too small measures of control volumes. Hence the composite method on Mesh 1
appears to be the most suitable method for this problem.
Tilted barrier
The domain Ω =]0, 1[×]0, 1[ is composed of 3 subdomains: Ω1 = {(x, y) ∈ Ω; ϕ1 (x, y) < 0}, with ϕ1 (x, y) =
y − δ(x − .5) − .475, Ω2 = {(x, y) ∈ Ω; ϕ1 (x, y) > 0, ϕ2 (x, y) < 0}, and ϕ2 (x, y) = ϕ1 (x, y) − 0.05,
Ω3 = {(x, y) ∈ Ω; ϕ2 (x, y) > 0}, and δ = 0.2 is the slope of the drain. We impose, at the boundary of the
domain, the values of the analytical solution given by u(x, y) = −ϕ1 (x, y) on Ω1 , u(x, y) = −ϕ1 (x, y)/10−2 on
Ω2 , u(x, y) = −ϕ2 (x, y) − 0.05/10−2 on Ω2 .
The permeability tensor Λ is heterogeneous and isotropic, given by Λ(x) = λ(x)Id, with λ(x) = 1 for a.e.
x ∈ Ω1 , λ(x) = 10−2 for a.e. x ∈ Ω2 and λ(x) = 1 for a.e. Ω3 . Mesh 3 (containing 10 × 25 control volumes)
is obtained from Mesh 1 by the addition of two layers of very thin control volumes around each of the two lines
of discontinuity of Λ: because of the very low thickness of these layers, equal to 1/10000, the picture representing
Mesh 3 is not different from that of Mesh 1.
We get the following results for the approximations of the four fluxes at the boundary.
nb. unknowns matrix size x=0 x=1 y=0 y=1

analytical -0.2 0.2 1. -1.
Ebar = Eint (mesh 1) 210 2424 -1.17 1.17 3.51 -3.51
Ebar = Eint (mesh 2) 1000 11904 -0.296 0.296 1.27 -1.27
Ebar = Eint (mesh 3) 250 2904 -0.208 0.208 1.02 -1.02
∅ ( Ebar ( Eint (mesh 1) 239 2583 -0.2 0.2 1. -1.
Ebar = ∅ (mesh 1) 599 4311 -0.2 0.2 1. -1.
Note that in the cases ∅ ( Ebar ( Eint and Ebar = ∅, the numerical solution is equal to the analytical one (this
holds under the only conditions that the interfaces located on the lines ϕi (x, y) = 0, i = 1, 2 are not included in
Ebar , and that, for all σ ∈ Ebar , all K ∈ Nσ are included in the same subdomain Ωi ). Note that Mesh 3, which
leads to acceptable results for the computation of the fluxes, is not well designed for such a coupled problem,
because of too small measures of control volumes. Hence the partly hybrid method appears to be the most suitable
method for this problem and this mesh.
2.3 Further analysis

We now give the abstract properties of the discrete fluxes which are sufficient to prove the convergence of the
schemes provided in this chapter. Let us first introduce some notations related to the mesh. Let D = (M, E, P) be
22
a discretization of Ω in the sense of definition 1.2. The size of the discretization D is defined by:
hD = sup{diam(K), K ∈ M},
and the regularity of the mesh by:

µ ¶
dK,σ hK
θD = max max , max . (2.60)
σ∈Eint ,K,L∈Mσ dL,σ K∈M,σ∈E K dK,σ
For a given set A ⊂ Eint and for a given family (βσK ) K∈M satisfying property (2.25), we introduce some measure
σ∈Eint
of the resulting regularity with
Ã P !
L∈M |βσL ||xL − xσ |2
θD,A = max θD , max . (2.61)
K∈M,σ∈EK ∩A h2K
Definition 2.2 (Continuous, coercive, consistent and symmetric families of fluxes) Let F be a family of dis-
D
cretizations in the sense of definition 1.2. For D = (M, E, P) ∈ F , K ∈ M and σ ∈ E, we denote by FK,σ a
D
linear mapping from XD to R, and we denote by Φ = ((FK,σ )K∈M )D∈F . We consider the bilinear form defined
σ∈E
by X X
D 2
hu, viF = FK,σ (u)(vK − vσ ), ∀(u, v) ∈ XD . (2.62)
K∈M σ∈EK
The family of numerical fluxes Φ is said to be continuous if there exists M > 0 such that
2
hu, viF ≤ M |u|X |v|X , ∀(u, v) ∈ XD , ∀D = (M, E, P) ∈ F. (2.63)
The family of numerical fluxes Φ is said to be coercive if there exists α > 0 such that
α|u|2X ≤ hu, uiF , ∀u ∈ XD , ∀D = (M, E, P) ∈ F. (2.64)
The family of numerical fluxes Φ is said to be consistent (with Problem (2.1)) if for any family (uD )D∈F satisfying:
• uD ∈ XD,0 for all D ∈ F,
• there exists C > 0 with |uD |X ≤ C for all D ∈ F,
• there exists u ∈ L2 (Ω) with lim kΠM uD − ukL2 (Ω) = 0 (we get from Lemma 5.7 that u ∈ H01 (Ω)),
hD →0
then the following holds

Z
lim huD , PD ϕiF = Λ(x)∇ϕ(x) · ∇u(x)dx, ∀ϕ ∈ Cc∞ (Ω). (2.65)
hD →0 Ω
Finally the family of numerical fluxes Φ is said to be symmetric if

2
hu, viF = hv, uiF , ∀(u, v) ∈ XD , ∀D = (M, E, P) ∈ F.
We may now state the general convergence theorem.
Theorem 2.1 Let F be a family of discretizations in the sense of definition 1.2; for any D ∈ F, let A ⊂ Eint and
(βσK ) K∈M satisfying property (2.25). We assume that there exists θ > 0 such that θD,A ≤ θ, for all D ∈ F, where
σ∈Eint
D
θD,A is defined by (2.61). Let Φ = ((FK,σ )K∈M )D∈F be a continuous, coercive and symmetric and consistent
σ∈E
family of numerical fluxes in the sense of definition 2.2. Let (uD )D∈F be the family of functions solution to (2.28)
for all D ∈ F. Then ΠM uD converges in L2 (Ω) to the unique solution u of (2.5) as hD → 0. Moreover ∇D uD
converges to ∇u in L2 (Ω)d as hD → 0.
23
Proof. We let v = uD in (2.28), we apply the Cauchy-Schwarz inequality to the right hand side. We get
Z
huD , uD iD = f (x)ΠM uD (x)dx ≤ kf kL2 (Ω) kΠM uD kL2 (Ω) .
Ω
We apply the Sobolev inequality (5.13) with p = 2, which gives in this case kΠM uD kL2 (Ω) ≤ C1 kΠM uk1,2,M .
Using (2.84) and the consistency of the family Φ of fluxes, we then have
α|ΠM uD |2X ≤ C1 kf kL2 (Ω) |uD |X .
This leads to the inequality

C1
kuk1,2,M ≤ |uD |X ≤ kf kL2 (Ω) . (2.66)
α
Thanks to lemma 5.7, we get the existence of u ∈ H01 (Ω), and of a subfamily extracted from F, such that
kΠM uD − ukL2 (Ω) tends to 0 as hD → 0. For a given ϕ ∈ Cc∞ (Ω), let us take v = PD,A ϕ in (2.28) (recall that
PD,A ϕ ∈ XD,A ). We get Z
huD , PD,A ϕiF = f (x)PM ϕ(x)dx.
Ω
Let us remark that, thanks to the continuity of the family Φ of fluxes, we have
C1
huD , PD,A ϕ − PD ϕiF ≤ M kf kL2 (Ω) |PD,A ϕ − PD ϕ|X .
α
Thanks to (2.25) and (2.61), we get the existence of Cϕ , only depending on ϕ such that, for all K ∈ M and all
σ ∈ A ∩ EK , X X
| βσL ϕ(xL ) − ϕ(xσ )| ≤ |βσL ||xL − xσ |2 Cϕ ≤ θD,A Cϕ h2K . (2.67)
L∈M L∈M
We can then deduce

lim |PD,A ϕ − PD ϕ|X = 0. (2.68)
hD →0
Thanks to the properties of subfamily extracted from F, we can apply the consistency hypothesis on the family Φ
of fluxes, which gives Z
lim huD , PD ϕiF = Λ(x)∇ϕ(x) · ∇u(x)dx.
hD →0 Ω
Gathering the two above results leads to
Z
lim huD , PD,A ϕiF = Λ(x)∇ϕ(x) · ∇u(x)dx,
hD →0 Ω
which concludes the proof that

Z Z
Λ(x)∇ϕ(x) · ∇u(x)dx = f (x)ϕ(x)dx.
Ω Ω
Therefore, u is the unique solution of (2.5), and we get that the whole family (uD )D∈F converges to u as hD → 0.
Let us now prove the second part of the theorem. Let ϕ ∈ C∞ c (Ω) be given (this function is devoted to approximate
u in H01 (Ω)). Thanks to the Cauchy-Schwarz inequality, we have
Z
|∇D uD (x) − ∇u(x)|2 dx ≤ 3 (T1D + T2D + T3 )
Ω
where Z
T1D = |∇D uD (x) − ∇D PD ϕ(x)|2 dx,
Ω
24
Z
T2D = |∇D PD ϕ(x) − ∇ϕ(x)|2 dx,
Ω
and Z
T3 = |∇ϕ(x) − ∇u(x)|2 dx.
Ω
We have, thanks to Lemma 2.7,
lim T2D = 0. (2.69)
hD →0
We have, thanks to Lemma 2.5 and to the coercivity of the family of fluxes, that there exists C2 such that
k∇D vk2L2 (Ω)d ≤ C12

2
|v|2X ≤ C2 hv, viF , ∀v ∈ XD ,
2
C12
with C2 = α . Taking v = uD − PD ϕ, we have
T1D ≤ C2 (huD , uD iF − 2huD , PD ϕiF + hPD ϕ, PD ϕiF ).
Using the result of convergence proved for uD and the consistency of the family of fluxes, we get
Z
lim huD , PD ϕiF = ∇u(x) · Λ(x)∇ϕ(x)dx. (2.70)
hD →0 Ω
The fact that |PD ϕ|X remains bounded, results from the regularity of ϕ and the regularity hypotheses of the family
of discretizations. Hence we can use the consistency of the family of fluxes, which writes in this case
Z
lim hPD ϕ, PD ϕiF = ∇ϕ(x) · Λ(x)∇ϕ(x)dx. (2.71)
hD →0 Ω
R
Remarking that passing to the limit hD → 0 in (2.28) with v = uD provides that huD , uD iF converges to Ω
∇u ·
Λ∇udx, we get that
Z Z
lim huD − PD ϕ, uD − PD ϕiF = ∇(u − ϕ) · Λ∇(u − ϕ)dx ≤ λ |∇u − ∇ϕ|2 dx,
hD →0 Ω Ω
which yields Z
lim sup T1D ≤ C2 λ |∇u − ∇ϕ|2 dx.
hD →0 Ω
From the above results, we obtain that there exists C3 , independent of D, such that
Z Z
2
|∇D uD (x) − ∇u(x)| dx ≤ C3 |∇ϕ(x) − ∇u(x)|2 dx + T4D ,
Ω Ω
with (noting that ϕ is fixed)

lim T4D = 0. (2.72)
hD →0
R
Let ε > 0. We can choose ϕ such that Ω |∇ϕ(x) − ∇u(x)|2 dx ≤ ε, and we can then choose hD small enough
such that T4D ≤ ε. This completes the proof that
Z
lim |∇D uD (x) − ∇u(x)|2 dx = 0 (2.73)
hD →0 Ω
in the case of a general continuous, coercive, consistent and symmetric family of fluxes. ¤
Let us write an error estimate, in the particular case that Λ = Id and that the solution of (2.5) is regular enough.
25
Theorem 2.2 (Error estimate, isotropic case) We consider the particular case Λ = Id, and we assume that the
solution u ∈ H01 (Ω) of (2.5) is in C 2 (Ω). Let D = (M, E, P) be a discretization in the sense of definition 1.2, let
A ⊂ Eint be given, let A = (βσK )σ∈A,K∈M be a family of real numbers such that (2.25) holds, and let θ ≥ θD,A
be given (see (2.61)). Let (FK,σ )K∈M,σ∈E be a family of linear mappings from XD to R, such that there exists
α > 0 with
α|u|2X ≤ hu, uiF , ∀u ∈ XD , (2.74)
defining hu, viF by (2.62). We denote by
Ã !1/2
X X dK,σ µ Z ¶2
Ru = FK,σ (PD,A u) + ∇u(x) · nK,σ dγ(x) . (2.75)
m(σ) σ
K∈M σ∈EK
Then the solution uD of (2.28) verifies that there exists C4 , only depends on α and on θ, such that
kΠM uD − PM ukL2 (Ω) ≤ C4 Ru, (2.76)
and verifies that there exists C5 , only depending on α, θ and u such that
k∇D uD − ∇ukL2 (Ω)d ≤ C5 (Ru + hD ) . (2.77)
Remark 2.11 The extension of Theorem 2.2 to the case u ∈ H 2 (Ω) could be studied in the case d = 2 or d = 3.
Such a study, which demands a rather longer and more technical proof, is not expected to provide more information
on the link between accuracy and the regularity of the mesh than the result presented here.
Proof. Let v ∈ XD , since −∆u = f , we get:

X Z Z
− vK ∆u(x)dx = f (x)ΠM v(x)dx. (2.78)
K∈M K Ω
Thanks to the following equality (recall that u ∈ C 2 (Ω) and therefore ∇u · nK,σ is defined on each edge σ)
X Z X X Z
− vK ∆u(x)dx = − (vK − vσ ) ∇u(x) · nK,σ dγ(x),
K∈M K K∈M σ∈EK σ
we get that
Z X X µ Z ¶
D
hPD,A u, viF = f (x)ΠM v(x)dx + FK,σ (PD,A u) + ∇u(x) · nK,σ dγ(x) (vK − vσ ).
Ω K∈M σ∈EK σ
Taking v = PD,A u − uD ∈ XD,A in this latter equality and using (2.78) we get
X X µ Z ¶
D
hv, viF = FK,σ (PD,A u) + ∇u(x) · nK,σ dγ(x) (vK − vσ ),
K∈M σ∈EK σ
which leads, using (2.74) and the Cauchy-Schwarz inequality, to
α|v|X ≤ Ru. (2.79)
Using (2.84) and the Sobolev inequality (5.13) with p = 2 provides the conclusion of (2.76). Let us now prove
(2.77). We have
k∇D uD − ∇ukL2 (Ω)d ≤ k∇D uD − ∇D PD,A ukL2 (Ω)d + k∇D PD,A u − ∇ukL2 (Ω)d .
The bound of the first term in the above right hand side is bounded thanks to Lemma 2.5 and (2.79). The inequality
k∇D PD,A u − ∇ukL2 (Ω)d ≤ C6 hD is obtained thanks to Lemma 2.7 and using a similar inequality to (2.67),
replacing ϕ by u. ¤
26
2.3.1 The convergence proof in the (harmonic averaging) isotropic case

Lemma 2.1 [Properties of the numerical fluxes in the harmonic averaging scheme for isotropic diffusion,
∆−adapted mesh]
Under hypothesis (1.5) and (2.19), let F be a family of pointed ∆−adapted polygonal finite volume space dis-
D
cretization of Ω in the sense of Definition 1.4. Let Φ = ((FK,σ )K∈M )D∈F be defined by (2.20). Then Φ is a
σ∈E
continuous, coercive and symmetric and consistent family of numerical fluxes in the sense of definition 2.2.
Proof. The only nonobvious property is the consistency of the family of numerical fluxes. Let (uD )D∈F be a
family satisfying:
• there exists u ∈ L2 (Ω) with lim kΠM uD − ukL2 (Ω) = 0 (we get from Lemma 5.7 that u ∈ H01 (Ω)),
hD →0
For any D ∈ F, we define the function GD ∈ LD (Ω)d (see Definition 1.3), by

uσ − uK
GD
K,σ = d nK,σ , ∀K ∈ M, ∀σ ∈ EK . (2.80)
dK,σ
Let us show that GD weakly converges to ∇u in L2 (Ω)d as h(M) → 0. We first prolong u and GD by 0 outside
Ω. Let us first remark that, thanks to the Cauchy-Schwarz inequality and using m(DK,σ ) = m(σ)dK,σ /d,
X X µ ¶2
uσ − uK
kGD k2L2 (Rd )d =d m(σ)dK,σ = d|uD |2X .
dK,σ
K∈M σ∈EK
Since kGD kL2 (Rd )d remains bounded, we can extract from F a subfamily, again denoted F, such that there exists
G ∈ L2 (Rd )d and GD weakly converges in L2 (Rd )d to G as h(M) → 0. Let us prove that G = ∇u, which
implies, by uniqueness of the limit, that the whole family converges to ∇u as h(M) → 0. R
1
Let ϕ ∈ Cc1 (Ω)d be given. Let us denote by ϕD ∈ LD (Ω)d the element defined by the value m(σ) σ
ϕ(x)dγ(x),
2 d d
in DK,σ for all K ∈ Mσ , again prolonged by 0 outside Ω. We then have R that ϕ D converges to ϕ R (R ) . We
in L
1 1
get, noticing that the terms uσ vanish for the interior edges since m(σ) σ ϕ(x)dγ(x)·nK,σ + m(σ) σ ϕ(x)dγ(x)·
nL,σ = 0, that
Z X X Z Z
GD (x) · ϕ(x)dx = − uK ϕ(x)dγ(x)nK,σ = − uD (x)div(ϕ)(x)dx,
Rd K∈M σ∈EK σ Rd
which implies, by passing to the limit h(M) → 0,

Z Z
G(x) · ϕ(x)dx = − u(x)div(ϕ)(x)dx.
Rd Rd
This proves that u ∈ H 1 (Rd ). Since u = 0 outside Ω, we get that u ∈ H01 (Rd ), and that
Z Z
G(x) · ϕ(x)dx = ∇u(x) · ϕ(x)dx,
Rd Rd
which proves that G = ∇u. For any D ∈ F, we now define the function H D ∈ LD (Ω)d (see Definition 1.3), by
D ϕ(xσ ) − ϕ(xK )
HK,σ = nK,σ + ∇ϕ(xK ) − (∇ϕ(xK ) · nK,σ )nK,σ , ∀K ∈ M, ∀σ ∈ EK . (2.81)
dσ
27
Then H D converges to ∇ϕ in L∞ (Ω)d as h(M) → 0. Indeed, we have, for K ∈ M and σ ∈ EK , using the
choice of xσ for ∆−adapted pointed finite volume discretizations,
d2K,σ
ϕ(xσ ) − ϕ(xK ) = dK,σ ∇ϕ(xK ) · nK,σ + RK,σ ,
2
with |RK,σ | is bounded by some L∞ norm of the second order partial derivatives of ϕ, denoted by C7 . We consider
h(M) small enough, such that ϕ vanishes on all K ∈ M such that EK ∩ Eext 6= ∅, hence for σ ∈ EK ∩ Eext ,
D
HK,σ = 0 = ∇ϕ(xK ).
We then get
D dσ
|HK,σ − ∇ϕ(xK )| ≤ C7 ≤ h(M)C7 , ∀K ∈ M, ∀σ ∈ EK .
2
Using that Z
huD , PD ϕiX = λD (x)GD (x) · H D (x)dx,
Ω
in which λD (x) denotes the piecewise function equal to λK in each K ∈ M, we get, by weak-strong convergence,
Z
lim huD , PD ϕiX = λ(x)∇u(x) · ∇ϕ(x)dx.
h(M)→0 Ω
Hence the conclusion of the consistency of the family of numerical fluxes. ¤

We then get the convergence of the scheme and an error estimate, using Theorem 2.1 with A = ∅.
2.3.2 The convergence proof in the case of the gradient schemes

Lemma 2.2 [Properties of the numerical fluxes in the gradient scheme on ∆−adapted meshes for anisotropic
diffusion]
Under hypothesis (1.5) and (2.3), let F be a family of pointed ∆−adapted polygonal finite volume space dis-
D
cretization of Ω in the sense of Definition 1.4. Let Φ = ((FK,σ )K∈M )D∈F be defined by (2.33).
σ∈E
Then Φ is a continuous, coercive and symmetric and consistent family of numerical fluxes in the sense of definition
2.2.
2
Proof. Let us first show that this family is coercive. We apply (2.88) to (RK,σ u) , defining βK > 0 below. We
get
µ ¶2
2 βK uσ − uK
(RK,σ u) ≥ αK d − αK dβK |∇K u|2 . (2.82)
1 + βK dK,σ
Hence Z
¡ ¢
Λ(x)∇D u(x) · ∇D u(x) + RD u(x)2 dx
Ω Ã µ ¶2 !
X βK X uσ − uK
2
≥ (λ − αK dβK )m(K)|∇K u| + αK m(σ)dK,σ
1 + βK dK,σ
K∈M σ∈EK
λ 2ab
Choosing βK = αK d and using a+b ≥ min(a, b) for a, b > 0, we get
Z
¡ ¢
Λ(x)∇D u(x) · ∇D u(x) + RD u(x)2 dx
Ω
X X µ ¶2
1 uσ − uK 1
≥ min( λ, α) m(σ)dK,σ = min( λ, α)|uD |2X .
d dK,σ d
K∈M σ∈EK
Hence we get the coerciveness property. The consistency property is resulting from the weak and strong conver-
gence properties (Lemmas 2.3 and 2.4) of ∇D u defined by (2.30). ¤
Let us state and prove the needed results on the discrete gradient defined by (2.30).
28
Lemma 2.3 Let F be a family of discretizations in the sense of definition 1.4 such that there exists θ > 0 with
θ ≥ θD for all D ∈ F. Let (uD )D∈F be a family of functions, such that:
• uD ∈ XD,Eint for all D ∈ F,
• there exists u ∈ L2 (Ω) with lim kΠM uD − ukL2 (Ω) = 0,
hD →0
Then u ∈ H01 (Ω) and ∇D uD weakly converge in L2 (Ω)d to ∇u as hD → 0, where the operator ∇D is defined by
(2.30).
Proof. Let us prolong ΠM uD and ∇D uD by 0 outside of Ω. Thanks to the Cauchy-Schwarz inequality, we easily
get that
k∇D uD k2L2 (Ω)d ≤ d|uD |2X .
Hence, up to a subsequence, there exists some function G ∈ L2 (Rd )d such that ∇D uD weakly converges in
L2 (Rd )d to G as hD → 0. Let us show that G = ∇u. Let ψ ∈ Cc∞ (Rd )d be given. We assume that h(M) is
small enough for ensuring that the support of ψ does not intersect any boundary control volume. Let us consider
the term T5D defined by Z
T5D = ∇D uD (x) · ψ(x)dx.
Rd
Using xL − xK = dσ nK,σ , we get that T5D = T6D + T7D , with

X Z
1 1
T6D = m(σ)(uL − uK )nK,σ · (ψK + ψL ), with ψK = ψ(x)dx,
2 σ∈Eint
m(K) K
Mσ ={K,L}
and X X uσ − uK
T7D = m(σ) (xσ − xK ) · (ψK − ψL ).
dK,σ
K∈M σ∈EK ∩Eint
Mσ ={K,L}
We compare T6D with T8D defined by

X
T8D = m(σ)(uL − uK )nK,σ · ψσ ,
σ∈Eint
Mσ ={K,L}
with Z
1
ψσ = ψ(x)dγ(x).
m(σ) σ
We get that
X m(σ) X X ψK + ψL
(T6D − T8D )2 ≤ (uL − uK )2 m(σ)dK,σ | − ψ σ |2 ,
dσ 2
σ∈Eint K∈M σ∈EK
Mσ ={K,L}
which leads to lim (T6D − T8D ) = 0.

hD →0
Since
X X Z
T8D =− m(σ)uK nK,σ · ψσ = − ΠM uD (x)divψ(x)dx,
K∈M σ∈EK Rd
R
we get that lim T8D = − Rd
u(x)divψ(x)dx. Let us now turn to the study of T7D . We have, using the Cauchy-
hD →0
Schwarz inequality and the regularity of the mesh
(T7D )2 ≤ C8 |u|2X h(M)2 ,

29
where C8 only depends on ϕ and θ. Hence

lim T7D = 0.
hD →0
This proves that the function G ∈ L2 (Rd )d is a.e. equal to ∇u in Rd . Since u = 0 outside of Ω, we get that
u ∈ H01 (Ω), and the uniqueness of the limit implies that the whole family ∇D uD weakly converges in L2 (Rd )d to
∇u as hD → 0.
¤
Let us now state some strong consistency property of the discrete gradient applied to the interpolation of a regular
function.
Lemma 2.4 Let D be a discretization of Ω in the sense of Definition 1.4, and let θ ≥ θD be given. Then, for any
function ϕ ∈ C 2 (Ω), there exists C9 only depending on d, θ and ϕ such that:
k∇D PD ϕ − ∇ϕk(L∞ (Ω))d ≤ C9 hD , (2.83)
where ∇D is defined by (2.30).
Proof. Taking into account definition (2.30) and (2.39), we write, for any K ∈ M,
X µ ¶
1 ϕ(xσ ) − ϕ(xK )
∇K PD ϕ − ∇ϕ(xK ) = m(σ) − ∇ϕ(xK ) · nK,σ (xσ − xK ).
m(K) dK,σ
σ∈EK
Thanks to ¯ ¯
¯ ϕ(xσ ) − ϕ(xK ) ¯
¯ − ∇ϕ(xK ) · nK,σ ¯¯ ≤ C10 h(M),
¯ dK,σ
where C10 only depends on ϕ, we conclude the proof.¤
We then get the convergence of the scheme and an error estimate, using Theorem 2.1 with A = Eint .
Convergence study in the case of general meshes

In this section, we denote by xσ the center of gravity of σ ∈ E.
For all v ∈ XD , we denote by ΠM v ∈ HM (Ω) the piecewise function from Ω to R defined by ΠM v(x) = vK for
a.e. x ∈ K, for all K ∈ M. Using the Cauchy-Schwarz inequality, we have for all σ ∈ Eint with Mσ = {K, L},
(vK − vL )2 (vK − vσ )2 (vσ − vL )2

≤ + , ∀v ∈ XD ,
dσ dK,σ dL,σ
which leads to the relation

kΠM vk21,2,M ≤ |v|2X , ∀v ∈ XD,0 . (2.84)
2
The following lemma provides an equivalence property between the L -norm of the discrete gradient, defined by
(2.38)-(2.42) and the norm | · |X .
Lemma 2.5 Let D be a discretization of Ω in the sense of Definition 1.3, and let θ ≥ θD be given (where θD is
defined by (2.60)). Then there exists C11 > 0 and C12 > 0 only depending on θ and d such that:
C11 |u|X ≤ k∇D ukL2 (Ω) ≤ C12 |u|X , ∀u ∈ XD , (2.85)
where ∇D is defined by (2.38)-(2.42).
Proof. By definition,
X X m(σ)dK,σ
k∇D uk2L2 (Ω)d = |∇K,σ u|2 .
d
K∈M σ∈EK
30
From the definition (2.41), thanks to (2.39) and to the definition (2.38), we get that
X m(σ)dK,σ
RK,σ u nK,σ = 0, ∀K ∈ M. (2.86)
d
σ∈EK
Therefore, Ã !
X X m(σ)dK,σ
k∇D uk2L2 (Ω)d = m(K)|∇K u| + 2
(RK,σ u)2 . (2.87)
d
K∈M σ∈EK
Let us now notice that the following inequality holds:
λ
(a − b)2 ≥ a2 − λb2 , ∀a, b ∈ R, ∀λ > −1. (2.88)
1+λ
2
We apply this inequality to (RK,σ u) for some λ > 0 and obtain:
µ ¶2 µ ¶2
2 λd uσ − uK |xσ − xK |
(RK,σ u) ≥ − λd|∇K u|2 . (2.89)
1+λ dK,σ dK,σ
This leads to
X m(σ)dK,σ µ ¶2
2 λ X uσ − uK
(RK,σ u) ≥ m(σ)dK,σ − λm(K)d|∇K u|2 θ2 .
d 1+λ dK,σ
σ∈EK σ∈EK
Choosing λ as
1
λ= , (2.90)
dθ2
we get that
λ
k∇D uk2(L2 (Ω))d ≥
|u|2 ,
1+λ X
which shows the left inequality of (2.85). Let us now prove the right inequality. Remark that, thanks to the
assumption that K is xK -star-shaped, the property
X
m(σ)dK,σ = d m(K), ∀K ∈ M (2.91)
σ∈EK
holds. On one hand, using the definition (2.38) of ∇K u and (2.91), the Cauchy–Schwarz inequality leads to:
1 X m(σ) X d X m(σ)
|∇K u|2 ≤ ( )2 (uσ − uK )2 m(σ)dK,σ = (uσ − uK )2 . (2.92)
m(K) dK,σ m(K) dK,σ
σ∈EK σ∈EK σ∈EK
On the other hand, by definition (2.41), and thanks to the regularity of the mesh (2.95), we have:
µ ¶ µ ¶
2 uσ − uK 2 2 xσ − xK 2 uσ − uK 2 2 2
(RK,σ u) ≤ 2d ( ) + |∇K u| | | ≤ 2d ( ) + θ |∇K u| . (2.93)
dK,σ dK,σ dK,σ
From (2.87), (2.92) et (2.93), we conclude that the right inequality of (2.85) holds. ¤
We can now state a result of weak convergence for the discrete gradient of a sequence of bounded discrete functions.
Lemma 2.6 Let F be a family of discretizations in the sense of definition 1.3 such that there exists θ > 0 with
θ ≥ θD for all D ∈ F. Let (uD )D∈F be a family of functions, such that:
31

• there exists u ∈ L2 (Ω) with lim kΠM uD − ukL2 (Ω) = 0,
hD →0
Then u ∈ H01 (Ω) and ∇D uD weakly converge in L2 (Ω)d to ∇u as hD → 0, where the operator ∇D is defined by
(2.38)-(2.42).
Remark 2.12 Note that the proof that u ∈ H01 (Ω) also results from (2.84), which allows to apply Lemma 5.7 of
the appendix in the particular case p = 2.
Proof. Let us prolong ΠM uD and ∇D uD by 0 outside of Ω. Thanks to Lemma 2.5, up to a subsequence, there
exists some function G ∈ L2 (Rd )d such that ∇D uD weakly converges in L2 (Rd )d to G as hD → 0. Let us show
that G = ∇u. Let ψ ∈ Cc∞ (Rd )d be given. Let us consider the term T9D defined by
Z
D
T9 = ∇D uD (x) · ψ(x)dx.
Rd
We get that T9D = T10

D D
+ T11 , with
X X Z
D 1
T10 = m(σ)(uσ − uK )nK,σ · ψK , with ψK = ψ(x)dx,
m(K) K
K∈M σ∈EK
and
X X Z
D
T11 = RK,σ u nK,σ · ψ(x)dx.
K∈M σ∈EK DK,σ
D D
We compare T10 with T12 defined by
X X
D
T12 = m(σ)(uσ − uK )nK,σ · ψσ ,
K∈M σ∈EK
with Z
1
ψσ = ψ(x)dγ(x).
m(σ) σ
We get that
X X m(σ) X X
D D 2
(T10 − T12 ) ≤ (uσ − uK )2 m(σ)dK,σ |ψK − ψσ |2 ,
dK,σ
K∈M σ∈EK K∈M σ∈EK
D D
which leads to lim (T10 − T12 ) = 0.
hD →0
Since
X X Z
D
T12 =− m(σ)uK nK,σ · ψσ = − ΠM uD (x)divψ(x)dx,
K∈M σ∈EK Rd
D
R D
we get that lim T12 = − Rd
u(x)divψ(x)dx. Let us now turn to the study of T11 . Noting again that (2.86)
hD →0
holds, we have: Z
X X
D
T11 = RK,σ u nK,σ · (ψ(x) − ψK )dx.
K∈M σ∈EK DK,σ
R
Since ψ is a regular function, there exists Cψ only depending on ψ such that | DK,σ
(ψ(x) − ψK )dx| ≤
m(σ)dK,σ
C ψ hD . From (2.93) and the Cauchy-Schwarz inequality, we thus get:
d
D
lim T11 = 0.
hD →0
32
This proves that the function G ∈ L2 (Rd )d is a.e. equal to ∇u in Rd . Since u = 0 outside of Ω, we get that
u ∈ H01 (Ω), and the uniqueness of the limit implies that the whole family ∇D uD weakly converges in L2 (Rd )d to
∇u as hD → 0.
¤
Let us now state some strong consistency property of the discrete gradient applied to the interpolation of a regular
function.
Lemma 2.7 Let D be a discretization of Ω in the sense of Definition 1.3, and let θ ≥ θD be given. Then, for any
function ϕ ∈ C 2 (Ω), there exists C13 only depending on d, θ and ϕ such that:
k∇D PD ϕ − ∇ϕk(L∞ (Ω))d ≤ C13 hD , (2.94)
where ∇D is defined by (2.38)-(2.42).
Proof. Taking into account definition (2.42), and using definition (2.40), we write:
|∇K,σ PD ϕ − ∇ϕ(xK )| ≤ |∇K PD ϕ − ∇ϕ(xK )| + |RK,σ PD ϕ|
From (2.38), we have, for any K ∈ M,
1 X
∇K PD ϕ = m(σ)(ϕ(xσ ) − ϕ(xK ))
m(K)
σ∈EK
1 X ¡ ¢
= m(σ) ∇ϕ(xK ) · (xσ − xK ) + h2K ρK,σ nK,σ ,
m(K)
σ∈EK
where |ρK,σ | ≤ Cϕ with Cϕ only depending on ϕ. Thanks to (2.39) and to the regularity of the mesh, we get:
1 X
|∇K PD ϕ − ∇ϕ(xK )| ≤ m(σ)h2K |ρK,σ | ≤ hK d Cϕ θ.
m(K)
σ∈EK
From this last inequality, using Definition 2.41, we get:

√
d
|RK,σ PD ϕ| = |ϕ(xσ ) − ϕ(xK ) − ∇K PD ϕ · (xσ − xK )|
dK,σ
√
d ¡ 2 ¢
≤ hK ρK,σ + h2K d Cϕ θ
dK,σ
√
≤ dθ(hK Cϕ + hK dCϕ θ),
which concludes the proof.¤
Lemma 2.8 Let F be a family of discretizations in the sense of definition 1.3. We assume that there exists θ > 0
with
θD ≤ θ, ∀D ∈ F, (2.95)
D
where θD is defined by (2.60). Let Φ = ((FK,σ )K∈MD )D∈F be the family of fluxes defined by (2.43)-(2.46). Then
σ∈EK
the family Φ is a continuous, coercive, consistent and symmetric family of numerical fluxes in the sense of definition
2.2.
Proof. Since the family of fluxes is defined by (2.43)-(2.46), it satisfies (2.37), and therefore we have:
Z
hu, viF = ∇D u(x) · Λ(x)∇D v(x)dx, ∀u, v ∈ XD .
Ω
Hence the property hu, viF = hv, uiF holds. The continuity and coercivity of the family Φ result from Lemma
2.5 and the properties of Λ, which give: hu, viF ≤ λk∇D ukL2 (Ω) k∇D vkL2 (Ω) and hu, viF ≥ λk∇D uk2L2 (Ω) for
any u, v ∈ XD . The consistency results from the weak and strong convergence properties of lemmas 2.6 and 2.7,
which give ∇D uD → ∇u weakly in L2 (Ω) and ∇D PD ϕ → ∇ϕ in L2 (Ω) as the mesh size tends to 0. ¤
Let us apply the error estimate, in the particular case that Λ = Id and that the solution of (2.5) is regular enough.
33
Theorem 2.3 (Error estimate, isotropic case) Under the hypotheses and notations of Theorem 2.2 and in the
particular case where (FK,σ )K∈M,σ∈E is defined by (2.43)-(2.46), there exists C14 , only depending on α, θ and
u, such that
Ru ≤ C14 hD . (2.96)
Proof. Let us assume that the family of fluxes is defined by (2.43)-(2.46). Indeed, we get in this case that, for all
v ∈ XD ,
X m(σ 0 )dK,σ0 σ0 σ
FK,σ (v) = − (∇K v + RK,σ0 v nK,σ0 ) · y ,
0
d
σ ∈EK
with  √ µ ¶

 m(σ) d m(σ)
 nK,σ + 1− nK,σ · (xσ − xK ) nK,σ if σ = σ 0
σ0 σ m(K) dK,σ√ m(K)
y =

 m(σ) d
 nK,σ − m(σ)nK,σ · (xσ0 − xK )nK,σ0 otherwise .
m(K) dK,σ0 m(K)
Using (2.39), we get that
X m(σ 0 )dK,σ0 0
y σ σ = m(σ)nK,σ .
0
d
σ ∈EK
Since it is easy to see that there exists C15 ∈ R+ such that |RK,σ0 PD,A u| ≤ C15 hK , we then obtain that there
exists some C16 ∈ R+ with
¯ Z ¯
¯ ¯
¯FK,σ (PD,A u) + ∇u(x) · nK,σ dγ(x)¯ ≤ C16 m(σ)hK .
¯ ¯
σ
This easily leads to the conclusion of (2.96). ¤
2.3.3 The convergence proof in the case of the mixed-type schemes

Lemma 2.9 Let F be a family of discretizations in the sense of definition 1.3. We assume that there exists θ > 0
D
such that (2.95) holds, where θD is again defined by (2.60). Let Φ = ((FK,σ )K∈MD )D∈F be the family of fluxes
σ∈EK
defined by (2.48)-(2.49)-(2.53). Then the family Φ is a continuous, coercive, consistent and symmetric family of
numerical fluxes in the sense of definition 2.2.
Proof. Let us remark that we easily get
X µ ¶2
2 FK,σ
m(K)|vK (F )| ≤ C17 m(σ)dK,σ .
m(σ)
σ∈EK
Using the Cauchy-Schwarz inequality, we then obtain
X X µ ¶2
2 FK,σ
m(σ)dK,σ |vK,σ (F )| ≤ C18 m(σ)dK,σ .
m(σ)
σ∈EK σ∈EK
m(σ)
Hence, letting GK,σ = dK,σ (uK − uσ ) in (2.48), we get
kuk2X ≤ (hF, F iY )1/2 (hG, GiY )1/2 ≤ (hF, F iY C18 λ)1/2 kukX ,
which is sufficient to prove the coercivity of the family of fluxes. We now remark that, thanks to (2.39), the property
X m(σ)dK,σ xσ − xK
HK,σ (F )Λ−1
K =0
d dK,σ
σ∈EK
34
holds, which allows to write

Z Ã ¯ ¯2 !
X X m(σ)dK,σ ¯ xσ − xK ¯
|v̄D F (x)|2 dx = m(K)|vK (F )|2 + HK,σ (F )2 ¯¯Λ−1 ¯ .
Ω d K
dK,σ ¯
K∈M σ∈EK
We then get, using again(2.88) in order to get a lower bound of HK,σ (F )2 , the existence of C19 such that
X µ ¶2
2 FK,σ
m(K)|vK (F )| ≥ C19 m(σ)dK,σ ,
m(σ)
σ∈EK
which yields the continuity of the family of fluxes (applying the Cauchy-Schwarz inequality to the right-hand side
of (2.48)). Note that we shall again get a strong convergence property for vD (F ) do ∇u.
¤
Chapter 3
Nonlinear convection - degenerate

diffusion problems

Let Ω be a bounded open subset of Rd , (d ∈ N? ) with boundary ∂Ω and let T ∈ R∗+ . One considers the following
problem.
³ ´
ut (x, t) + div q f (u) (x, t) − ∆ϕ(u)(x, t) = 0, for (x, t) ∈ Ω × (0, T ). (3.1)
The initial condition is formulated as follows:
u(x, 0) = u0 (x) for x ∈ Ω. (3.2)

The boundary condition is the following nonhomogeneous Dirichlet condition:
u(x, t) = u(x, t), for (x, t) ∈ ∂Ω × (0, T ). (3.3)

This problem arises in different physical contexts. One of them is the problem of two phase flows in a porous
medium, such as the air-water flow of hydrological aquifers. In this case, (3.1)-(3.3) represents the conservation
of the incompressible water phase, described by the water saturation u, submitted to convective flows (first order
space terms q(x, t) f (u)) and capillary effects (∆ϕ(u)). The expression q(x, t) f (u) for the convective term in
(3.1) appears to be a particular case of the more general expression F (u, x, t), but since it involves the same tools
as the general framework, the results of this chapter could be extended to some other problems.
One supposes that the following hypotheses, globally referred to in the following as Assumption 3.1, are fulfilled.
Assumption 3.1
(H1) We assume hypothesis (1.5) on Ω,
(H2) u0 ∈ L∞ (Ω) and u ∈ L∞ (∂Ω×(0, T )), u being the trace of a function of H 1 (Ω×(0, T ))∩L∞ (Ω×(0, T ))
(also denoted u); one sets uI = min(infess u0 , infess ū) and uS = max(supess u0 , supess ū),
(H3) ϕ is a nondecreasing
√ Lipschitz-continuous function, with Lipschitz constant Φ, and one defines a function ζ
such that ζ 0 = ϕ0 ,
(H4) f ∈ C 1 (R, R), f 0 ≥ 0; one sets F = maxs∈[uI ,uS ] f 0 (s),
(H5) q is the restriction to Ω × (0, T ) of a function of C 1 (Rd × R, Rd ),
35
36
Xd
∂qi
(H6) div(q(x, t)) = 0 for all (x, t) ∈ Rd ×(0, T ), where div(q(x, t)) = (x, t), (qi is the i-eth component
i=1
∂x i
of q) and
q(x, t).n(x) = 0, for a.e. (x, t) ∈ ∂Ω × (0, T ), (3.4)
(for x ∈ ∂Ω, n(x) denotes the outward unit normal to Ω at point x).
Remark 3.1 The function f is assumed to be non decreasing in Hypothesis (H3) of Assumption 3.1 for the sake
of simplicity. In fact, the convergence analysis which we present here would also hold without this monotonicity
assumption using for instance a flux splitting scheme for the treatment of the convective term qf (u).
Under Assumption 3.1, (3.1)-(3.3) does not have, in the general case, strong regular solutions. Because of the pres-
ence of a non-linear convection term, the expected solution is an entropy weak solution in the sense of Definition
3.2 given below.
3.1.1 Finite volume approximation

We may now define the finite volume discretization of (3.1)-(3.3). Let D be a ∆−adapted pointed polygonal finite
volume space-time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7. The initial condition is
discretized by:
Z
1
u0K = u0 (x)dx, ∀K ∈ M. (3.5)
m(K) K
In order to introduce the finite volume scheme, we need to define:
Z tn+1 Z
1
ūn+1
σ = n ū(x, t)dγ(x)dt, ∀σ ∈ Eext , ∀n ∈ [[0, N ]], (3.6)
δt m(σ) tn σ
Z tn+1 Z
n+1 1
qK,σ = n q(x, t) · nK,σ dγ(x)dt, ∀K ∈ M, ∀σ ∈ EK , ∀n ∈ [[0, N ]]. (3.7)
δt tn σ
For all n ∈ N, un+1 ∈ HM (Ω) and g ∈ C 0 (R), we introduce the notations
un+1 n+1
K,σ = uL , ∀K ∈ M, ∀σ ∈ EK ∩ Eint , Mσ = {K, L}
(3.8)
un+1 n+1
K,σ = ūσ , ∀K ∈ M, ∀σ ∈ EK ∩ Eext , Mσ = {K},
and
n+1
δK,σ g(u) = g(un+1 n+1
K,σ ) − g(uK ), ∀K ∈ M, ∀σ ∈ EK . (3.9)
An implicit finite volume scheme for the discretization of (3.1)-(3.3) is given by the following set of nonlinear
equations, the discrete unknowns of which are u = (un+1
K )K∈M,n∈[[0,N ]] :
un+1 − unK X · m(σ) n+1

¸
K n+1 + n+1 n+1 − n+1
m(K) + (qK,σ ) f (uK ) − (qK,σ ) f (uK,σ ) − δ ϕ(u) = 0,
δtn dσ K,σ (3.10)
σ∈EK
∀K ∈ M, ∀n ∈ [[0, N ]],
n+1 + n+1 − n+1 n+1 + n+1
where (qK,σ ) and (qK,σ ) denote the positive and negative parts of qK,σ (i.e. (qK,σ ) = max(qK,σ , 0) and
n+1 − n+1
(qK,σ ) = − min(qK,σ , 0)).
37
Remark 3.2 The upwind discretization of the flux qf (u) in (3.10) uses the monotonicity of f and should be
replaced in the general case by, for instance, a flux splitting scheme.
Remark
X X to Hypothesis (H6) of Assumption 3.1, one gets for all K ∈ M and n ∈ [[0, N ]],
3.3 Thanks
n+1 n+1 + n+1 −
qK,σ = [(qK,σ ) − (qK,σ ) ] = 0. This leads to
σ∈EK σ∈EK
X¡ ¢ X
n+1 +
(qK,σ ) f (un+1 n+1 − n+1
K ) − (qK,σ ) f (uK,σ ) = −
n+1 − n+1
(qK,σ ) δK,σ f (u). (3.11)
σ∈EK σ∈EK
This property will be used in the following.
In Section (3.2.2) we shall prove the existence (Lemma 3.1) and the uniqueness (Lemma 3.4) of the solution
u = (un+1
K )K∈M,n∈[[0,N ]] to (3.5)-(3.10). We may then define the approximate solution to (3.1)-(3.3) associated
to a ∆−adapted pointed polygonal finite volume space-time discretization of D of Ω × (0, T ) by:
Definition 3.1 Let D be a ∆−adapted pointed polygonal finite volume space-time discretization of Ω × (0, T ) in
the sense of Definitions 1.4 and 1.7. The approximate solution of (3.1)-(3.3) associated to the discretization D is
defined almost everywhere in Ω × (0, T ) by:
uD (x, t) = un+1 n n+1

K , ∀x ∈ K, ∀t ∈ (t , t ), ∀K ∈ M, ∀n ∈ [[0, N ]], (3.12)
where (un+1
K )K∈M,n∈[[0,N ]] is the unique solution to (3.5)-(3.10).
3.1.2 A numerical example

We finally present some numerical results which we obtained by implementing the scheme which was studied
above in a prototype code.
The domain Ω is the unit square (0, 1) × (0, 1). We define two subregions Ω1 = (0.1, 0.3) × (0.4, 0.6) and
Ω2 = (0.7, 0.9) × (0.4, 0.6). The initial data is given by 0.5 in Ω \ (Ω1 ∪ Ω2 ), 1 in Ω1 and 0 in Ω2 . It is represented
on upper left corner of the figure below. The boundary value is the constant 0.5.
The function ϕ is defined by ϕ(s) = 0 if s ∈ [0, 0.5] and ϕ(s) = 0.2(s − 0.5) if s ∈ [0.5, 1], so that the diffusion
effect only takes place in the areas where the saturation u is greater than .5. The function f is defined by f (s) = s
and the field q is defined by q(x, y) = (10(x − x2 )(1 − 2y), −10(y − y 2 )(1 − 2x)). Hence there is a linear
rotating convective transport.
We define a coarse mesh of 14 admissible triangles on the unit square, from which we obtain a fine mesh of 12 600
triangles by refining these 14 triangles uniformly 30 times. This fine mesh is used for the computations.
The figure below presents the obtained results at times 0.000, 0.007, 0.028 and 0.112. The black points correspond
to the value 1, the white ones to the value 0, with a continuous scale of greys between these values. One observes
that the initial value 0 is transported, only modified by the numerical diffusion due to the convective upstream
weighting, and that, on the contrary, the initial value 1 is rapidly smoothed, due to the effect of the parabolic term
which is active on the range [0.5, 1].
38
Computed solution at time t = 0 (initial condition), t = 0.007, t = 0.028 and t = 0.112.
3.2 Further analysis

3.2.1 Continuous definitions and main convergence result
Definition 3.2 (Entropy weak solution) Under Assumption 3.1, a function u is said to be an entropy weak solu-
tion to (3.1)-(3.3) if it verifies:
u ∈ L∞ (Ω × (0, T )), (3.13)

2
ϕ(u) − ϕ(u) ∈ L (0, T ; H01 (Ω)), (3.14)
+
and u satisfies the following Kruzkov entropy inequalities: ∀ ψ ∈ D (Ω × [0, T )), ∀κ ∈ R,
 
Z |u(x, t) − κ| ψt (x, t)+ Z
 (f (u(x, t)>κ) − f (u(x, t)⊥κ)) q(x, t) · ∇ψ(x, t)  dxdt + |u0 (x) − κ|ψ(x, 0)dx ≥ 0,
Ω×(0,T ) −∇|ϕ(u)(x, t) − ϕ(κ)| · ∇ψ(x, t) Ω
(3.15)
where one denotes by a>b the maximum value between two real values a and b, and by a⊥b their minimum value
and where D+ (Ω × [0, T )) = {ψ ∈ Cc∞ (Ω × R, R+ ), ψ(·, T ) = 0} .
39
This notion has been introduced by several authors ([8], [29]), who proven the existence of such a solution in
bounded domains. In [29], the proof of existence uses strong BV estimates in order to derive estimates in time and
space for the solution of the regularized problem obtained by adding a small diffusion term. In [8], the existence
of a weak solution is proven using semigroup theory (see [2]), and the uniqueness of the entropy weak solution is
proven using techniques which have been introduced by S.N. Krushkov and extended by J. Carrillo.
In the present study, thanks to condition (3.4), boundary conditions are entirely taken into account by (3.14) and
do not appear in the entropy inequality (3.15). For studies of the continuous problem, one can refer to [29], which
uses the classical Bardos-Leroux-Nédélec formulation [1], or [8] in the case of a homogeneous Dirichlet boundary
condition on ∂Ω without condition (3.4).
Let us mention some related work in the case of infinite domains (Ω = Rd ): In [3], the authors prove the existence
in the case Ω = Rd , regularizing the problem with the “general kinetic BGK” framework to yield estimates on
translates of the approximate solutions. Continuity of the solution with respect to the data for a more general
equation was studied by Cockburn and Gripenberg [13], and convergence of the discretization with an implicit
finite volume scheme was recently studied by Ohlberger [31].
We shall deal here with the case of a bounded domain. The aim of the present work is then to prove the convergence
of approximate solutions obtained using a finite volume method with general unstructured meshes towards the
entropy weak solution of (3.1)-(3.3) as the mesh size and time step tend to 0. We state this result in Theorem 3.1
in Section 3.1.1, after presenting the finite volume scheme. Then in Section 3.2.2, the existence and uniqueness
of the solution to the nonlinear set of equations resulting from the finite volume scheme is proven, along with
some properties of the discrete solutions. In Section 3.2.3 we show some compactness properties of the family of
approximate solutions. We show in Section 3.2.4 that there exists some subsequence of sequences of approximate
solutions which tends to a so-called “entropy process solution”, and in Section 3.2.5 we prove the uniqueness of
this entropy process solution, which allows us to conclude to the convergence of the scheme in Section 3.2.6. We
finally give an example of numerical implementation in Section 3.1.2.
Theorem 3.1 (Convergence of the approximate solution towards the entropy weak solution)
Let ξ ∈ R, consider a family of ∆−adapted pointed polygonal finite volume space-time discretization of Ω×(0, T )
in the sense of Definitions 1.4 and 1.7 such that, for all D in the family, one has ξ ≥ reg(D). For a given
∆−adapted pointed polygonal finite volume space-time discretization D of this family, let uD denote the associated
approximate solution as defined in Definition 3.1. Then:
uD → u ∈ Lp (Ω × (0, T )) as h(D) → 0, ∀p ∈ [1, +∞),
where u is the unique entropy weak solution to (3.1)-(3.3).
The proof of this convergence theorem will be concluded in Section 3.2.6 after we lay out the properties of the
discrete solution (sections 3.2.2 and 3.2.3), its convergence towards an “entropy process solution” (Section 3.2.4)
and a uniqueness result on this entropy process solution (Section 3.2.5).
Remark 3.4 All the results of this chapter also hold for explicit schemes, under a convenient CFL condition on
the time step and mesh size.
3.2.2 Existence, uniqueness and discrete properties

We state here the properties and estimates which are satisfied by the scheme which we introduced in the previous
section and prove existence and uniqueness of the solution to this scheme. All the discrete properties which we
address here correspond to natural estimates which are satisfied, at least formally, by regular continuous solutions.
Let us first start by an L∞ estimate:
Lemma 3.1 (L∞ estimate) Under Assumption 3.1, let D be a ∆−adapted pointed polygonal finite volume space-
time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7 and let (un+1
K )K∈M,n∈[[0,N ]] be a solution
of scheme (3.5)-(3.10). Then
40
uI ≤ un+1
K ≤ uS , ∀K ∈ M, ∀n ∈ [[0, N ]].
Proof.
Let uM = max um+1
L and let n ∈ [[0, N ]] and K ∈ M such that un+1
K = uM . Equations (3.10) and (3.11)
L∈M,m∈[[0,N ]]
yield
µ ¶
δtn X m(σ) n+1
uM = un+1
K = unK + n+1 − n+1
(qK,σ ) δK,σ f (u) + δK,σ ϕ(u) . (3.16)
m(K) dσ
σ∈EK
If one assumes that

n+1
uM ≥ max ūm+1
σ , using the monotonicity of ϕ (which implies δK,σ ϕ(u) ≤ 0) and that of f (which
σ∈Eext ,m∈[[0,N ]]
n+1
implies δK,σ f (u) ≤ 0, one gets uM ≤ unK , and therefore uM ≤ u0K .
This shows that
uM ≤ max( max ūm+1

σ , max u0L ),
σ∈Eext ,m∈[[0,N ]] L∈M
yielding uM ≤ uS . By the same method, one shows that min um+1

L ≥ uI . ¤
L∈M,m∈[[0,N ]]
A corollary of Lemma 3.1 is the existence of a solution (un+1
K )K∈M,n∈[[0,N ]] to (3.5)-(3.10). (Uniqueness is proven
in Lemma (3.4) below).
Corollary 3.1 (Existence of the solution to the scheme) Under Assumption 3.1, let D be a ∆−adapted pointed
polygonal finite volume space-time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7. Then there
exists a solution (un+1
K )K∈M,n∈[[0,N ]] to the scheme (3.5)-(3.10).
The proof of this corollary is an adaptation of the technique which was used in [17] for the existence of the solution
to an implicit finite volume scheme for the discretization of a pure hyperbolic equation.
The two following lemmas express the monotonicity of the scheme. Both are used to derive continuous entropy
inequalities.
Lemma 3.2 (Regular convex discrete entropy inequalities)

Under Assumption 3.1, let D be a ∆−adapted pointed polygonal finite volume space-time discretization of Ω ×
(0, T ) in the sense of Definitions 1.4 and 1.7 and let u = (un+1
K )K∈M,n∈[[0,N ]] be a solution to (3.5)-(3.10).
Then, for all η ∈ C 2 (R, R), with η 00 ≥ 0, for all µ and ν in C 1 (R, R) with µ0 = η 0 (ϕ) and ν 0 = η 0 (ϕ)f 0 , for
un+1
all K ∈ M, and n ∈ [[0, N ]], there exist (eK,σ )σ∈EK with uen+1 n+1 n+1 n+1 n+1
K,σ ∈ (min(uK , uK,σ ), max(uK , uK,σ )) for all
σ ∈ EK satisfying
µ(un+1 n X µ ¶
K ) − µ(uK ) n+1 + n+1 n+1 − n+1 m(σ) n+1
m(K) + (qK,σ ) ν(uK ) − (qK,σ ) ν(uK,σ ) − δ η(u)
δtn dσ K,σ
σ∈EK
(3.17)
1 X m(σ) 00
+ η (ϕ(eun+1 n+1 2
K,σ ))(δK,σ ϕ(u)) ≤ 0
2 dσ
σ∈EK
Proof.
In order to prove (3.17), one multiplies Equation (3.10) by η 0 (ϕ(un+1
K )).
The convexity of µ yields
41
un+1
K − unK 0 n+1 µ(un+1 n+1
K ) − µ(uK )
m(K) η (ϕ(u )) ≥ m(K) . (3.18)
δtn K
δtn
Using the convexity of ν and Remark 3.3, one gets
X X
n+1 − n+1
− (qK,σ ) δK,σ f (u)η 0 (ϕ(un+1
K )) ≥ −
n+1 − n+1
(qK,σ ) δK,σ ν(u)
σ∈EK σ∈EK
X¡ ¢
n+1 +
≥ (qK,σ ) ν(un+1 n+1 − n+1
K ) − (qK,σ ) ν(uK,σ ) .
σ∈EK
The Taylor-Lagrange formula gives, for all σ ∈ EK , the existence of

en+1
u n+1 n+1 n+1 n+1
K,σ ∈ (min(uK , uK,σ ), max(uK , uK,σ )) such that
n+1 1 00
−δK,σ ϕ(u) η 0 (ϕ(un+1 n+1
K )) = −δK,σ η(ϕ(u)) + η (ϕ(e un+1 n+1 2
K,σ ))(δK,σ ϕ(u)) .
2
Then collecting the previous inequalities gives Inequality (3.17) . ¤
Lemma 3.3 (Kruzkov’s discrete entropy inequalities) Under Assumption 3.1, let D be a ∆−adapted pointed
polygonal finite volume space-time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7 and let
u = (un+1
K )K∈M,n∈[[0,N ]] be a solution of the scheme (3.5)-(3.10).
Then, for all κ ∈ R, K ∈ M and n ∈ [[0, N ]],
|un+1 − κ| − |unK − κ| X · (q n+1 )+ |f (un+1 ) − f (κ)| ¸

K K,σ K
m(K) + n+1 −
δtn −(qK,σ ) |f (un+1
K,σ ) − f (κ)|
σ∈EK
X m(σ) (3.19)
− δ n+1 |ϕ(u) − ϕ(κ)| ≤ 0
dσ K,σ
σ∈EK
Proof. In order to prove Kruzkov’s entropy inequalities, one follows [17]. Equation (3.10) is rewritten as
B(un+1 n n+1 n+1

K , uK , (uL )L∈NK , (ūσ )σ∈EK,ext ) = 0, (3.20)
where B is nonincreasing with respect to each of its arguments except un+1

K . Consequently,
B(un+1 n n+1 n+1

K , uK >κ, (uL >κ)L∈NK , (ūσ >κ)σ∈EK,ext ) ≤ 0. (3.21)
Since B(κ, κ, (κ)L∈NK , (κ)σ∈EK,ext ) = 0, one gets
B(κ, unK >κ, (un+1 n+1

L >κ)L∈NK , (ūσ >κ)σ∈EK,ext ) ≤ 0. (3.22)
Using the fact that un+1 n+1

K >κ = uK or κ, (3.21) and (3.22) give
B(un+1 n n+1 n+1

K >κ, uK >κ, (uL >κ)L∈NK , (ūσ >κ)σ∈EK,ext ) ≤ 0. (3.23)
In the same way one obtains
B(un+1 n n+1 n+1

K ⊥κ, uK ⊥κ, (uL ⊥κ)L∈NK , (ūσ ⊥κ)σ∈EK,ext ) ≥ 0. (3.24)
Substracting (3.24) from (3.23) and remarking that for any nondecreasing function g and all real values a, b,
g(a>b) − g(a⊥b) = |g(a) − g(b)| yields Inequality (3.19). ¤
Let us now prove the uniqueness of the solution to (3.5)-(3.10) and define the approximate solution.
42
Lemma 3.4 (Uniqueness of the approximate solution) Under Assumption 3.1, let D be a ∆−adapted pointed
polygonal finite volume space-time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7. Then there
exists a unique solution (un+1
K )K∈M,n∈[[0,N ]] to (3.5)-(3.10).
Proof.
The existence of (un+1
K )K∈M,n∈[[0,N ]] was established in Corollary 3.1. There only remains to prove the uniqueness
of the solution. Let (un+1 n+1 0 0
K )K∈M,n∈[[0,N ]] and (vK )K∈M,n∈[[0,N ]] (setting vK = uK ) be two solutions to the
scheme (3.5)-(3.10). Following the proof of Lemma 3.3, one gets, for all K ∈ M and all n ∈ [[0, N ]],
B(un+1 n+1 n n n+1 n+1 n+1

K >vK , uK >vK , (uL >vL )L∈NK , (ūσ )σ∈EK,ext ) ≤ 0, (3.25)
and
B(un+1 n+1 n n n+1 n+1 n+1

K ⊥vK , uK ⊥vK , (uL ⊥vL )L∈NK , (ūσ )σ∈EK,ext ) ≥ 0, (3.26)
which by substraction give
|un+1 n+1
− vK | − |unK − vK
n
| X · (q n+1 )+ |f (un+1 ) − f (v n+1 )| ¸
K K,σ K K
m(K) + n+1 −
δt n
−(qK,σ ) |f (un+1
K,σ ) − f (v n+1
K,σ )|
σ∈EK
X m(σ) (3.27)
− δ n+1 |ϕ(u) − ϕ(v)| ≤ 0.
dσ K,σ
σ∈EK
For a given n ∈ [[0, N ]], one sums (3.27) on K ∈ M and multiplies by δtn . All the exchange terms between
neighbouring control volume disappear, and because of the sign of the boundary terms (for σ ∈ EK,ext , since
un+1 n+1 n+1
K,σ = vK,σ , we get δK,σ |ϕ(u) − ϕ(v)| ≤ 0), one gets
X X
|un+1
K
n+1
− vK | m(K) ≤ |unK − vK
n
| m(K). (3.28)
K∈M K∈M
X
Since u0K = vK
0
, one concludes |un+1
K
n+1
− vK | m(K) = 0, for all n ∈ [[0, N ]], which concludes the proof of
K∈M
uniqueness. ¤
Let us now give two discrete estimates on the approximate solution uD which will be crucial in the convergence √
analysis. The first estimate (3.32) is a discrete L2 (0, T, H 1 (Ω)) estimate on the function ζ(uD ) where ζ 0 = ϕ0 .
This estimate will yield some compactness on ζ(uD ).
The second estimate is the weak BV inequality (3.32) on f (uD ). Such an inequality also holds for the continuous
problem with an additional diffusion term −ε∆f (u). This inequality does not give any compactness property (to
our knowledge, no BV estimate is known in the case of unstructured meshes); however it it plays an essential
role in the proof of convergence, where it is used to control the numerical diffusion introduced by the upstream
weighting scheme (see Section 3.2.4 and references [10], [12], [17] and [9]. Let us introduce a measure for the
regularity of the scheme.
µ ¶
dK,σ hK
reg(D) = max max , max . (3.29)
σ∈Eint ,K,L∈Mσ dL,σ K∈M,σ∈E K dK,σ
Lemma 3.5 (Discrete H 1 estimate and weak BV inequality)

Under Assumption 3.1, let D be a ∆−adapted pointed polygonal finite volume space-time discretization of Ω ×
(0, T ) in the sense of Definitions 1.4 and 1.7. Let ξ ∈ R be such that ξ ≥ reg(D); let (un+1
K )K∈M,n∈[[0,N ]] be the
solution of the scheme (3.5)-(3.10).
43
For all n ∈ N, un+1 ∈ HM (Ω) and g ∈ C 0 (R), we introduce the notation
δσn+1 g(u) = |g(uL ) − g(uK )|, ∀σ ∈ Eint , Mσ = {K, L},

(3.30)
δσn+1 g(u) = |g(ūn+1
σ ) − g(uK )|, ∀σ ∈ Eext , Mσ = {K}.
We also define
¯ R n+1 R ¯
¯ t ¯
qσn+1 = ¯ δt1n tn σ
q(x, t) · nK,σ dγ(x)dt ¯ , ∀σ ∈ Eint , Mσ = {K, L}. (3.31)
Then there exists a real number C > 0, only depending on Ω, T, u0 , ū, f, q, ϕ and ξ such that
N
X X m(σ)
(ND (ζ(uD )))2 = δtn (δσn+1 ζ(u))2 ≤ C
n=0
dσ
σ∈E
N
X X
(BD (f (uD )))2 = δtn n+1
qσn+1 (δK,σ f (u))2 ≤ C (3.32)
n=0 σ∈Eint
Proof. One first defines discrete values by averaging, in each control volume, the function ū, whose trace
on ∂Ω defines the Dirichlet boundary condition. Note that this proof uses ū ∈ H 1 (Ω × (0, T )) and not only
ū ∈ L2 (0, T ; H 1 (Ω)) and ūt ∈ L2 (0, T ; H −1 (Ω)), since we use below the fact that ūt ∈ L2 (0, T ; L1 (Ω)). Let
Z
1
ū0K = ū(x, 0)dx, ∀K ∈ M, (3.33)
m(K) K
Z tn+1 Z
1
ūn+1
K = n ū(x, t)dxdt, ∀K ∈ M, ∀n ∈ [[0, N ]], (3.34)
δt m(K) tn K
Setting v = u − ū, one multiplies (3.10) by δtn vK

n+1
and sums over K ∈ M and n ∈ [[0, N ]]. This yields
T1 + T2 + T3 = 0 with
N X
X
T1 = m(K)(un+1
K
n+1
− unK )vK , (3.35)
n=0 K∈M
N
X X X
T2 = δtn n+1 +
((qK,σ ) f (un+1 n+1 − n+1 n+1
K ) − (qK,σ ) f (uK,σ ))vK , (3.36)
n=0 K∈Mσ∈EK
N
X X X m(σ)
T3 = δtn δ n+1 ϕ(u) vK
n+1
. (3.37)
n=0
dσ K,σ
K∈Mσ∈EK
Using u = v + ū yields T1 = T4 + T5 with

N
1 X N +1 2 0 2 1X X n+1 n 2
T4 = m(K)((vK ) − (vK ) )+ m(K)(vK − vK ) (3.38)
2 2 n=0
K∈M K∈M
N X
X
T5 = m(K)(ūn+1
K
n+1
− ūnK )vK . (3.39)
n=0 K∈M
Setting
44
Z Z
1 1
An,K = ūn+1
K − n
ū(x, t )dx and Bn,K = ū(x, tn ) − ūnK ,
m(K) K m(K) K
one has
N X
X N X
X
n+1 n+1
T5 = m(K)An,K vK + m(K)Bn,K vK .
n=0 K∈M n=0 K∈M
By a classical density argument one gets:

1
|An,K | ≤ kūt kL1 (K×(tn ,tn+1 )) , ∀n ∈ [[0, N ]], ∀K ∈ M
m(K)
and
1
|Bn,K | ≤ kūt kL1 (K×(tn−1 ,tn )) , ∀n ∈ [[1, N ]], ∀K ∈ M
m(K)
(note that B0,K = 0 for all K ∈ M). Using these two inequalities and the L∞ stability of the scheme (Lemma
3.1) yields:
|T5 | ≤ 2kūt kL1 (Ω×(0,T )) (uS − uI ).

Now remarking that
1 X 0 2 1
T4 ≥ − m(K)vK ≥ − ku0 − ū(·, 0)k2L2 (Ω)
2 2
K∈M
the previous inequality allows us to obtain the existence of T6 > 0, only depending on Ω, T, u0 and ū, such that
T1 ≥ T6 .
The term T2 can be decomposed in T2 = T7 + T8 with
N
X X X
T7 = δtn n+1 +
((qK,σ ) f (un+1 n+1 − n+1 n+1
K ) − (qK,σ ) f (uK,σ ))uK ,
n=0 K∈Mσ∈EK
N
X X X
T8 = − δtn n+1 +
((qK,σ ) f (un+1 n+1 − n+1 n+1
K ) − (qK,σ ) f (uK,σ )) ūK ,
n=0 K∈Mσ∈EK
Using Remark 3.3, one gets
N
X X X
T7 = δtn n+1 − n+1
(qK,σ ) δK,σ f (u) un+1
K . (3.40)
n=0 K∈Mσ∈EK
Let g̃ be a primitive of f and g(s) = sf (s) − g̃(s) for all real s. The following inequality holds for all pairs of real
values (a, b) (see [20] and [9]).
1
g(b) − g(a) ≤ b(f (b) − f (a)) − (f (b) − f (a))2 (3.41)
2F
Using (3.41) for (a, b) = (un+1 n+1
L , uK ) and (3.40) yield
45
N
X X X 1
T7 ≥ δtn n+1 − n+1
(qK,L ) δK,σ g(u) + (BD (f (uD )))2 .
n=0
2F
K∈Mσ∈EK
Using Remark 3.3 with g instead of f gives
N
X X X
δtn n+1 − n+1
(qK,L ) δK,σ g(u) = 0, (3.42)
n=0 K∈Mσ∈EK
and therefore
1
T7 ≥ (BD (f (uD )))2 . (3.43)
2F
A discrete space integration by parts in T8 does not yield any boundary term since q · n = 0 on ∂Ω, and gives,
using the Cauchy-Schwarz inequality,
N
X X
T8 = − δtn n+1 +
((qK,L ) f (un+1 n+1 − n+1 n+1
K ) − (qK,L ) f (uL ))(ūK − ūn+1
L )
n=0 K|L∈Eint
N
X X
≥ −kqkL∞ (Ω×(0,T )) max |f (s)| δtn m(σ)δσn+1 ū
s∈[uI ,uS ]
n=0 σ∈Eint
Ã N
! 21
X X
n
≥ −kqkL∞ (Ω×(0,T )) max |f (s)|ND (ūD ) δt m(σ)dσ
s∈[uI ,uS ]
n=0 σ∈Eint
1
≥ −ND (ūD )kqkL∞ (Ω×(0,T )) max |f (s)|(d m(Ω) T ) 2 .
s∈[uI ,uS ]
The following estimate for ND (ūD ) holds:
ND (ūD ) ≤ F (ξ)kūkL2 (0,T,H 1 (Ω)) , (3.44)
where F ≥ 0 only depends on ξ (Inequality (3.44) is proven in [19], with a different definition of the regularity
factor of the mesh), leading to a lower bound of T8 denoted by C1 , only depending on Ω, T, u0 , ū, f, q and ξ.
There only remains to deal with T3 . A discrete space integration by parts, using the fact that
Vσn+1 = 0, ∀σ ∈ Eext , ∀n ∈ [[0, N ]], yields
N
X X X n+1 n+1
δK,σ ϕ(u) δK,σ v
T3 = δtn m(σ)dK,σ . (3.45)
n=0
dσ dσ
K∈Mσ∈EK
Writing again v into u − ū leads to T3 = T9 + T10 where
N
X X X n+1 n+1
δK,σ ϕ(u) δK,σ u
T9 = δtn m(σ)dK,σ (3.46)
n=0
dσ dσ
K∈Mσ∈EK
N
X X X n+1 n+1
δK,σ ϕ(u) δK,σ ū
T10 = − δtn m(σ)dK,σ (3.47)
n=0
dσ dσ
K∈Mσ∈EK
46
One has
√ for0 all pairs of real numbers (a, b) the inequality (ζ(a) − ζ(b))2 ≤ (a − b)(ϕ(a) − ϕ(b)). Also using
0 0
ϕ ≤ Φζ (recall that Φ = kϕ k∞ ), one gets
T9 ≥ (ND (ζ(uD )))2 , (3.48)

√
T10 ≥ − ΦND (ζ(uD ))ND (ūD ). (3.49)
Using the Young inequality and (3.44), one gets the existence of C2 only depending on Ω, T, u0 , ū, f, q, ϕ and ξ
such that
1
T10 ≥ − (ND (ζ(uD )))2 + C2 . (3.50)
2
Gathering the previous inequalities, one gets
1 1
T6 + (BD (f (uD )))2 + T1 + (ND (ζ(uD )))2 + C2 ≤ 0, (3.51)
2F 2
which completes the proof. ¤
√
Remarking that from the estimate of Lemma 2 in [19], one has ND (ζ(ūD )) ≤ ΦCkūkL2 (0,T,H 1 (Ω)) , where
C ≥ 0 only depends on ξ, one gets
Corollary 3.2 (Discrete H01 estimate) Under Assumption 3.1, let D be a ∆−adapted pointed polygonal finite
volume space-time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7. Let ξ ∈ R be such that ξ ≥
reg(D), let u = (un+1 n+1
K )K∈M,n∈[[0,N ]] be the solution of the scheme (3.5)-(3.10) and let ū = (ūK )K∈M,n∈[[0,N ]]
0
be defined by (3.34). Then, setting z = ζ(u) − ζ(ū), there exists C ∈ R+ , only depending on Ω, T, u0 , ū, ϕ, q, f
and ξ such that
N
Ã n+1
!2 N
X X X δK,σ z X X m(σ) ¡ ¢2
n
δt m(σ)dK,σ = δtn δσn+1 z ≤ C 0 (3.52)
n=0
dσ n=0
dσ
K∈Mσ∈EK σ∈E
3.2.3 Compactness of a family of approximate solutions

From Lemma 3.1, we know that for any sequence of ∆−adapted pointed polygonal finite volume space-time dis-
cretizations (Dm )m∈N , of Ω×(0, T ) in the sense of Definitions 1.4 and 1.7, the associated sequence of approximate
solutions (uDm )m∈N is bounded in L∞ (Ω × (0, T )). Therefore one may extract a subsequence which converges
for the weak star topology of L∞ (Ω × (0, T )) as m tends to infinity. This convergence is unfortunately insufficient
to pass to the limit in the nonlinearities. In order to pass to the limit, we shall use two tools:
1. the nonlinear weak star convergence which was introduced in [17] and which is equivalent to the notion of
convergence towards a Young measure as developped in [15].
2. Kolmogorov’s compactness theorem, which was used in [19] in the case of a semilinear elliptic equation.
Let us now show that we are in position to apply the Kolmogorov’s compactness theorem to (ζ(uDm ))m∈N . From
the discrete estimates Lemma 3.5 and Corollary 3.2, one can state the following continuous estimates on zD , where
zD is defined almost everywhere in Ω × (0, T ) by
zD (x, t) = ζ(un+1 n+1 n n+1

K ) − ζ(ūK ) for x ∈ K and t ∈ (t , t ) (3.53)
where (un+1 n+1
K )K∈M,n∈[[0,N ]] is the solution to (3.5)-(3.10) and (ūK )K∈M,n∈[[0,N ]] is defined by (3.34).
47
Corollary 3.3 (Space and time translates estimates) Under Assumption 3.1, let D be a ∆−adapted pointed
polygonal finite volume space-time discretization of Ω × (0, T ) in the sense of Definitions 1.4 and 1.7. Let ξ
be a real number such that ξ ≥ reg(D); let u be the solution of scheme (3.5)-(3.10), and let uD be defined by
(3.12). Let ū be defined by (3.34), let zD be defined by (3.53), and be prolonged by zero on (0, T ) × Ωc . Then there
exists C1 only depending on Ω, T, u0 , ū, ϕ, q, f and ξ, and there exists C0 , only depending on Ω, such that
Z T Z
d
∀ξ ∈ R , (zD (x + ξ, t) − zD (x, t))2 dxdt ≤ C1 |ξ|(|ξ| + C0 h(M)), (3.54)
0 Rd
and there exists C2 only depending on Ω, T, u0 , ū, ϕ, q, f and ξ such that
Z T −s Z
∀s > 0, (ζ(uD )(x, t + s) − ζ(uD )(x, t))2 dxdt ≤ C2 s. (3.55)
0 Rd
The use of space translate estimates for the study of numerical schemes for elliptic problems was recently intro-
duced in [19]. The technique of [19] may easily be adapted here to prove (3.54), using the estimates of Corollary
3.2. A time translate estimate was introduced in [16] to obtain some compactness in the study of finite volume
schemes for parabolic equations. The proof of (3.55) follows the technique of [16] and uses estimate (3.32) and
the discrete equation (3.10).
From Kolmogorov Copmpactness Theorem and the estimates (3.54) and (3.55) of Corollary 3.3 we deduce the
following compactness result:
Corollary 3.4 (Compactness of a family of approximate solutions)

Let (Dm )m∈N be a sequence of ∆−adapted pointed polygonal finite volume space-time discretization of Ω×(0, T )
in the sense of Definitions 1.4 and 1.7 such that there exists ξ ∈ R with ξ ≥ reg(Dm ) for all m ∈ N. For all m ∈ N,
let uDm be defined by the scheme (3.5)-(3.10) and (3.12) with D = Dm , and let zDm be defined by (3.53) with
D = Dm and (3.34). Then there exists u ∈ L∞ (Ω × (0, T ) × (0, 1)) and z ∈ L2 (Ω × (0, T )) such that, up to a
subsequence, uDm tends to u in the nonlinear weak star sense and zDm tends to z in L2 (Ω × (0, T )) as m → ∞.
Furthermore one has z ∈ L2 (Ω × (0, T ))(0, T, H01 (Ω)), ζ(u) = z + ζ(ū), and ζ(u) = ζ(ū) a.e. on ∂Ω.
Proof. The convergence of uDm towards u ∈ L∞ (Ω × (0, T ) × (0, 1)) in the nonlinear weak star sense is a
consequence of Lemma 3.1 and Theorem 3.2, stated at the end of the present proof. The convergence of zDm to z
in L2 (Ω × (0, T ) is a consequence of Kolmogorov’s compactness theorem and the estimates (3.54) and (3.55) of
Corollary 3.3.
Following [20] or [19], one then deduces from (3.55) that Di z ∈ L2 (Ω × (0, T )) for i = 1, . . . , d and since
zDm (x, t) = 0 on Ωc × (0, T ) for all m ∈ N, one has z ∈ L2 (Ω × (0, T ))(0, T, H01 (Ω)).
Now since uDm converges to u in the nonlinear weak star sense and that the function ūDm defined a.e. by
ūDm (x, t) = ūn+1
K for (x, t) in K × (tn , tn+1 ) converges uniformly to ū, one deduces that ζ(uDm ) converges
to ζ(u) in the nonlinear weak star sense and to z + ζ(ū) in L2 (Ω × (0, T )) as m tends to infinity. Therefore,
by Lemma 3.6 below, one obtains that ζ(u) = z + ζ(ū) and ζ(u) does not depend on α. Furthermore, since
z ∈ L2 (Ω × (0, T ))(0, T, H01 (Ω)), it follows that ζ(u) = ζ(ū) a.e. on ∂Ω which ends the proof of the corollary.
¤
Theorem 3.2 (Nonlinear weak star convergence) Let Q be a Borelian subset of Rk , k ∈ N? , and (un )n∈N be a
bounded sequence in L∞ (Ω × (0, T ))(Q). Then there exists u ∈ L∞ (Ω × (0, T ))(Q × (0, 1)), such that up to a
subsequence, un tends to u “in the nonlinear weak star sense” as n → ∞, i.e.:
Z 1
∀g ∈ C(R, R), g(un ) * g(u(·, α))dα for the weak star topology of L∞ (Ω × (0, T ))(Q) as n → ∞. (3.56)
0
We refer to [15; 17] for details and proof of Theorem 3.2.

48
Lemma 3.6 Let Q be a Borelian subset of Rk , k ∈ N? , and let (un )n∈N ⊂ L∞ (Q) be such that un converges to
u ∈ L∞ (Ω × (0, T ))(Q × (0, 1)) in the nonlinear weak star sense, and to w in L2 (Q), as n tends to infinity, then
u(x, α) = w(x), for a.e. (x, α) ∈ Q × (0, 1) and u does not depend on α.
Proof. With the notations of the lemma, we have
Z 1Z Z 1Z Z 1 Z Z 1 Z
(u(x, α) − w(x))2 dxdα = (u(x, α))2 dxdα − 2 u(x, α)w(x)dxdα + w(x)2 dxdα.
0 Q 0 Q 0 Q 0 Q
Since un tends to u in the nonlinear weak star sense , one has

Z 1Z Z Z 1Z Z
(u(x, α))2 dxdα = lim (un (x))2 dx and u(x, α)w(x)dxdα = lim un (x)w(x)dx,
0 Q n→+∞ Q 0 Q n→+∞ Q
and since un tends to w in L2 (Q), one deduces that u(x, α) = w(x), for a.e. (x, α) ∈ Q × (0, 1) and u does not
depend on α. ¤
3.2.4 Convergence towards an entropy process solution

This section is mainly devoted to the proof of the convergence theorem 3.3, which states the convergence of the
approximate solution to a measure valued solution as introduced in [15], which is also called entropy process
solution [17], and defined as follows.
Definition 3.3 Under Assumption 3.1, an entropy process solution to (3.1)-(3.3) is a function u such that,
u ∈ L∞ (Ω × (0, T ) × (0, 1)), (3.57)

2
ϕ(u) − ϕ(u) ∈ L (0, T ; H01 (Ω)), (3.58)
(note that ϕ(u) does not depend on α), and u satisfies the following inequalities :
1. Regular convex entropy inequalities :
 R1 
Z µ(u(x, t, α))dα ψt (x, t)+ Z
 R 1 ν(u(x, t, α))dα q(x, t) · ∇ψ(x, t)
0

 0  dxdt + µ(u0 (x))ψ(x, 0)dx ≥ 0,
 −∇η(ϕ(u)(x, t)) · ∇ψ(x, t) 
Ω×(0,T ) Ω (3.59)
−η 00 (ϕ(u)(x, t))(∇ϕ(u)(x, t))2 ψ(x, t)
∀ ψ ∈ D+ (Ω × [0, T )), ∀η ∈ C 2 (R, R), η 00 ≥ 0, µ0 = η 0 (ϕ(·)), ν 0 = η 0 (ϕ(·))f 0 (·).
2. Kruzkov’s entropy inequalities :

 R1 
Z |u(x, t, α) − κ|dα ψt (x, t)+
 R 1 (f (u(x, t, α)>κ) − f (u(x, t, α)⊥κ))dα q(x, t) · ∇ψ(x, t)  dxdt
0
0
Ω×(0,T )
Z −∇|ϕ(u)(x, t) − ϕ(κ)| · ∇ψ(x, t)
+ |u0 (x) − κ|ψ(x, 0)dx ≥ 0,
Ω
∀ ψ ∈ D+ (Ω × [0, T )), ∀κ ∈ R.
In the previous definition, we use two types of entropies, since in the proof (given below) of the uniqueness theorem
one should make use of terms η 00 (ϕ(u)). In [8], these terms are obtained from the equation satisfied by a weak
solution, which itself can be obtained from the Krushkov entropy inequalities. We have prefered here to keep this
slightly more complex definition since the following theorem shows that (3.59) and (3.60) are both obtained by the
natural limit of the approximate solutions.
49
Theorem 3.3 (Convergence towards an entropy process solution) Under Assumption 3.1, let (Dm )m∈N be a
sequence of ∆−adapted pointed polygonal finite volume space-time discretization of Ω × (0, T ) in the sense of
Definitions 1.4 and 1.7, with h(Dm ) → 0 as m → ∞, such that there exists ξ ∈ R with ξ ≥ reg(Dm ) for all
m ∈ N. For all m ∈ N, let uDm be defined by the scheme (3.5)-(3.10) and (3.12) with D = Dm .
Then, there exists an entropy processus solution of (3.1)-(3.3) in the sense of Definition 3.3 and a subsequence of
(uDm )m∈N , again denoted by (uDm )m∈N , such that (uDm )m∈N converges to u in the nonlinear weak star sense
and (ζ(uDm ))m∈N converges in L2 (Ω × (0, T )) to ζ(u) ∈ L2 (0, T ; H 1 (Ω)) as m tends to ∞.
Proof. By Lemma 3.4, there exist u ∈ L∞ (Ω × (0, T ) × (0, 1)) and a subsequence of (uDm )m∈N , again denoted
(uDm )m∈N , such that (uDm )m∈N converges to u in the nonlinear weak star sense and (ζ(uDm ))m∈N converges in
L2 (Ω × (0, T )) to ζ(u) ∈ L2 (0, T ; H 1 (Ω)). There remains to show that the function u ∈ L∞ (Ω × (0, T ) × (0, 1))
is an entropy process solution.
A number of the arguments involved in order to do so may be found in [17] or [16] and therefore will be given
with
Z few details. The main new argument introduced here concerns the term
η 00 (ϕ(u)(x, t))(∇ϕ(u)(x, t))2 ψ(x, t)dxdt in equation (3.59). The passage to the limit to obtain this
Ω×(0,T )
nonlinearity motivates the use of Lemma 5.13 (a related technique was used in [27] in the case of a variational
inequality).
The idea of the proof is to derive the continuous inequalities (3.59) and (3.60) for the limit u by mutliplying
the discrete entropy inequalities (3.17) and (3.19) by regular test functions and passing to the limit. Indeed, let
ψ ∈ D+ (Ω × [0, T )) = {ψ ∈ Cc∞ (Ω × R, R+ ), ψ(·, T ) = 0}. For a given m, let us denote D = Dm , and let
(un+1 n
K )K∈M,n∈[[0,N ]] be the solution of the scheme (3.5)-(3.10) associated to D. Let Ψ = (ΨK )K∈M,n∈[[0,N +1]]
be defined by
ΨnK = ψ(xK , tn ) ∀K ∈ M, ∀n ∈ [[0, N + 1]]. (3.60)
Remark 3.5 One cannot use for ΨnK the mean value of ψ on K × (tn , tn+1 ); indeed, in order to pass to the limit
D Ψn −Ψn
on the term T13 below (see (3.65) and (3.66)), we shall use the consistency of the approximation d(xKK ,xLL) to the
normal derivative ∇ψ · nK,L . This consistency holds if ΨnK = ψ(xK , tn ) thanks to the assumption on the family
(xK )K∈M in Definition 1.4, but does not generally hold if ΨnK is the mean value of ψ on K × (tn , tn+1 ). Note
that discrete values using the mean values were used for ū when studying an upper bound of ND (ū) with respect
to the L2 (0, T ; H 1 (Ω)) norm of ū. However we did not have to use the consistency of the flux on ū.
With the notations of lemmas 3.2 and 3.3, let us multiply the discrete entropy inequalities (3.17) and (3.19) by
δtn ΨnK and sum over K ∈ M and n ∈ [[0, N ]]. From (3.17), one gets
D D D D
T11 + T12 + T13 + T14 ≤0 (3.61)
D D D D
where T11 , T12 , T13 and T14 are defined by
N
X X µ(un+1 n
K ) − µ(uK ) n
D
T11 = δtn m(K) n ΨK
n=0
δt
K∈M
XN X X
D
T12 = − δtn n+1 − n+1
(qK,σ ) δK,σ ν(u) ΨnK
n=0 K∈M σ∈EK
XN X X m(σ)
D
T13 = − δtn δ n+1 η(ϕ(u)) ΨnK
n=0
dσ K,σ
K∈M σ∈EK
XN X 1 X m(σ) 00
D
T14 = δtn un+1
η (ϕ(e n+1 2 n
K,σ ))(δK,σ ϕ(u)) ΨK
n=0
2 dσ
K∈M σ∈EK
50
Each of these terms will be shown to converge to the corresponding continuous terms of Inequality (3.59) by
passing to the limit on the space and time steps, i.e. letting m → ∞.
Since ψ(·, T ) = 0, one has ΨN
K
+1
= 0 and therefore:
N
X X ΨnK − Ψn+1
D
T11 = δtn m(K)µ(un+1
K )
K
n=0
δtn
K∈M
X
− m(K)Ψ0K µ(u0K )
K∈M
R1
The sequence µ(uD ) converges weakly to 0
µ(u(·, α))dα as m → ∞. Let χD be the function defined almost
n+1
ΨnK −ΨK
everywhere on Ω × (0, T ) by χD (x, t) = (x, t) ∈ K × (tn , tn+1 ); then χD converges to ψt in
δtn if
1 0
L (Ω × (0, T )) as m → +∞. Furthermore, let ψM (resp 0
u0M ) be defined almost everywhere on Ω by ψM = Ψ0K
0 0 0 p 0
(resp. uM = uK ) if x ∈ K. Then, µ(uM ) converges to µ(u0 ) in L (Ω) for any p ∈ [1, +∞) and ψM converges
D
to ψ(., 0) uniformly as m → +∞. Hence passing to the limit as m → +∞ in T11 yields:
Z T Z Z 1 Z T Z
Dm
lim T11 =− µ(u(x, t, α))dα ψt (x, t)dxdt − µ(u0 (x))ψ(x, 0)dx. (3.62)
m→∞ 0 Ω 0 0 Ω
D
Let us now rewrite T12 as:
N
X X X
D
T12 = − δtn ν(un+1
K )
n+1 + n
((qK,σ n+1 − n
) ΨK,σ − (qK,σ ) ΨK ). (3.63)
n=0 K∈M σ∈EK
n+1 + n n+1 − n
R tn+1 R
We replace the term (qK,σ ) ΨK,σ − (qK,σ ) ΨK by δt1n tn σ
ψ(x, t) q(x, t) · nK,σ dγ(x)dt. When doing so,
we commit an error which may be controlled (see the details in [17]) thanks to the consistency and the conserva-
tivity of the scheme and thanks to the the weak BV inequality (3.32). Using the weak convergence of ν(uM ) to
R1
0
ν(u(·, α))dα as m → ∞, we then obtain:
Z TZ Z 1
lim T Dm = − ν(u(x, t, α))dα∇(q(x, t) ψ(x, t))dxdt
m→∞ 12 0 Ω 0
Z TZ Z 1
= − ν(u(x, t, α))dα q(x, t) · ∇ψ(x, t)dxdt. (3.64)
0 Ω 0
D
Turning now to the study of T13 , one remarks that for h(M) small enough, the support of ψ does not intersect the
control volumes with edges on ∂Ω. Then for all control volumes K ∈ M the sum over σ ∈ EK,ext vanishes and
thus
N
X X X δK,σ η(ϕ(u)) ΨnK,σ − ΨnK
D
T13 = − δtn m(σ)dK,σ (3.65)
n=0
dσ dσ
K∈Mσ∈EK
Using the straightforward extension of Lemma 5.13 to time-dependent problems, one gets with computations
similar as in [19]:
Z TZ Z TZ
lim T Dm =− η(ϕ(u))(x, t)∆ψ(x, t)dxdt = ∇η(ϕ(u))(x, t) · ∇ψ(x, t)dxdt. (3.66)
m→∞ 13 0 Ω 0 Ω
51
D
One now deals with T14 . We remark that
1 X X m(σ) 00
η (ϕ(eun+1 n+1 2 n
K,σ ))(δK,σ ϕ(u)) ΨK
2 dσ
K∈Mσ∈EK
Ã n+1 !2
1 X X 00 n+1
δK,σ ϕ(u)
= m(σ)dσ η (ϕ(e uK,σ )) ΨnK
2 dσ
K∈Mσ∈EK
Ã n+1 !2
X X δK,σ ϕ(u) ΨnK + ΨnK,σ
00 n+1
= m(σ)dK,σ η (ϕ(e uK,σ ))
dσ 2
K∈Mσ∈EK
Ψn n
K +ΨK,σ
Thanks to the strong convergence of the function defined in LD (Ω) × (0, T ) by the value η 00 (ϕ(un+1
K,σ )) 2
to η 00 (ϕ(u))ψ, the application of Lemma 5.13 provides
Z TZ
Dm
lim inf T14 ≥ (∇ϕ(u)(x, t))2 η 00 (ϕ(u)(x, t))ψ(x, t)dxdt. (3.67)
m→∞ 0 Ω
Gathering (3.61), (3.62), (3.64), (3.66) and (3.67), the proof that u verifies (3.59) is therefore complete.
The same steps are completed in a similar way in order to show that u satisfies (3.60), without the difficult problem
of the treatment of η 00 . This also completes the proof of Theorem 3.3. ¤
To complete the proof of Theorem 3.1 there only remains to show the uniqueness of an entropy process solution.
This is the aim of Section 3.2.5.
3.2.5 Uniqueness of the entropy process solution.

One proves in this section the following theorem.
Theorem 3.4 (Uniqueness of the entropy process solution) Under Assumption 3.1, let u and v be two entropy
process solutions to (3.1)-(3.3) in the sense of Definition 3.3. Then there exists a unique function w ∈ L∞ (Ω ×
(0, T )) such that u(x, t, α) = v(x, t, β) = w(x, t), for almost every (x, t, α, β) ∈ Ω × (0, T ) × (0, 1) × (0, 1).
Proof.
This proof uses on the one hand Carrillo’s handling of Krushkov entropies, on the other hand the concept of
entropy process solution, which allows the use of the theorem of continuity in means, necessary to pass to the limit
on mollifiers. Note that the hypothesis (3.4) makes it easier to handle the boundary conditions.
In order to prove Theorem 3.4, one defines for all ε > 0 a regularization Sε ∈ C 1 (R, R) of the function sign given
by
Sε (a) = −1, ∀a ∈ (−∞, −ε],

3ε2 a−a3
Sε (a) = 2 ε3 , ∀a ∈ [−ε, ε], (3.68)
Sε (a) = 1, ∀a ∈ [ε, +∞).
One defines Rϕ = {a ∈ R, ∀b ∈ R \ {a}, ϕ(b) 6= ϕ(a)}. Note that ϕ(R \ Rϕ ) is countable, because for all
s ∈ ϕ(R \ Rϕ ), there exists (a, b) ∈ R2 with a < b and ϕ((a, b)) = {s}, and therefore there exists at least one
r ∈ Q with r ∈ (a, b) verifying ϕ(r) = s.
52
Let κ ∈ RRϕ . Let ε > 0 and let u be an entropy processus

R a 0 solution. One introduces Rina (3.59) the function
a 0
ηε,κ (a) = ϕ(κ) Sε (s − ϕ(κ))ds. One defines µε,κ (a) = κ ηε,κ (ϕ(s))ds and νε,κ (a) = κ ηε,κ (ϕ(s))f 0 (s)ds,
for all a ∈ R. Using the dominated convergence theorem, one gets for all a ∈ R that lim ηε,κ (a) = |a − ϕ(κ)|,
ε→0
and, since κ ∈ Rϕ , lim µε,κ (a) = |a − κ| and lim νε,κ (a) = f (a>κ) − f (a⊥κ). One gets for all ψ ∈ D+ (Ω ×
ε→0 ε→0
[0, T )),
 R1 
Z |u(x, t, α) − κ|dα ψt (x, t)
0R
 + 1 (f (u(x, t, α)>κ) − f (u(x, t, α)⊥κ))dα q(x, t) · ∇ψ(x, t)  dxdt
0
Ω×(0,T )
Z −Sε (ϕ(u)(x, t) − ϕ(κ))∇ϕ(u)(x, t) · ∇ψ(x, t)
£ 0 ¤
− Sε (ϕ(u)(x, t) − ϕ(κ))(∇ϕ(u))2 (x, t)ψ(x, t) dxdt
ZΩ×(0,T )
+ |u0 (x) − κ|ψ(x, 0)dx ≥ A(ε, u, κ, ψ),
Ω
(3.69)
where for any entropy process solution u, any ψ ∈ D+ (Ω × [0, T )), any κ ∈ Rϕ and any ε > 0, A(ε, u, κ, ψ) is
defined by
 R ³ ´ 
1
Z 0
|u(x, t, α) − κ| − µε,κ (u(x, t, α)) dα ψt (x, t)+
 R ³ ´ 
A(ε, u, κ, ψ) =  1  dxdt
 0 (f (u(x, t, α)>κ) − f (u(x, t, α)⊥κ)) − νε,κ (u(x, t, α)) dα 
Ω×(0,T )
Z ³ q(x, t) · ∇ψ(x, t)
´
+ |u0 (x) − κ| − µε,κ (u0 (x)) ψ(x, 0)dx.
Ω
(3.70)
Thanks to the dominated convergence theorem, one has
lim A(ε, u, κ, ψ) = 0. (3.71)

ε→0
This convergence is not uniform w.r.t. κ (even if κ remains bounded), but A(ε, u, κ, ψ) remains bounded (for a
given u) if κ, ψ, ψt and ∇ψ remain bounded and if the support of ψ remains in a fixed compact set of Rd × [0, T ).
Using (3.60), one now remarks that, for all κ ∈ R, one has for all ψ ∈ D+ (Ω × [0, T )),
 R1 
Z |u(x, t, α) − κ|dα ψt (x, t)+
 R 1 (f (u(x, t, α)>κ) − f (u(x, t, α)⊥κ))dα
0

 0  dxdt
 q(x, t) · ∇ψ(x, t) 
Ω×(0,T ) (3.72)
−Sε (ϕ(u)(x, t) − ϕ(κ))∇ϕ(u)(x, t) · ∇ψ(x, t)
R
+ Ω
|u0 (x) − κ|ψ(x, 0)dx ≥ B(ε, u, κ, ψ),
+
where for an entropy process solution u, all ψ ∈ D (Ω × [0, T )), all κ ∈ R and all ε > 0, B(ε, u, κ, ψ) is defined
by
Z h ³ ´ i
B(ε, u, κ, ψ) = ∇ |ϕ(u)(x, t) − ϕ(κ)| − ηε,κ (ϕ(u)(x, t)) · ∇ψ(x, t) dxdt. (3.73)
Ω×(0,T )
+
For all ψ ∈ D (Ω × [0, T )), one has
Z h³ ´ i
B(ε, u, κ, ψ) = − |ϕ(u)(x, t) − ϕ(κ)| − ηε,κ (ϕ(u)(x, t)) ∆ψ(x, t) dxdt, (3.74)
Ω×(0,T )
and
53
lim B(ε, u, κ, ψ) = 0, (3.75)

ε→0
for all ψ ∈ D+ (Ω × [0, T )), ε > 0 and κ ∈ R.

As for the study of A, the quantity B(ε, u, κ, ψ) remains bounded (for a given u) if κ and ∆ψ remain bounded and
if the support of ψ remains in a fixed compact set of Rd × [0, T ).
Let u and v be two entropy process solutions in the sense of Definition 3.3. One defines the sets Eu = {(x, t) ∈
Ω × (0, T ), u(x, t, α) ∈ Rϕ , for a.e. α ∈ (0, 1)} and Ev = {(x, t) ∈ Ω × (0, T ), v(x, t, α) ∈ Rϕ , for
a.e. α ∈ (0, 1)}. Indeed, recall that ϕ(u) and ϕ(v) do not depend of α ∈ (0, 1). Then, Ω × (0, T ) \ Eu =
∪s∈ϕ(R\Rϕ ) Es,u with Es,u = {(x, t) ∈ Ω × (0, T ), ϕ(u)(x, t) = s} (the same property is available for v). Let
ξ ∈ Cc∞ (Rd × R × Rd × R, R+ ) such that, for all (x, t) ∈ Ω × [0, T ), ξ(x, t, ·, ·) ∈ D+ (Ω × [0, T )) and for all
(y, s) ∈ Ω × [0, T ), ξ(·, ·, y, s) ∈ D+ (Ω × [0, T )). One introduces in (3.69), for (y, s) ∈ Ev , and a.e. β ∈ (0, 1),
κ = v(y, s, β) and ψ = ξ(·, ·, y, s). One integrates the result on Ev × (0, 1). One then gets
 R1R1 
Z Z |u(x, t, α) − v(y, s, β)|dαdβ ξt (x, t, y, s)+
 R 1 R 1 (f (u(x, t, α)>v(y, s, β)) − f (u(x, t, α)⊥v(y, s, β)))dαdβ 
0 0
 0 0  dxdtdyds
 q(x, t) · ∇ ξ(x, t, y, s) 
Ev Ω×(0,T ) x
Z Z −Sε (ϕ(u)(x, t) − ϕ(v)(y, s))∇ϕ(u)(x, t) · ∇x ξ(x, t, y, s)

£ 0 ¤
− Sε (ϕ(u)(x, t) − ϕ(v)(y, s))(∇ϕ(u))2 (x, t)ξ(x, t, y, s) dxdtdyds (3.76)
ZEv ZΩ×(0,T
Z 1 )
+ |u0 (x) − v(y, s, β)|ξ(x, 0, y, s)dβdxdyds
ZE1v Z Ω 0
≥ A(ε, u, v(y, s, β), ξ(·, ·, y, s))dydsdβ.
0 Ev
One introduces in (3.72), for (y, s) ∈ Ω × (0, T ) \ Ev , and any β ∈ (0, 1), κ = v(y, s, β) and ψ = ξ(·, ·, y, s).
One integrates the result on (Ω × (0, T ) \ Ev ) × (0, 1). One then gets
 R1R1 
Z Z |u(x, t, α) − v(y, s, β)|dαdβ ξt (x, t, y, s)
 + R 1 R 1 (f (u(x, t, α)>v(y, s, β)) − f (u(x, t, α)⊥v(y, s, β)))dαdβ
0 0

 q(x, t) · ∇x ξ(x, t, y, s) 
Ω×(0,T )\Ev Ω×(0,T )
−Sε (ϕ(u)(x, t) − ϕ(v)(y, s))∇ϕ(u)(x, t) · ∇x ξ(x, t, y, s)
Z Z Z 1
ZΩ×(0,T
1Z
)\Ev Ω 0
≥ B(ε, u, v(y, s, β), ξ(·, ·, y, s))dydsdβ.

0 Ω×(0,T )\Ev
(3.77)
Adding (3.76) and (3.77) gives
54
 R1R1 
Z Z |u(x, t, α) − v(y, s, β)|dαdβ ξ t (x, t, y, s)
 + R 1 R 1 (f (u(x, t, α)>v(y, s, β)) − f (u(x, t, α)⊥v(y, s, β)))dαdβ 
0 0
 q(x, t) · ∇x ξ(x, t, y, s) 
Ω×(0,T ) Ω×(0,T )
Z Z −Sε (ϕ(u)(x, t) − ϕ(v)(y, s))∇ϕ(u)(x, t) · ∇x ξ(x, t, y, s)

£ 0 ¤
− Sε (ϕ(u)(x, t) − ϕ(v)(y, s))(∇ϕ(u))2 (x, t)ξ(x, t, y, s) dxdtdyds
Z Ev Ω×(0,T
Z Z) 1
ZΩ×(0,T
1 Z ) Ω 0
Z 1Z
≥ A(ε, u, v(y, s, β), ξ(·, ·, y, s))dydsdβ + B(ε, u, v(y, s, β), ξ(·, ·, y, s))dydsdβ
0 Ev 0 Ω×(0,T )\Ev
(3.78)
One now exchanges the roles of u and v, and add the resulting equations. It gives
T15 + T16 + T17 (ε) + T18 (ε) + T19 (ε) ≥ T20 (ε), (3.79)
where T15 , T16 , T17 , T18 , T19 and T20 are defined by
T15 =
 R1R1 
Z Z 0R 0R
|u(x, t, α) − v(y, s, β)|dαdβ (ξt (x, t, y, s) + ξs (x, t, y, s))
 + 1 1 (f (u(x, t, α)>v(y, s, β)) − f (u(x, t, α)⊥v(y, s, β)))dαdβ  (3.80)
 ³ 0 0 ´  dxdtdyds,
Ω×(0,T ) Ω×(0,T )
q(x, t) · ∇x ξ(x, t, y, s) + q(y, s) · ∇y ξ(x, t, y, s)
T =
Z16 Z Z 1
|u0 (x) − v(y, s, β)|ξ(x, 0, y, s)dβdxdyds
Ω×(0,T ) Ω (3.81)
Z Z 0Z 1
+ |u0 (y) − u(x, t, α)|ξ(x, t, y, 0)dαdydxdt,
Ω×(0,T ) Ω 0
T3Z(ε) = Z · ¸
Sε (ϕ(u)(x, t) − ϕ(v)(y, s))∇ϕ(u)(x, t)·
− dxdtdyds
(∇x ξ(x, t, y, s) + ∇y ξ(x, t, y, s)) (3.82)
ZΩ×(0,T ) ZΩ×(0,T ) · ¸
Sε (ϕ(v)(y, s) − ϕ(u)(x, t))∇ϕ(v)(y, s)·
− dxdtdyds,
Ω×(0,T ) Ω×(0,T ) (∇x ξ(x, t, y, s) + ∇y ξ(x, t, y, s))
T
Z4 (ε) = Z
[Sε (ϕ(u)(x, t) − ϕ(v)(y, s))∇ϕ(u)(x, t) · ∇y ξ(x, t, y, s)] dxdtdyds
Z
Ω×(0,T Z
) Ω×(0,T ) (3.83)
+ [Sε (ϕ(v)(y, s) − ϕ(u)(x, t))∇ϕ(v)(y, s) · ∇x ξ(x, t, y, s)] dxdtdyds,
Ω×(0,T ) Ω×(0,T )
T19Z(ε) Z=
£ 0 ¤
− Sε (ϕ(u)(x, t) − ϕ(v)(y, s))(∇ϕ(u))2 (x, t)ξ(x, t, y, s) dxdtdyds
Z Ev Z )
Ω×(0,T (3.84)
£ 0 ¤
− Sε (ϕ(u)(x, t) − ϕ(v)(y, s))(∇ϕ(v))2 (y, s)ξ(x, t, y, s) dxdtdyds,
Ω×(0,T ) Eu
55
and
T20 (ε) =
Z 1Z
A(ε, u, v(y, s, β), ξ(·, ·, y, s))dydsdβ
0Z EZ v
1
+ B(ε, u, v(y, s, β), ξ(·, ·, y, s))dydsdβ
(3.85)
Z0 1 ZΩ×(0,T )\Ev
+ A(ε, v, u(x, t, α), ξ(x, t, ·, ·))dxdtdα
0
Z 1Z E u
+ B(ε, v, u(x, t, α), ξ(x, t, ·, ·))dxdtdα.

0 Ω×(0,T )\Eu
By an integration by parts in (3.83) and using the fact that ξ vanishes on ∂Ω × (0, T ) × Ω × (0, T ) and on
Ω × (0, T ) × ∂Ω × (0, T ) one gets
T
Z4 (ε) = Z
[Sε0 (ϕ(u)(x, t) − ϕ(v)(y, s))ξ(x, t, y, s)∇ϕ(u)(x, t) · ∇ϕ(v)(y, s)] dxdtdyds
Z
Ω×(0,T Z
) Ω×(0,T ) (3.86)
+ [Sε0 (ϕ(v)(y, s) − ϕ(u)(x, t))ξ(x, t, y, s)∇ϕ(v)(y, s) · ∇ϕ(u)(x, t)] dxdtdyds.
Ω×(0,T ) Ω×(0,T )
Recall that Es,u = {(x, t) ∈ Ω × (0, T ), ϕ(u)(x, t) = s} for all s ∈ R. One has ∇ϕ(u) = 0 a.e. on Es,u (see [5]
for instance). Since Ω × (0, T ) \ Eu = ∪s∈ϕ(R\Rϕ ) Es,u , and since ϕ(R \ Rϕ ) is countable, the following equations
hold.
∇ϕ(u) = 0, a.e. on Ω × (0, T ) \ Eu (3.87)

and
∇ϕ(v) = 0, a.e. on Ω × (0, T ) \ Ev . (3.88)

It leads to
Z
T4 (ε) = [Sε0 (ϕ(u)(x, t) − ϕ(v)(y, s))ξ(x, t, y, s)∇ϕ(u)(x, t) · ∇ϕ(v)(y, s)] dxdtdyds
EZ
u ×Ev (3.89)
+ [Sε0 (ϕ(v)(y, s) − ϕ(u)(x, t))ξ(x, t, y, s)∇ϕ(v)(y, s) · ∇ϕ(u)(x, t)] dxdtdyds
Eu ×Ev
and
Z
£ 0 ¤
T19 (ε) = − Sε (ϕ(u)(x, t) − ϕ(v)(y, s))(∇ϕ(u))2 (x, t)ξ(x, t, y, s) dxdtdyds
ZEu ×Ev (3.90)
£ 0 ¤
− Sε (ϕ(u)(x, t) − ϕ(v)(y, s))(∇ϕ(v))2 (y, s)ξ(x, t, y, s) dxdtdyds.
Eu ×Ev
Therefore ∀ε > 0,
T18 (ε)
Z + ZT5 (ε)
· ³ ´2 ¸
=− Sε0 (ϕ(u)(x, t) − ϕ(v)(y, s))ξ(x, t, y, s) ∇ϕ(u)(x, t) − ∇ϕ(v)(y, s) dxdtdyds (3.91)
Ev Eu
≤ 0.
One thus gets ∀ε > 0,
T15 + T16 + T17 (ε) ≥ T20 (ε). (3.92)
56
One can now let ε → 0 in (3.92). This gives, since T20 (ε) → 0 (thanks to the dominated convergence theorem),
 R1R1 
|u(x, t, α) − v(y, s, β)|dαdβ (ξt (x, t, y, s) + ξs (x, t, y, s))+
 R01 R01 
Z Z  ³0 0 (f (u(x, t, α)>v(y, s, β)) − f (u(x, t, α)⊥v(y,´s, β)))dαdβ 
 
 q(x, t) · ∇x ξ(x, t, y, s) + q(y, s) · ∇y ξ(x, t, y, s)  dxdtdyds
Ω×(0,T ) Ω×(0,T )  
 −(∇ |ϕ(u)(x, t) − ϕ(v)(y, s)| + ∇ |ϕ(u)(x, t) − ϕ(v)(y, s)|) 
x y
·(∇x ξ(x, t, y, s) + ∇y ξ(x, t, y, s))
Z Z Z 1
ZΩ×(0,T ) ZΩ Z0 1
+ |u0 (y) − u(x, t, α)|ξ(x, t, y, 0)dαdydxdt
Ω×(0,T ) Ω 0
≥ 0.
(3.93)
RT
Now, let us consider the analog of (3.60) for v instead of u, with κ = u0 (x) and ψ(y, s) = s
ξ(x, 0, y, τ )dτ and
integrate the result on x ∈ Ω. One then gets
 R1 
− 0 |v(y, s, β) − u0 (x)|dβ ξ(x, 0, y, s)+
R
 1 
Z  0 (f (v(y, s, β)>u0 (x)) − f (v(y, s, β)⊥u0 (x)))dβ q(y, s)· 
R  R 
Ω  ∇y sT ξ(x, 0, y, τ )dτ  dydsdx +
Ω×(0,T )  
 −∇y |ϕ(v)(y, s) − ϕ(u0 (x))|·  (3.94)
RT
∇y ξ(x, 0, y, τ )dτ
Z Z s Z T
|u0 (x) − u0 (y)| ξ(x, 0, y, τ )dτ dxdy ≥ 0.
Ω Ω 0
A sequence of mollifiers in R and Rd is now introduced. Let ρ ∈ Cc∞ (Rd , R+ ) and ρ̄ ∈ Cc∞ (R, R+ ) be such that
{x ∈ Rd ; ρ(x) 6= 0} ⊂ {x ∈ Rd ; |x| ≤ 1},
{x ∈ R; ρ̄(x) 6= 0} ⊂ [−1, 0] (3.95)

and
Z Z
ρ(x)dx = 1, ρ̄(x)dx = 1. (3.96)
Rd R
For n ∈ N? , define ρn = nd ρ(nx) for all x ∈ Rd and ρ̄n = nρ̄(nx) for all x ∈ R.
One sets ξ(x, t, y, s) = ψ(x, t)ρn (x − y)ρ̄m (t − s), where ψ ∈ Cc∞ (Ω × [0, T ), R+ ) and n and m are large
enough to ensure, for all (x, t) ∈ Ω × [0, T ), ξ(x, t, ·, ·) ∈ D+ (Ω × [0, T )) and for all (y, s) ∈ Ω × [0, T ),
ξ(·, ·, y, s) ∈ D+ (Ω × [0, T )). This choice is not symmetrical in (x, t) and (y, s), which gives an easier way to
take the limit as n → ∞ and m → ∞. One gets, from (3.93),
57
 
ρn (x − y)ρ̄m (t − s)
 R1R1 
 0 0
|u(x,
µ t, α) − v(y, s, β)|dαdβ ψt (x,
¶ t) 
 R 1 R 1 f (u(x, t, α)>v(y, s, β)) 
Z Z  − 0 0 dαdβ 
 −f (u(x, t, α)⊥v(y, s, β)) 
  dxdtdyds
 (ρn (x − y)ρ̄m (t − s)q(x, t) · ∇ψ(x, t) 
Ω×(0,T ) Ω×(0,T )  
  (3.97)
 −ψ(x, t)ρ̄m (t − s)(q(x, t) − q(y, s)) · ∇ρn (x − y)) 
 −ρn (x − y)ρ̄m (t − s)(∇x |ϕ(u)(x, t) − ϕ(v)(y, s)| 
+∇y |ϕ(u)(x, t) − ϕ(v)(y, s)|) · ∇ψ(x, t)
Z Z Z 1
+ |u0 (x) − v(y, s, β)|ψ(x, 0)ρn (x − y)ρ̄m (−s)dβdxdyds ≥ 0.
Ω×(0,T ) Ω 0
The second of the two initial terms vanishes because of the asymmetric choice of ρ̄m . Using the same test function
in (3.94), at t = 0, i.e. ξ(x, 0, y, s) = ψ(x, 0)ρn (x − y)ρ̄m (−s) and (3.96), we get
 R1 
− |v(y, s, β) − u0 (x)|dβ ψ(x, 0)ρn (x − y)ρ̄n (−s)
 Z0 1 
 − (f (v(y, s, β)>u0 (x)) − f (v(y, s, β)⊥u0 (x)))dβ q(y, s)· 
 
Z Z  0 Z T 
 
 ψ(x, 0)∇ρn (x − y) ρ̄m (−τ )dτ  dydsdx
 
Ω Ω×(0,T )  s  (3.98)
 +∇ |ϕ(v)(y, s) − ϕ(u 
 y 0 (x))|· 
 Z T 
ψ(x, 0)∇ρn (x − y) ρ̄m (−τ )dτ
Z Z s
+ |u0 (x) − u0 (y)|ψ(x, 0)ρn (x − y)dxdy ≥ 0.
Ω Ω
One can now add (3.97) and (3.98) let m tend to ∞ and use the theorem of continuity in means. Since the function
RT
s → s ρ̄m (−τ )dτ is bounded and tends to zero as m → ∞ for all s ∈ (0, T ), one gets
 R1R1 
ρn (y − x)
µ 0 0
|u(x, t, α) − v(y, t, β)|dαdβ
¶ ψt (x, t)
 R 1 R 1 f (u(x, t, α)>v(y, t, β)) 
 + dαdβ 
Z Z  0 0 −f (u(x, t, α)⊥v(y, t, β)) 
 
 (ρn (y − x)q(x, t) · ∇ψ(x, t)+  dxdtdy
 
Ω Ω×(0,T )  
 ψ(x, t)(q(y, t) − q(x, t)) · ∇ρn (y − x))  (3.99)
 −ρn (x − y)(∇x |ϕ(u)(x, t) − ϕ(v)(y, t)| 
Z Z +∇y |ϕ(u)(x, t) − ϕ(v)(y, t)|) · ∇ψ(x, t)
+ |u0 (x) − u0 (y)|ψ(x, 0)ρn (x − y)dxdy ≥ 0.
Ω Ω
Remarking that
Z Z · ¸
ρn (x − y)(∇x |ϕ(u)(x, t) − ϕ(v)(y, t)|
dxdtdy
Ω Z Z )
Ω×(0,T +∇y |ϕ(u)(x, t) − ϕ(v)(y, t)|) · ∇ψ(x, t)
£ ¤ (3.100)
=− ρn (x − y)|ϕ(u)(x, t) − ϕ(v)(y, t)|∆ψ(x, t) dxdtdy,
Ω Ω×(0,T )
it is possible to let n → ∞ in (3.99). Using divq = 0 and the theorem of continuity in means again, one gets
 R1R1 
Z |u(x, t, α) − v(x, t, β)|dαdβ ψt (x, t)
 + R 1 R 1 (f (u(x, t, α)>v(x, t, β)) − f (u(x, t, α)⊥v(x, t, β)))dαdβ
0 0

 0 0  dxdt ≥ 0. (3.101)
 q(x, t) · ∇ψ(x, t) 
Ω×(0,T )
−∇|ϕ(u)(x, t) − ϕ(v)(x, t)| · ∇ψ(x, t)
58
One notices that (3.101) holds for any ψ ∈ H 1 (Ω×(0, T )), with ψ ≥ 0 and ψ(., T ) = 0, using a density argument.
Therefore one can now take, in (3.101), for ψ the functions ψε (x, t) = (T − t) min( d(x,∂Ω)
ε , 1), for ε > 0.
Assume momentarily that for all w ∈ H01 (Ω) with w ≥ 0,
Z
d(x, ∂Ω)
lim inf ∇w(x) · ∇ min( , 1)dx ≥ 0 (3.102)
ε→0 Ω ε
(The proof of (3.102) is given below).
The expression q(x, t) · ∇ min( d(x,∂Ω)

ε , 1) verifies
d(x, ∂Ω)
lim q(x, t) · ∇ min( , 1) = 0, for a.e. (x, t) ∈ Ω × (0, T ),
ε→0 ε
and under condition (3.4) (and (H5) of Assumption 3.1) remains bounded independently of ε for a.e. (x, t) ∈
Ω × (0, T ). Letting ε → 0, (3.101), with ψ = ψε , gives
Z ·Z 1 Z 1 ¸
− |u(x, t, α) − v(x, t, β)|dαdβ dxdt ≥ 0,
Ω×(0,T ) 0 0
which finally proves that u = v and that u is a classical function of space and time (it does not depend on α).
Proof of (3.102)
Let ε > 0. Let (∂Ωi )i=1,...,N be the faces of Ω, ni their normal vector outward to Ω, and for i = 1, ...N , let Ωi be
the subset of Ω such that, for all x ∈ Ωi , d(x, ∂Ωi ) < ε and d(x, ∂Ωi ) < d(x, ∂Ωj ) for all j 6= i. One has
Z N Z
X ∇w(x) · ni
∇w(x) · ∇ min(d(x, ∂Ω)/ε, 1)dx = dx.
∪N
i=1 Ωi i=1 Ωi ε
For each Ωi , let Ω̃i be the largest cylinder generated by ni included in Ωi . One denotes by ∂Ω0i the face of Ω̃i
parallel to ∂Ωi . Let Ωε be defined by Ωε = Ω \ ∪N 2
i=1 Ω̃i . One has meas(Ωε ) ≤ C(Ω)ε and
Z N Z
X Z
w(x) |∇w(x)|
∇w(x) · ∇ min(d(x, ∂Ω)/ε, 1)dx ≥ dγ(x) − dx.
Ω i=1 ∂Ω0i ε Ωε ε
Thanks to the Cauchy-Schwarz inequality, one gets
Z Z
( |∇w(x)|dx)2 ≤ m(Ωε ) (∇w(x))2 dx.
Ωε Ωε
Z
2
One concludes, using lim (∇w(x)) dx = 0.
ε→0 Ωε
Remark 3.6 Inequation (3.102) could also be proven in the case where Ω is regular instead of polygonal, with a
slightly different method. Let Ωε = {x ∈ Ω, d(x, ∂Ω) < ε} and let ∂Ω0ε be the other face of Ωε . The normal
vector to ∂Ω0ε at any point x is equal to ∇d(x, ∂Ω). Therefore one has
Z Z Z
w(x) ∆d(x, ∂Ω)
∇w(x) · ∇ min(d(x, ∂Ω)/ε, 1)dx = dγ(x) − w(x) dx.
Ω ∂Ωε0 ε Ωε ε
Since Hardy’s inequality leads to
Z µ ¶2 Z
w(x)
dx ≤ C(Ω) (∇w(x))2 dx,
Ωε d(x, ∂Ω) Ωε
Z
one concludes using lim (∇w(x))2 dx = 0.
ε→0 Ωε
59
3.2.6 Conclusion
Let us finally prove the convergence theorem by way of contradiction:
Assume that the convergence stated in the Theorem 3.1 does not hold. Then there exist ε > 0, p ∈ [1, +∞) and
a sequence (uDm )m∈N such that kuDm − ukLp (Ω×(0,T ) ≥ ε, for any m ∈ N. Then by Theorem 3.3, there exists
a subsequence of the sequence (uDm )m∈N , still denoted by (uDm )m∈N which converges to an entropy process
solution of (3.1)-(3.3). By Theorem 3.4 this entropy process solution is the unique entropy weak solution to (3.1)-
(3.3), and from Lemma 3.7 which is stated below, the convergence of (uDm )m∈N is strong in any Lq (Ω × (0, T )).
This is in contradiction with the fact that kuDm − ukLp (Ω×(0,T ) ≥ ε, for any m ∈ N.
Lemma 3.7 Let Q be a Borelian subset of Rk , k ∈ N? , and let (un )n∈N ⊂ L∞ (Q) be such that un converges to
u ∈ L∞ (Ω × (0, T ))(Q × (0, 1)) in the nonlinear weak star sense where u does not depend on α, then (un )n∈N
converges to u in Lploc (Q) for any p ∈ [1, ∞).
Proof. Let K be a compact subset of Q, since un converges to u in the nonlinear weak star sense , one has
Z Z Z Z
2 2
|un (x) − u(x)| dx = un (x)dx − 2 un (x)u(x)dx + u(x)2 dx → 0 as n → +∞;
K K K K
since K is bounded, one also has:

Z
|un (x) − u(x)|p dx → 0 as n → +∞, ∀p ∈ [1, 2]
K
and since the sequence (un )n∈N is bounded in L∞ (Q),

Z
|un (x) − u(x)|p dx → 0 as n → +∞, ∀p > 2.
K
Remark 3.7 An interesting (and open to our knowledge) question is to find the convergence rate of the finite
volume approximations. In the case of a pure hyperbolic equation, i.e. ϕ = 0, it was proven by several authors
(under varying assumptions, see e.g. [12], [34], [17], [9]) that the error between the approximate finite volume
solution and the entropy weak solution is of order less than h1/4 where h is the size of the mesh, under a usual CFL
condition for the explicit schemes which are considered in [12], [34], [17], [9], and of order less than h1/4 + k 1/2
where k is the time step in the case of the implicit scheme considered in [17]. However, it is also known that these
estimates are not sharp, since numerically the order of the error behaves as 1/2.
In the case of a pure linear parabolic equation, estimates of order 1 were obtained in [28] (see also [20])
We made a first attempt in the direction of an error estimate in the case of the present degenerate parabolic equation
by looking at the analogous continuous problem [21]: let uε be the unique solution to
³ ´
ut (x, t) + div q f (u) (x, t) − ∆ϕ(u)(x, t) − ε∆u(x, t) = 0, for (x, t) ∈ Ω × (0, T ), (3.103)
with initial condition (3.2) and boundary condition (3.3) and let u be the unique entropy weak solution solution
of (3.1)-(3.3), then under Assumption 3.1, we are able to prove that kuε − ukL1 (QT ) ≤ Cε1/5 where C ∈ R+
depends only on the data. This estimate is however probably not optimal and we have not yet been able to
transcribe its proof to the discrete setting (the term −ε∆u being the continuous diffusive representation of the
diffusive perturbation introduced by the finite volume scheme).
Chapter 4
Navier-Stokes equations

The Stokes problem
We first study the following linear steady problem: find an approximation of u and p, weak solution to the gener-
alized Stokes equations with homogeneous boundary conditions on ∂Ω, which read
−ν∆u + ∇p = f in Ω,
divu = 0 in Ω, (4.1)
u = 0 on ∂Ω.
For this problem, in addition to (1.5), the following assumptions are made:
ν ∈ (0, +∞), (4.2)
f ∈ L2 (Ω)d . (4.3)
We then consider the following weak sense for problem (4.1).
Definition 4.1 (Weak solution to the steady Stokes equations) Under hypotheses ((1.5),(4.2),(4.3)), let E(Ω) be
defined by:
E(Ω) := {v = (v̄ (i) )i=1,...,d ∈ H01 (Ω)d , divv = 0 a.e. in Ω}. (4.4)
Then (u, p) is called a weak solution of (4.1) (see e.g. [33] or [4]) if
 Z

 u ∈ E(Ω), p ∈ L 2
(Ω) with p(x)dx = 0,



 Z Ω
ν ∇u(x) : ∇v(x)dx (4.5)

 Ω Z Z



 − p(x)divv(x)dx = f (x) · v(x)dx, ∀v ∈ H01 (Ω)d .
Ω Ω
where, for all u, v ∈ H01 (Ω)d and for a.e. x ∈ Ω, we use the following notation:
d
X
∇u(x) : ∇v(x) = ∇u(i) (x) · ∇v (i) (x) .
i=1
The existence and uniqueness of the weak solution of (4.1) in the sense of the above definition is a classical result
(see e.g. [33] or [4]).
60
61
The steady Navier-Stokes problem

A solution of the incompressible steady Navier-Stokes equations is given by the fields u = (ū(i) )i=1,...,d : Ω → Rd ,
and p : Ω → R, such that
−ν∆u + ∇p + (u · ∇)u = f in Ω,
(4.6)
divu = 0 in Ω.
with homogeneous Dirichlet boundary conditions for the velocity, we define the following weak sense.
Definition 4.2 (Weak solution to the steady Navier-Stokes equations) Under hypotheses ((1.5),(4.2),(4.3)), let
E(Ω) be defined by (4.4). Then (u, p) is called a weak solution of (4.6) if
 Z

 2
u ∈ E(Ω), p ∈ L (Ω) with p(x)dx = 0,



 Ω
 Z
ν ∇u(x) : ∇v(x)dx (4.7)

 ΩZ Z




 − p(x)divv(x)dx + b(u, u, v) = f (x) · v(x)dx ∀v ∈ H01 (Ω)d ,
Ω Ω
where the trilinear form b(., ., .) is defined, for all u, v, w ∈ (H01 (Ω))d , by
Z d Z
d X
X
b(u, v, w) = (u(x) · ∇)v(x) · w(x)dx = u(i) (x)∂i v (k) (x)w(k) (x) dx.
Ω k=1 i=1 Ω
Our concern is to propose some finite volume schemes, in addition to the proof of their convergence. The notion of
finite volume scheme will be linked with the mass conservation equation, the gradient of pressure and the nonlinear
terms.
4.2 Finite volume scheme with staggered variables

The purpose of this section is to propose a family of finite volume schemes, whose unknowns are the pressures in
the control volumes of a mesh, called the ”pressure mesh”, and the normal velocities to the edges of this mesh.
Let D be a finite volume discretization of Ω in the sense of Definition 1.3, devoted to play the role of the ”pressure
mesh”. In this section, we consider that, for all σ ∈ E, the point xσ is the center of gravity of σ. We define
the set XE of all u ∈ RE such that uσ = 0 for all σ ∈ Eext , which contains the discrete velocities, and we define
the set PE = {xσ , σ ∈ E}. The scheme first consists in finding u ∈ XE and p ∈ HM (Ω) such that the mass
conservation is discretized by the following finite volume scheme:
divK u = 0, ∀K ∈ M, (4.8)
where X
1
divK v = m(σ)vK,σ , ∀K ∈ M, (4.9)
m(K)
σ∈EK
and where we denote

vK,σ = vσ nσ · nK,σ , ∀σ ∈ EK , ∀K ∈ M, (4.10)
We now discretize the momentum equation, using a variational formulation. We define the set LD,0 (Ω) ⊂ LD (Ω)
of the piecewise constant functions in DK,σ , such that, for all u ∈ LD,0 (Ω), then uK,σ = 0 for all K ∈ M and
σ ∈ EK ∩ Eext . We then assume that the following hypotheses, denoted below by Hypotheses (HS), are fulfilled:
62
1. There exists a linear mapping ΠD : XE → LD (Ω) (see definition 1.3), such that, for ϕ ∈ C 2 (Ω)d ∩
H01 (Ω)d , defining v ∈ Xσ by vσ = ϕ(xσ ) · nσ , then ΠK,σ v, the constant value of ΠD v in DK,σ , is a
second order approximation at the point xσ of ϕ(xσ ). This is completed, assuming the relation
X 0
ΠK,σ v = aσσ vσ0 , ∀v ∈ XE , ∀K ∈ M, ∀σ ∈ EK , (4.11)
σ 0 ∈E
0
where the coefficients aσσ are assumed to satisfy
X 0
aσσ (v + A(xσ0 − xσ )) · nσ0 = v, ∀v ∈ Rd , ∀A ∈ Rd×d , ∀K ∈ M, ∀σ ∈ EK . (4.12)
σ 0 ∈E
(i)
We then denote ΠD v = (ΠD v)i=1,...,d .
2. The following consistency property holds:
ΠK,σ v · nσ = vσ , ∀K ∈ M, ∀σ ∈ EK , ∀v ∈ XE . (4.13)
d
R h·, ·iD is defined on LD (Ω) , such that, under suitable hypotheses, it approximates the
3. An inner product
inner product Ω ∇u : ∇vdx.
Then we approximate the equation (4.7) by

Z Z
νhΠD u, ΠD vi − p(x)divD v(x)dx + bD (u, u, v) = b
f (x) · Πv(x)dx, ∀v ∈ XE , (4.14)
Ω Ω
b ∈ HM (Ω)d defined by
with Πv
1 X
b K=
Πv m(σ)vK,σ (xσ − xK ), ∀K ∈ M, (4.15)
m(K)
σ∈EK
X X b L − Πv
Πv b K
bD (u, v, w) = m(σ)uK,σ b K , ∀u, v, w ∈ XE .
· Πw (4.16)
2
K∈M σ∈EK
Mσ ={K,L}
Hence the number of unknowns is #Eint + #M. Note that the following property holds
1 X b K |2 divK u, ∀u, v ∈ XE ,
bD (u, v, v) = − m(K)|Πv (4.17)
2
K∈M
R
which mimicks the continuous property b(u, v, v) = − 21 Ω
|v|2 divudx.
4.2.1 The MAC scheme on regular grids

Let us briefly show why the MAC scheme can be analyzed following the line of the schemes presented here. We
consider a grid, whose control volumes are orthogonal parallelepipedic, without grid refinement. In such a case, it
is easy to define the inner product h·, ·iD as the sum of the contributions of each direction i = 1, . . . , d of scalar
products, such as the one defined by (2.18), available in the case of ∆−adapted pointed meshes.
63
4.2.2 An example, built on the pressure grid

0
We aim in this section to use the scheme described in section 2.1. We use the matrix Aσσ
K defined by (2.45), for
setting, for i = 1, . . . , d,
X X X 0
hu, viD = Aσσ
K (uσ − u bK ), ∀u, v ∈ LD,0 (Ω)d ,
b K ) · (vσ0 − v (4.18)
with P P σσ 0
σ∈EK uσ σ 0 ∈EK AK
bK =
u P σσ 0
, ∀K ∈ M. (4.19)
2 AK
(σ,σ 0 )∈EK
We then get X X X 0
hu, viD = bσσ
A d
K uσ · vσ 0 , ∀u, v ∈ LD,0 (Ω) , (4.20)
with
bσσ0 = Aσσ0 − aK,σ aK,σ0 SK
A (4.21)
K K
P 0
where we define SK = 2
(σ,σ 0 )∈EK Aσσ
K and we set
1 X σσ0
aK,σ = AK . (4.22)
SK 0
σ ∈EK
(i)
We define coefficients ασσ0 such that the mapping Πi is defined by
X (i)
Πi v = v̄σ · e(i) = ασσ0 vσ0 , ∀σ ∈ Eint (4.23)
σ 0 ∈E
with the properties

d
X
(i) (i)
ασσ e · nσ = 1 (4.24)
i=1
and
d X
X (i)
ασσ0 e(i) · nσ = 0, ∀σ 0 6= σ, (4.25)
i=1 σ 0 ∈E
and second order approximation, that is: if there exists x and a matrix B with vσ0 = (x + B(xσ0 − xσ )) · nσ0 for
P (i)
all σ 0 ∈ E, then σ0 ∈E ασσ0 vσ0 = x · e(i) .
Numerical example
We show in the following figure the stream lines, obtained with Re = 5000 on a 80 × 80 grid.
4.2.3 Further analysis

Let us provide the mathematical analysis of the convergence of the above method. To this purpose, we introduce
the following definition for the regularity of the mesh:
Ã Ã P !!
σ0 2
dK,σ hK σ 0 ∈E |aσ ||xσ − xσ |
0
θD = max max , max , . (4.26)
σ∈Eint ,K,L∈Mσ dL,σ K∈M,σ∈EK dK,σ h2K
Then we define a norm, suitable for deriving compactness properties:

We then introduce the notion of continuous, coercive, consistent and symmetric families of inner products on
LD (Ω).
64
Figure 4.1: Streamlines for the lid driven cavity with Re = 5000
Definition 4.3 (Continuous, coercive, consistent and symmetric families of inner products) Let F be a family
of discretizations in the sense of definition 1.2. For D = (M, E, P) ∈ F, K ∈ M and σ ∈ E, we denote by
hu, viD an inner product on LD (Ω)d . The family of inner products (h·, ·iD )D∈F is said to be continuous if there
exists M > 0 such that
1/2 M
(hu, uiD ) ≤ |u| 2 d , ∀u ∈ LD (Ω)d , ∀D = (M, E, P) ∈ F. (4.27)
hD L (Ω)
The family of inner products (h·, ·iD )D∈F is said to be coercive if for any sequence (Dn )n∈N of elements of F
satisfying that limn→∞ hDn = 0 and that there exists un ∈ LDn (Ω)d for all n ∈ N and there exists C > 0 with
hun , un iDn ≤ C for all n ∈ N, then there exists u ∈ H01 (Ω)d and a subsequence family of (Dn )n∈N , again
denoted (Dn )n∈N , with lim kuD − ukL2 (Ω) = 0
hD →0
The family of inner products (h·, ·iD )D∈F is said to be consistent (with the viscous operator) if for any family
(uD )D∈F satisfying:
• uD ∈ LD (Ω)d for all D ∈ F ,
• there exists C > 0 with huD , uD iD ≤ C for all D ∈ F,
• there exists u ∈ H01 (Ω)d with lim kuD − ukL2 (Ω) = 0,
hD →0
then, for all ϕ ∈ Cc∞ (Ω)d , the following holds

Z
lim huD , vD iD = ∇u(x) : ∇ϕ(x)dx, (4.28)
hD →0 Ω
1
R
where, omitting for the simplicity of notation the index n, we define vD ∈ LD (Ω)d defined by vσ = m(σ) σ
ϕ(x)dγ(x)
for all σ ∈ E. Finally the family of inner products (h·, ·iD )D∈F is said to be symmetric if
hu, viD = hv, uiD , ∀u, v ∈ LD (Ω)d , ∀D = (M, E, P) ∈ F.
We then have the following lemma.

Lemma 4.1 (Discrete Poincaré inequality) Under hypotheses (1.5), let F be a family of discretizations in the
sense of definition 1.2 and let (h·, ·iD )D∈F be a coercive family of inner products, in the sense of Definition 4.3.
Then, for all sequence (Dn )n∈N of elements of F such that limn→∞ hDn = 0, there exists C1 > 0 such that
kvk2L2 (Ω)d ≤ C1 hv, viDn , ∀v ∈ LDn (Ω)d , ∀n ∈ N. (4.29)
Proof. For any n ∈ N, since the dimension of LDn (Ω) is finite, one can set
kvk2L2 (Ω)d
ζn = sup .
v∈LDn (Ω)d hv, viDn
65
Morover, there exists vn ∈ LDn (Ω)d such that hvn , vn iDn = 1 and kvk2L2 (Ω)d = ζn . Let us assume that the
lemma is wrong. It means that the sequence (ζn )n∈N is not bounded. Then there exists a strictly increasing
function ψ from N to N, such that the sequence (ζψ(n) )n∈N tends to +∞.
Then, using the coercivity of the family of inner products, we get from hvψ(n) , vψ(n) iDψ(n) = 1 and from
limn→∞ hDψ(n) = 0 the existence of a subsequence of (Dψ(n) )n∈N , denoted (Dψ(ϕ(n)) )n∈N , where ϕ is strictly
increasing from N to N, and an element v ∈ H01 (Ω)d , such that
lim kvψ(ϕ(n)) − vkL2 (Ω)d = 0.

n→∞
We then get a contradiction with kvψ(ϕ(n)) k2L2 (Ω)d = ζψ(ϕ(n)) , which tends to +∞ as n → +∞. ¤
We then have the following lemma.
Lemma 4.2 (Estimate) Under hypotheses (1.5) and (4.2), let D be a finite volume discretization of Ω in the sense
of Definition 1.3. Let ζD be defined by
kvk2L2 (Ω)d
ζD = sup . (4.30)
v∈LD (Ω)d hv, viDn
Let (uE , pM ) ∈ XE × HM (Ω) be a solution to (4.8)-(4.9)-(4.14). Then we have:
ν(hΠD u, ΠD uiD )1/2 ≤ d θD ζD kf kL2 (Ω)d . (4.31)
where θD is defined by (4.26).

Proof.
Let us take v = u in (4.14). We get
Z
νhΠu, ΠuiD = b
f (x) · Πu(x)dx,
Ω
1
R
thanks to (4.8), which, together with (4.17), also implies b(u, u, u) = 0. Denoting by fK = m(K) K
f (x)dx, we
have Z X X
b
f (x) · Πu(x)dx = fK m(σ)uK,σ (xσ − xK ).
Ω K∈M σ∈EK
Applying the Cauchy-Schwarz inequality, we get

µZ ¶2 X X X X
b
f (x) · Πu(x)dx ≤ |fK |2 m(σ)|xσ − xK | m(σ)u2K,σ |xσ − xK |,
Ω K∈M σ∈EK K∈M σ∈EK
which gives, thanks to the definition of the regularity of the mesh,

¯Z ¯
¯ ¯
b
¯ f (x) · Πu(x)dx ¯ ≤ d θD kf kL2 (Ω)d kΠD vkL2 (Ω)d .
¯ ¯
Ω
We can then write

νhΠD u, ΠD uiD ≤ d θD ζD kf kL2 (Ω)d (hΠD u, ΠD uiD )1/2 ,
which implies (4.31). ¤
We then get the following lemma.
Lemma 4.3 (Existence of a discrete solution) Under hypotheses (1.5) and (4.2), let D be a finite volume dis-
cretization of Ω in the sense of Definition 1.3. Then there exists at least one (uE , pM ) ∈ XE × HM (Ω), solution
to (4.8)-(4.9)-(4.14).
66
Proof. It suffices to remark that one can estimate the pressure, using Nečas procedure. ¤
Let us now state the convergence of the scheme.
Theorem 4.1 Under hypotheses (1.5) and (4.2), let F be a family of discretizations in the sense of definition
1.3 and let (h·, ·iD )D∈F be a coercive family of inner products, in the sense of Definition 4.3. Let (Dn )n∈N
be a sequence of elements of F, such that limn→∞ hDn = 0 and such that there exists θ > 0 with θ ≥ θDn
(defined by (4.26)), for all n ∈ N. For all n ∈ N, let (un , pn ) ∈ XEn × HMn (Ω) be a solution to (4.8)-(4.9)-
(4.14) with D = Dn . Then there exists a subsequence of (Dn )n∈N , again denoted (Dn )n∈N , and a solution
(u, p) ∈ E(Ω) × L2 (Ω) of (4.7) such that ΠDn un tends to u in L2 (Ω)d .
Proof. Let us first remark that ζDn defined by (4.30) is bounded thanks to Lemma 4.1 by some value, denoted
by C1 . Thanks to Lemma 4.2, we get that the sequence hΠDn un , ΠDn un iDn is bounded, which shows, from the
coercivity of the family of inner products (h·, ·iD )D∈F , that we can extract from (Dn )n∈N a subsequence, again
denoted (Dn )n∈N , such that there exists u ∈ H01 (Ω)d with ΠDn un tends to u in L2 (Ω)d . We then see, using (4.8)
and the consistency property (4.13), that divu = 0 a.e. in Ω, which implies u ∈ E(Ω). Let ϕ ∈ Cc∞ (Ω)d , with
divϕ = 0 in R Ω be given. Let n ∈ N. Omitting for the simplicity R index n, we define vD ∈ XE by
of notation the
1 d 1
vσ = m(σ) σ
ϕ(x) · nσ dγ(x) for all σ ∈ E. and v D ∈ L D (Ω) by vσ = m(σ) σ ϕ(x)dγ(x) for all σ ∈ E. Note
that we have, thanks to the definition of θD ,
2
max |∂ij ϕb |
kvD − ΠD vD kL2 (Ω)d ≤ h2D θD .
2
Using the consistency of the family of inner products (h·, ·iD )D∈F , we get
Z
lim hΠDn un , vDn iDn = ∇u(x) : ∇ϕ(x)dx.
n→∞ Ω
Using the continuity of the family of inner products (h·, ·iD )D∈F , we have
lim hvD − ΠD vD , vD − ΠD vD iDn = 0,

n→∞
which implies Z
lim hΠDn u, ΠDn vDn iDn = ∇u(x) : ∇ϕ(x)dx.
n→∞ Ω
Since we have divD vD = 0 in this case, we can write, taking vD as test function in (4.14)
Z
νhΠD u, ΠD vD iD + bD (u, u, vD ) = b D (x)dx.
f (x) · Πv
Ω
We get, using the results of chapter (5), that bDn (un , un , vDn ) converges to b(u, u, ϕ), using the convergence in
b (i) u−Π
b (i) u
b n to u, the weak convergence of the piecewise constant function with value d Π
L2 (Ω)d of Πu L K
nK,σ in
2dK,σ
b b
(i) ∞ d
DK,σ to ∇u and the convergence in L (Ω) of the piecewise constant function with value ΠL vD +2 ΠK vD in
DK,σ to ϕ. The proof is concluded thanks to the convergence of the right hand side.
¤ We then have the following lemma.
Lemma 4.4 (Application) The scheme defined by (4.18)-(4.22) provides a continuous, coercive, consistent and
symmetric families of inner products.
We let the proof of the lemma to the reader, using the results of Chapter 2.
67
4.3 Finite volume scheme with collocated variables

4.3.1 Discrete scheme
We denote by D = (M, E, P) a pointed discretization in the sense of Definition 1.2.
For all function ψ ∈ C 0 (Ω), we recall that PM ψ is the element v ∈ HM (Ω) such that vK = ψ(xK ) for all
K ∈ M. Let us follow the scheme presented in Chapter 2, in which we want to directly define the unknowns in
HM (Ω) instead of XD,0 . To this purpose, for any edge σ ∈ Eint , we define a linear mapping Πσ : HM (Ω) → R
such that for all regular function ψ, Πσ PM ψ is a second order approximation of ψ(xσ ). In such a case, following
the idea of (2.24), we use coefficients (βσL )L∈M such that
X X X
∀u ∈ HM (Ω), Πσ u = βσL uL with βσL = 1, xσ = βσL xL . (4.32)
L∈M L∈M L∈M
It is always possible to restrict in 3D (resp. 2D) the number of nonzero βσL to four (resp. three). In practice,
the scheme is shown to be robust with respect to the choice of these four control volumes if they are taken close
enough to the considered edge.
We introduce some notations related to the mesh. The size of the discretization D is defined by:
hD = sup{hK , K ∈ M},
and the regularity of the mesh by:

 X 
|βσL ||xL − xσ |2
 dK,σ hK L∈M 
θD = max 
 max , max , max .
 (4.33)
σ∈Eint dL,σ K∈M dK,σ K∈M h2K
K,L∈Mσ σ∈EK σ∈EK ∩Eint
Discretization of the viscous terms

We begin with defining an approximate of the gradient of elements of HM (Ω) on cell K ∈ M. We set, for any
u ∈ HM (Ω) and K ∈ M:
1 X
∇K u = m(σ)(Πσ u − uK )nK,σ . (4.34)
m(K)
σ∈EK
Then, for all σ ∈ EK , we define RK,σ u ∈ R, corresponding to some local residual of the second order interpolation
of a regular function, by: √
d
RK,σ u = (Πσ u − uK − ∇K u · (xσ − xK )) . (4.35)
dK,σ
We then give the following expression for the discrete gradient of u ∈ HM (Ω) in the cone CK,σ :
∇K,σ u = ∇K u + RK,σ u nK,σ . (4.36)
We can then define the function ∇D u by
∇D u(x) = ∇K,σ u, for a.e. x ∈ CK,σ , ∀K ∈ M, ∀σ ∈ EK . (4.37)
We then get that

Z X X m(σ)dK,σ
∇D u(x) · ∇D v(x)dx = ∇K,σ u · ∇K,σ v, ∀u, v ∈ HM (Ω). (4.38)
Ω d
K∈M σ∈EK
R
Notice that Ω ∇RD u(x) · ∇D v(x)dx defines a symmetric inner product on HM (Ω) which provides a good ap-
proximation for Ω ∇u(x) · ∇v(x)dx and that the following lemma holds (see lemma 2.5):
68
Lemma 4.5 Let D be a space discretization of Ω in the sense of Definition 1.2 and let θ ≥ θD . Then, defining
Ã !1/2
X X m(σ)
|u|D = (Πσ u − uK )2 , ∀u ∈ HM (Ω), (4.39)
dK,σ
K∈M σ∈EK
there exist α > 0 and β > 0, only depending on θ, such that

α|u|D ≤ k∇D ukL2 (Ω)d ≤ β|u|D , ∀u ∈ HM (Ω). (4.40)
Pressure-velocity coupling, mass balance and convective contributions

For all v̄ ∈ (HM (Ω))d , we define a discrete divergence operator by:
X Xd
1
divK v̄ = m(σ)Πσ v̄ · nK,σ = (∇K v (i) )(i) , ∀K ∈ M. (4.41)
m(K) i=1
σ∈EK
We then define the function divD v̄ by the relation

divD v̄(x) = divK v̄, for a.e. x ∈ K, ∀K ∈ M.
As recalled in the introduction of this paper, a pressure stabilization method must be implemented in the mass
conservation equation in order to prevent from oscillations of the pressure. To this aim, we first define a stabilized
mass flux across σ ∈ Eint with Mσ = {K, L}, by
ΦλK,σ (u, p) = m(σ) (Πσ u · nK,σ + λσ (pK − pL )) , (4.42)
where (λσ )σ∈Eint is a given family of positive reals, the choice of which is discussed in [11], in the case of meshes
satisfying an orthogonality property. In this paper we set
hγD
λσ = λ , ∀σ ∈ Eint with Mσ = {K, L}, (4.43)
dK,σ + dL,σ
for given values λ > 0 and γ ∈ (0, 4) (this expression, chosen in [7] in the framework of a collocated finite element
method, makes the mathematical study easier than the ”cluster” choice, which is shown in [11] to be more precise).
Note that, for all σ ∈ Eint with Mσ = {K, L}, ΦλK,σ (u, p) + ΦλL,σ (u, p) = 0. We then use this modified flux,
in order to define a stabilized transport operator. The centred transport operator is defined, for all v̄ ∈ (HM (Ω))d ,
w ∈ HM (Ω) and K ∈ M, by
1 X wK + wL
divλK (w, v̄, p) = ΦλK,σ (u, p) .
m(K) 2
σ∈EK ∩Eint ,Mσ ={K,L}
Then the function divλD (w, v̄, p) is defined by

divλD (w, v̄, p)(x) = divλK (w, v̄, p), for a.e. x ∈ K, ∀K ∈ M.
Resulting discrete equations

By considering the previous definitions, the discrete approximation of equations (4.6) therefore reads: find u and
p such that Z X
u ∈ (HM (Ω))d and p ∈ HM (Ω) with p(x)dx = m(K)pK = 0,
Ω
Z ³ ´K∈M
ν∇D u : ∇D v̄dx − pdivD v̄ + divλD (u, u, p) · v̄ dx (4.44a)

ΩZ
= f · v̄dx, ∀v̄ ∈ (HM (Ω))d ,
Ω
divλD (1, u, p) = 0 a.e. in Ω. (4.44b)

69
4.3.2 Further analysis

Existence of a discrete solution and estimates
The system of discrete equations (4.44) appears as a system of nonlinear equations, for which we have to prove the
existence of at least one solution, satisfying suitable estimates. In this direction, we have the following lemma.
Lemma 4.6 (Discrete H01 (Ω) estimate on the velocities) Let D be a space discretization of Ω, let θ ≥ θD and let
λ ∈ (0, +∞) and γ ∈ (0, 4) be given. Let ρ ∈ [0, 1] be given and let (u, p), be a solution to the following system
of equations (which reduces to (4.44) as ρ = 1)
Z X
u ∈ (HM (Ω))d and p ∈ HM (Ω) with p(x)dx = m(K)pK = 0,
Ω
Z ³ K∈M
´
ν∇D u : ∇D v̄dx − pdivD v̄ + ρdivλD (u, u, p) · v̄ dx (4.45a)
ΩZ
= f · v̄dx, ∀v̄ ∈ (HM (Ω))d ,
Ω
divλD (1, u, p) = 0 a.e. in Ω. (4.45b)

Then there exists C2 , only depending on θ such that u and p satisfy the following estimates:
νk∇D uk
X (L2 (Ω)d )d ≤ C2 kf k(L2 (Ω))d
λσ m(σ)(pL − pK )2 ≤ C2 kf k2(L2 (Ω))d (4.46)
σ∈Eint ,Mσ ={K,L}
R
Proof. We let v̄ = u in (4.45a). Let us first remark that, thanks to (4.45b), we get Ω divλD (u, u, p) · udx = 0
and Z X
pdivD udx = − λσ m(σ)(pL − pK )2 .
Ω σ∈Eint ,Mσ ={K,L}
We apply the Cauchy-Schwarz inequality, and the discrete Poincaré inequality. We then conclude the proof of the
lemma. ¤
We can now state the existence of at least one solution to (4.44).
Lemma 4.7 (Existence of at least one solution) Let D be a space discretization of Ω, let λ ∈ (0, +∞) and γ ∈
(0, 4) be given. Then there exists at least one (u, p) solution to (4.44), which therefore satisfies (4.46).
Proof. We remark that (4.45) with ρ = 0 is a linear system. Thanks to Lemma 4.6, we can then apply the results
on the topological degree [32]. This shows the existence of at least one solution of (4.45) with ρ = 1, that is (4.44).
¤
As in [32], we could also state an L2 estimate on the pressure, under some additionnal regularity hypotheses on
the mesh. For the sake of shortness, we focus here on the convergence of the scheme to the solution of (4.6).
Passing to the limit

Let us now state the convergence theorem.
Theorem 4.2 Let (D(m) )m∈N be a sequence of space discretizations of Ω in the sense of Definition 1.2, such that
hD(m) tends to 0 as m → ∞ and such that there exists θ > 0 with θD(m) ≤ θ, for all m ∈ N. Let λ ∈ (0, +∞)
and γ ∈ (0, 4) be given. Let, for all m ∈ N, (u(m) , p(m) ) ∈ (XD(m) )d × HM(m) (Ω), be a solution to (4.44)
with D = D(m) . Then there exists a weak solution u of (4.6) and a subsequence of (D(m) )m∈N , again denoted
(D(m) )m∈N , such that the corresponding subsequence of solutions (u(m) )m∈N converges to u in L2 (Ω)d .
70
Proof. Thanks to (4.46), we can apply the results of chapter 2 concerning the relative compactness of the
family (u(m) )m∈N and the regularity of the limit of a subsequence. Hence we get the existence of a subsequence
of (D(m) )m∈N , such that the corresponding sequence (u(m) )m∈N converges to some function u ∈ H01 (Ω)d in
L2 (Ω)2 . Using (4.41), we get that divD(m) u(m) converges for the weak topology of L2 (Ω)d to divu. Moreover,
the corresponding sequence (p(m) )m∈N satisfies for all m ∈ N the inequality (4.46). Hence, multiplying (4.44b)
by a regular test function and summing on K ∈ M(m) , we get that the term in pressure tends to 0 as m → ∞,
using (4.46) and the Cauchy-Schwarz inequality. This proves that divu = 0 a.e. in Ω, and therefore that u ∈ V .
We then consider a function ψ ∈ V such that ψ ∈ Cc∞ (Ω)d and we denote, for some m ∈ N, D = D(m) . We then
set v̄ = PM ψ in (4.44). We define the terms
Z Z
(m) (m)
T1 = ν∇D u : ∇D PM ψdx, T2 = − pdivD PM ψdx,
ZΩ ZΩ
(m) (m)
T3 = divλD (u, u, p) · PM ψdx, T4 = f · PM ψdx.
Ω Ω
Z
(m)
Using the results of chapter 2, we get lim T1 = ν∇u : ∇ψdx. We have
m→∞ Ω
Z X
pdivD PM ψdx = m(σ)(pL − pK )Πσ PM ψ · nK,σ .
1
R
Using divψ = 0, we get, setting ψσ = m(σ) σ
ψds, that
Z X
pdivD PM ψdx = m(σ)(pL − pK )(Πσ PM ψ − ψσ ) · nK,σ .
Thanks to the definition of the regularity of the mesh, we have the existence of C3 , only depending on ψ and θ,
such that |Πσ PM ψ − ψσ | ≤ C3 h2D . Therefore, thanks to the Cauchy-Schwarz inequality, γ ∈ (0, 4) and to (4.46),
(m)
we get limm→∞ T2 = 0.
Remark 4.1 In [32], the exponent γ had to be taken lower than 2 in the pressure stabilization term, since the
consistency of the divergence operator was only satisfied with an order 1.
(m) (m) (m) (m)
We now turn to T3 . Using (4.44b), we rewrite this term as T3 = T5 + T6 with
(m)
X (m) (m) ψ(xK ) + ψ(xL )
T5 = m(σ)Πσ u(m) · nK,σ (uL − uK ) ·
2
and
(m)
X (m) (m) ψ(xK ) + ψ(xL )
T6 = m(σ)λσ (pL − pK )(uL − uK ) · .
2
d (m)
We use the fact that, for all i = 1, . . . , d, the function, defined in CK,σ by the constant value dK,σ (Πσ uL −
(m)
uK )(i) nK,σ weakly converges in L2 (Ω)2 to ∇u(i) and the function, defined in CK,σ by the value Πσ u (m)
converges in L2 (Ω)2 to u, we get

Z
(m)
lim T5 = div(u ⊗ u) · ψdx.
m→∞ Ω
Using the Cauchy-Schwarz inequality, we get that there exists C4 > 0, only depending on θ such that
(m)
X
(T6 )2 ≤ hγD λC4 λσ m(σ)(pL − pK )2 |u(m) |2D kψk∞ ,
(m) (m) R
which shows, thanks to (4.46) and to (4.40), that limm→∞ T6 = 0. We easily get that limm→∞ T4 = Ω
f·
ψdx. Gathering the above results, we get that u is a solution to (4.6), which concludes the proof. ¤
71
4.3.3 Numerical example

We assume that the analytical solution is given by
pref (x) = cos(πx(1) ) cos(πx(2) ) cos(πx(3) )

Pd
uref (x) = curl( i=1 (4x(1) (x(1) − 1))3 (4x(2) (x(2) − 1))4 (4x(3) (x(3) − 1))5 e(i) )
in the case ν = 1 and Ω = (0, 1)3 . We consider three types of meshes with n3i control volumes, for ni =
10, . . . , 60. The first ones are regular cubic, in the second ones (”smooth meshes”), the vertices are slightly moved
by a regular nonlinear mapping. In this case, the control volumes become ”hexahedra” with non planar faces
splitted into two triangles, which means that the interior control volumes have six neighbors, but that any pair
of neighboring control volumes share two faces, and the control volumes are no longer convex. In the ”random
meshes”, the vertices of cubic meshes are randomly moved, only preserving that the faces can be defined without
excluding the initial center of the control volumes. Table 4.1 shows that the convergence orders of the gradients
Cubic meshes Smooth meshes Random meshes

L2 u 2 1.91 1.71
p 2 1.08 0.80
H1 u 1.98 1.77 1.34
p 1.91 −− −−
Table 4.1: Numerical orders of convergence.
are better than the expected first order. Unsurprisingly, the H 1 -norm of the pressure does not tend to zero with the
size of the mesh.
Chapter 5
Discrete functional analysis
The contents of this chapter concern mathematical analysis tools which are needed for the in depth mathematical
studies of the following finite volume schemes. It can be bypassed by the reader who is mainly interested in the
implementation features and the numerical efficiency of the finite volume schemes.
5.1 The topological degree argument

This argument is used several times in this book in order to prove the existence of a solution to systems of nonlinear
equations provided by finite volume schemes. For the sake of completeness, we recall this argument (which was
first used for numerical schemes in [17]) in the finite dimensional case in the following theorem and refer to [14]
for the general case.
Theorem 5.1 (Application of the topological degree, finite dimensional case)
Let V be a finite dimensional vector space on R and g be a continuous function from V to V . Let us assume that
there exists a continuous function F from V × [0, 1] to V satisfying:
1. F (·, 1) = g, F (·, 0) is an affine function.
2. There exists R > 0, such that for any (v, ρ) ∈ V × [0, 1], if F (v, ρ) = 0, then kvkV 6= R.
3. The equation F (v, 0) = 0 has a solution v ∈ V such that kvkV < R.
Then there exists at least a solution v ∈ V such that g(v) = 0 and kvkV < R.
5.2 Discrete Sobolev embedding

Note that these properties hold for general finite volume discretizations.
?
5.2.1 Discrete embedding of W 1,1 (Ω) in L1 (Ω)
Let us first give the discrete counterpart of the norm in W 1,1 (Ω).
Definition 5.1 Under hypothesis (1.5), let D be a polygonal finite volume space discretization of Ω in the sense of
Definition 1.1. We define a norm on HM (Ω) by
X
kuk1,1,D = m(σ)δσ u, ∀u ∈ HM (Ω) (5.1)
σ∈M
defining δσ u for all u ∈ HM (Ω) and σ ∈ E by

δσ u = |uL − uK |, ∀σ ∈ Eint , Mσ = {K, L},
(5.2)
δσ u = |uK |, ∀σ ∈ Eext , Mσ = {K}.
72
73
?
We can now state the inequality corresponding to the discrete embedding of HM (Ω) in L1 (Ω).
Lemma 5.1 Under hypothesis (1.5), let D be a polygonal finite volume space discretization of Ω in the sense of
Definition 1.1. Then, with the notation of Definition 5.1 :
1
kukL1? (Ω) ≤ √ kuk1,1,D , ∀u ∈ HM (Ω), (5.3)
2 d
d
where 1? = d−1 .
Proof.
Different proofs of this lemma are possible. A first one consists to adapt to this discrete
√ setting the classical proof
of the Sobolev embedding due to L. Nirenberg (actually, it gives 1/2 instead of 1/(2 d) in (5.3)). It is done using
an induction on d. This proof is essentially given in [20]. Indeed, the hypotheses given in [20] are√slightly less
general but a quite easy adaptation of the proof leads to the present lemma (with 1/2 instead of 1/(2 d) in (5.3)).
We give here another proof using directly the result of L. Nirenberg, namely:
1
kukL1? (Rd ) ≤ kukW 1,1 (Rd ) , ∀u ∈ W 1,1 (Rd ), (5.4)
2d
Pd
where kukW 1,1 (Rd ) = i=1 kDi ukL1 (Rd ) and Di u is the weak derivative (or derivative in the sense of distribu-
tions) of u in the direction xi (with x = (x1 , . . . , xd ) ∈ Rd ).
Pd R ∂ϕ
For u ∈ L1 (Rd ), one sets kukBV = i=1 kDi ukM with, for i = 1, . . . , d, kDi ukM = sup{ Rd u(x) ∂x i
(x)dx,
∞ d 1 d
ϕ ∈ Cc (R ), kϕk∞ ≤ 1}. One says that the function u is in the space BV if u ∈ L (R ) and kukBV < ∞.
We first remark that (5.4) is true with kukBV (Rd ) instead of kukW 1,1 (Rd ) , and if u ∈ BV instead of W 1,1 (Rd ).
R
Indeed, to prove this result (which is classical), let ρ ∈ Cc∞ (Rd , R+ ) with Rd ρ(x)dx = 1. For n ∈ N? , define
ρn = nd ρ(n·). Let u ∈ BV and un = u ? ρn so that, with (5.4):
d
1 X
kun kL1? (Rd ) ≤ kDi un kL1 (Rd ) . (5.5)
2d i=1
But, kDi un kL1 (Rd ) = kDi un kM , and, for ϕ ∈ Cc∞ (Rd ), using Fubini’s theorem:
Z Z
∂ϕ ∂
un (x) (x)dx = u(x) (ϕ ? ρn )(x)dx ≤ kDi ukM kϕkL∞ (Rd ) .
Rd ∂xi Rd ∂xi
This leads to kDi un kL1 (Rd ) ≤ kDi ukM . Since un → u a.e., as n → ∞, at least for a subsequence, Fatou’s lemma
gives, from (5.5):
1
kukL1? (Rd ) ≤ kukBV ∀u ∈ BV. (5.6)
2d
Let now u ∈ HM (Ω). One sets u = 0 outside Ω so that u ∈ L1 (Rd ). One has
Z
kukBV = sup{ u(x)divϕ(x)dx, ϕ ∈ Cc∞ (Rd , Rd ), kϕkL∞ (Rd ) ≤ 1}, (5.7)
Rd
with kϕkL∞ (Rd ) = supdi=1 kϕi kL∞ (Rd ) and ϕ = (ϕ1 , . . . , ϕd ). But, for ϕ ∈ Cc∞ (Rd , Rd ) such that kϕkL∞ (Rd ) ≤
1, an integration by parts on each element of M gives (where nσ is a normal vector to σ and γ is the (d −
1)−Lebesgue measure on σ):
Z X Z √
u(x)divϕ(x)dx = δσ u |ϕ · nσ |dγ(x) ≤ dkuk1,1,D .
Rd σ∈E σ
√
Then, one has kukBV ≤ dkuk1,1,D and (5.6) leads to (5.3).
¤
74
?
5.2.2 Discrete embedding of W 1,p (Ω) in Lp (Ω), 1 < p < d
We now prove a discrete Sobolev embedding for 1 < p < d and for meshes in sense of Definition 1.3.
Lemma 5.2 Under hypothesis (1.5), let D be a pointed strictly star-shaped polygonal finite volume space dis-
cretization of Ω in the sense of Definition 1.3. Let η > 0 such that
η ≤ dK,σ /dL,σ ≤ 1/η, ∀σ ∈ Eint , m(σ) = {K, L}. (5.8)
Then, there exists C1 , only depending on d, p and η such that (see Definition 5.1):
kukLp? (Ω) ≤ C1 kuk1,p,D ∀u ∈ HM (Ω), (5.9)

pd
where p? = d−p and
X X µ ¶p
δσ u
kukp1,p,D = m(σ)dK,σ , (5.10)
dσ
K∈M σ∈EK
where dσ is defined by (1.12).
Proof. Note first that definition (5.10) is compatible with (5.1) for p = 1, since in (5.10), for an internal edge
σ ∈ Eint with Mσ = {K, L}, we have
dK,σ dL,σ
+ = 1.
dσ dσ
We follow here also the proof of Sobolev embedding due to L. Nirenberg. Let α be such that α1? = p? (that is
α = p(d − 1)/(d − p) > 1). Let u ∈ HM (Ω). Inequality (5.3) applied with |u|α instead of u leads to:
µZ ¶ d−1
d X
?
|u(x)|p dx ≤ m(σ)δσ |u|α .
Ω σ∈E
Let us remark that, for all σ ∈ Eint with Mσ = {K, L}, we have δσ |u|α ≤ α(|uK |α−1 + |uL |α−1 )δσ u, and for all
σ ∈ Eext , we have δσ |u|α ≤ α|uK |α−1 δσ u. This gives
µZ ¶ d−1
d X X
?
|u(x)|p dx ≤α m(σ)|uK |α−1 δσ u. (5.11)
Ω K∈M σ∈EK
For all σ ∈ Eint with Mσ = {K, L}, we can write
δσ u
δσ u ≤ (1 + η)dK,σ .
dσ
Since this also holds for σ ∈ Eext , Hölder Inequality applied with q = p/(p − 1) to (5.11) yields:
µZ ¶ d−1 Ã ! q1
? d X X
|u(x)|p dx ≤ α(1 + η) m(σ)dK,σ |uK |(α−1)q kuk1,p,D . (5.12)
Ω K∈M σ∈EK
Since (α − 1)q = p? , we have

X X Z
?
m(σ)dK,σ |uK |(α−1)q = d |u(x)|p dx.
K∈M σ∈EK Ω
Then, noticing that (d − 1)/d − 1/q = 1/p? , we deduce (5.9) from (5.12) with C1 = α(1 + η)d1/q only depending
on d, p and η. ¤
75
5.2.3 Discrete embedding of W 1,p (Ω) in Lq (Ω), for some q > p

Let 1 ≤ p < ∞. Lemma 5.3 gives easily a discrete embedding of W 1,p in Lq , for some q > p, this is given in the
following lemma.
Lemma 5.3 Under hypothesis (1.5), let D be a pointed strictly star-shaped polygonal finite volume space dis-
cretization of Ω in the sense of Definition 1.3. Let η > 0 such that property (5.8) holds. Then, there exists q > p
only depending on p and there exists C2 , only depending on d, Ω, p and η such that (see Definition 5.1):
kukLq (Ω) ≤ C2 kuk1,p,D ∀u ∈ HM (Ω), (5.13)
where kukp1,p,D is defined by (5.10).
Proof. If p = 1, one takes q = 1? and the result follows from lemma 5.1 (in this case C2 does not depend on η).
If 1 < p < d, one takes q = p? and the result is given by lemma 5.2.
If p ≥ d, one chooses any q ∈]p, ∞[ and p1 < d such that p?1 = q (this is possible since p?1 tends to ∞ as p1
tends to d). lemma 5.9 gives, for some C1 only depending on p, d and η, kukLq (Ω) ≤ C1 kuk1,p1 ,D . But, using
Hölder inequality, there exists C3 , only depending on d, p, Ω, such that kuk1,p1 ,D ≤ C3 kuk1,p,D . Inequality (5.13)
follows with C2 = C1 C3 . ¤
5.3 Compactness results for bounded families in discrete W 1,p (Ω) norm
Note that these properties again hold for general finite volume discretizations.
5.3.1 Compactness in Lp (Ω)

We prove in this section that bounded families in the discrete W 1,p (Ω) norms are relatively compact in Lp (Ω).
We begin here also with the case p = 1, giving in this case a crucial inequality which holds for general polygonal
partitions of Ω.
Lemma 5.4 Under hypothesis (1.5), let D be a polygonal finite volume space discretization of Ω in the sense of
Definition 1.1. Then, with the notation of Definition 5.1 :
√
ku(· + h) − ukL1 (Rd ) ≤ |h| dkuk1,1,D , ∀u ∈ HM (Ω), ∀h ∈ Rd , (5.14)
where u is defined on the whole Rd , taking u = 0 outside Ω, and |h| is the eucllidean norm of h ∈ Rd .
Proof. As in lemma 5.1, a proof of this result is possible with a method similar to the method for proving
compactness
√ results for bounded families in the discrete W 1,p norms, given in [20] (and we obtain (5.14) without
d). Indeed, this proof of [20] holds here in this case of a general partition, thanks to the fact that p = 1. More
restrictive assumptions are needed for the case p > 1. We give here another proof, using the BV −space, as in
lemma 5.1.
Let u ∈ Cc∞ (Rd ). For x, h ∈ Rd , one has:
Z 1 Z 1
|u(x + h) − u(x)| = | ∇u(x + th) · hdt| ≤ |h| |∇u(x + th)|dt.
0 0
Integrating with respect to x and using Fubini’s Theorem gives the well kown result
Z d
X
ku(· + h) − ukL1 ≤ |h| |∇u(x)|dx ≤ |h| kDi ukL1 , (5.15)
Rd i=1
where ∇u = (D1 u, . . . , Dd u). By density of Cc∞ (Rd ) in W 1,1 (Rd ), Inequality (5.15) is also true for u ∈
W 1,1 (Rd ).
76
We proceed now as in lemma 5.1, using the same notations. Let u ∈ BV and un = u ? ρn . Since un ∈ W 1,1 (Rn ),
Pd
Inequality (5.15) gives, for all h ∈ Rd , kun (· + h) − un kL1 ≤ |h| i=1 kDi un kL1 . But, for i = 1, . . . , d, as in
lemma 5.1, kDi un kL1 ≤ kDi ukM . Then, since un → u in L1 , as n → ∞, we obtain:
d
X
ku(· + h) − ukL1 ≤ |h| kDi ukM = |h|kukBV , ∀u ∈ BV, ∀h ∈ Rd . (5.16)
i=1
√
Let now u ∈ HM (Ω). One sets u = 0 outside Ω so that u ∈ L1 (Rd ). lemma 5.1 gives kukBV ≤ dkuk1,1,D ,
then: √
ku(· + h) − ukL1 ≤ |h| dkuk1,1,D .
¤
An easy consequence of lemmas 5.1 and 5.4 is a compactness result in L1 given in the following lemma.
Lemma 5.5 Under hypothesis (1.5), let F be a family of polygonal finite volume space discretizations of Ω in the
sense of Definition 1.1. For M ∈ F , let uM ∈ HM (Ω) and assume that there exists C ∈ R such, for all M ∈ F ,
kuM k1,1,D ≤ C. Then the family (uM )M∈F is relatively compact in L1 (Ω) and also in L1 (Rd ) taking uM = 0
outside Ω.
?
Proof. The proof is quite easy. lemma 5.1 gives that the family (uM )M∈F is bounded in L1 (Ω). Then, since
Ω is bounded, the family (uM )M∈F is bounded in L1 (Ω) and also in L1 (Rd ) taking uM = 0 outside Ω. Then,
thanks to Kolmogorov Copmpactness Theorem, lemma 5.4 gives that the family (uM )M∈F is relatively compact
in L1 (Ω) and also in L1 (Rd ) taking uM = 0 outside Ω. ¤
For p > 1, we need some an additional hypothese on the meshes, actually given by Definition 1.3 with a “uniform
η”.
Lemma 5.6 Let d ≥ 1, 1 ≤ p < ∞ and Ω be a polygonal open bounded connected subset of Rd . Let F be a
family pointed strictly star-shaped polygonal finite volume space discretizations of Ω in the sense of Definition 1.3.
Let η > 0 such that the property (5.8) holds for all D ∈ F . For D ∈ F , let uD ∈ HM (Ω) and assume that there
exists C ∈ R such, for all D ∈ F , kuD k1,p,D ≤ C. Then the family (uD )D∈F is relatively compact in Lp (Ω) and
also in Lp (Rd ) taking uD = 0 outside Ω.
Proof. Here also the proof is quite simple. Thanks to lemma 5.3 and the fact that Ω is bounded, the family
(uD )D∈F is bounded in L1 (Ω) and also in L1 (Rd ) taking uD = 0 outside Ω. Thanks, once again, to the fact
that Ω is bounded the family (kuD k1,1,D )D∈F is bounded in R. Then, as in the preceding lemma, Kolmogorov
Copmpactness Theorem gives that the family (kuD k1,1,D )D∈F is relatively compact in L1 (Ω) and also in L1 (Rd )
taking uD = 0 outside Ω.
In order to conclude we use, once again, lemma 5.3. It gives that the family (uD )D∈F is bounded in Lq (Ω) for
some q > p. With the relative compactness in L1 (Ω), this leads to the fact that the family (uD )D∈F is relative
compact in Lp (Ω) (and then also in Lp (Rd ) taking uD = 0 outside Ω). ¤
5.3.2 Regularity of the limit

With the hypotheses of lemma 5.6, assume that uD → u in Lp as h(D) → 0 (lemma 5.6 gives that this is possible,
at least for subsequences of sequences of meshes whose size goes to 0). We prove below that u ∈ W01,p (Ω).
Lemma 5.7 Under hypothesis (1.5), let (Dn )n∈N be a family of pointed strictly star-shaped polygonal finite vol-
ume space discretizations of Ω in the sense of Definition 1.2. Let η > 0 such that, for all n ∈ N, the property (5.8)
holds with D = Dn . For n ∈ N, let u(n) ∈ HMn (Ω) and assume that there exists C4 ∈ R such, for all n ∈ N,
ku(n) k1,p,Dn ≤ C4 . Assume also that h(Dn ) → 0 as n → ∞. Then:
1. There exists a subsequence of (u(n) )n∈N , still denoted by (u(n) )n∈N , and u ∈ Lp (Ω) such that u(n) → u in
Lp (Ω) as n → ∞.
77
2. u ∈ W01,p (Ω) and

1 + η p−1
k∇ukLp (Ω)d = k|∇u|kLp (Ω) ≤ d p C4 . (5.17)
η
Proof. The fact that there exists a subsequence of (u(n) )n∈N , still denoted by (u(n) )n∈N , and u ∈ Lp (Ω) such
that u(n) → u in Lp (Ω) as n → ∞ is a consequence of the relative compactness of (u(n) )n∈N in Lp given in
lemma 5.6. Assuming that u(n) → u in Lp (Ω) as n → ∞, we have now to prove that u ∈ W01,p (Ω).
Letting u(n) = 0 and u = 0 outside Ω, one also has u(n) → u in Lp (Rd ). The method consists to construct some
˜ D u(n) , bounded in Lp (Ω), equal to 0 outside Ω and converging, at least in the
approximate gradient, namely ∇ n
distribution sense, to ∇u.
Step 1 Construction of ∇˜ D u, for u ∈ HM (Ω), and properties
Let n ∈ N and D = Dn . For this step, one sets u = u(n) (not to be confused with the limit of the sequence
(u(n) )n∈N ). For σ ∈ E, one sets uσ = 0 if σ is on the boundary of Ω. Otherwise, one has Mσ = {K, L} and we
choose a value uσ between uK and uL . (it is possible to choose, for instance; uσ = 12 (uK + uL ) but any other
choice between uK and uL is possible). Then, one defines ∇ ˜ D u on K ∈ D on the following way:
1 X
˜ Du =
∇ m(σ)nK,σ (uσ − uK ).
m(K)
σ∈EK
˜ D u is constant on each K ∈ M and, on K, using Hölder Inequality:

The function ∇
1 X 1 X X δσ u p
˜ D u|p ≤
|∇ ( m(σ)nK,σ (uσ − uK ))p ≤ ( m(σ)dK,σ )p−1 m(σ)dK,σ ( ) .
(m(K)) p (m(K))p dK,σ
σ∈EK σ∈EK σ∈EK
P
Since σ∈EK m(σ)dK,σ = dm(K), one deduces
˜ D u|p ≤ dp−1 X δσ u p
|∇ m(σ)dK,σ ( ) .
m(K) dK,σ
σ∈EK
˜ D u in (Lp (Ω))d (or in (Lp (Rd ))d , setting ∇

This gives a Lp − estimate on ∇ ˜ D u = 0 outside Ω), in terms of
kuk1,p,D , namely:
˜ D u|kLp ≤ 1 + η d p kuk1,p,D .
p−1
k|∇ (5.18)
η
In order to prove, in the next step, the convergence of this approximate gradient, we compute now the integral of
this gradient against a test function. Let ϕ ∈ Cc∞ (Rd ; Rd ), ϕK the mean value of ϕ on K ∈ D and ϕσ the mean
value of ϕ on σ. Then:
Z X X X X
∇˜ D u · ϕdx = m(σ)nK,σ (uσ − uK )ϕK = m(σ)nK,σ (−uK )ϕσ + R(u, ϕ), (5.19)
Rd K∈D σ∈EK K∈D σ∈EK
with X X
R(u, ϕ) = m(σ)nK,σ (uσ − uK )(ϕK − ϕσ ).
K∈D σ∈EK
Then there exists Cϕ only depending on ϕ, d, p and Ω such that |R(u, ϕ)| ≤ Cϕ h(D)kuk1,p,D . Equation (5.19)
can also be written as:
Z XZ Z
˜
∇D u · ϕdx = (−uK )div(ϕ)dx + R(u, ϕ) = − udiv(ϕ)dx + R(u, ϕ). (5.20)
Rd K∈D K Rd
˜ D u(n) to ∇u and proof of u ∈ W 1,p (Ω) .

Step 2 Convergence of ∇ n 0
78
We consider now the sequence (u(n) )n∈N . Inequality (5.18) gives
˜ D u(n) |kLp ≤ 1 + η p−1

k|∇ d p ku(n) k1,p,D .
η
˜ D u(n) )n∈N is bounded in Lp (Rd )d and we can assume, up to a subsequence, that ∇
Then, the sequence (∇ ˜ D u(n)
p−1
converges to some w weakly in Lp (Rd )d , as n → ∞ and k|w|kLp ≤ 1+η η d
p C.
Let ϕ ∈ Cc∞ (Rd ; Rd ), Equation (5.20) gives

Z Z
∇˜ D u(n) · ϕdx = − u(n) div(ϕ)dx + R(u(n) , ϕ). (5.21)
Rd Rd
Thanks to |R(u(n) , ϕ)| ≤ Cϕ h(Dn )ku(n) k1,p,Dn , one has R(u(n) , ϕ) → 0, as n → ∞. Since u(n) → u in Lp (Rd )
as n → ∞, passing to the limit in (5.21) gives:
Z Z
w · ϕdx = − udiv(ϕ)dx.
Rd Rd
1+η p−1
Since ϕ is arbitrary in Cc∞ (Rd , Rd ), one deduces that ∇u = w. Then u ∈ W 1,p (Rd ) and k|∇u|kLp ≤ η d
p C.
Finally, since u = 0 outside Ω, one has u ∈ W01,p (Ω). ¤
5.4 Properties in the case of the ∆−adapted discretizations

We now focus on some properties which use the orthogonality property assumed in Definition 1.4.
Let us first give the two following results, already proven in [20]. Nevertheless, since their proof is quite short, we
provide it for the sake of completeness, and in order to make clear the differences with the general framework of
pointed finite volume discretizations.
Lemma 5.8 Under hypothesis (1.5), let D be a pointed ∆−adapted polygonal finite volume space discretization
of Ω in the sense of Definition 1.4. We prolong all the elements of HM (Ω) by 0 in Rd \ Ω. Then
ku(· + h) − uk2L2 (Rd ) ≤ |h|(|h| + 4h(M))kuk21,2,D , ∀u ∈ HM (Ω). (5.22)
Proof. For σ ∈ E, define χσ from Rd × Rd to {0, 1} by

1. χσ (x, y) = 1 if ]x, y[∩σ 6= ∅ and there exists z ∈]x, y[∩σ such that [x, z] ⊂ Ω or [z, y] ⊂ Ω,
2. χσ (x, y) = 0 otherwise.
Let h ∈ Rd , h 6= 0. One has
X X X δσ u
|u(x + h) − u(x)| ≤ χσ (x, x + h)δσ u = χσ (x, x + h)dK,σ , for a.e. x ∈ Ω
dσ
σ∈E K∈M σ∈EK
(see Definition (5.2) for the definition of δσ u).

This gives, using the Cauchy-Schwarz inequality,
X X (δσ u)2 X X
|u(x + h) − u(x)|2 ≤ dK,σ χσ (x, x + h) χσ (x, x + h)dK,σ cK,σ , for a.e. x ∈ Rd ,
d2σ cK,σ
K∈M σ∈EK K∈M σ∈EK
(5.23)
h
where cK,σ = |nK,σ · |h| |.
Let us now prove that
79
X X
χσ (x, x + h)dK,σ cK,σ ≤ |h| + 4 h(M), (5.24)
K∈M σ∈EK
for a.e. x ∈ Rd .
Let x ∈ Rd such that σ ∩ [x, x + h] contains at most one point, for all σ ∈ E, and [x, x + h] does not contain any
vertex of M (proving (5.25) for such points x gives (5.25) for a.e. x ∈ Rd , since h is fixed).
Let y, z ∈ [x, x + h] such that (z − y) · h > 0 and [y, z] ⊂ Ω. Let us assume that there exists σ ∈ Eint
such that χσ (y, z) = 1. We then consider the sequence (Ki ), i = 1, . . . , k of the elements of M such that
]y, z[∩Ki 6= ∅, y ∈ K1 , z ∈ Kk , and for any σ ∈ E such that χσ (y, z) = 1, there exists i = 1, . . . k − 1 such
xKi+1 −xKi
that Mσ = {Ki , Ki+1 }. Definition 1.4 implies that nKi ,σ = dσ and nKi+1 ,σ = −nKi ,σ . Thanks to
(z − y) · h > 0, we have
xKi+1 − xKi
nKi ,σ · h = · h ≥ 0.
dσ
Hence we get
X X k−1
X h h
χσ (y, z)dK,σ cσ = (xKi+1 − xKi ) · = (xKk − xK1 ) · .
i=1
|h| |h|
K∈M σ∈EK
Since |y − xK1 | ≤ h(M) and |z − xKk | ≤ h(M), this leads to

X X
χσ (y, z)dK,σ cσ ≤ |y − z| + 2 h(M). (5.25)
K∈M σ∈EK
In the case where there does not exist σ ∈ Eint such that χσ (y, z) = 1, (5.25) remains true. Note that (5.25)
implies (5.24) if [x, x + h] ⊂ Ω.
We now consider the case where [x, x + h] 6⊂ Ω. In the case where x ∈
/ Ω and x + h ∈ / Ω, then χσ (x, x + h) = 0
for all σ ∈ E, and (5.24) holds. In the case where x ∈ Ω and x + h ∈/ Ω, there exists z ∈ [x, x + h] ∩ ∂Ω such
that [x, z] ⊂ Ω. We then have χσ (x, x + h) = χσ (x, z) for all σ ∈ E. Then (5.25), with y = x, implies (5.24).
In the case where x ∈ / Ω and x + h ∈ Ω, there exists y ∈ [x, x + h] ∩ ∂Ω such that [y, x + h] ⊂ Ω. We then
have χσ (x, x + h) = χσ (y, x + h) for all σ ∈ E. Then (5.25), with z = x + h, implies (5.24).
Finally, in the case where x ∈ Ω and x + h ∈ Ω, there exists z ∈ [x, x + h] ∩ ∂Ω such that [x, z] ⊂ Ω and
y ∈ [x, x + h] ∩ ∂Ω such that [y, x + h] ⊂ Ω. We then have χσ (x, x + h) = χσ (x, z) + χσ (y, x + h) for all
σ ∈ E, and we get from (5.25):
X X
χσ (x, x + h)dK,σ cσ ≤ |x − z| + 2 h(M) + |x + h − y| + 2 h(M),
K∈M σ∈EK
which implies (5.24).

In order to conclude the proof of Lemma 5.8, remark that, for all K ∈ M and σ ∈ EK ,
Z
χσ (x, x + h)dx ≤ m(σ)cK,σ |h|.
Rd
d
Therefore, integrating (5.23) over R yields, with (5.24),
Ã µ ¶2 !
X X δσ u
2
ku(· + h) − ukL2 (Rd ) ≤ m(σ)dK,σ |h|(|h| + 4 h(M)),
dσ
K∈M σ∈EK
which is (5.22). ¤
Applying the previous lemma with any h ∈ Rd such that |h| = diam(Ω) yields the following one.
Lemma 5.9 (Discrete Poincaré inequality) Under hypothesis (1.5), let D be a pointed ∆−adapted polygonal
finite volume space discretization of Ω in the sense of Definition 1.4. Then the following inequality holds:
kukL2 (Ω) ≤ (diam(Ω) + 2 h(M))kuk1,2,D ≤ 3 diam(Ω)kuk1,2,D , ∀u ∈ HM (Ω). (5.26)
80
Remark 5.1 Lemma 5.2 provides a similar inequality to (5.26), but the constant C1 given in this lemma depends
on some regularity factor of the discretization which is not required here. Note that in [20], under additional
hypotheses on the points xK for any K ∈ M such that EK ∩ Eext 6= ∅, the inequality (5.26) can be replaced by
kukL2 (Ω) ≤ diam(Ω)kuk1,2,D , ∀u ∈ HM (Ω).
The next lemma provides the weak convergence of a discrete gradient (this lemma has been first given in [24]),
defined by constant values by cone DK,σ .
Lemma 5.10 (Weak convergence of a discrete gradient) Under hypothesis (1.5), let F be a family of pointed
∆−adapted polygonal finite volume space discretization of Ω in the sense of Definition 1.4.
We assume that, for all D ∈ F, there exists uD ∈ HM (Ω) such that kuD k1,2,D remains bounded by a value C5 ,
and such that there exists u ∈ L2 (Ω) such that uD → u in L2 (Ω). We again define, for all u ∈ HM (Ω)
δK,σ u = uL − uK , ∀σ ∈ Eint , Mσ = {K, L},
(5.27)
δK,σ u = −uK , ∀σ ∈ Eext , Mσ = {K}.
For any D ∈ F , we define the function GD ∈ LD (Ω)d (see Definition 1.3), by

δK,σ u
GD
K,σ = d nK,σ , ∀K ∈ M, ∀σ ∈ EK . (5.28)
dσ
Then u ∈ H01 (Ω) holds, and GD weakly converges to ∇u in L2 (Ω)d as h(M) → 0.
Proof. We first prolong u and GD by 0 outside Ω. Let us first remark that, thanks to the Cauchy-Schwarz
inequality and using m(DK,σ ) = m(σ)dK,σ /d,
X X µ ¶2
δσ u
kGD k2L2 (Rd )d = d m(σ)dK,σ = dkuD k21,2,D .
dσ
K∈M σ∈EK
Since kGD kL2 (Rd )d remains bounded, we can extract from F a subfamily, again denoted F, such that there exists
G ∈ L2 (Rd )d and GD weakly converges in L2 (Rd )d to ∇u in G as h(M) → 0. Let us prove that G = ∇u,
which implies, by uniqueness of the limit, that the whole family converges to ∇u as h(M) → 0. R
1
Let ϕ ∈ Cc1 (Ω)d be given. Let us denote by ϕD ∈ LD (Ω)d the element defined by the value m(σ) σ
ϕ(x)dγ(x),
in DK,σ for all K ∈ Mσ , again prolonged by 0 outside Ω. We then have that ϕD converges to ϕ in L2 (Rd )d . We
get, by reordering the terms in the summation, that
Z X X Z Z
GD (x) · ϕ(x)dx = − uK ϕ(x)dγ(x)nK,σ = − uD (x)div(ϕ)(x)dx,
Rd K∈M σ∈EK σ Rd
which implies, by passing to the limit h(M) → 0,

Z Z
G(x) · ϕ(x)dx = − u(x)div(ϕ)(x)dx.
Rd Rd
This proves that u ∈ H 1 (Rd ). Since u = 0 outside Ω, we get that u ∈ H01 (Rd ), and that
Z Z
G(x) · ϕ(x)dx = ∇u(x) · ϕ(x)dx,
Rd Rd
which proves that G = ∇u, hence completing the proof of the lemma. ¤
We can now state a compactness lemma, which holds without the regularity hypothesis on the mesh (5.8).
Lemma 5.11 (Compactness in L2 (Ω)) Under hypothesis (1.5), let F be a family of pointed ∆−adapted polyg-
onal finite volume space discretization of Ω in the sense of Definition 1.4. For M ∈ F , let uM ∈ HM (Ω) and
assume that there exists C ∈ R such, for all M ∈ F , kuM k1,2,D ≤ C. Then the family (uM )M∈F is relatively
compact in L2 (Ω) and also in L2 (Rd ) taking uM = 0 outside Ω. Moreover, the limit u ∈ L2 (Ω) of any converging
sequence extracted from the family as h(M) → 0 is such that u ∈ H01 (Ω).
81
Proof. The proof is quite easy. lemma 5.9 gives that the family (uM )M∈F is bounded in L2 (Ω) and in L2 (Rd )
taking uM = 0 outside Ω. Thanks to en, thanks to Kolmogorov Copmpactness Theorem, lemma 5.4 gives that the
family (uM )M∈F is relatively compact in L2 (Ω) and also in L2 (Rd ). The regularity of the limit is an immediate
consequence of Lemma 5.10. Note that it can also be proven by passing to the limit h(M) → 0 in (5.22) (this is
the method used in [20]). ¤
Lemma 5.12 (Strong convergence of an interpolated approximate gradient) Under hypothesis (1.5), let F be
a family of pointed ∆−adapted polygonal finite volume space discretization of Ω in the sense of Definition 1.4.
Let ϕ ∈ Cc2 (Ω) be given.
For any D ∈ F , we define the function H D ∈ LD (Ω)d (see Definition 1.3), by
D δK,σ ΠM ϕ
HK,σ = nK,σ + ∇ϕ(xK ) − (∇ϕ(xK ) · nK,σ )nK,σ , ∀K ∈ M, ∀σ ∈ EK , (5.29)
dσ
which can also be developped in
D ϕ(xL ) − ϕ(xK )
HK,σ = nK,σ + ∇ϕ(xK ) − (∇ϕ(xK ) · nK,σ )nK,σ ,
dσ
∀K ∈ M, ∀σ ∈ EK ∩ Eint , Mσ = {K, L}
D 0 − ϕ(xK )
HK,σ = nK,σ + ∇ϕ(xK ) − (∇ϕ(xK ) · nK,σ )nK,σ ,
dσ
∀K ∈ M, ∀σ ∈ EK ∩ Eext .
Then H D converges to ∇ϕ in L∞ (Ω)d as h(M) → 0.
Proof. We have, for K ∈ M and σ ∈ EK ∩ Eint with Mσ = {K, L},
d2σ
ϕ(xL ) − ϕ(xK ) = dσ ∇ϕ(xK ) · nK,σ + RK,σ ,
2
with |RK,σ | is bounded by some L∞ norm of the second order partial derivatives of ϕ, denoted by C6 . We consider
h(M) small enough, such that ϕ vanishes on all K ∈ M such that EK ∩ Eext 6= ∅, hence for σ ∈ EK ∩ Eext ,
D
HK,σ = 0 = ∇ϕ(xK ).
We then get
D dσ
|HK,σ − ∇ϕ(xK )| ≤ C6 ≤ h(M)C6 , ∀K ∈ M, ∀σ ∈ EK .
2
This concludes the proof of the lemma. ¤
Let us now give the following lemma [27; 18].
Lemma 5.13 (Convergence of discrete scalar product, ∆−adapted mesh) Under hypothesis (1.5), let F be a
family of pointed ∆−adapted polygonal finite volume space discretization of Ω in the sense of Definition 1.4.
We assume that, for all D ∈ F, there exists uD ∈ HM (Ω) such that kuD k1,2,D remains bounded by a value C7 ,
and such that there exists u ∈ H01 (Ω) such that uD → u in L2 (Ω). We also assume that, for all D ∈ F , there
exists g D ∈ LM (Ω), such that there exists g ∈ L2 (Ω) with g D → g in L2 (Ω).
For any D ∈ F , we define
X X δK,σ v δK,σ w
[v, w]D,α = αK,σ m(σ)dK,σ , ∀v, w ∈ HM (Ω), ∀α ∈ LD (Ω). (5.30)
dσ dσ
K∈M σ∈EK
Then the following holds:

Z
lim [uD , ΠM ϕ]D,gD = g(x)∇u(x) · ∇ϕ(x)dx, ∀ϕ ∈ Cc2 (Ω). (5.31)
h(M)→0 Ω
Moreover, we have, if g ∈ L∞ (Ω),

Z
lim inf [uD , uD ]D,|gD | ≥ |g(x)||∇u(x)|2 dx. (5.32)
h(M)→0 Ω
82
Remark 5.2 Under the hypotheses of the lemma, one can derive the existence of u from results obtained in Section
5.2.
Proof. Let us first show (5.31). Let GD ∈ LD (Ω)d be defined by (5.28) and H D ∈ LD (Ω)d be defined by (5.29).
We deduce from Lemma 5.12, that g D H D tends to g∇ϕ in L2 (Ω), which implies, by weak/strong convergence,
that Z Z
lim g D (x)GD (x) · H D (x)dx = g(x)∇u(x) · ∇ϕ(x)dx.
h(M)→0 Ω Ω
We now remark that, thanks to definition (5.30) of [·, ·]D,· ,

Z
g D (x)GD (x) · H D (x)dx = [uD , ΠM ϕ]D,gD ,
Ω
which concludes the proof of (5.31).

Let us now prove (5.32). We now assume that g ∈ L∞ (Ω). We have
[uD − ΠM ϕ, uD − ΠM ϕ]D,|gD | ≥ 0, ∀D ∈ F .
Hence we get
[uD , uD ]D,|gD | ≥ 2[uD , ΠM ϕ]D,|gD | − [ΠM ϕ, ΠM ϕ]D,|gD | , ∀D ∈ F .
Applying (5.31), we get
Z
lim [uD , ΠM ϕ]D,gD = |g(x)|∇u(x) · ∇ϕ(x)dx.
h(M)→0 Ω
Using that kΠM ϕk1,1,D remains bounded, we can again apply (5.31), substituting ΠM ϕ to uD . It leads to
Z
lim [ΠM ϕ, ΠM ϕ]D,gD = |g(x)|∇ϕ(x) · ∇ϕ(x)dx.
h(M)→0 Ω
Hence we get Z
lim inf [uD , uD ]D,|gD | ≥ |g(x)|(2∇ϕ(x) − ∇u(x)) · ∇ϕ(x)dx.
h(M)→0 Ω
Since the above inequality holds for all ϕ ∈ Cc2 (Ω), letting ϕ → u in H 1 (Ω) provides the convergence in L1 (Ω)
of (2∇ϕ − ∇u) · ∇ϕ to ∇u. Hence we get (5.32). ¤
In the same spirit as the previous lemma, let us give the following result, given in [30].
Lemma 5.14 (Weak-Strong convergence for space-time problems)

Under hypothesis (1.5), let T > 0 be given. Let (Dm )m∈N be a sequence of finite volume discretizations of
Ω × (0, T ) in the sense of Definitions 1.4 and 1.7 such that lim size(Dm ) = 0 and such that the regularity
m→+∞
property
X
∃θ ∈ R+ such that ∀K ∈ Mm , mK|L dK|L ≤ θm(K). (5.33)
L∈NK
is satisfied. Let (vDm )m∈N ⊂ L2 (Ω × (0, T )) (resp. (wDm )m∈N ⊂ L2 (Ω × (0, T ))) be a sequence of piecewise
constant functions corresponding to a sequence of discrete functions (vDm )m∈N (resp. (wDm )m∈N ) from Mm ×
[[0, Nm + 1]] to R. Assume that there exists a real value C8 > 0 independent on m verifying
N
X X X m(K|L)
δtn n+1
(vL n+1 2
− vK ) ≤ C8 ,
n=0
dK|L
K∈ML∈NK
83
and that the sequence (vDm )m∈N (resp. (wDm )m∈N ) converges to some function v ∈ L2 (Ω × (0, T )) (resp.
w ∈ L2 (Ω × (0, T ))) weakly (resp. strongly) in L2 (Ω × (0, T )), as m → +∞. Let ϕ ∈ C ∞ (Rd × R) such that
ϕ(·, T ) = 0 and ∇ϕ · n = 0 on ∂Ω × (0, T ). For all m ∈ N, let Am be defined by
N
X X X m(K|L)
Am = − δtn ϕ(xK , tn+1 ) n+1 n+1
wK,L n+1
(vL − vK ),
n=0
dK|L
K∈M L∈NK
n+1 n+1 n+1 n+1 n+1

where, for all (K, L) ∈ E and n ∈ [[0, N ]], wK,L = wL,K , and wK,L is either equal to wK or to wL . Then
2 1
v ∈ L (0, T ; H (Ω)) and
Z TZ
lim Am = w(x, t)∇v(x, t)∇ϕ(x, t)dxdt.
m→+∞ 0 Ω
5.5 Lemma used for time translates

Lemma 5.15 Let (tn )n∈Z be a stricly increasing sequence of real values such that δtn := tn+1 − tn is uniformly
bounded, lim tn = −∞ and lim tn = ∞. For all t ∈ R, we denote by n(t) the element n ∈ Z such that
n→−∞ n→∞
t ∈ [tn , tn+1 ). Let (an )n∈Z be a family of non negative real values with a finite number of non zero values. Then
Z n(t+τ )
X X
(δtn an+1 )dt = τ (δtn an+1 ), ∀τ ∈ (0, +∞), (5.34)
R n=n(t)+1 n∈Z
and
 
Z n(t+τ )
X X
 δt n  n(t+ζ)+1
a dt ≤ (τ + max δtn ) (δtn an+1 ), ∀τ ∈ (0, +∞), ∀ζ ∈ R. (5.35)
R n∈Z
n=n(t)+1 n∈Z
Proof. Let us define the function χ(t, n, τ ) by χ(t, n, τ ) = 1 if t < tn and t + τ ≥ tn , else χ(t, n, τ ) = 0. We
have
Z n(t+τ
X) Z X Xµ Z ¶
(δtn an+1 )dt = (δtn an+1 χ(t, n, τ ))dt = δtn an+1 χ(t, n, τ )dt .
R n=n(t)+1 R n∈Z n∈Z R
R R tn
Since R χ(t, n, τ )dt = tn −τ dt = τ , thus (5.34) is proven.
We now turn to the proof of (5.35). We define the function χ̃(n, t) by χ̃(n, t) = 1 if n(t) = n, else χ̃(n, t) = 0.
We have    
Z n(t+τ )
X Z n(t+τ )
X X
 δt  an(t+ζ)+1 dt =
n  δt 
n
am+1 χ̃(m, t + ζ)dt,
R n=n(t)+1 R n=n(t)+1 m∈Z
which yields
   
Z n(t+τ )
X X Z tm+1 −ζ n(t+τ )
X
 δtn  an(t+ζ)+1 dt = am+1  δtn  dt. (5.36)
R n=n(t)+1 m∈Z tm −ζ n=n(t)+1
Since we have
n(t+τ )
X X
δtn = (tn+1 − tn ) ≤ τ + max δtn ,
n∈Z
n=n(t)+1 n∈Z, t<tn ≤t+τ
we can write from (5.36)

 
Z n(t+τ )
X X Z tm+1 −ζ X
 δtn  an(t+ζ)+1 dt ≤ (τ + max δtn ) am+1 dt = (τ + max δtn ) am+1 δtm ,
R n∈Z tm −ζ n∈Z
n=n(t)+1 m∈Z m∈Z
84
which is exactly (5.35). ¤

We have the following corollary of Kolmogorov’s theorem.
Corollary 5.1 (Consequence of Kolmogorov’s theorem) Let Ω be a bounded open subset of Rd , d ∈ N∗ , and let
(wn )n∈N be a sequence of functions such that:
for all n ∈ N, wn ∈ L∞ (Ω) and there exists a real value Cb > 0 which does not depend on n such that
kwn kL∞ (Ω) ≤ Cb ,
there exists a real value CK > 0 and a sequence of non negative real values (µn )n∈N verifying lim µn = 0
n→∞
and
Z
(wn (x + ξ) − wn (x))2 dx ≤ CK (|ξ| + µn ), ∀n ∈ N, ∀ξ ∈ Rd .
Ωξ
where, for all ξ ∈ Rd , we set Ωξ = {x ∈ Rd , [x, x + ξ] ⊂ Ω}.
Then there exists a subsequence of (wn )n∈N , again denoted (wn )n∈N , and a function w ∈ L∞ (Ω) such that
wn → w in Lp (Ω) as n → ∞ for all p ≥ 1.
Proof. We first extend the definition of wn on Rd by the value 0 on Rd \ Ω. We then prove that
Z
lim sup (wn (x + ξ) − wn (x))2 dx = 0. (5.37)
|ξ|→0 n∈N Rd
Indeed, let ε > 0. Since ∪η>0 {x ∈ Ω, B(x, η) ⊂ Ω} = Ω, then there exists η̄ > 0 such that
meas(Ω \ {x ∈ Ω, B(x, η̄) ⊂ Ω}) ≤ ε.
Thus, for all ξ ∈ Rd verifying |ξ| ≤ η̄, we have
{x ∈ Ω, B(x, η̄) ⊂ Ω} ⊂ Ωξ ∪ Ω−ξ ⊂ Ω
and therefore meas(Ω \ (Ωξ ∪ Ω−ξ )) ≤ ε. Let n0 ∈ N such that, for all n > n0 , µn ≤ ε. Thanks to the theorem
of continuity
R in means, for all n = 0, . . . , n0 , there exists ηn > 0 such that, for all ξ ∈ Rd verifying |ξ| ≤ ηn , we
have Rd (wn (x + ξ) − wn (x))2 dx ≤ ε.
d
We now takeR ξ ∈ R verifying |ξ| 2 ≤ min(η̄, (ηn )n=0,...,n0 , ε). We then get that, for all n = 0, . . . , n0 , the
inequality Rd (wn (x + ξ) − wn (x)) dx ≤ ε holds, and for all n ∈ N such that n > n0 , then
Z Z
(wn (x + ξ) − wn (x))2 dx ≤ 4Cb2 ε + (wn (x + ξ) − wn (x))2 dx,
Rd Ωξ
and Z
(wn (x + ξ) − wn (x))2 dx ≤ CK (|ξ| + µn ) ≤ 2CK ε.
Ωξ
Gathering the previous results gives (5.37). Then applying Kolmogorov’s theorem gives the conclusion of Corol-
lary 5.1. ¤
Let us now give the following corollary of Ascoli’s theorem, useful for convergence at any time.
Theorem 5.2 Let T > 0 be given, let E be a separable Banach space and E 0 be its dual. Let (un )n∈N be a
sequence of applications from [0, T ] to E 0 such that there exists C > 0 with kun (t)kE 0 ≤ C for all n ∈ N and
t ∈ [0, T ]. We assume that there exists a supspace F ⊂ E with the following properties:
1. F is dense in E
85
2. for all ϕ ∈ F , the family (un,ϕ )n∈N of functions un,ϕ : [0, T ] → R defined for all t ∈ [0, T ] by un,ϕ (t) =
hun (t), ϕi is equicontinuous.
Then there exists a subsequence of (un )n∈N , again denoted (un )n∈N , and an element u ∈ C 0 (0, T ; E 0 ) (E 0 being
embedded with the weak-? topology) such that for all t ∈ [0, T ], the sequence (un (t))n∈N converges to u for the
weak-? topology of E 0 .
Proof. Let (tm )m∈N be a sequence of elements of [0, T ], dense in [0, T ]. For all m ∈ N, the sequence
(un (tm ))n∈N is bounded in E 0 . Therefore there exists a subsequence of (un (tm ))n∈N which converges for the
weak-? topology of E 0 . Using the diagonal process, we extract from (un )n∈N a subsequence, again denoted
(un )n∈N , such that for all m ∈ N, the sequence (un (tm ))n∈N converges for the weak-? topology of E 0 .
Let t ∈ [0, T ]. For ϕ ∈ F , let us prove that the sequence (hun (t), ϕi)n∈N is a Cauchy sequence. Let ε > 0 be
given. Let η > 0 be such that, for all t0 ∈ [0, T ] with |t − t0 | ≤ η and for all n ∈ N, |hun (t) − un (t0 ), ϕi| ≤ ε.
Let m ∈ N be such that |t − tm | ≤ η. Since the sequence (un (tm ))n∈N converges for the weak-? topology of
E 0 , the sequence (hun (tm ), ϕi)n∈N is a Cauchy sequence. Hence there exists n0 ∈ N such that, for all n, p ≥ n0 ,
|hun (tm ) − up (tm ), ϕi| ≤ ε. For such n, p ≥ n0 , we get
|hun (t) − up (t), ϕi| ≤ |hun (t) − un (tm ), ϕi| + |hun (tm ) − up (tm ), ϕi| + |hup (tm ) − up (t), ϕi| ≤ 3 ε.
This shows that the sequence (hun (t), ϕi)n∈N is a Cauchy sequence.
Let ψ ∈ E. Let us prove that the sequence (hun (t), ψi)n∈N is a Cauchy sequence. Let ε > 0 be given. We first
choose ϕ ∈ F such that kψ − ϕkE ≤ ε. Since the sequence (hun (t), ϕi)n∈N is a Cauchy sequence, there exists
n0 ∈ N such that, for all n, p ≥ n0 , |hun (t) − up (t), ϕi| ≤ ε. For such n, p ≥ n0 , we get
|hun (t) − up (t), ψi| ≤ |hun (t), ψ − ϕi| + |hun (t) − up (t), ϕi| + |hup (t), ϕ − ψi| ≤ (2 C + 1)ε.
This shows that the sequence (hun (t), ψi)n∈N is a Cauchy sequence. Since for all t ∈ [0, T ], n ∈ N and ψ ∈ E,
we have |hun (t), ψi| ≤ CkψkE , we get that
¯ ¯
¯ ¯
¯ lim hun (t), ψi¯ ≤ CkψkE .
n→∞
Hence the application u defined, for all t ∈ [0, T ] by u(t)(ψ) = limn→∞ hun (t), ψi for all ψ ∈ E is such that
u(t) ∈ E 0 .
Let us show that u ∈ C 0 (0, T ; E 0 ), E 0 being embedded with the weak-? topology. Let t ∈ [0, T ] and ψ ∈ E. Let
ε > 0 be given. We first choose ϕ ∈ F such that kψ − ϕkE ≤ ε. Let η > 0 be such that, for all t0 ∈ [0, T ] with
|t−t0 | ≤ η and for all n ∈ N, |hun (t)−un (t0 ), ϕi| ≤ ε. Passing to the limit n → ∞, we get |hu(t)−u(t0 ), ϕi| ≤ ε.
We then get
|hu(t) − u(t0 ), ψi| ≤ |hu(t), ψ − ϕi| + |hu(t) − u(t0 ), ϕi| + |hu(t0 ), ϕ − ψi| ≤ (2 C + 1)ε.
This shows that the function uψ : [0, T ] → R defined for all t ∈ [0, T ] by uψ (t) = hu(t), ψi is continuous, hence
proving u ∈ C 0 (0, T ; E 0 ).
¤
Bibliography
[1] C. Bardos, A.Y. Leroux, and J.C. Nédélec. First order quasilinear equations with boundary conditions. Comm.
P. D. E., 4(9):1017–1034, 1979.
[2] Ph. Bénilan. Equations d’évolution dans un espace de banach quelconque et applications. Thèse d’état,
Université d’Orsay, France, 1972.
[3] F. Bouchut, F. Guarguaglini, and R. Natalini. Diffusive bgk approximations for nonlinear multidimensional
parabolic equations. preprint, 1999.
[4] F. Boyer and P. Fabrie. Eléments d’analyse pour l’étude de quelques modèles d’écoulements de fluides
visqueux incompressibles, volume 52. Springer, 2006.
[5] H. Brezis. Analyse fonctionnelle. Collection Mathématiques Appliquées pour la Maı̂trise. [Collection of
Applied Mathematics for the Master’s Degree]. Masson, Paris, 1983. Théorie et applications. [Theory and
applications].
[6] F. Brezzi, K. Lipnikov, and M. Shashkov. Convergence of the mimetic finite difference method for diffusion
problems on polyhedral meshes. SIAM J. Numer. Anal., 43(5):1872–1896, 2005.
[7] F. Brezzi and J. Pitkäranta. On the stabilization of finite element approximations of the Stokes equations.
In Efficient solutions of elliptic systems (Kiel, 1984), volume 10 of Notes Numer. Fluid Mech., pages 11–19.
Vieweg, Braunschweig, 1984.
[8] J. Carrillo. Entropy solutions for nonlinear degenerate problems. Arch. Rat. Mech. Anal., 147:269–361, 1999.
[9] Claire Chainais-Hillairet. Finite volume schemes for a nonlinear hyperbolic equation. Convergence towards
the entropy solution and error estimate. M2AN Math. Model. Numer. Anal., 33(1):129–156, 1999.
[10] S. Champier, T. Gallouët, and R. Herbin. Convergence of an upstream finite volume scheme for a nonlinear
hyperbolic equation on a triangular mesh. Numer. Math., 66(2):139–157, 1993.
[11] E. Chenier, R. Eymard, and O. Touazi. Numerical results using a colocated finite-volume scheme on unstruc-
tured grids for incompressible fluid flows. Numerical Heat Transfer Part B: Fundamentals, 49(3):259–276,
September 2006.
[12] B. Cockburn, F. Coquel, and Ph. LeFloch. An error estimate for finite volume methods for multidimensional
conservation laws. Math. Comput., 63(207):77–103, 1994.
[13] B. Cockburn and G. Gripenberg. Continuous dependence on the nonlinearities of solutions of degenerate
parabolic equations. J. Differential Equations, 151(2):231–251, 1999.
[14] K. Deimling. Nonlinear functional analysis. Springer-Verlag, Berlin, 1985.
[15] R. DiPerna. Measure-valued solutions to conservation laws. Arch. Rat. Mech. Anal., 88:223–270, 1985.
[16] E. Eymard, T. Gallouët, D. Hilhorst, and Y. Slimane. Finite volumes and nonlinear diffusion equations.
RAIRO Modél. Math. Anal. Numér., 32(6):747–761, 1998.
86
87
[17] R. Eymard, T. Gallouët, M. Ghilani, and R. Herbin. Error estimates for the approximate solutions of a
nonlinear hyperbolic equation given by finite volume schemes. IMA J. Numer. Anal., 18(4):563–594, 1998.
[18] R. Eymard, T. Gallouët, R. Herbin, and A. Michel. Convergence of a finite volume scheme for nonlinear
degenerate parabolic equations. Numer. Math., 92(1):41–82, 2002.
[19] R. Eymard, T. Gallouët, and R. Herbin. Convergence of finite volume schemes for semilinear convection
diffusion equations. Numer. Math., 82(1):91–116, 1999.
[20] R. Eymard, T. Gallouët, and R. Herbin. Finite volume methods. In P. G. Ciarlet and J.-L. Lions, editors,
Techniques of Scientific Computing, Part III, Handbook of Numerical Analysis, VII, pages 713–1020. North-
Holland, Amsterdam, 2000.
[21] R. Eymard, T. Gallouët, and R. Herbin. Error estimate for approximate solutions of a nonlinear convection-
diffusion problem. Adv. Differential Equations, 7(4):419–440, 2002.
[22] R. Eymard, T. Gallouët, and R. Herbin. A cell-centered finite-volume approximation for anisotropic diffusion
operators on unstructured meshes in any space dimension. IMA J. Numer. Anal., 26(2):326–353, 2006.
[23] R. Eymard, T. Gallouët, and R. Herbin. A discretization scheme for anisotropic heterogeneous diffusion
problems. Benchmark on discretization schemes for anisotropic diffusion problems on general grids, FVCA5,
http://www.latp.univ-mrs.fr/fvca5, 2008.
[24] R. Eymard and T. Gallouët. H-convergence and numerical schemes for elliptic equations. SIAM Journal on
Numerical Analysis, 41(2):539–562, 2000.
[25] R. Eymard and R. Herbin. A new colocated finite volume scheme for the incompressible navier-stokes
equations on general non matching grids. C. R. Math. Acad. Sci. Paris, 344(10):659–662, 2007.
[26] T. Gallouët, R. Herbin, and M.-H. Vignal. Error estimates on the approximate finite volume solution of
convection diffusion equations with general boundary conditions. SIAM J. Numer. Anal., 37(6):1935–1972
(electronic), 2000.
[27] R. Herbin and E. Marchand. Finite volume approximation of a class of variational inequalities. IMA J. Numer.
Anal., 21(2):553–585, 2001.
[28] R. Herbin. Finite volume methods for diffusion convection equations on general meshes. Finite volumes for
complex applications, Problems and Perspectives,, pages 153–160, 1996.
[29] A. Lagha-Benabdallah and F. Smadhi. Existence de solutions faibles pour un problème aux limites associé à
une équation parabolique dégénérée. Maghreb Math. Rev., 2:201–222, 1993.
[30] A. Michel. A finite volume scheme for two-phase immiscible flow in porous media. SIAM J. Numer. Anal.,
41:1301–1317, 2003.
[31] M. Ohlberger. A posteriori error estimates for vertex centered finite volume approximations of convection-
diffusion-reaction equations. preprint 12, 2000.
[32] R. R. Eymard, Herbin R., and J.-C. Latché. Convergence analysis of a colocated finite volume scheme for
the incompressible Navier-Stokes equations on general 2 or 3d meshes. SIAM J. Numer. Anal., 45(1):1–36,
2007.
[33] Roger Temam. Navier-Stokes equations, volume 2 of Studies in Mathematics and its Applications. North-
Holland Publishing Co., Amsterdam, third edition, 1984. Theory and numerical analysis, With an appendix
by F. Thomasset.
[34] J.P. Vila. Convergence and error estimate in finite volume schemes for general multidimensional conservation
laws, i. explicit monotone schemes. Model. Math. Anal. Numer., 28(3):267–285, 1994.
[35] M. Vohralı́k. Equivalence between mixed finite element and multi-point finite volume methods. C. R. Acad.
Sci. Paris., Ser. I, 339:525–528, 2004.
Index
notation ∆−adapted pointed polygonal finite volume space

|D| Def. 1.7, 8 discretization, 7
δK,σ u (2.17), 11 ∆−super-adapted pointed polygonal finite volume
n+1
δK,σ g(u) (3.9), 36 space discretization, 8
δσ u (5.2), Def. 5.1, 72 Pointed polygonal finite volume space discretiza-
δσn+1 g(u) (3.30), Lemma 3.5, 43 tion, 6
∂K Def. 1.1, 6 Pointed star-shaped polygonal finite volume space
DK,σ (1.11), Def. 1.3, 7 discretization, 7
dK,σ Def. 1.2, 6 Space-time discretization, 8
Dσ Def. 1.3, 7 Time discretization, 8
dσ (1.12), Def. 1.3, 7
δtn Def. 1.6, 8 Polygonal finite volume space discretization, 6
E Def. 1.1, 6
EK Def. 1.1, 6 Star shaped, 7
∇D u (2.30), 14
Topological degree, 72
h(M) (1.7), Def. 1.1, 6
HM (Ω) Def. 1.1, 6
LD (Ω) Def. 1.3, 7
LD,0 (Ω) 4.2, 61
M Def. 1.1, 6
m(K) Def. 1.1, 6
m(σ) Def. 1.1, 6
Mσ Def. 1.1, 6
NK Def. 1.1, 6
nK,σ Def. 1.1, 6
kuk1,1,D (5.1), Definition 5.1, 72
kukBV (5.7), 73
P Def. 1.2, 6
PE 4.2, 61
PD (1.9), Def. 1.2, 6
PM (1.8), Def. 1.2, 6
ΠM u Def. 1.1, 6
[v, w]D,α (5.30), Lemma 5.13, 81
n+1
uK,σ (3.8), 36
XD Def. 1.1, 6
XD,0 Def. 1.1, 6
XE 4.2, 61
xK Def. 1.2, 6
xσ Def. 1.2, 6
YD (2.47), 17
discretization
88

Finite Volume Methods for Solving PDEs

Uploaded by

Document Information

Original Description:

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

Finite Volume Methods for Solving PDEs

Uploaded by

Copyright:

Available Formats

Finite Volume Methods

Schemes and Analysis

Robert Eymard1 , Thierry Gallouët2 and Raphaèle Herbin3

1 Finite volume discretization

2 Linear diffusion problems 9

3 Nonlinear convection - degenerate diffusion problems 34

5 Discrete functional analysis 70

Finite volume discretization

1.1 A generic example

ut + div(−Λ∇u + vu) + bu = 0 on Ω, (1.4)

the flux through a given edge then reads:

1.2 Finite volume meshes and space-time discretizations

Definition 1.1 (Polygonal finite volume space discretization of Ω)

NK = {L ∈ M \ {K}, ∃σ ∈ Eint , Mσ = {K, L}}. (1.6)

Definition 1.2 (Pointed polygonal finite volume space discretization of Ω)

1. (M, E) is a polygonal finite volume discretization of Ω in the sense of definition 1.1.

We then define PM : C(Ω) → HM (Ω) by

PM ϕ(x) = ϕ(xK ) for a.e. x ∈ K, ∀K ∈ M, ∀ϕ ∈ C(Ω). (1.8)

PD ϕK = ϕ(xK ), ∀K ∈ M, PD ϕσ = ϕ(xσ ), ∀σ ∈ E, ∀ϕ ∈ C(Ω). (1.9)

Remark 1.1 The conditions xK ∈ K and xσ ∈ σ can easily be relaxed.

Figure 1.1: Pointed mesh

Thanks to the algebraic definition of dK,σ , we have the relation

Definition 1.3 (Pointed star-shaped polygonal finite volume space discretization of Ω)

DK,σ = {txK + (1 − t)y, t ∈ (0, 1), y ∈ σ}, (1.11)

dσ = dL,σ + dK,σ , ∀σ ∈ Eint , Mσ = {K, L},

Definition 1.4 (∆−adapted pointed polygonal finite volume space discretization of Ω)

An example of two neighbouring control volumes K and L of M is depicted in Figure 1.2.

Figure 1.2: Notations for a ∆-adapted mesh

|D| = max(h(M), (δtn )n∈[[0,N ]] )

Linear diffusion problems

2.1 Actual problems and schemes

Ω is an open bounded connected polygonal subset of Rd , d ∈ N? , (2.2)

Λ is a measurable function from Ω to Md (R), where Md (R) denotes the set of

Common features of the schemes

The case of the Laplace operator on ∆−adapted pointed meshes

Recall that the above scheme also writes

The case of harmonic averaging of an isotropic tensor

denoting by λσ , for all σ ∈ E, the harmonic averaged value defined by:

Elimination of uσ in the general case

The composite scheme which we consider in this work reads:

Then all the mathematical properties shown below still hold.

The case of an anisotropic tensor and of a ∆−adapted mesh

A first natural scheme, inspired by the finite element framework, would be

Gradient schemes on general non matching grids

Figure 2.1: Non conforming subsurface flow modelling

Mixed-type schemes on general non matching grids

YD = {((FK,σ )σ∈EK )K∈M , FK,σ ∈ R}. (2.47)

to connect an element F ∈ YD with an element u ∈ XD,0 by

Note that the definition X

leads to a non-invertible matrix. We then introduce the linear form

2.2 Numerical results

 with FK,σ (u) = K (uσ 0 − uK ), ∀K ∈ M, ∀σ ∈ EK .

Figure 2.2: Regular conforming coarse and fine triangular grids

nb. unknowns matrix size x=0 x=1 y=0 y=1

nb. unknowns matrix size x=0 x=1 y=0 y=1

2.3 Further analysis

and the regularity of the mesh by:

α|u|2X ≤ hu, uiF , ∀u ∈ XD , ∀D = (M, E, P) ∈ F. (2.64)

then the following holds

Finally the family of numerical fluxes Φ is said to be symmetric if

We may now state the general convergence theorem.

α|ΠM uD |2X ≤ C1 kf kL2 (Ω) |uD |X .