Homogenization of p-Laplacian in perforated domain
B. Amaziane
a,∗, S. Antontsev
b, L. Pankratov
a,c, A. Piatnitski
d,eaLaboratoire de Mathématiques et leur Applications, CNRS-UMR5142, Université de Pau, Av. de l’Université, 64000 Pau, France bCMAF, Universidade de Lisboa, Av. Prof. Gama Pinto, 2, 1649-003, Lisboa, Portugal
cDepartment of Mathematics, B.Verkin Institut for Low Temperature Physics and Engineering, 47, av. Lenin, 61103, Kharkov, Ukraine dNarvik University College, Postbox 385, Narvik, 8505, Norway
eLebedev Physical Institute RAS, leninski prospect 53, Moscow, 119991, Russia Received 4 February 2007; accepted 10 June 2009
Available online 4 July 2009
Abstract
We study the homogenization of the following nonlinear Dirichlet variational problem:
inf Ωε
1
pε(x)|∇u|pε(x)+ 1
pε(x)|u|pε(x)−f (x)u
dx:u∈W01,pε(·) Ωε
in a perforated domainΩε=Ω\Fε⊂Rn, n2, where εis a small positive parameter that characterizes the scale of the microstructure. The non-standard exponentpε(x)is assumed to be an oscillating continuous function inΩ¯ such that, for any ε >0, 1< pε(x)ninΩ; for anyx, y∈Ω,|pε(x)−pε(y)|ωε(|x−y|)with limτ→0ωε(τ )ln(1/τ )=0; and converges uniformly inΩto a functionp0which satisfies the same properties. Moreover, we assume thatpε(x)p0(x)inΩ. Denotinguε a minimizer in the above variational problem, without any periodicity assumption, for a large range of perforated domains we find, by means of the variational homogenization technique, the global behavior ofuεasεtends to zero. It is shown thatuεextended by zero inFε, converges weakly inW1,p0(·)(Ω)to the solution of the following nonlinear variational problem:
min Ω
1
p0(x)|∇u|p0(x)+ 1
p0(x)|u|p0(x)+c(x, u)−f (x)u
dx:u∈W01,p0(·)(Ω)
,
where the functionc(x, u)is defined in terms of the local characteristic ofΩε. This result is then illustrated with a periodic and a non-periodic examples.
©2009 Elsevier Masson SAS. All rights reserved.
Résumé
Nous étudions l’homogénéisation du problème variationnel de Dirichlet nonlinéaire suivant : inf
Ωε 1
pε(x)|∇u|pε(x)+ 1
pε(x)|u|pε(x)−f (x)u
dx:u∈W01,pε(·) Ωε
* Corresponding author.
E-mail addresses:brahim.amaziane@univ-pau.fr (B. Amaziane), antontsevsn@mail.ru (S. Antontsev), leonid.pankratov@univ-pau.fr (L. Pankratov), andrey@sci.lebedev.ru (A. Piatnitski).
0294-1449/$ – see front matter ©2009 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.anihpc.2009.06.004
dans un domaine perforéΩε=Ω\Fε⊂Rn,n2, oùε >0 est un petit paramètre qui caractérise la taille des perforations.
La fonction puissancepε(x)est nonstandard et supposée être une fonction continue et oscillante dansΩ. Elle vérifie, pour tout¯ ε >0, 1< pε(x)ndansΩ, pour toutx, y∈Ω,|pε(x)−pε(y)|ωε(|x−y|)avec limτ→0ωε(τ )ln(1/τ )=0 ; et elle est uniformément convergente dansΩvers une fonctionp0qui vérifie les mêmes propriétés. De plus, on suppose quepε(x)p0(x) dansΩ. On noteuε une solution du problème de minimisation variationnel ci-dessus, sans hypothèse de périodicité et pour différents milieux perforés, on trouve le problème limite décrivant le comportement global deuε lorsqueε tend vers zéro, en utilisant la technique de l’homogénéisation variationnelle. On montre queuε, prolongée par zéro dansFε, converge faiblement dansW1,p0(·)(Ω), quandεtend vers zéro, vers la solutionudu problème variationel nonlinéaire suivant :
min Ω
1
p0(x)|∇u|p0(x)+ 1
p0(x)|u|p0(x)+c(x, u)−f (x)u
dx:u∈W01,p0(·)(Ω)
,
où la fonctionc(x, u)est définie à partir des caractéristiques géométriques locales du domaineΩε. Enfin, nous présentons deux exemples, un périodique et l’autre nonpériodique, pour illustrer les résultats obtenus.
©2009 Elsevier Masson SAS. All rights reserved.
MSC:35B40; 35J60; 46E35; 74Q05; 76M50
Keywords:Homogenization; Nonlinear; Non-standard growth
1. Introduction
In this paper we study the homogenization of the following nonlinear problem:
−div∇uεpε(x)−2∇uε
+uεpε(x)−2uε=f (x) inΩε, uε∈W01,pε(·) Ωε
, (1.1)
whereεis a small positive parameter,Ωε=Ω\Fε is a perforated domain inRn (n2) withΩ being a bounded Lipschitz domain, and pε is a smooth positive oscillating function in Ω satisfying some conditions which will be specified in Section 3, and uniformly converging inΩ to a smooth functionp0.f is a given function. Equations of such type are calledpε(x)-Laplacian equations with non-standard growth conditions.
In recent years, there has been an increasing interest in the study of such equations (in the case where there is no dependence on the small parameter) motivated by their applications to the mathematical modeling in continuum mechanics. These equations arise, for example, from the modeling of non-Newtonian fluids with thermo-convective effects (see, e.g., [7,9]), the modeling of electro-rheological fluids (see, e.g., [30,31]), the thermistor problem (see, e.g., [39]), the problem of image recovery (see, e.g., [24]), and the motion of a compressible fluid in a heterogeneous anisotropic porous medium obeying to the nonlinear Darcy law (see, e.g., [8,11]).
Eq. (1.1) is an idealized model for a variety of interesting physical problems; we motivate our work by describing one of them. We consider a steady flow of a compressible barotropic gaz through a porous medium. The nonlinear Darcy law with the continuity equation lead to the equation given by [10]
−div
K(x)|∇u|p(x)−2∇u
+R(x)|u|p(x)−2u=f (x, t ). (1.2)
ustands for the fluid pressure,f is a source term andK,p,Rare characteristic functions of the heterogeneous porous medium. For more details on the formulation of such problems see for instance [10,13]. We refer to [10,11,17,18] and the references therein for a detailed analysis of such equations.
In the present paper we deal with the Dirichlet boundary value problem for the nonlinear equation (1.1). More precisely, we consider the corresponding variational problem:
inf
Ωε
1
pε(x)|∇u|pε(x)+ 1
pε(x)|u|pε(x)−f (x)u
dx: u∈W01,pε(·) Ωε
. (1.3)
The homogenization of the Dirichlet boundary value problem was studied for the first time in [25] and then it was revisited by many authors (see, e.g., [12,15,16,20,26,33], and the references therein). Note also that the homogeniza- tion of nonlinear elliptic equations is a long-standing problem and a number of methods have been developed. There is an extensive literature on this subject. We will not attempt a review of the literature here, but merely mention a few ref- erences, see for instance [2,14,16,29], and the references therein. Let us mention that the homogenization problems for
the Lagrangians with variable exponents were first studied in [22,34–37] (see also the book [38]) which focus on the variational functionals with non-standard growth conditions. In particular, the homogenization and Γ-convergence problems for Lagrangians with variable rapidly oscillating exponentsp(x)were considered in [35,36]. Variational functionals with non-standard growth conditions have also been considered in the book [14], namely Chapter 21 of this book focuses on theΓ-convergence of such functionals inLp spaces. The Dirichlet homogenization problem and related questions for Lagrangians ofpε(x)growth inW1,pε(·)(Ωε), whereΩε is a perforated domain, have been studied recently in [3–6].
Following the approach developed in [20], instead of a classical periodicity assumption on the structure of the perforated domainΩε, we impose certain conditions on the so-calledlocal energy characteristicsassociated with the boundary value problem (1.1). It will be shown that the asymptotic behavior, asε→0, of the solutionuεis described by the following variational problem:
inf
Ω
1
p0(x)|∇u|p0(x)+ 1
p0(x)|u|p0(x)+c(x, u)−f (x)u
dx: u∈W01,p0(·)(Ω)
, (1.4)
where the functionc(x, u)is calculated by the local energy characteristic ofΩε.
The proof of the main result is based on the variational homogenization technique which is nowadays widely used in the homogenization theory (see, e.g., [14,26,38] and the references therein). Let us also mention that another non-periodic homogenization approach was proposed recently in [28] for nonlinear monotone operators.
The paper is organized as follows. In Section 2, for the sake of completeness, we recall the definition and the main results on the Lebesgue and Sobolev spaces with variable exponents which will be used in the sequel. In Section 3 we state the problem and formulate the main result which will be proved in Section 4. Two examples of periodic and locally periodic structures are considered in Section 5.
2. Sobolev spaces with variable exponents
In this section we introduce the function spaces used throughout the paper and describe their basic properties, see for instance [19,21,27,32].
We assume thatΩ is a bounded Lipschitz domain inRnand the functionp(x)satisfies the following conditions:
1< p(−)=inf
Ω p(x)p(x)sup
Ω
p(x)=p(+)<+∞ withp(+)n. (2.1)
For allx, y∈Ω,
p(x)−p(y)ω
|x−y|
with lim
τ→0ω(τ )ln 1
τ =0. (2.2)
1. ByLp(·)(Ω)we denote the space of measurable functionsf inΩ such that Ap(·)(f )=
Ω
f (x)p(x)dx <+∞.
The spaceLp(·)(Ω)equipped with the norm fLp(·)(Ω)=inf
λ >0:Ap(·) f
λ 1
(2.3) becomes a Banach space.
2. The following inequalities hold
⎧⎪
⎨
⎪⎩ min
fpL(p(·)−)(Ω),fpL(p(·)+)(Ω)
Ap(·)(f )max
fpL(p(·)−)(Ω),fpL(p(·)+)(Ω) , min
A
p(−)1
p(·) , A
p(+)1
p(·)
fLp(·)(Ω)max A
p(−)1
p(·) , A
p(+)1
p(·)
.
(2.4)
3. Letf∈Lp(·)(Ω),g∈Lq(·)(Ω)with 1
p(x)+ 1
q(x)=1, 1< p(−)p(x)p(+)<∞, 1< q(−)q(x)q(+)<+∞. Then the Hölder’s inequality holds
Ω
|f g|dx2fLp(·)(Ω)gLq(·)(Ω). (2.5)
4. According to (2.5), for every 1q=const< p(−)p(x) <+∞
fLq(Ω)CfLp(·)(Ω) with the constantC=21
L
p(·) p(·)−q(Ω)
. (2.6)
It is straightforward to check that for domainsΩsuch that measΩ <+∞, 1Lp(·)(Ω)2 max
[measΩ]2/p(−),[measΩ]1/2p(+)
. (2.7)
5. The spaceW1,p(·)(Ω),p(·)∈ [p(−), p(+)] ⊂ ]1,+∞[, is defined by W1,p(·)(Ω)=
f ∈Lp(·)(Ω): |∇f| ∈Lp(·)(Ω) .
If condition (2.2) is satisfied,W01, p(·)(Ω)is the closure of the setC0∞(Ω)with respect to the norm ofW1, p(·)(Ω).
If the boundary ofΩ is Lipschitz-continuous andp(x)satisfies (2.2), thenC0∞(Ω)is dense inW01, p(·)(Ω). The norm in the spaceW01,p(·)is defined by
uW1,p(·)
0 =
i
DiuLp(·)(Ω)+ uLp(·)(Ω).
If the boundary ofΩ is Lipschitz andp∈C0(Ω), then the norm · W1,p(·)
0 (Ω)is equivalent to the norm uW1,p(x)
0 (Ω)=
i
DiuLp(·)(Ω). (2.8)
6. Ifp∈C0(Ω), thenW1,p(·)(Ω)is separable and reflexive.
7. Ifp, q∈C0(Ω), p∗(x)=
p(x)n
n−p(x) ifp(x) < n,
+∞ ifp(x) > n, and 1< q(x)sup
Ω
q(x) <inf
Ω p∗(x), then the embeddingW01,p(·)(Ω) →Lq(·)(Ω)is continuous and compact.
8. Friedrich’s inequality is valid in the following form: ifp(x)satisfies conditions (2.1)–(2.2), then there exists a constantC >0 such that for everyf ∈W01,p(·)(Ω)
fLp(·)(Ω)C∇fLp(·)(Ω). (2.9)
3. Statement of the problem and the main result
LetΩ be a bounded domain inRn (n2) with sufficiently smooth boundary. LetFε be a closed subset inΩ. Here εis a small parameter characterizing the scale of the microstructure. We assume thatFε is distributed in an asymptotically regular way in Ω, i.e., for any ball V (y, r)of radius r centered aty∈Ω andε >0 small enough (εε0(r)),V (y, r)∩Fε= ∅andV (y, r)∩(Ω\Fε)= ∅. We set
Ωε=Ω\Fε. (3.1)
Letpε=pε(x)be a continuous function defined in the domainΩ. We assume that, for anyε >0, it satisfies the following conditions:
(i) this function is bounded in the following sense:
1<p(−)p(ε−)≡min
x∈Ω
pε(x)pε(x)max
x∈Ω
pε(x)≡pε(+)p(+)n inΩ; (3.2) (ii) for anyx, y∈Ω, we have
pε(x)−pε(y)ωε
|x−y|
with lim
τ→0ωε(τ )ln 1
τ =0; (3.3)
(iii) the functionpεconverges uniformly inΩ to a functionp0, i.e.,
εlim→0pε−p0C0(Ω)=0, (3.4)
where the limit functionp0is assumed to be bounded in the sense of the condition (2.1) and satisfies (2.2);
(iv) the functionpεis such that
pε(x)p0(x) inΩ. (3.5)
We consider the following variational problem:
min
Jε[u]: u∈W01,pε(·) Ωε
, Jε[u] =
Ωε
1
pε(x)|∇u|pε(x)+ 1
pε(x)|u|pε(x)−f (x)u
dx, (3.6)
wheref ∈C1(Ω). It is known from [1,10,11,17] that, for eachε >0, there exists a unique solution (minimizer) uε∈W1,pε(·)(Ωε)of problem (3.6). Let us extenduεinFεby zero (keeping for it the same notation). Then we obtain the family{uε} ⊂W1,pε(·)(Ω). We study the asymptotic behavior ofuεasε→0.
Instead of the classical periodicity assumption on the microstructure of the perforated domainΩε, we impose certain conditions on the local energy characteristic of the setFε. To this end we introduceKhzan open cube centered atz∈Ω with length equal toh(0< εh <1) and we set
cε,h(z, b)=inf
vε
Khz
1
pε(x)∇vεpε(x)+h−p(+)−γS
vε−b
dx, (3.7)
whereγ >0, S
vε−b
=vε−bpε(x)+vε−bp0(x), (3.8)
and the infimum is taken overvε∈W1,pε(·)(Ω)that equal zero inFε. We assume that:
(C.1) there exists a continuous functionc(x, b)such that for anyx∈Ω, and anyb∈R, and a certainγ=γ0>0,
hlim→0lim
ε→0h−ncε,h(z, b)=lim
h→0lim
ε→0
h−ncε,h(z, b)=c(x, b); (C.2) there exists a constantAindependent ofεsuch that, for anyx∈Ω,
hlim→0lim
ε→0h−ncε,h(z, b)A
1+ |b|p0(x) .
The examples of the functionspε(x)and the domainsΩε which satisfy all the above conditions, will be given in Section 5.
The main result of the paper is the following.
Theorem 3.1. Let conditions(i)–(iv)on the functionpε and conditions(C.1)–(C.2)on the local characteristic be satisfied. Thenuεthe solution(minimizer)of the variational problem(3.6) (extended by zero inFε)converges weakly inW1,p0(·)(Ω)touthe solution(minimizer)of
inf
Ω
1
p0(x)|∇u|p0(x)+ 1
p0(x)|u|p0(x)+c(x, u)−f (x)u
dx: u∈W01,p0(·)(Ω)
. (3.9)
Remark 1. The condition (C.1) and the definition of the local energy characteristic cε,h(z, b) imply that meas[Fε∩Khx] =o(hn)for sufficiently smallε(εε(h)), uniformly with respect to˜ z∈Ω. Therefore, measFε→0 asε→0.
Notation.In what followsC,C1,C2, etc. are generic constants independent ofε.
4. Proof of Theorem 3.1
It follows from (3.6), (2.4), and the regularity properties of the functionsf,pεthat uε
W1,pε (·)(Ωε)C. (4.1)
We extenduε by zero to the setFε and consider{uε}as a sequence in the spaceW1,pε(·)(Ω). It follows from (4.1) that uε
W1,pε (·)(Ω)C. (4.2)
Condition (iv) and (4.2) immediately imply that uε
W1,p0(·)(Ω)C. (4.3)
Hence, one can extract a subsequence{uε, ε=εk→0}that converges weakly to a functionu∈W1,p0(·)(Ω). We will show thatu=u(x)is the solution of the variational problem (3.9). The proof will be done in two mains steps.
4.1. Step 1. Upper bound
Let{xα}be a periodic grid inΩ with a periodh=h−h1+γ /p(+) (εh1, 0< γ <p(+)). Let us cover the domainΩ by the cubes Khα of lengthh >0 centered at the pointsxα. We associate with this covering a partition of unity{ϕα}: 0ϕα(x)1;ϕα(x)=0 forx /∈Khα;ϕα(x)=1 forx∈Khα\
β=αKhβ;
αϕα(x)=1 forx∈Ω;
|∇ϕα(x)|Ch−1−γ /p(+).
Now letvαε=vεα(x)be a function minimizing the functional (3.7)–(3.8) withb=bαandz=xα, wherebα will be specified later. It follows from condition (C.1) that, ash→0,
εlim→0
Kαh∩Ωε
1
pε(x)∇vαεpε(x)dx=O hn
; lim
ε→0
Khα∩Ωε
S vαε−bα
dx=O
hn+p(+)+γ
. (4.4)
Moreover, condition (iv) implies that
εlim→0
Kαh∩Ωε
∇vεαp0(x)dx=O hn
ash→0. (4.5)
Denote byKhα andΠhα the cube of lengthh centered at the pointxα, and the setKhα\Khα, respectively. It follows from condition (C.1) of Theorem 3.1 that, ash→0,
εlim→0
Πhα∩Ωε
1
pε(x)∇vαεpε(x)dx=o hn
; lim
ε→0
Πhα∩Ωε
S
vαε−bα dx=o
hn+p(+)+γ
. (4.6)
Moreover, condition (iv) implies that
εlim→0
Πhα∩Ωε
∇vαεp0(x)dx=o hn
ash→0. (4.7)
Fig. 1. The setBα(ε, h;ϑ)and the functionvεα.
Fig. 2. The functionVαε.
Now letwbe a smooth function inΩ such thatw(x)=0 on∂Ω and letKθ denotes a subset of the cubesKhα coveringΩsuch that|w(x)|> θ >0 for anyx∈Khα. We set
bα=w xα
forKhα∈Kθ and bα=1 forKhα∈/Kθ. For anyKhα, we also define the set (see Fig. 1)
Bα(ε, h;ϑ )=
x∈Khα: vεα(x)signbα|bα| −ϑ
(4.8) and the function (see Fig. 2)
Vαε(x)=
vαε(x) inBα(ε, h;ϑ );
bϑα ≡(|bα| −ϑ )signbα inKhα\Bα(ε, h;ϑ ), (4.9) where 0< ϑθ/21.
Now let us estimate measBα(ε, h;ϑ ). Forεsufficiently small, from (4.5), we have ϑp(−)measBα(ε, h;ϑ )
Bα(ε,h;ϑ)∩Ωε
vαε−bαpε(x)dx
Khα∩Ωε
vεα−bαpε(x)dxChn+p(+)+γ. We setϑ=h. Then
εlim→0measBα(ε, h;ϑ )=O
hn+(p(+)−p(−))+γ
=o hn
ash→0. (4.10)
In the domainΩε we introduce the function wεh(x)=w(x)+
α
w(x) bϑα
Vαε(x)−bϑα
ϕα(x). (4.11)
From the definition of the functions{ϕα}and (4.9) we have thatwhε∈W01,pε(·)(Ωε).
Sinceuεis the solution of the variational problem (3.6) then we have Jε
uε Jε
whε
. (4.12)
Let us estimate the right-hand side of the inequality (4.12). It is clear that Jε
whε
α
Khα∩Ωε Fε
x, wεh,∇wεh
dx+
α,β
(Khα∩Khβ)∩Ωε
Fε
x, wεh,∇whεdx, (4.13)
where
Fε(x, u,∇u)= 1
pε(x)|∇u|pε(x)+ 1
pε(x)|u|pε(x)−f (x)u. (4.14)
First, we consider the second term on the right-hand side of (4.13). It follows from the definition of the partition of unity{ϕα}that for any intersectionKhα∩Khβ the number of terms in the sum overα, βis finite and does not depend onε. Then to estimate the second term on the right-hand side of (4.13) it is sufficient to consider the following integral:
jε whε
=
(Khα∩Khβ)∩Ωε
1 pε
∇
w+ w bαϑ
Vαε−bϑα
ϕα pε+ 1 pε
w+ w bϑα
Vαε−bϑα ϕα
pε
−f (x)
w+ w bϑα
Vαε−bϑα
ϕα dx
≡jε1 wεh
+jε2 whε
+jε3 wεh
. (4.15)
For the first term on the right-hand side of (4.15) we have jε1
wεh
=
(Khα∩Khβ)∩Ωε
1 pε(x)
∇
w+ w bϑα
Vαε−bϑα
ϕα pε(x)dx C1
(Khα∩Khβ)∩Ωε
|∇w|pε(x)dx+C1
(Khα∩Kβh)∩Ωε
∇w 1 bϑα
Vαε−bϑα ϕα
pε(x)dx +C1
(Khα∩Khβ)∩Ωε
1 pε(x)
w
bϑα∇vαεϕα
pε(x)dx+C1
(Khα∩Khβ)∩Ωε
w bαϑ
Vαε−bϑα
∇ϕα
pε(x)dx. (4.16)
First, it is clear that
εlim→0
(Khα∩Khβ)∩Ωε
|∇w|pε(x)dx=o hn
ash→0. (4.17)
For the second term on the right-hand side of (4.16), from (4.6), we have, ash→0,
εlim→0
(Khα∩Khβ)∩Ωε
∇w 1 bϑα
Vαε−bϑα ϕα
pεdxC2lim
ε→0
(Khα∩Khβ)∩Ωε
vεα−bαpεdx=o hn
. (4.18)
For the third term on the right-hand side of (4.16), from (4.6), we have, ash→0,
εlim→0
(Khα∩Khβ)∩Ωε
1 pε
w bϑα∇vαεϕα
pεdxC3lim
ε→0
(Kαh∩Khβ)∩Ωε
1
pε∇vεαpεdx=o hn
. (4.19)
Finally, for the fourth term on the right-hand side of (4.16), from (4.6) and the properties ofϕα, we have
εlim→0
(Khα∩Khβ)∩Ωε
w bϑα
Vαε−bϑα
∇ϕα pε(x)dx C4h−p(+)−γ lim
ε→0
(Khα∩Khβ)∩Ωε
vαε−bαpε(x)dx=o hn
ash→0. (4.20)
Thus, from (4.15)–(4.20) we get
hlim→0lim
ε→0jε1 whε
=0. (4.21)
In a similar way we can estimate the integralsjε2[whε],jε3[wεh]. Therefore, for the second term on the right-hand side of (4.13), we get
hlim→0lim
ε→0
α,β
(Khα∩Khβ)∩Ωε
Fε
x, whε,∇whεdx=0. (4.22)
Consider now the first term on the right-hand side of (4.13). First, let us denote:
Bα1(ε, h)=
Khα∩Ωε
∩Bα(ε, h;ϑ ) and Bα2(ε, h)=
Khα∩Ωε
\B1α(ε, h), (4.23)
where the setBα(ε, h;ϑ )is defined in (4.8) withϑ=h. Thenwεh(x)=w(x)inB2α(ε, h)and
Bα2(ε,h) Fε
x, whε,∇wεh dx=
Bα2(ε,h)
Fε(x, w,∇w) dx=
Bα2(ε,h)
F0(x, w,∇w) dx+IFε, (4.24) where
F0(x, w,∇w)= 1
p0(x)|∇w|p0(x)+ 1
p0(x)|w|p0(x)−f (x)w (4.25)
and IFε=
Bα2(ε,h)
Fε(x, w,∇w)−F0(x, w,∇w)
dx. (4.26)
Moreover, it follows from (3.4) that
εlim→0
IFε=0. (4.27)
Therefore, from (4.24)–(4.27) and the regularity properties of the functionsw,f, we have
εlim→0
Bα2(ε,h) Fε
x, whε,∇wεh dx
Khα
F0(x, w,∇w) dx+o hn
ash→0. (4.28)
Let us consider now the integral over the setBα1(ε, h)(Khα∈Kθ). In the setBα1(ε, h)the functionwhεhas the form:
wεh(x)=w(x)+w(x) bαϑ
vεα−bαϑ
inB1α(ε, h). (4.29)
Therefore, we have
Bα1(ε,h) Fε
x, whε,∇wεh dx=
Bα1(ε,h)
1
pε∇wεhpεdx+
B1α(ε,h)
1
pεwεhpε−whεf
dx. (4.30)
Now it follows from the regularity properties of the functionsw,f, the estimate for the measure of the setBα(ε, h;ϑ ) (see (4.10)) and the boundedness of the functionvαε on the setBα(ε, h;ϑ )that