Nonlinear Analysis 54 (2003) 9 – 37

www.elsevier.com/locate/na

Existence of solutions for nonlinear elliptic

degenerate equations
A. Benkirane∗ , J. Bennouna
Departement de Mathematique, Faculte des Sciences Dhar Mehraz, BP 1796 Atlas Fes, Morocco

Received 6 September 2001; accepted 15 September 2002

Abstract
In this paper, we study the problem
−div a(x; u; ∇u) − div (u) + g(x; u) = f in

in the setting of the weighted sobolev space W01; p ( ; ). The main novelty of our work is L∞
estimates on the solutions, and the existence of a weak and renormalized solution.
? 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Degenerate equations; Weighted Sobolev space; Weak solutions; Renormalized solutions

1. Introduction

Let be a bounded open subset of RN and let A(u)=−div a(x; u; ∇u) be a degenerate
operator of Leray–Lions type, de ned in the weighted Sobolev space W 1; p ( ; ), where
 is the weight function de ned on .
We consider the following nonlinear elliptic problem:
− div a(x; u; ∇u) − div (u) + g(x; u) = f in ; (1.1)

u = 0 on @ ; (1.2)
where g(x; t) is a Caratheodory function such that for a:e: x ∈ and all t ∈ R,
g(x; t)t ¿ 0; (1.3)

∗ Corresponding author. Fax: +212-5642500.

E-mail addresses: abdelmoujib@iam.net.ma (A. Benkirane), jbennouna@hotmail.com (J. Bennouna).

0362-546X/03/$ - see front matter ? 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0362-546X(03)00031-2
10 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

sup |g(:; t)| = hn (:) ∈ L1 ( ) ∀n; (1.4)

|t|6n

 ∈ (C 0 (R))N (1.5)
and
N

@fi
f= with fi ∈ Lr ( ; −r=p ) and r ¿ p′ : (1.6)
@xi
i=1
Note that no growth hypothesis is assumed on the function  which implies that the
term div (u) may be meaningless, even as a distribution. The notion of renormalized
solution (see De nition 4.1) gives a meaning to a possible solution of (1.1) and (1.2).
In the case where  = 0, existence of a weak solutions to (1.1) and (1.2) is proved
in the usual sense by Rakotoson and Temam [12].
The notion of renormalized solutions in the usual case (i.e. not degenerated operator)
was introduced by Diperna and Lions [5] for the study of the Boltzmann equations. This
notion was then adapted to the study of the problems (1.1) and (1.2) by Boccardo et al.
′
[3] when the right-hand side is in W −1; p ( ), by Rakotoson [11] when the right-hand
side is in L1 ( ), and nally by Dal Maso et al. [4] for the case where the right-hand
side is a general measure data.
Atik and Rakotoson studied in [1,10] degenerate problems containing the type of
degeneracy given here, in local T-set subset of W01; 1 ( ). Their equations do not contain
the term div (u). The second member f is a L1 -datum in [1], and a distribution
in [10].
It is our purpose, in this paper, to study the problems (1.1) and (1.2) in the setting
of the weighted Sobolev space W01; p ( ; ). The main novelty of the paper is the proof
of L∞ estimates on the solutions (see Theorem 3.1), the existence of a weak and
renormalized solution, and the fact that the main operator is degenerate in the space
variable. The proof of Theorem 3.1 is based on the method of relative rearrangement
performed by Rakotoson and Temam [12]. We shall distinguish two cases:

(1) In the rst case, we suppose that r ¿ rc with rc = q=(q − N ):N=(p − 1) and
q ¿ N . We prove in Theorem 3.1, that any possible solution u of (1.1) and (1.2) is
bounded, and then the term div (u) is meaningful as a distribution. Finally we prove
in Theorem 3.2 the existence of a weak solution.
(2) In the second case, that is in the case p′ 6 r 6 rc , the solution u is in general
not bounded and the term div (u) have nonmeaning even as a distribution. We prove
in Theorem 4.1 the existence of renormalized solution. Our proof is di erent from the
one given in [3].
For some results of renormalized solution, see for example [2].

2. Preliminaries

2.1 Weighted Lebesgue spaces Lp ( ; !)

Let 1 6 p ¡ ∞, and ! = !(x) be a weighted function, i.e., a function which is
measurable and positive a.e. in , Lp ( ; !) = {u; u!1=p ∈ Lp ( )}.
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 11

It is a Banach space (uniformly convex and hence re exive if p ¿ 1) equipped with

the norm
 1=p
u p; ! = |u(x)|p !(x) d x :

2.2 Weighted Sobolev spaces W k; p ( ; !)

(a) Let k ∈ N; 1 6 p 6 ∞. Let ! = {! (x); x ∈ ; | | 6 k}, be a given family of

weighted function ! ; | | 6 k.
We denote by W k; p ( ; !) the set of all functions u ∈ Lp ( ; ! ) (where  is the
zero-multiindex,  =(0; : : : ; 0)) for which the weak derivatives D u with | | 6 k belong
to Lp ( ; ! ). The weighted Sobolev space W k; p ( ; !) is a normed linear space if
equipped with the norm
 1=p


u k; p; ! =  D u pp; !  :
| |6k

Lemma 2.1 (Drabek et al. [6]): (i) Let 1 ¡ p ¡ ∞ and suppose that the weight func-
tions ! satisfy

!−1=(p−1) ∈ L1loc ( ); | | 6 k; (2.1)

then W k; p ( ; !) is a uniformly convex (and hence re exive) Banach Space.

(ii) if we suppose that also

! ∈ L1loc ( ); | | 6 k; (2.2)

then C0∞ ( ) is a subset of W k; p ( ; !), and we can introduce the space W0k; p ( ; !)
as the closure of C0∞ ( ) with respect to the norm : k; p; ! .

(b) The imbedding theorem is much more complicated than the case of nonweighted
space W k; p ( ). For simplicity, we deal with the case k=1, that is, the space W 1; p ( ; !),
with a special choice of the family ! : !0 (x) = 1; !1 (x) = · · · = !N (x) = (x).
In this case, the space W 1; p ( ; !) = W 1; p ( ; ) is normed by
  1=p
p p
u 1; p;  = |u| d x + |∇u| (x) d x : (2.3)

Let us suppose that the weight function  satis es (2.1) and (2.2) and
  
−s 1 N 1
 ∈ L ( ) with s ∈ ;∞ ∩ ;∞ ; (2.4)
P p−1

(2.4) guarantees some important imbeddings of W 1; p ( ; ) into Lebesgue space

W 1; p ( ; ) ,→,→ Lp ( ): (2.5)
12 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

(c) Let us consider the space W 1; p ( ; ) and its subspace W01; p ( ; ), the imbedding
derived for W 1; p ( ; ) hold also for W01; p ( ; ). The expression
 1=p
p
u 1; p;  = |∇u| (x) d x ; (2.6)

is a norm on the space W01; p ( ; ) equivalent with the norm (2.3).

2.1. Properties of the relative rearrangement

Let be a bounded open subset of RN , we denote by |E| the Lebesgue measure

of a set E. Let u : → R be a measurable function, we denote by u∗ the decreasing
rearrangement of u the function de ned on ∗ = ]0; | |[ by
u∗ (s) = inf{ ∈ R=|u ¿ | 6 s};
u∗ (0) = ess sup u:

We recall also that the functions u and u∗ are equimeasurable, i.e.

|u ¿ s| = |u∗ ¿ s|
|u ¿ s| = |u∗ ¿ s| ∀s ∈ R;
which implies that for any Borel function f; f ¿ 0,
  | |
f(u(x)) d x = f(u∗ (s)) ds:
0
p
Let now v ∈ L ( ) and u a measurable function, we associate to them the function !
de ned on ]0; | |[ by



 v(x) d x if |u = u∗ (s)| = 0;
 u¿u∗ (s)
!(s) =   s−|u¿u∗ (s)|

v(x) d x + [v|P(s) ]∗ () d if not;



u¿u∗ (s) 0

where [v|P(s) ]∗ is the decreasing rearrangement of the restriction of v to P(s) =

{u = u∗ (s)}.

De nition 2.1. We recall that the relative rearrangement of v with respect to u, the
function d!=ds denoted by
d!
v∗u = :
ds

Theorem 2.1 (Mossino and Temam [9], see Theorem 0 of Rakotoson [10]): Let v ∈
Lp ( ), 16p6∞, u a measurable function from to R, then
∗ ∗
(i) ! ∈ W 1; p ( ) with = ]0; | |[,
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 13

(ii) v∗u Lp ( ∗ ) = d!
ds L ( ) 6 v L ( ) ,
p ∗ p

(iii) if v ¿ 0 a.e. in , then v∗u ¿ 0 a.e. in .

Theorem 2.2 (Rakotoson and Temam [12, Theorem 1.1]). Let be a bounded open
set of RN and let us assume that u ∈ W01; 1 ( ), u ¿ 0 and v ∈ L1 ( ). Then, for almost
every t ∈ R

d
v(x) d x = v∗u ((t)):′ (t);
dt u¿t

where  is the distribution function of u, (t) = |u ¿ t|.

Theorem 2.3 (Rakotoson and Temam [12], see (1.15)). If f1 ¿ 0 is locally integrable
function on ∗ , then for every s; s′ ; 0 6 s 6 s′ 6 | |,
 u∗ (s)  s′
′
f1 (())(− ()) d 6 f1 () d:
u∗ (s′ ) s

3. Regularity and existence of weak solutions

Let be a bounded open subset of RN . A is a nonlinear operator of the Leray–

Lions type from a weighted Sobolev space W 1; p ( ; ) (where =(x) is weight function
de ned on ). A is de ned by A(u)=−div a(x; u; ∇u) where a is Caratheodory function
from × R × RN → RN satisfying for a.e. x ∈ , and for all 0 ; ,

|a(x; 0 ; )| 6 K(|0 |)1=p (x){d(x) + |0 |p−1 + 1−1=p ||p−1 }; (3.1)
′
where K veri es sup ess|s|6k K(|s|) ¡ ∞, ∀k ¿ 0, and d(x) ∈ Lp ( ).

[a(x; 0 ; ) − a(x; 0 ; )][ − ] ¿ 0 where  = : (3.2)

Let  be a continuous function such that  ¿ 0 and de ned on R+ , the degeneracy of

the operator A is expressed by the assumption

a(x; 0 ; ) ¿ (x)(|0 |)||p (3.3)

for all 0 ;  and let 1 such that

 s
′
1 (s) = [(t)]p =p dt; s ¿ 0 with 1 (∞) = ∞: (3.4)
0

Furthermore, we shall assume that for some m0 ¿ 0,

 
p:(1−p′ =r)
inf m inf (t) ¿ 0: (3.5)
m¿m0 m6t6m+1

p=p′
For instance (t) = t − for t ¿ 1, 0 ¡ ¡ 1 (small).
14 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

We will suppose that

 ∈ L1loc ( ); −1=(p−1) ∈ L1 ( ); (3.6)
  
−s 1 N 1
 ∈L ( ) with s ∈ ;∞ ∩ ;∞ : (3.7)
p p−1
We consider the nonlinear elliptic problem
− div a(x; u; ∇u) − div (u) + g(x; u) = f; (3.8)

u = 0 on @ ; (3.9)
where  = (1 ; : : : ; N ) satisfy
 ∈ (C 0 (R))N : (3.10)
Note that no growth hypothesis is assumed on the function . Let g(x; t) be a
Caratheodory function such that for a:e: x ∈ and all t ∈ R,
g(x; t) t ¿ 0; (3.11)

sup |g(:; t)| = hn (:) ∈ L1 ( ) ∀n: (3.12)

|t|6n

The right-hand side in (3.8) is of the form

N

@fi
f= ; (3.13)
@xi
i=1

where the family {fi i = 1; : : : ; N } satis es the conditions

′ ′
fi ∈ Lr ( ; −r=p ) ,→ Lp ( ; −p =p ) = (Lp ( ; ))∗ for i = 1; : : : ; N

with r ¿ p′ :
We distinguish two cases.
The rst case if r ¿ rc with rc = q=(q − N ):N=(p − 1) and q ¿ N . In this case we
prove in Theorem 3.1 that the solution u of problems (3.8) and (3.9) is bounded, and
then the term div (u) is meaningful as a distribution. In Theorem 3.2, we prove that
the last problem admits a weak solution (see De nition 3.1).
The second case if p′ 6 r 6 rc the solution u is not bounded and the term div (u)
may be meaningless, even as a distribution. As in [3] we de ne the notion of renor-
malized solution (see De nition 4.1), which gives a meaning to a possible solution of
(3.8) and (3.9). In Theorem 4.1 we prove the existence of renormalized solution.

De nition 3.1. Assume that (3.1)– (3.7), (3.10) – (3.13) hold true. A function u is a
weak solution of the problem (3.8) and (3.9) if
u ∈ W01; p ( ; ) ∩ L∞ ( ); g(x; u) ∈ L1 ( ); u g(x; u) ∈ L1 ( ); (3.14)
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 15

  
a(x; u; ∇u)∇’ d x + (u)∇’ d x + g(x; u)’ d x

N 

@’
=− fi d x ∀’ ∈ D( ): (3.15)
@xi
i=1

In Theorem 3.1 we prove the boundedness of the solution of problems (3.8) and (3.9),
in Theorem 3.2 we prove the existence of weak solution if r ¿ rc .

Lemma 3.1. Let be a bounded open subset of RN . If u ∈ (W01; p ( ; ))N , then


div u d x = 0:

Proof of Lemma 3.1. Let u = (u1 ; : : : ; uN ) ∈ (W01; p ( ; ))N . Since D( ) = W01; p ( ; )

we have for each ui , there exist a sequence (uki ) ∈ D( ) such that uki → ui for the
norm : 1; p;  .
Since uki ∈ D( ) ⊂ D(  ), Green’s Theorem gives
@uki
 
dx = uki :n ds = 0:
@xi @

Let ps such that 1 6 ps = p:s=(s + 1) ¡ p and s de ned for s ¿ N=p we have

  i p   i p
 @uk @ui  s  @uk @ui  s ps =p −ps =p

 @xi − d x =  −   dx
@xi   @xi @xi 
  i p ps =p  1=(s+1)
 @uk @ui  −s
6 
 @xi − (x) d x  d x
@xi 
 i ps
 @uk @ui 
6
 −  :C by (3:6):
@xi @xi 1; p; 

Hence @uki =@xi → @ui =@xi for the norm of Lps ( ) and L1 ( ) so that
N N
@ui @ui



k
0= dx → dx
@xi @xi
i=1 i=1

giving that div u d x = 0.

Lemma 3.2. Let F : R → R be uniformly lipschitzian with F(0)=0. Let u ∈ W01; p ( ; ),
then F(u) ∈ W01; p ( ; ).

For the proof of Lemma 3.2 it suces to approximate u by smooth functions as for
the proof of Lemma 3.1.
16 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

Let u ∈ W01; p ( ; ) and let S; h a real Lipschitzian function de ned for  ¿ 0, h ¿ 0 by


 1 if  ¿  + h;



 −
if  6  6  + h;


 h



S; h () = 0 if || 6 ;



 +

 if −  − h 6  6 − ;
h





−1 if  6 −  − h;


we have S; h (u) ∈ W01; p ( ; ) by Lemma 3.2 and we suppose that u satis es ∀ ∈ ]0;
ess sup |u|[; ∀h ∈ ]0; ess sup |u| − [,
  
a(x; u; ∇u)∇S; h (u) d x + (u)∇S; h (u) + g(x; u)S; h (u) d x

N 

@S; h (u)
=− fi d x: (3.16)
@xi
i=1

Remark 3.1. It is easy to see that each possible solution of problems (3.8) and (3.9)
veri es relation (3.16).

Theorem 3.1. Under assumptions (3.1)– (3.7),(3.10) – (3.13). Let u ∈ W01; p ( ; ) which
satis es (3.16), and we suppose furthermore fi ∈ Lr ( ; −r=p ), with r ¿ rc where rc =
q=(q−N ):N=(p−1) and q ¿ N . Then u is bounded and we have the following estimate

1 ( u ∞ ) 6 CN 1 2 3 ⇔ u ∞ 6 1−1 (CN 1 2 3) = M; (3.17)

where

1 = −1=p Lq ( ) ;

 N 1=(p−1)


2 = fi Lr ( ; −r=p ) ;
i=1
 1=m
| |
m(1−1=N ) 1 1 1
3 =  d ; with + + = 1;
0 m r(p − 1) q

CN = (1=N N1=N ), N is the measure of the unit ball of RN , 1−1 is the inverse of the
function 1 (see (3.4)).

To be able to give a proof of the last Theorem, we need the following Lemma.
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 17

Lemma 3.3. Under assumptions of Theorem 3.1 we have

′ dv∗ 1=p′
−(v∗ (t))p =p (t) 6 CN (−1=(p−1) )∗v (t)(F∗v )1=p (t)t 1=N −1 ;
dt
N ′ ′
where v = |u|, F = ( i=1 |fi |2 )p =2 −p =p , v∗ is the decreasing rearrangement of v, b∗v
is the relative rearrangement of b with respect to v.

Proof of Lemma 3.3. We consider the function S; h , we have by Lemma 3.2 S; h (u) ∈
W01; p ( ; ) ∩ L∞ ( ) and a.e. x ∈

@  1 @u if  ¡ |u| 6  + h;

S; h (u) = h @xi
@xi 
0 if not:

Then relation (3.16) can be written with v = |u| as

  
1 1
a(x; u; ∇u)∇u d x + (u)∇u d x + g(x; u)S; h (u) d x
h {¡v6+h} h {¡v6+h}
N 
1
@u
=− fi d x:
h {¡v6+h} @xi
i=1

We shall verify that


{¡v6+h} (u)(u)∇u d x = 0:

t
Indeed consider 1 (t) = (t){¡|t|6+h} (t), ˜1 (t) = 0 1 () d, we have by Lemma
3.2 ˜1 (u) ∈ (W01; p ( ; ))N . Then we obtain by Lemma 3.1
  
(u){¡v6+h} (u)∇u d x = (u)∇u d x = ˜
div((u)) d x = 0:

On the other hand since S; h (0) = 0 we have

 
S; h (u)
g(x; u)S; h (u) d x = ug(x; u) d x;
u=0 u

since S; h is increasing, S; h (u)=u ¿ 0 and by (3.11) we have g(x; u)S; h (u) d x ¿ 0.
Then (3.3) gives that
  N  
1 1
 @u 
(x)(|u|)|∇u|p d x 6  @xi  d x:
 fi 
h {¡v6+h} h {¡v6+h}
i=1

Since  is bounded on [;  + h] we have

 
p
min () (x)|∇u| d x 6 (x)(|u|)|∇u|p d x:
¡6+h {¡v6+h} {¡v6+h}
18 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

By the Holder inequality we have

 N  

 @u 
 @xi  d x
 fi 
{¡v6+h} i=1

  N p′ =2 1=p′

′
6 |fi |2 :−p =p d x
{¡v6+h} i=1

 1=p
p
× (x)|∇u| d x :
{¡v6+h}
N ′ ′
Let F = ( i=1 |fi |2 )p =2 −p =p , we have F ∈ L1 ( ), and
 N    1=p′

 @u 
 fi  dx 6 F(x) d x
 @xi 
{¡v6+h} i=1 {¡v6+h}

 1=p
p
× (x)|∇u| d x :
{¡v6+h}

When h → 0 we have ∀ ∈ ]0; ess sup v[,

     1=p′
d d
() − (x)|∇u|p d x 6 − F(x) d x
d {v¿} d {v¿}
  1=p
d p
× − (x)|∇u| d x :
d {v¿}

We simplify and we use Theorem 2.3 (since u ∈ W01; p ( ; ) ,→ W01; 1 ( ), and F ∈L1 ( )),
to obtain
  1=p′
d ′ ′
() − (x)|∇u|p d x 6 (−′ ())1=p (F∗v (()))1=p
d {v¿}
then
  1=p
′ d
()p =p − (x)|∇u|p d x 6 (−′ ())1=p (F∗v (()))1=p : (3.18)
d {v¿}

On the other hand we have

   1=p
d d
− |∇u| d x 6 − |∇u|p d x
d {v¿} d {v¿}
  1=p′
d −1=(p−1)
× −  dx : (3.19)
d {v¿}
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 19

We use Theorem 2.3 since −1=(p−1) ∈ L1 ( ) (see (3.6)) then


d
− −1=(p−1) d x = (−1=(p−1) )∗v (())(−′ ()):
d {v¿}
Using the Fleming–Rishel formula (see [8]) we can write

d
− |∇u| d x ¿ N N1=N (())1−1=N
d {v¿}
the inequalities (3.18) and (3.19) give with the last inequality
1=N 1−1=N ′
N N (()) ()p =p
′ ′
6 (−′ ())1=p (F∗v (()))1=p (−′ ())1=p (−1=(p−1) )∗v
1=p
(()):
1=N
Since  =v∗ (()), the last inequality gives for all  ∈ ]0, ess sup v[, with CN =1=N N
1
1 6 CN (())(1=N )−1 (F∗v (()))1=p
(v∗ (()))p′ =p
′
×(−1=(p−1) )∗v
1=p
(())(−′ ()):
′
1=p 1=p
Since  ¿ 0 is continuous and the function F∗v and (−1=(p−1) )∗v are integrable (see
(ii) of Theorem 2.2) we deduce that the mapping
1
 ∈ ∗ = ]0; | |[ → (1=N )−1 :(F∗v ())1=p (−1=(p−1) )1=p∗v ()
(v∗ ())p′ =p
is locally integrable.
We apply Theorem 2.4 to obtain that, for 0 ¡ t ¡ t +  ¡ | |
v∗ (t) − v∗ (t + )

CN t+ 1=N −1

1 1=p −1=(p−1) 1=p′
6  p ′ =p (F∗v ()) ( )∗v () d:
 t (v∗ ())
When  tends to zero, we obtain for almost every t ∈ ∗
′ dv∗ 1=p′
−(v∗ (t))p =p (t) 6 CN :t 1=N −1 (−1=(p−1) )∗v (t):(F∗v )1=p (t):
dt
Proof of Theorem 3.1. By the de nition of 1 (see (3.4)) we have 1 (v∗ (t)) =
 v∗ (t) ′ 1; p 1; p
0
[(s)]p =p ds, since v∗ ∈ Wloc ( ∗ ), it seems 1 (v∗ (t)) ∈ Wloc ( ∗ ) and we have
a.e. t
d ′ dv∗
1 (v∗ (t)) = (v∗ (t))p =p :
dt dt
Then since v∗ (| |) = ess inf v = 0 and the function 1 is continuous and increasing
from R+ to R+ thus invertible, we have by Lemma 3.3
  
| |
′
v∗ (t) 6 1−1 CN (−1=(p−1) )∗v
1=p
(t)(F∗v )1=p (t)t 1=N −1 dt :
t
−1=p
Since  ∈ L ( ), for q ¿ N see (3.7), and fi ∈ Lr ( ; −r=p ) for r ¿ N=(p−1):q=(q−
q

N ), we apply Holder inequality for the second member of the right-hand side of the
20 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

last inequality with 1=m + 1=r(p − 1) + 1=q = 1, and we use (ii) in Theorem 2.2 to
obtain
 | |
1=p′
(−1=(p−1) )∗v (t)(F∗v )1=p (t)t 1=N −1 dt
t

1=q 

N
1=(p−1)  1=m
 | |
−q=p m(1−1=N )
6  fi Lr ( ; −r=p ) t dt :
i=1 0

The last integral is de ned since r ¿ N=(p − 1):q=(q − N ) then the two last inequalities
give as ∀t ∈  ∗
v∗ (t) 6 1−1 (CN 1 2 3) = M;
i.e. u ∞ 6 M , (since v∗ (0) = sup essv ).

Theorem 3.2. Under assumptions (3.1)– (3.7), (3.10) – (3.13). There exist a bounded
weak solution u ∈ W01; p ( ; )∩L∞ ( ) (in the sense of De nition 3.1) of problem (3.8)
and (3.9).

Proof of Theorem 3.2. Step 1: In order to avoid the diculties related to the possible
growth of a(x; 0 ; ) as |0 | → ∞, we introduce a new coecients b(x; 0 ; ) de ned
for, a.e. x ∈ , ∀0 ; , by the formula


 a(x; −M; ) if 0 ¡ − M;

b(x; 0 ; ) = a(x; 0 ; ) if |0 | 6 M; (3.20)


a(x; M; ) if 0 ¿ M;


where M is the constant given by Theorem 3.1.

The function b veri es the following properties, b is Caratheodory function and for
a.e. x ∈ , ∀0 ; , we have
|b(x; 0 ; )| 6 C1=p (x){d(x) + |0 |p−1 + 1−1=p ||p−1 };
where C = C(M ), it is a consequence of (3.1).
We de ne the truncation of  by
  |s|
′
 (|s|) if |s| 6 M and ′ (|s|) =

[′ (t)]p =p dt:
1
′ (|s|) = 0

(M ) if |s| ¿ M:


And for a.e. x ∈ , ∀0 ∈ R; ;  ∈ RN , we have

[b(x; 0 ; ) − b(x; 0 ; )][ − ] ¿ 0; where  = :
We deduce from (3.3) that for a.e. x ∈ , ∀0 ∈ R,  ∈ RN , we have
b(x; 0 ; ) ¿ (x)′ (|0 |)||p :
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 21

We introduce the following problem

u ∈ W01; p ( ; ) ∩ L∞ ( )




−div b(x; u; ∇u) − div (u) + g(x; u) = f in D′ ( ); (3.21)


u = 0 on @ :


We claim that any solution of (3.21) is a solution of (3.8) and (3.9) and conversely.
Indeed let u be a solution of (3.21). As in Theorem 3.1 one can check, replacing
1 by 1′ that
1′ ( u ∞ ) 6 CN 1 2 3 = 1 (M ): (3.22)
By the de nition of 1′
we remark that 1 (M ) = 1′ (M ), and since 1′ is invertible we
deduce from (3.22) that
u ∞ 6 M:
Then b(x; u; ∇u) = a(x; u; ∇u) for almost every x ∈ . We deduce that u is a solution
of (3.8) and (3.9).
The proof of the converse is similar. Which gives the end of the claim.
Step 2: Let us de ne for each k ¿ 0, the truncation
s
Tk (s) = s if |s| 6 k; Tk (s) = k if |s| ¿ k
|s|
and for each  ¿ 0, the approximations
 (s) = (T1= (s)); g (x; t) = T1= (g(x; t)):
Consider the nonlinear elliptic problem
u ∈ W01; p ( ; ) ∩ L∞ ( )

(3.23)
−div b(x; u ; ∇u ) − div  (u ) + g (x; u ) = f in :
By b̃(x; t; ) = b(x; t; ) +  (t), (3.23) is equivalent to
u ∈ W01; p ( ; ) ∩ L∞ ( )

(3.24)
−div b̃(x; u ; ∇u ) + g (x; u ) = f in :
The existence of u ∈ W01; p ( ; ) is proved by applying the Theorem 2.8 (see [7]).
Then for all , there exists at least one solution u of (3.24) and of it’s equivalent
(3.23). Since each solution u of the problem (3.23) veri es Eq. (3.16), then we obtain
by applying Theorem 3.1,
u ∞ 6 M:
Using in (3.23) the test function u , we have
  
a(x; u ; ∇u )∇u d x +  (u )∇u d x + g (x; u )u d x

N 

@u
=− fi d x: (3.25)
@xi
i=1
22 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37


Since  (u )∇u = 0, and g (x; u )u ¿ 0, we get
 N 

@u
a(x; u ; ∇u )∇u d x 6 − fi d x:
@xi
i=1

Let 1 = min 060 6M (|0 |), by the degeneracy assumption (3.3) and (3.13) we have
 N   
p

 @u  1=p −1=p
1 (x) |∇u | d x 6  fi
 @xi   
 dx
i=1

 N  1=p′  1=p
p


p′ −p′ =p
6 |fi |  dx |∇u |  d x
i=1
 N  1=r


r −r=p
6 C u 1; p;  |fi |  dx
i=1

which implies that

p−1
u 1; p;  6 C1 :
Passing to a subsequence if necessary, we can assume that
u * u weakly in W01; p ( ; ) and a:e: in

(3.26)
u * u weakly-star in L∞ ( ):
Since u ∞ 6 M , we can use in the next a(x; 0 ; ) in place of b(x; 0 ; ).
On the other hand, we have by (3.25)

g (x; u )u d x 6 C2 :

Since g (x; u ) → g(x; u) a.e. in , by the sign condition (3.11) and Fatou’s Theorem
we obtain
g(x; u)u ∈ L1 ( ): (3.27)
Then by Vitali’s Theorem we have
g (x; u ) → g(x; u) strongly in L1 ( ): (3.28)
Step 3: De ne, for each k ¿ 0, the set Ek
= {x ∈ : |u (x)| ¿ k}.
Using in (3.23) the test function u − Tk (u ), we obtain by an easy computation
N 
@(u − Tk (u))

p


lim sup |∇u | (x) d x 6 − fi d x: (3.29)
→0 
Ek @xi
i=1

Step 4: De ne for i; j ¿ 0, Fij

= {x ∈ : |u (x) − Tj u(x)| 6 i}. Using in (3.23) the
test function w = Ti (u − Tj (u)) ∈ W01; p ( ; ), since
∇(u − Tj (u)) on Fij ;

∇w = (3.30)
0 on \ Fij :
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 23

We obtain, as in [3]

lim a(x; u ; ∇u )∇(u − Tj (u))
→0 Fij

N 
@Ti (u − Tj (u))


=− fi − g(x; u)Ti (u − Tj (u)) d x: (3.31)
@xi
i=1

Step 5: We prove in this step that


[a(x; u ; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] d x → 0 as  → 0: (3.32)

Let us write (3.32) in the form


[a(x; u ; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] d x = Iij + IIij ;

where

Iij = [a(x; u; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] d x;
Fij

IIij = [a(x; u ; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] d x:
\Fij

Note that
  

 I
 ij
 = a(x; u ; ∇u  )[∇u  − ∇T j (u)] d x − a(x; u ; ∇u)[∇u − ∇Tj (u)]
 Fij Fij
 (3.33)



 + [a(x; u  ; ∇u  ) − a(x; u ; ∇u)][∇T j (u) − ∇u] d x:
Fij

The rst term of the right-hand side of (3.33) is bounded by a function of j which
goes to zero as j tends to ∞ (see (3.31)). Indeed
N 
 N  1=r

@Ti (u − Tj (u))

r −r=p
− fi dx 6 |fi |  dx u − Tj (u) 1; p; 
@xi
i=1 i=1

and
  
 
 g(x; u)Ti (u − Tj (u)) d x 6 g(x; u)u d x; since g(x; u):u ¿ 0:
 
{|u|¿j}

The second term of the right-hand side of (3.33) is equal by (3.30) to a(x; u ; ∇u)
∇w d x, and we claim that
 
lim a(x; u ; ∇u)∇w d x = a(x; u; ∇u)∇(Ti (u − Tj (u))) d x: (3.34)
→0

Indeed since w remains in a bounded set of W01; p ( ; ) and w (x) → w(x) = Ti (u −

Tj (u)) a.e. in . Thus, w converges weakly to w in W01; p ( ; ), ∇w * ∇w in
24 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

(Lp ( ; ))N -weak. By the Growth condition and the fact that u (x) → u(x) a.e., we
′
have a(x; u ; ∇u) → a(x; u; ∇u) a.e. and strongly in (Lp ( ; −1=(p−1) ))N . So we deduce
 
a(x; u ; ∇u)∇w d x → a(x; u; ∇u):∇w d x:

Relation (3.34) is bounded by a function of j which goes to zero as j tends to ∞,

which allows to estimate the second term of (3.33). Indeed
 
 a(x; u; ∇u)∇(Ti (u − Tj (u))) d x 6 [ d p′ + u p−1 ] u − Tj (u) 1; p; 
 
  1; p; 

6 C u − Tj (u) 1; p;  :

For the third term of the right-hand side of (3.33), we have

 
 
[a(x; u ; ∇u ) − a(x; u ; ∇u)][∇Tj (u) − ∇u] d x
 

 Fij 

p−1 p−1 p−1

6 [2 d p′ + 2 u p + u 1; p;  + u 1; p;  ]: u − Tj (u) 1; p; 

6 C: u − Tj (u) 1; p;  :

Which allows us to estimate the third term of the right-hand side of (3.33), and gives
nally

lim sup Iij 6 F(j) where F(j) → 0 as j → ∞:

→0

Estimate now IIij .

For each i ¿ j ¿ 0 we have \ Fij ⊆ Ei−j 
, and by young’s Inequality, we have also

0 6 IIij 6 [a(x; u ; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] d x

Ei−j

6 1=p {2d(x) + 2|u |p−1 + 1−1=p |∇u |p−1

Ei−j

+ 1−1=p |∇u|p−1 }|∇u − ∇u| d x

  
p p p p′
6 K4 (x)|∇u | d x + [(x)|∇u| + |u | + |d| ] d x : (3.35)

Ei−j 
Ei−j

By estimate (3.29) the rst term of the right-hand side of (3.35) is small provided
i − j is suciently large.
On the other hand, since lim sup→0 Ei−j  (x)|∇u|p 6 {|u|¿i−j} (x)|∇u|p a.e. in ,
p p
and lim sup→0 Ei−j
 |u | 6 {|u|¿i−j} |u| a.e. in then by Fatou’s Theorem we have
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 25

for the second term of the right-hand side of (3.35)


′
lim sup [(x)|∇u|p + |u |p + |d|p ] d x
→0 
Ei−j

′
6 [(x)|∇u|p + |u|p + |d|p ] d x
{|u|¿i−j}

which is small if i − j is suciently large. Thus IIij is small if i − j is large.

Since Iij and IIij is small if j, i − j is large we deduce that (3.32) holds true.
Step 6: We claim that
u → u strongly in W01; p ( ; ): (3.36)
Relation (3.32) gives that for a subsequence
[a(x; u ; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] → 0 a:e: in : (3.37)
Then arguing as in Leray–Lions, we have for a subsequence
∇u (x) → ∇u(x) a:e: in : (3.38)
On the other hand

(x)|∇u |p d x

6 a(x; u ; ∇u )∇u

6 [a(x; u ; ∇u ) − a(x; u ; ∇u)][∇u − ∇u] d x
 
+ a(x; u ; ∇u)[∇u − ∇u] d x + a(x; u ; ∇u )∇u d x: (3.39)

The rst integral of (3.39) goes to 0 by (3.32).

For the second integral of (3.39), we have ∇u − ∇u * 0 in (Lp ( ; ))N -weak and
′
a(x; u ; ∇u) converges strongly to a(x; u; ∇u) in (Lp ( ; −1=(p−1) ))N . So we deduce

a(x; u ; ∇u)[∇u − ∇u] d x → 0:

For the third integral of (3.39), we have ∇u ∈ (Lp ( ; ))N and a(x; u ; ∇u ) converges
′
weakly to a(x; u; ∇u) in (Lp ( ; −1=(p−1) ))N . So we deduce
 
a(x; u ; ∇u )∇u d x → a(x; u; ∇u)∇u:

We then have
 
p
lim sup (x)|∇u | d x 6 a(x; u; ∇u)∇u d x:
→0
26 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

Consequently, since |∇u|p is integrable over


lim sup (x)|∇u − ∇u|p d x = 0;
→0

hence (3.36).
Step 7: Using in (3.23) the test function ’ ∈ D( ), we obtain (3.15) after passing
to the limit. Theorem 3.2 is then proved.

4. Existence of renormalized solutions

If p′ 6 r 6 rc , the solution of problems (3.8) and (3.9) may not be bounded and
the term div((u)) may have no meaning even a distribution. Then we consider the
following notion of renormalized solution which gives a meaning to a possible solution
of (3.8) and (3.9).

De nition 4.1. Assume that (3.1)–(3.7), (3.10)–(3.13) hold true. A function u is a

renormalized solution of the problem (3.8) and (3.9) if

u ∈ W01; p ( ; ); g(x; u) ∈ L1 ( ); ug(x; u) ∈ L1 ( ) (4.1)

 
′



 a(x; u; ∇u)[h (u)∇u’ + h(u)∇’] d x + (u)[h′ (u)∇u’ + h(u)∇’] d x



N 
 
@h(u)’ (4.2)
+ g(x; u)h(u)’ d x = − fi d x:∀h ∈ Cc1 (R);
@x

i

i=1




1; p
 ∞
’ ∈ W0 ( ; ) ∩ L ( ):


In Theorem 4.1 we prove the existence of renormalized solution of (3.8) and (3.9).

Theorem 4.1. Under assumptions (3.1)– (3.7), (3.10) – (3.13), and furthermore for
p′ 6 r 6 rc . Then there exists a renormalized solution u (in the sense of De nition
4.1) of problems (3.8) and (3.9).

Proof of Theorem 4.1. Step 1: Let  ¿ 0, and let fi = T1= (fi ) the truncation of the
function fi , we have

fi → fi in Lr ( ; −r=p ) when  → 0:

We consider the nonlinear elliptic problem

u ∈ W01; p ( ; )




N (4.3)

@f i
 −div a(x; u ; ∇u ) − div  (u ) + g (x; u ) = f = in :


@xi
i=1
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 27

′ ′
We have by Theorem 3.1 (with fi = fi and since fi ∈ Lr ( ; −r =p ), ∀r ′ ¿ rc ) each
solution u veri es
  N 1=(p−1) 


u ∞ 6 M () since 2 = fi Lr ( ; −r=p ) :
i=1

We apply Theorem 3.2 and we prove that the problem (4.3) has at least one solution
u . We have for all k ¿ 0, with the notation Tk (u ) = uk ,
uk 1; p;  6 C2 :
Passing to a subsequence if necessary, we can assume that
Tk (u ) = uk * Tk (u) = uk weakly in W01; p ( ; ) and a:e: in : (4.4)
Which implies that
u → u a:e: in :
Step 2:
′
Lemma 4.1. Let m = inf m6t6m+1 (t). Assume that 0 = inf m¿m0 (mp:(1−p =r) :m ) ¿ 0
for some m0 ¿ 0. Then there exists a constant c ¿ 0 depending only on xed data
(p; r; ), such that for all m ¿ m0 , all  ¿ 0
   N  p′ =r


|∇u |p d x 6 C |fi |r −r=p d x : (4.5)
{m6|u |6m+1} {m6|u |6m+1} i=1

Proof of Lemma 4.1. Let us consider as a test function ’ = Tm+1 (u ) − Tm (u ).
Since the map  → Tm+1 () − Tm () is nondecreasing, the sign condition on g
implies that:

[Tm+1 (u ) − Tm (u )]g (x; u ) d x ¿ 0:

Furthermore

[Tm+1 (u ) − Tm (u )] div  (u ) d x = 0:

From these properties of ’ , the relation (4.3) give us

 
inf (t) |∇u |p d x 6 f ∇u d x: (4.6)
m6t6m+1 {m6|u |6m+1} {m6|u |6m+1}

From Holder inequality, we have




 f ∇u d x


 {m6|u |6m+1}

  1=r



1−p′ =r −r=p
6 mes{x : m 6 |u | 6 m + 1} |f | dx (4.7)

 {m6|u |6m+1}

  



 p


 |∇u |  d x :
{m6|u |6m+1}
28 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

From the two last relations (4.6) and (4.7), we derive, from m ¿ m0

|∇u |p d x
{m6|u |6m+1}

 ′
p ′
mes{x : m 6 |u | 6 m + 1}1−p =r
6
m
 p′ =r
r −r=p
: |f | : dx : (4.8)
{m6|u |6m+1}

By the Chebychev inequality and the Sobolev imbedding (2.5), we have


1 c
mes{x : m 6 |u | 6 m + 1} 6 p |u (x)|p d x 6 p : (4.9)
m m

Thus, from relation (4.9), one has

′
mes{x : m 6 |u | 6 m + 1}1−p =r c
6 ¡ + ∞: (4.10)
m 0
From relations (4.8) and (4.10), we then deduce
   N  p′ =r


|∇u |p d x 6 C |fi |r −r=p d x :
{m6|u |6m+1} {m6|u |6m+1} i=1

Step 3: We prove that for all k ¿ 0


[a(x; uk ; ∇uk ) − a(x; uk ; ∇uk )][∇uk − ∇uk ] d x → 0 as  → 0: (4.11)

Indeed, using in (4.3) the test function Hm (u )(uk − uk ), k ¿ 0 xed, and Hm the
lipschitzian function associated to a real number m ¿ 0, such that


 1 if || 6 m;


0 if || ¿ m + 1;

Hm () =


 − + m + 1 if m 6  6 m + 1;


+m+1 if − m − 1 6  6 − m:


We obtain
 
Hm′ (u )(uk − uk )a(x; u ; ∇u )∇u d x + Hm (u )a(x; u ; ∇u )∇(uk − uk ) d x
 
+ Hm′ (u )(uk − uk ) (u )∇u + Hm (u ) (u )∇(uk − uk )
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 29

 N 

@u
+ Hm (u )(uk − uk ):g (x; u ) d x = − Hm′ (u )fi (uk − uk )
@xi
i=1

N 

@(uk − uk )
− Hm (u )fi : (4.12)
@xi
i=1

Fix m ¿ k and let  goes to 0, we obtain the rst integral of the right-hand side of
(4.12) tends to 0. Indeed
@um+1
 
′ k @u
 k
Hm (u )fi (u − u ) dx = Hm′ (u )fi (uk − uk )  d x:
@xi {m¡|u |¡m+1} @xi

Since for all k ¿ 0, we have uk → uk weakly in W01; p ( ; ) and a.e. in , then
@um+1 @um+1
* weakly in Lp ( ; ):
@xi @xi
Since Hm′ (u )fi (uk − uk ) → 0 a.e. in and
 1=p′  1=p′
−p′ =p ′  k k p′ −p′ =p  p′
 |Hm (u )fi (u − u )| d x 6 2k  |fi | d x
E E
 1=r
−r=p r
6 2Ck  |fi | d x ;
E

for all measurable set E ⊂ . Then the equiintegrability gives

′ ′
Hm′ (u )fi (uk − uk ) → 0 strongly in Lp ( ; −p =p ):
The second integral of the right-hand side of (4.12) tends to 0. Indeed @(uk −uk )=@xi * 0
′ ′
weakly in Lp ( ; ), Hm (u )fi → Hm (u)fi strongly in Lp ( ; −p =p ) since itself con-
verges to Hm (u)fi a.e. in and
 1=p′  1=p
−p′ =p  p′ −p′ =p  p′
 |Hm (u )fi | d x 6  |fi | d x
E E
 1=r
−r=p r
6C  |fi | d x
E

which gives the equiintegrabilite of Hm (u )fi .

The last integral of the left-hand side of (4.12) tends to 0. Indeed Hm (u )(uk −
k
u )g (x; u ) → 0 a.e. in and
 
k k
|Hm (u )(u − u )g (x; u )| d x 6 |Hm (u )(uk − uk )g(x; um+1 )| d x
E E∩{|u |6m+1}

6 2k sup |g(x; t)| dt
E |t|¡m+1

by (3.12) we have the equiintegrability of Hm (u )(uk − uk )g (x; u ).

30 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

The fourth integral of the left-hand side of (4.12) tends to 0. Indeed ∇(uk − uk ) * 0
weakly in (Lp ( ; ))N , Hm (u ) (u ) → Hm (u)(u) a.e. in and
 
′ ′ ′ ′
|Hm (u ) (u )|p −p =p d x = |Hm (u ) (u )|p −p =p d x:
E E∩{|u |6m+1}

On the other hand, | (u )| 6 Cm by the continuity of  and since |u | 6 m + 1 then
 1=p′  1=p′
p′ −p′ =p −p′ =p
|Hm (u ) (u )|  dx 6 Cm  dx :
E∩{|u |6m+1} E

′
Since −p =p = −1=(p−1) ∈ L1 ( ) by (3.6), we have Hm (u ) (u ) is equiintegrable and
′
tends strongly to Hm (u)(u) in (Lp ( ; −1=(p−1) ))N .
The third integral of the left-hand side of (4.12) tends to 0, indeed ∇u * ∇u in
(Lp ( ; ))N , Hm′ (u )(uk − uk ) (u ) → 0 a.e. in and
 1=p′
p′ −p′ =p
|Hm′ (u )(uk k
− u ) (u )|  dx
E
 1=p′
p′ −p′ =p
6 2k | (u )|  dx
E∩{m6|u |6m+1}

 1=p′
−p′ =p
6 2kCm  dx
E

which implies that Hm′ (u )(uk − uk ) (u ) is equiintegrable and tends strongly to 0 in
′
(Lp ( ; −1=(p−1) ))N .
For the rst integral of the left-hand side of (4.12) we use (3.1) and ess sup|s|¡m+1
K(|s|) ¡ C,

Hm′ (u )(uk − uk )a(x; u ; ∇u )∇u

= Hm′ (u )(uk − uk )a(x; um+1 ; ∇um+1 )∇um+1
{m6|u |6m+1}

Hm′ (u )|uk − uk |1=p d(x) + |um+1 |p−1 |∇um+1 |
 
6C
{m6|u |6m+1}

 
+ Hm′ (u )|uk −u k
||∇um+1 |p dx : (4.13)
{m6|u |6m+1}

The rst integral of the relation (4.13) tends to 0, indeed ∇um+1 * ∇um+1 in
(L ( ; ))N , {m6|u |6m+1} Hm′ (u )(uk − uk )1=p {d(x) + |um+1 |p−1 } → 0 strongly in
p
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 31

′
(Lp ( ; −1=(p−1) ))N since itself converges to 0 a.e. in , and
 1=p′
′
|Hm′ (u )(uk −uk )1=p {d(x)+|um+1 |p−1 }|p −1=(p−1) d x dx
E∩{m6|u |6m+1}

 1=p′
p′
6 2Ck (d(x) + |um+1 |p ) d x
E∩{m6|u |6m+1}

 1=p′
p′ p
6 2Ck (d(x) + (m + 1) ) d x
E
−1
which gives the equiintegrability of {m6|u |6m+1} Hm′ (u )(uk −uk )1=p {d(x)+|um+1 |p }.
For the second integral of the relation (4.13) we have

Hm′ (u )|uk − uk ||∇um+1 |p d x
{m6|u |6m+1}

6 2k |∇um+1 |p d x
{m6|u |6m+1}

=2k |∇u |p d x
{m6|u |6m+1}
or by (4.5) we have
   N  p′ =r


p
|∇u | d x 6 C |fi |r −r=p d x :
{m6|u |6m+1} {m6|u |6m+1} i=1
Which gives by the absolute continuity of the integral

lim sup |∇u |p d x = o(1) (m → ∞)
→0 {m6|u |6m+1}
and

lim sup Hm′ (u )(uk − uk )a(x; u ; ∇u )∇u d x = o(1) (m → ∞):
→0
For the second integral of the left-hand side of (4.12), for m ¿ k and since Hm (u ) = 1
for |u | 6 m, we can write
 
 H (u )a(x; u ; ∇u )∇(uk −uk ) d x = − Hm (u )a(x; u ; ∇u )∇uk d x

 m    
{|u |¿k}

  (4.14)
 k k k k k k k


 + a(x; u  ; ∇u  )∇(u  −u ) d x + a(x; u  ; ∇u  )∇u d x:
{|u |¿k}

For the rst integral of the right-hand side of the relation (4.14) we have,

Hm (u )a(x; u ; ∇u )∇uk d x
{|u |¿k}

= {k6|u |¡m+1}∩{|u|¡k} Hm (u )a(x; um+1 ; ∇um+1 )∇uk d x:
32 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

Since u → u a.e. in then {k6|u |¡m+1}∩{|u|¡k} :Hm (u )a(x; um+1 ; ∇um+1 )∇uk → 0
a.e. in , and we obtain, since ess sup|u |¡m+1 K(|u |) ¡ C,

{|u |¿k} Hm (u )a(x; u ; ∇u )∇uk d x
E
 1=p′  1=p
′
6C {d(x) + (m + 1)p−1 }p |∇uk |p
E E
 1=p  1=p 
+ |∇um+1 |p d x |∇uk |p
E E

 1=p′  1=p
p−1 p′
6C {d(x) + (m + 1) } |∇uk |p
E E
 1=p 
+ C(m + 1) |∇uk |p
E

the equiintegrability, which gives


Hm (u )a(x; u ; ∇u )∇uk d x → 0 as  → 0:
{|u |¿k}

For the third integral of the right-hand side of the relation (4.14), we have since
u → u a.e. in

lim a(x; uk ; ∇uk )∇uk d x
→0 {|u |¿k}

=lim a(x; sgn(k)k; 0)∇uk d x = 0:
→0 {|u |¿k}∩{|u|¡k}

Which gives from (4.12) and (4.14)


lim sup a(x; uk ; ∇uk )∇(uk − uk ) d x 6 o(1) (m → ∞):
→0

If we tend m → ∞, we obtain

lim sup a(x; uk ; ∇uk )∇(uk − uk ) d x 6 0: (4.15)
→0

On the other hand, we have


lim a(x; uk ; ∇uk )∇(uk − uk ) d x = 0: (4.16)
→0

Indeed since ∇(uk −∇uk ) * 0 weakly in (Lp ( ; ))N and a(x; uk ; ∇uk ) → a(x; uk ; ∇uk )
′
strongly in (Lp ( ; −1=(p−1) ))N .
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 33

Then (4.15) and (4.16) give


lim [a(x; uk ; ∇uk ) − a(x; uk ; ∇uk )][∇uk − ∇uk ] d x = 0:
→0

Step 4: We claim that for all k ¿ 0

uk → uk strongly in W01; p ( ; ): (4.17)
Relation (4.11) gives for a subsequence
[a(x; uk ; ∇uk ) − a(x; uk ; ∇uk )][∇uk − ∇uk ] → 0 a:e: in :
Then arguing as in Leray–Lions [7], we have for the subsequence
∇uk (x) → ∇uk (x) a:e: in :
Let E be a measurable subset of , we shall prove the equiintegrability of ∇uk

min (|u |) (x)|∇uk |p d x
|u |¡k E

6 (x)(|uk |)|∇uk |p d x
E

6 a(x; uk ; ∇uk )∇uk
E

6 [a(x; uk ; ∇uk ) − a(x; uk ; ∇uk )][∇uk − ∇uk ] d x
E
 
+ a(x; uk ; ∇uk )[∇uk − ∇uk ] d x + a(x; uk ; ∇uk )∇uk d x (4.18)
E E

The rst and the second integral of (4.18) goes to 0 by (4.11) and (4.16).
On the other hand, we have since ess sup||¡k K(||) ¡ C,
  1=p′  1=p
k k k p−1 p′ k p
a(x; u ; ∇u )∇u d x 6 C {d(x) + |k| } |∇u | d x
E E E
 1=p  1=p
+ |∇uk |p dx k p
|∇u | d x :
E E

Then the function a(x; uk ; ∇uk )∇uk is equiintegrable and since itself converge to
a(x; uk ; ∇uk )∇uk a.e. in , we have for the third integral of (4.18)
 
lim sup a(x; uk ; ∇uk )∇uk d x = a(x; uk ; ∇uk )∇uk d x
→0

which implies that

 
lim sup (x)|∇u | d x 6 C a(x; uk ; ∇uk )∇uk d x:
k p
→0 E E
34 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

Consequently

lim sup (x)|∇u − ∇u|p d x = 0:
→0

Hence (4.17).
Step 5: Using in (4.3) the test function h(u )’ with h ∈ Cc1 (R) and ’ ∈ W01; p ( ; ) ∩
∞
L ( ), we obtain

a(x; u ; ∇u )[h′ (u )’∇u + h(u )∇’] d x

+  (u )[h′ (u )’∇u + h(u )∇’] d x

 N 

@h(u )’
+ g (x; u )h(u )’ d x = − fi d x: (4.19)
@xi
i=1

Since h and h′ have compact support on R, we have for  suciently small

 (t)h(t) = (T1= (t))h(t) = (t)h(t);

 (t)h′ (t) = (T1= (t))h′ (t) = (t)h′ (t)

and the functions h and h′ belongs to (C 0 (R) ∩ L∞ (R))N .

Let s ¿ 0 such that supph ⊂ ] − s; s[. For k ¿ s, relation (4.19) becomes

a(x; uk ; ∇uk )[h′ (uk )’∇uk + h(uk )∇’] d x

+  (uk )[h′ (uk )’∇uk + h(uk )∇’] d x

N 
@h(uk )’


+ g (x; uk )h(uk )’ d x = − fi d x: (4.20)
@xi
i=1
N 
The integral of the right-hand side of (4.20) tends to i=1 fi (@h(uk )’=@xi ). Since
N  k
N k
i=1 fi (@h(u )’=@xi ) tends to i=1 fi (@h(u )’=@xi ) a.e. in and
N N N
@h(uk )’ @uk
  


@’

fi dx = fi h(uk ) dx + fi ’h′ (uk )  d x
E @xi E @xi E @xi
i=1 i=1 i=1

N  1=p′   p 1=p


 p′ −p′ =p
 @’ 
6 h ∞ |fi |  dx  @xi  (x) d x
 
i=1 E E

N  1=p′   k p 1=p
′

′ ′  @u 
+ h ∞ ’ ∞ |fi |p −p =p dx   (x) d x
E @xi
 
i=1 E
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 35

 N  1=p′   1=p

′ ′
−p =p
6C |fi |p  dx p
|∇’| (x) d x
i=1 E E

 1=p 
+ |∇uk |p (x) d x
E

 N  1=r


r −r=p
6C |fi |  dx [ ’ 1; p;  + uk 1; p;  ]
i=1 E

which gives the equiintegrabilite since uk → uk strongly in W01; p ( ; ).

For the rst integral of the left-hand side of (4.20), since ess sup||¡k K(||) ¡ C we
have

a(x; u ; ∇u )[h′ (uk )’∇uk + h(uk )∇’] d x
E
 1
1=p′
1− p ′
6C {d(x) + |uk |p−1 +  |∇uk |p−1 }p
E
 1=p
: |h′ (uk )’∇uk + h(uk )∇’|p
E
 1=p′  (p−1)=p 
p′
6C |d(x)| + |uk |p + |∇uk |p
E E

  1=p
× h′ p∞ ’ p
∞ |∇uk |p dx + h p
∞ |∇’| p
E E

p−1
6 C( d p′ + uk 1; p;  ):( uk 1; p;  + ’ 1; p; v ):

Since uk → uk strongly in W01; p ( ; ) and ’ ∈ W01; p ( ; ), then a(x; uk ; ∇uk )[h′ (uk )
’∇uk + h(uk )∇’] is equiintegrable and itself converges to a(x; uk ; ∇uk )[h′ (uk )’∇uk +
h(uk )∇’] a.e. in . Which gives

a(x; uk ; ∇uk )[h′ (uk )’∇uk + h(uk )∇’] d x

→ a(x; uk ; ∇uk )[h′ (uk )’∇uk + h(uk )∇’] d x:

The second integral of the left-hand side of (4.20) tends to


(uk )[h′ (uk )’∇uk + h(uk )∇’] d x:
36 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37

Indeed  (uk )[h′ (uk )’∇uk + h(uk )∇’] is equiintegrable and converges to (uk )[h′ (uk )
’∇uk + h(uk )∇’] a.e. in . The equiintegrability can be deduced by the
inequality

 (uk )[h′ (uk )’∇uk + h(uk )∇’] d x
E

 1=p′
6 h′ ∞ ’ ∞ −1=(p−1) d x uk 1; p; 
E

 1=p′
−1=(p−1)
+ h ∞  dx ’ 1; p;  ;
E

with −1=(p−1) ∈ L1 ( ) and the convergence uk → uk in W01;p ( ; ).

The third integral of the left-hand side of (4.20) tends to g(x; uk )h(uk )’ d x, since
1
k k k
g (x; u ) tends to g(x; u ) strongly in L ( ), and h(u )’ tends to h(uk )’ weakly in
L∞ ( ).
After passing to the limit in each term of (4.20) and using the fact that supph ⊂
] − s; s[ and by taking k ¿ s, we obtain (4.2). Theorem 4.1 is proved.

Acknowledgements

The authors is grateful to the Professor J.M. Rakotoson for suggestions about the
problem studied here, the proof of Theorem 3.1 and Lemma 4.1.

References

[1] Y. Atik, J.M. Rakotoson, Local T-sets and renormalized solutions of degenerate quasi-linear equations
in L1 -datum, Adv. Di erential Equations 6 (1996) 956–988.
[2] A. Benkirane, J. Bennouna, Existence of Renormalized Solutions for Some Elliptic Problems Involving
Derivatives of Nonlinear Terms in Orlicz Spaces, Lecture Note in Pure and Applied Mathematics,
Volume 229, Marcel Dekker, New York, 2002.
[3] L. Boccardo, D. Giachetti, J.I. Diaz, F. Murat, Existence and regularity of renormalized solutions of
some elliptic problems involving derivatives of nonlinear terms, J. Di erential Equations 106 (1993)
215–237.
[4] G. Dal Maso, F. Murat, L. Orsina, A. Prignet, De nition and existence of renormalized solutions of
elliptic equations with general measure data, C.R. Acad. Sci. Paris, 325 Serie I (1997), 481–486.
[5] R.J. Diperna, P.L. Lions, On the cauchy problem for Boltzmann equations: global existence and weak
stability, Ann. Math. 130 (1989) 321–366.
[6] P. Drabek, A. Kufner, F. Nicolosi, Quasilinear Elliptic Equations with Degenerations and Singularities,
Walter de Gruyter, Berlin, New York, 1997.
[7] P.L. Lions, Quelques methodes de resolution des problemes aux limites non-lineaires, Dunod, Paris,
1984.
[8] J. Mossino, Inegalites isoperimetriques et applications en physique, Collection travaux en cours, Herman,
Paris, 1984.
[9] J. Mossino, R. Temam, Directorial derivations of the increasing rearrangement mapping and application
to a queer di erential equation in plasma physics, Duke Math. J. 41 (1981) 475–495.
 A. Benkirane, J. Bennouna / Nonlinear Analysis 54 (2003) 9 – 37 37

[10] J.M. Rakotoson, Rearrangement relatif dans les e quations quasi-lineaires avec un second membre
distribution: application a un theoreme d’existence et de regularite, J. Di erential Equations 66 (1987)
391–419.
[11] J.M. Rakotoson, Uniqueness of renormalized solutions in a T -set for L1 data problems and the link
between various formulations, Indiana Univ. Math. J. 43 (1994) 2.
[12] J.M. Rakotoson, R. Temam, Relative rearrangement in quasilinear elliptic variational inequalities, Indiana
Univ. Math. J. 36 (1987) 4.