Anda di halaman 1dari 24

Lucio Boccardo Benedetta Pellacci

Bounded positive critical points of some


1
multiple integrals of the Calculus of Variations

1 Introduction
Let be a bounded domain in IRN , (N > 2) and a : IR IR be a
bounded smooth real function. We recall that simple functionals as
1Z
J(v) = a(x, v)|v|2 ,
2

are derivable only along directions of W01,2 () L (). The directional


derivative is given by
Z
1Z
hJ 0 (u), vi = a(x, u)uv + as (x, u)|u|2 v,
2

where the lower order term as (x, u)|u|2 , in general, belongs to L1 () and
not to W 1,2 (). Existence results of critical points of such functionals have
been proved in [2], [3], [4], [5], [7], [8], [9], [10], [11]. In [7], [9], [10] some
multiplicity result are proved by using a weak notion of derivative for
continuous functions in complete metric spaces. In [2], [3], [4], some existence
results are proved for bounded and unbounded coefficients a(x, s). Here we
will present some existence results for this problem, proved in [2], [5] and
[11].
We will present new proofs, which hinge on a different (simpler) order of
the steps in the proof of the PalaisSmale sequences compactness.
We consider the following class of integral functionals
1Z 1Z + p
I(v) = a(x, v)|v|2 (v ) ,
2 p

1
This work contains the unpublished part of the lecture held by the first author at
Bergamo 2001 Conference.

Siede Peschiera, bello e forte arnese


da fronteggiar Bresciani e Bergamaschi (Inf. XX)

1
but we want to point out that our methods works for general functionals as
Z Z
(v) = 1 (x, v, v) 2 (x, v),

and we assume that


2N
2 < p < 2 = , (1)
N 2
: IR+ IR,

a(x, s)
a(x, s)
= as (x, s) : IR+ IR
s
are functions, which are measurable with respect to x and continuous with
respect to s.
The paper is organized as follows.
In section 2 we will present a new proof of two existence results proved in
[2] in the case of bounded coefficients (studied in Subsection 2.1) and in [5]
in the case of unbounded coefficients (studied in Subsection 2.3). Finally, in
Section 3 we will consider nonincreasing coefficients and we will obtain a new
proof of an existence result proved in [11].

2 Nondecreasing coefficients
2.1 Bounded coefficients
We assume that the following conditions are satisfied for almost every x in
and for every s in IR+ (for , , , IR+ )

0 < a(x, s) , (2)

as (x, s) s 0, (3)
|as (x, s)| , (4)
(p 2)a(x, s) as (x, s)s . (5)
For s < 0, we define
a(x, s) a(x, 0) (6)
so that as (x, s) 0 for every s < 0.
Let us recall the definition of a critical point.

2
Definition 2.1 A function u is a critical point of I if

W01,2 () L ()


u :
1

Z Z Z
as (x, u)|u|2 = (u+ )p1 ,

a(x, u)u +
2


W01,2 () L ()

In this section we will present a new proof of an existence result proved in


[2]. In order to study the existence of critical points, we need a version of the
Ambrosetti and Rabinowitz Theorem [1] for functionals not differentiable in
all directions. The proof can be found in [2].

Theorem 2.2 Assume (1), (2), (3), (4), (5). Then the functional I has at
least one nontrivial (positive) critical point.

Proof. The proof (see [2]) is done by dividing it into two steps. In the
first one, only the geometric hypotheses are used to deduce the existence of
a sequence {un } in W01,2 () L () satisfying the following condition (9).
The second (longer) step is the proof of the compactness of this sequence
{un } in W01,2 (). In this way, condition (9) can be considered as a compact-
ness condition on our functional, which substitutes in the nondifferentiable
case the role done by the well-known Palais-Smale condition in the regular
classical (semilinear) case.
As in the semilinear case (i.e. a(x, s) = a(x)), assumptions (1) and (2)
imply that u0 = 0 is a strict local minimum of I: there exist , R IR+ such
that
I(v) > 0, for kvkW 1,2 () = R > 0. (7)
0

Remark that I(T 1 ) < 0, for some T IR+ , where 1 is the first eigen-
function of the Dirichlet problem for the laplacian in . Moreover T can be
chosen large enough, such that

kT 1 kW 1,2 () 2R. (8)


0

From now on, we follow the proof of Theorem 3.3 of [2]. We apply Theorem
2.1 of [2]. Consider X = W01,2 () and Y = W01,2 () L (), endowed with
the norm k kY = k kW 1,2 () + k k . Let
0

n   o
= : [0, 1] Y, k kY : continuous and (0) = 0, (1) = T 1 .

3
Observe that every is continuous from [0, 1] to W01,2 (), so that, by
(7) and (8), for every there exists t [0, 1] such that
k(t)kW 1,2 () = R.
0

Thus, I(T 1 ) < 0 implies that


inf max I((t)) > max{I(0), I(T 1 )} = 0.
t[0,1]

Let {n } be a sequence of paths for which


1
max I(n (t)) + , n IN.
t[0,1] 2n
For fixed n IN , consider Mn = max kn (t)k . Since n (1) = T 1 ,
t[0,1]
Mn T k1 k .
k k
Note that + k k is a norm in W01,p () L () which is equivalent
Mn
to k k + k k. By applying Theorem 2.1 of [2] we deduce the existence of
a path n and a function un = n (tn ) n ([0, 1]) satisfying
!
k n (t)n (t)k

L ()
q
1
maxt[0,1] + k n (t) n (t)kW 1,2 () ,


Mn n

0



1 1
n
I(u ) +
n 2n
, !


q kvk
0
|hI (un ), vi| n
1 L ()
+ kvkW 1,2 () , v W01,2 () L (),



Mn
0

(9)
and for n IN large enough,
kun kL () = k n (tn )kL () k n (tn ) n (tn )kL () + kn (tn )kL () 2Mn .

With this setting, we shall use the results of the following section concerning
the sequence {un }. Proposition 2.5 implies that {un } possesses a subsequence
{unk } which (thanks to theorem 2.10) is strongly convergent in W01,2 () to
some u W01,2 () L (), I(u) = > 0 and u is a nontrivial critical point
of I.

Remark 2.3 If p < 2, assumption (2) implies that I is coercive, so that there
exists a (nontrivial) bounded minimum. If assumption (3) is not satisfied,
some existence results of critical points can be found in [4] and [11].

4
2.2 Compactness of the sequence {un }
This section is the core of the paper, since our aim is the new proof of the
compactness of PalaisSmale sequences.
Recall that, on the functional I, we assume (1), (2), (3), (4), (5). Let un
satisfy (9).
We write again our framework as:

un W01,2 () L (), kun kL () 2Mn ,

|I(un )| R,
(10)
kvkL ()

|hI 0 (un ), vi| n + kvkW 1,2 () , v W01,2 () L (), (11)


Mn 0

where R is a positive constant, {Mn } IR+ {0} is any sequence and


{n } IR+ is a sequence converging to zero.
In the first step the boundedness of the sequence {un } is proved (unk *
u). In the second step, even if {un } is not bounded in L (), the bound-
edness of u is proved. In the third step, it is proved that the sequence {u
n}
1,2
converges strongly to zero in W0 (). Thus, u 0. Then, roughly speak-
ing, it is not negligible only the contribution of the sequence {un } on the
subsets {x : 0 un (x) u(x)}, that is where {un } is positive and bounded
in L ().

Proposition 2.4 The sequence {un } is bounded in W01,2 ().

Proof. Take p1 un as test function in (11). From (10) and (11) we derive

1 1Z + p
Z
2
a(x, u )|u | (un ) R


n n
2 p



Z


1Z 1 1Z + p

2 2

p
a(x, u n )|u n | u a
n s (x, u n )|u n | + (un )
2p p




n
[2 + kun kW 1,2 () ].



p

0

We sum the two previous inequalities and we get


Z
[(p 2)a(x, un ) un as (x, un )]|un |2 2n (2 + kun kW 1,2 () ) + 2pR.
0

5
Then assumptions (1) and (5) imply that there exists a positive constant L,
such that
kun kW 1,2 () L. (12)
0

As a consequence of the previous Proposition, we get that there exist


a function u W01,2 () and a subsequence of {un } (still denoted by {un })
such that un converges to u weakly in W01,2 (), strongly in L2 () and almost
everywhere. Moreover, we take v = un in (11) and we use conditions (2) and
(12) in order to deduce that for some c0 IR+ ,
Z
un as (x, un )|un |2 c0 .

Now we shall prove that the function u is bounded.


For every k > 0, we define

Tk (s) = max(k, min(k, s)) , Gk (v) = v Tk (v). (13)

Proposition 2.5 The function u belongs to L ().

Proof. Taking v = Gk (un ) as test function in (11), we deduce that


Z
1 Z
a(x, un ) |Gk (un )|2 + as (x, un ) Gk (un ) |Gk (un )|2
2
Z
(u+
n)
p1
Gk (un ) + (2 + L)n .

From (2) and (3), Sobolev inequality we obtain that


2
Z 2 Z
2 0
S |Gk (un )| (u+
n)
p1
Gk (un ) + n

Z Z
0
2p2 |Gk (un )|p + 2p2 k p1 |Gk (un )| + n ,

6
0
where n 0 and S is the Sobolev constant. Now Hoelder inequality implies
that
2 p2
Z Z 2 Z 2
p 2 1 2p 2
|Gk (un )| |Gk (un )| meas (An,k ) |Gk (un )| ,

where An,k = {x : |un (x)| > k}. Thanks to estimate (12) we have that
p2

Z 2
2
|Gk (un )| c1

and that there exists k > 0 such that, for k > k = k (L),

p
c1 2p2 meas (An,k )1 2 S.
2
Moreover (thanks again to estimate (12) and Hoelder and Young inequalities)
2 1
Z 2 Z 2
1 1
2 p1 2 0
c2 |Gk (un )| k |Gk (un )| [meas (An,k )] 2 + N + n

2
2
c 2 Z 2 0
|Gk (un )|2 + c3 k 2(p1) [meas (An,k )]1+ N + n
2

So 1
Z 2
1 1
2
c4 |Gk (un )| k (p1) [meas (An,k )] 2 + N + n ,

q
where n = 0n . Since kukL2 () c(L), we obtain
1
Z 2
p1 1 1 p1
c4 |Gk (un )|2
k (p1) [meas (An,k )] 2 [meas (An,k )] 2 + N 2 + n

1 1 p1
c5 (L)[meas (An,k )] 2 + N 2 + n .
1 1 1 p1 N +2
Remark that
< + , since p < . Thus, a classical result
2 2 N 2 N 2
due to Stampacchia ([13]) implies that
kukL () c(kukW 1,2 () ) = M. (14)
0

7
1,2
Proposition 2.6 The sequence {u n } converges strongly to zero in W0 ().
Moreover
1 Z
as (x, un ) u 2
n |un | 0.
2

Proof. Taking v = u
n as test function in (11), we deduce that
Z
1 Z
a(x, un ) |u
n|
2
as (x, un ) u
n |un |
2
2

(2 + L)n .
Then the assumptions on a(x, s) and as (x, s) imply the conclusion.

Remark 2.7 As a consequence of the previous proposition we get that u 0.

Lemma 2.8 The sequence {[un u]+ } converges strongly to zero in W01,2 ().

Proof. We take v = Gk (u+ n ) as test function in (11). In virtue of the


Sobolev embedding theorem, we deduce that
Z Z
0
|Gk (u+ 2
n )| (u+
n)
p1
Gk (u+
n ) + n

1 N +2

2 2N
Z Z
+ 2 (p1) N2N 0
|Gk (un )| (u+
n) + n .

+2

{x:un (x)k}

1 N +2
2N
Z 2 Z
(p1) N2N 0
c1 |Gk (u+ )|2
(u+
n) + n .

+2
n
{x:un (x)k}

Now, by the Young inequality, we obtain


N +2
N
Z Z
(p1) N2N 0
|Gk (u+ 2
n )| c2

(u+
n)
+2

+ n .
2

{x:un (x)k}

8
Since {un } is bounded in W01,2 () and as (p 1) N2N
+2
< 2 , we deduce that,
for any fixed  > 0, there exists k1 > 0 such that,
N +2
N
Z
(p1) N2N
2c2 (u+
n) , k k1 and n IN .

+2

{x:un (x)k}

So that Z
00
|Gk (u+ 2
n )|  + n , k k1 , n IN. (15)

2 1
Since |u| L (), there exists k2 > 0 such that, for every k k2 ,
Z
|u|2 . (16)
{x:kun (x)}

Now we study the behavior of the positive part of un u.


The positivity of u yields un (x) k in the subset {x : k un (x)u(x)}.
Therefore, we have Z
|Gk [un u]+ |2

Z Z
2
= |[un u]| |[un u]|2
{x:kun (x)u(x)} {x:kun (x)}
Z Z
2
2 |un | + 2 |u|2 .
{x:kun (x)} {x:kun (x)}

Fix k0 = max{k1 , k2 }. From by (15) and (16) we deduce that


Z
00
|Gk [un u]+ |2 2( + n ) + 2, k k0 , n IN. (17)

The use of Tk [un u]+ as test function in (11) leads to


Z
1 Z +
a(x, un ) un Tk [un u] + as (x, un ) Tk [un u]+ |un |2
2

Z
(u+
n)
p1
Tk [un u]+ + (2 + 4L)n .

9
From (2) and (3) we obtain that
Z Z
+ 2



|T [u
k n u] | (u+
n)
p1
Tk [un u]+


Z (18)



(2 + 4L)n + a(x, un ) uTk [un u]+ .

We point out that the last integral is different from zero only on the subset

{x : u(x) un (x) u(x) + k};

and here the sequence {un } is bounded, because u L (). Thus, there
exists n IN such that
Z
a(x, un ) uTk [un u]+ , n > n . (19)

Then (17), (18) and (19) imply that


Z
000 000
|[un u]+ |2 n + 5, k k0 , n > n (n 0),

that is
k[un u]+ kW 1,2 () 0.
0

Now we study the behaviour of the negative part of un u.


Lemma 2.9 The sequence {[un u] } converges strongly to zero in W01,2 ().
Proof. First, we will show that

k[u+
n u] kW 1,2 () 0.
0

Define
s2 2
(s) = se , = .
162

Recall that kukL () M . Note that


(TM [u+
n u] ) =

10

0
in {x : un (x) > u(x)},

= (T [u+
M n
u] ) in {x : 0 un (x) u(x)},

(u) in {x : un (x) < 0}.
Since u belongs to L (), the sequence {un } is bounded in the set {x : 0

un (x) u(x)}. Let zn = TM [u+ +
n u] . We use (TM [un u] ) as test
function in (11), from (1) and (3) we get
Z
1 Z
a(x, un )un zn 0 (zn ) + as (x, un )|un |2 (zn ) c1 n .
2
{x:un (x)0}

But
1 Z
2 1 Z
as (x, un )|un | (zn ) = as (x, un )|un |2 (zn )
2 2
{x:un (x)0} {x:0un (x)u(x)}

Z
2 1 Z
|(un u)| (zn )+ as (x, un ) (zn )|u|2
2 2
{x:0un (x)u(x)} {x:0un (x)u(x)}
Z
+ as (x, un )un u (zn )
{x:0un (x)u(x)}

and the last two integrals converge to zero. From now on we will denote with
= ni , for i = 1, 2, . . ., quantities converging to zero.
It results
Z Z
a(x, un )un zn 0 (zn ) = a(x, un )|(u un )|2 0 (zn )
{x:0un (x)u(x)}
Z Z
a(x, un )u(u un )0 (zn ) a(x, un )un u0 (u)
{x:0un (x)u(x)} {x:un 0}
Z Z
= a(x, un )|zn |2 0 (zn ) a(x, un )u(u un )0 (zn )
{x:0un (x)u(x)}
Z Z
a(x, un )un u0 (u) a(x, un )|u|2 0 (u).
{x:un 0} {x:un 0}

11
1,2 1,2
Since u L (), u
n 0 strongly in W0 () and un * u in W0 (), we
have Z
a(x, un )un u0 (u) 0,
{x:uZ
n 0}

a(x, un )|u|2 0 (u) 0.


{x:un 0}Z

a(x, un )u(u un )0 (zn ) 0.


{x:0un (x)u(x)}

Then the choice of implies that 0 (zn ) 2 (zn )


2
and
Z
2
|TM [u+ 1 2 3 4 5
n u] | c1 n + n + n + n + n + n .
2

Finally the inequality


Z Z Z
2 2
|[u+
n u] | = 2
|u| + |TM [u+
n u] |
{x:un (x)0} {x:0un (x)u(x)}
Z Z
2
|u|2 + |TM [u+
n u] |
{x:un (x)0}

gives the result Z


2
lim
n
|[u+
n u] | = 0. (20)

Moreover, we have
Z Z Z
|[un u] |2 |(u un )|2 + |[u+ 2
n u] |
{x:un (x)0}
Z Z Z
2 |u| + 22
|u
n|
2
+ |[u+ 2
n u] |
{x:un (x)0} {x:un (x)0}

We take into account (20) and that un converges to the positive function
u W01,2 () strongly in L2 () and almost everywhere joint with the fact
1,2
that u
n converges strongly to 0 in W0 (). Thus

k[un u] kW 1,2 () 0.
0

Lemma 2.8 and Lemma 2.9 give the following theorem.

12
Theorem 2.10 The sequence {un } converges strongly to u in W01,2 (). Mo-
reover 12 as (x, un )|un |2 converges to 12 as (x, u)|u|2 and {a(x, un )|un |2 }
converges to a(x, u)|u|2 in L1 () (which implies that I(un ) converges to
I(u)).

2.3 Unbounded coefficients


Here, we will study the existence of critical points of the integral functionals
with unbounded (from above) coefficients a(x, s) whose model example is
1Z 1Z + p
I2 (v) = (1 + |v|m )|v|2 (v ) , m > 0.
2 p

In this section we will present a new proof of an existence result proved in


[3], [5]. Notice that when we consider unbounded coefficients the functional
is well defined only on a subset of W01,2 (). More precisely, let us consider a
function a(x, s) : IR+ IR+ measurable with respect to x , derivable
with respect to s IR+ and such that hypotheses (3) and (5) are satisfied.
Moreover, we will suppose that there exist positive constants , satisfying
the following hypotheses

a(x, s) (s), (21)

|as (x, s)| (s), (22)


lim ((s) sp2 ) < 0, (23)
s+

where , are continuous, increasing (possibly unbounded) functions of a


real variable.
Remark 2.11 If we suppose (22) and that a(x, R0) 0 we deduce that (21)
is satisfied with (s) = 0 + (s), where (s) = 0s (t)dt.
We consider an exponent p that satisfies (1) and we define the functional
I : W01,2 () IR {+} by
1 1Z + p
Z
2

a(x, v)|v| (v ) , if I(v) < +,
I(v) = 2 p

+, otherwise,

We will prove the following theorem.

13
Theorem 2.12 Assume (1), (3), (5), (21), (22), (23). Then the functional
I has at least one nontrivial (positive) critical point.
Proof. We follow the outline of the proof of Theorem 2.2 and we notice
that, thanks to conditions (1), (21) and (23), the functional I satisfies the
geometrical hypotheses of Mountain Pass type, i.e. u0 = 0 is a strict local
minimum in the topology of W01,2 () and I(T 1 ) < 0 for T sufficiently large.
As before we set
X = W01,2 (), Y = W01,2 () L ().
Condition (21) and (22) imply that for every u, v Y there exists hJ 0 (u), vi,
moreover for every v Y the map v hJ 0 (u), vi is continuous for every fixed
u Y and for every u Y the map u hJ 0 (u), vi is continuous for every
fixed v Y . This regularity properties of the functional I are enough to apply
Theorem 2.1 in [2]. Then, we get that there exists un W01,2 ()L () that
satisfies (10) and (11). Propositions 2.13 and 2.14 imply that {un } possesses
a subsequence {unk }, which (thanks to Theorem 2.17) is strongly convergent
in W01,2 () to some u W01,2 () L (). Lemma 2.19 implies that u is a
critical point of I and from Lemma 2.20 we deduce that I(u) = > 0 so
that u is nontrivial.

Proposition 2.13 The sequence {un } is bounded in W01,2 (). Moreover, it


results Z
un as (x, un )|un |2 c1 , (24)

+
for some c1 IR .
Proof. The proof is the same of Proposition 2.4.

Thus, there exist a positive function u W01,2 () and a subsequence of


{un } (still denoted by {un }) such that un converges to u weakly in W01,2 ()
and strongly in L2 ().
Now we shall prove that the function u is bounded.
Lemma 2.14 The function u belongs to L ().
Proof. The proof is the same of Proposition 2.5. Indeed, note that for
every n un L (). So that we can take v = Gk (un ) as test function in
(11) in order to obtain (14).

14
1,2
Proposition 2.15 The sequence {u
n } converges strongly to zero in W0 ()
(thus u 0).

Proof. The proof is the same of Proposition 2.6.

Remark 2.16 Notice that, even if u


n 0, we cannot deduce that
Z Z
a(x, un )|u 2
n | 0, as (x, un )u 2
n |un | 0

as in Proposition 2.6, because the coefficients a(x, s) and as (x, s) are not
uniformly bounded in L ().

Theorem 2.17 The sequence {un } converges strongly to u in W01,2 ().

Proof. In order to prove that un converges to u in W01,2 () we can follow the


same procedure of Subsection 2.1. Indeed, notice that all the test functions
taken in Lemma 2.9 and in Lemma 2.8 are continuous function of un and
u. Moreover, they are different from zero only in subsets of where un is
uniformly, with respect to n, bounded in L (). This permits us to pass to
the limit and conclude that un u in W01,2 ().

Now we want to show that u is a critical point of the functional I.


Lemma 2.18 The sequence 21 as (x, un )|un |2 converges to 12 as (x, u)|u|2 in
L1 ().
We want to point out that, since as (x, un ) is not bounded in L (), the
strong convergence in L1 () of as (x, un )|un |2 to as (x, u)|u|2 is not a con-
sequence of the strong convergence in L1 () of |un |2 . Now we will prove
that as (x, un )|un |2 is uniformly equiintegrable. For any measurable subset
E of and for any m IR+ we have (thanks to (4) and (24))
Z Z
2
as (x, un )|un | = as (x, un )|un |2
E {xE:0un (x)<m}

Z Z
1 Z
+ as (x, un )|un |2 (m) |un |2 + un as (x, un )|un |2
m
{x:mun (x)} E

15
Z
1
(m) |un |2 + c1 ,
m
E

which proves the uniform equiintegrability of as (x, un )|un |2 as a conse-


quence of the uniform equiintegrability of |un |2 , i.e.
Z
lim as (x, un )|un |2 = 0, uniformly with respect to n,
|E|0
E

Therefore, we have as (x, un )|un |2 as (x, u)|u|2 strongly in L1 ().

Lemma 2.19 The function u is a solution of the Dirichlet problem


Z Z Z

a(x, u)u + as (x, u)|u|2 = up1 ,
(25)
W01,2 ()

L ().

Proof. Notice that, since a(x, un ) is not bounded in L (), up to now we


are not able to prove that a(x, un )|un |2 converges to a(x, u)|u|2 in L1 ().
Now we take v = Tk [un ] in (11) (k IR+ , W01,2 ()L ()) and pass
to the limit. Thanks to Theorem 2.17, Lemma 2.18 and since u L (),
we obtain
Z Z Z
2
a(x, u)u[u ] + as (x, u)|u| [u ]
up1 [u ],

W01,2 () L ().

Then, it is enough to take = u + and = u ( W01,2 () L ())


in order to obtain (25).

Finally, let us prove that u 6 0. This will be a consequence of the


following Lemma.
Lemma 2.20 The sequence {a(x, un )|un |2 } converges to a(x, u)|u|2 in
L1 ().
Proof. Take v = un in (11) and use Fatou Lemma. We have
Z
2 1 Z
a(x, u) |u| + as (x, u) u |u|2
2

16
Z
2 1 Z
lim inf a(x, un ) |un | + lim inf as (x, un ) un |un |2
2


Z 1 Z
lim inf a(x, un ) |un |2 + as (x, un ) un |un |2
2

Z Z
p
lim (un ) + (2 + L)n = up =

(thanks to (25), with = u)


Z
1 Z
2
= a(x, u) |u| + as (x, u) u |u|2 .
2

So we proved that {a(x, un )|un |2 } converges to a(x, u)|u|2 in L1 () and


that {un as (x, un )|un |2 } converges to uas (x, u)|u|2 in L1 ().

3 Nonincreasing coefficients
In this section we will consider nonincreasing coefficients a(x, s). We will
present one of the existence results proved in [11]. Our proof will be different
from [11] and we will follow the same outline of the previous section.
We will assume that a(x, s) satisfies (2), (4). Moreover, we will suppose the
following hypotheses on as (x, s) (for 0 and R0 IR+ )
as (x, s)s 0, s IR+ , (26)
|as (x, s)|s 0 , for every s R0 , (27)

(s) := sup |as (x, s)|,
(28)
(s) L1 ().

Remark 3.1 Notice that condition (2) implies that


Z +
sup |as (x, s)| ,
0

while condition (28) is slightly stronger and it is always satisfied in the model
case a(x, s) = a(s).

17
We will prove the following theorem.

Theorem 3.2 Assume (1), (2), (4) (26), (27) and (28). Then the functional
I has at least one nontrivial (positive) critical point.

Proof. We follow the outline of the proof of Theorem 2.2 and we notice
that, thanks to conditions (1) and (2), the functional I satisfies the geo-
metrical hypotheses of Mountain Pass type, i.e. u0 = 0 is a strict local
minimum in the topology of W01,2 () and I(T 1 ) < 0 for T sufficiently large.
Moreover, note that, as before, I satisfies all the assumptions requested by
Theorem 2.1 in [2]. Then we get that there exists un W01,2 ()L () that
satisfies (10) and (11). Propositions 3.3 and 3.5 imply that {un } possesses a
subsequence {unk }, which (thanks to Theorem 3.11) is strongly convergent
in W01,2 () to some u W01,2 () L (), I(u) = > 0 and u is a nontrivial
critical point of I.

Proposition 3.3 The sequence {un } is bounded in W01,2 ().

Proof. As in Proposition 2.4 we take p1 un as test function in (11). From


(10) and (11) we get
1 1Z + p
Z
2
a(x, u )|u | (un ) R


n n
2 p



Z


1Z 1 1Z + p

2 2
a(x, un )|un | un as (x, un )|un | + (un ) (29)
p
2p p




n
[2 + kun kW 1,2 () ].



p

0

Notice that (6) and (26) imply


1 Z
un as (x, un )|un |2 0,
2p

so that, when we sum the two inequalities in (29) we obtain


!
1 1 Z
a(x, un )|un |2 2n (2 + kun kW 1,2 () ) + 2pR.
2 p 0

18
Then assumption (26) implies that there exists a positive constant L, such
that
kun kW 1,2 () L. (30)
0

Thus, there exist a function u W01,2 () and a subsequence of {un } (still


denoted by {un }) such that un converges to u weakly in W01,2 () and strongly
in L2 ().
Remark 3.4 Notice that in the previous proposition we did not use hypoth-
esis (5) because we took advantage of condition (26).

Lemma 3.5 The function u belongs to L ().

Proof. Note that (27) implies that


(s)s 0 , for every s IR. (31)
We follow the argument of Lemma 3.5 in [11] and we define
1 Zs
(s) = (t) dt. (32)
2 0
We notice that condition (28) implies that (s) is a bounded function. Let
us consider v = e(un ) Gk (un ), where Gk (un ) is defined in (13). Conditions
(4) and (31) imply that v L () W01,2 (), then we can take v as test
function in (11) and we obtain
Z
1 Z (un )
e(un ) a(x, un )un Gk (un ) + e a(x, un )|un |2 Gk (un )(un )
2

1 Z (un ) Z
+ e as (x, un )|un | Gk (un ) c0 (u+
2
n)
p1
Gk (un ) + c1 n .
2

Condition (26) and the definition of (s) imply that


" #
Z
1 (un ) a(x, un )
e Gk (un )|un |2 (un ) + as (x, un ) 0.
2

Thereofre, we obtain
Z Z
a(x, un )un Gk (un ) c0 (u+
n)
p1
Gk (un ) + c1 n .

19
From now on it is possible to follow the same argument of Proposition 2.5 in
order to obtain that u belongs to L ().

Remark 3.6 The use of the function combined with exponential functions
has been introduced in [12] and in [6] in order to study problems of this kind
without assuming any sign condition on the quadratic gradient term.
1,2
Proposition 3.7 The sequence {u
n } converges strongly to zero in W0 ().
Moreover
Z
1 Z
a(x, un ) |u
n|
2
0, as (x, un ) u 2
n |un | 0.
2

Proof. The same of Proposition 2.6.

Remark 3.8 As a consequence of the previous Proposition we get that u 0.

Lemma 3.9 The sequence {[un u]+ } converges strongly to zero in W01,2 ().

Proof. We take v = e(un ) Gk (u+


n ) as test function in (11). As in Proposi-
tion 3.5 we get
Z Z
0
|Gk (u+
n )|
2
(u+
n)
p1
Gk (u+
n ) + n,

where {0n } IR+ is a sequence converging to zero. From now on it is possible


to follow the same argument of Lemma 2.8 to obtain (17). Then, in order to
show that (un u)+ 0, it is left to prove that Tk (un u)+ 0, as n tends
to infinity. Let us consider v = [Tk (un u)+ ], where k k0 (k0 is fixed in
(17)) is fixed and
2 2
(s) := ses , = .
162
Note that

0
in {x : un (x) < u(x)},
+
(Tk [un u] ) = (un u) in {x : u(x) un (x) u(x) + k},

(k) in {x : un (x) > u(x) + k}.

20
Therefore, v W01,2 () L () and we can take v as test function in (11).
We set zn = Tk (un u)+ and we obtain
Z
1Z
a(x, un )un zn 0 [zn ] + as (x, un )|un |2 [zn ] n,k
1
,
2

i
where lim n,k = 0 for fixed k and for i = 1, 2, . . .. Since un * u and from
n
condition (2) we get
Z Z
a(x, un )un zn 0 [zn ] = a(x, un )(un u)zn 0 [zn ]
Z
+ a(x, un )uTk (un u)+ 0 [zn ]

Z
= a(x, un )|Tk (un u)+ |2 0 [zn ] + n,k
2
.

Thus, condition (4) yields


Z
Z
a(x, un )|Tk (un u)+ |2 0 [zn ] |un |2 [zn ] + n,k
1 2
+ n,k . (33)
2

From (17) we deduce


Z Z Z
|un |2 [zn ] = |(un u)|2 [zn ] |u|2 [zn ]
2 2 2

Z
3 (k)
+ un u [zn ] n,k + ( + n )


Z
+ |Tk (un u)+ |2 [zn ],
2

The last inequality together with (33) imply


Z
(k)
|Tk (un u)+ |2 {0 [zn ] [zn ]} n,k
1 2
+ n,k 3
+ n,k + ( + n ).
2

From the choice of we deduce


Z
(k)
|Tk (un u)+ |2 n,k
4
+ ( + n ). (34)

Finally, (17) and (34) yields the conclusion.

21
Lemma 3.10 The sequence {[un u] } is strongly convergent to zero in
W01,2 ().

Proof. Let us take as test function v = (u+
n u) . From conditions (1)
and (26) we get Z
0
a(x, un )un (u+ n u) n ,

where {0n } +
IR is a sequence converging to zero. Note that
Z Z

a(x, un )un (u+
n u) = a(x, un )u+ +
n (un u)

Z Z
+ a(x, un )u +
n (un u) = a(x, un )|(u+ 2
n u) |

Z Z
a(x, un )u(u+
n
2
u) + a(x, un )u +
n (un u)

Proposition 3.7 joint with the fact that u+n * u implies that the last two
integrals tend to zero. Then, from condition (2) we have
Z
2 000
|(u+
n u) | n . (35)

Finally from (35) we have


Z Z Z
|(un u) |2 = |(u+ 2
n u) | + |(u 2
n u) |
{x : un 0}Z Z {x : un 0}
000
n + 2 |u
n|
2
+2 |u|2
{x : un 0}

Proposition 3.7 yields the conclusion.

Lemma 3.9 and Lemma 3.10 give the following Theorem.

Theorem 3.11 The sequence {un } tends strongly to u in W01,2 (). In addi-
tion, 21 as (x, un )|un |2 tends to 12 as (x, u)|u|2 in L1 () and {a(x, un )|un |2 }
converges to a(x, u)|u|2 in L1 () (so that implies that I(un ) converges to
I(u) ).

22
References
[1] A. Ambrosetti, P. H. Rabinowitz: Dual variational methods in critical
point theory and applications. J. Funct. Anal. 14 (1973), 349-381.
[2] D. Arcoya, L. Boccardo: Critical points for multiple integrals of the
Calculus of Variations. Arch. Rational Mech. Anal. 134 (1996), 249274.
[3] D. Arcoya, L. Boccardo: Some remarks on critical point theory for non-
differentiable functionals. NoDEA 6 (1999), 79100.
[4] D. Arcoya, L. Boccardo, L. Orsina: Existence of critical points for some
noncoercive functionals. Ann. Inst. H. Poincare Anal. Non Lineaire, 18
(2001), 437457.
[5] L. Boccardo: The Bensoussan & Co. technique for the study of some
critical points problems. In Optimal Control and PDE, J.L. Menaldi et
al. ed. Book in honour of Pressor Alain Bensoussans 60th birthday. IOS
Press (2000).
[6] L. Boccardo, S. Segura, C. Trombetti: Existence of bounded and un-
bounded solutions for a class of quasilinear elliptic problems with a
quadratic gradient term. J. Math. Pures et Appl., 80 (2001), 919940.
[7] A. Canino, M. Degiovanni: Nonsmooth critical point theory and quasi-
linear elliptic equations. Topological methods in differential equations
and inclusions (Montreal, PQ, 1994), 150, NATO Adv. Sci. Inst. Ser.
C Math. Phys. Sci., 472, Kluwer Acad. Publ., Dordrecht, 1995.
[8] J.N. Corvellec, M. Degiovanni: Nontrivial solutions of quasilinear equa-
tions via nonsmooth Morse theory. J. Differential Equations 136 (1997),
no. 2, 268293.
[9] J.N. Corvellec, M. Degiovanni, M. Marzocchi: Deformation properties
for continuous functionals and critical point theory. Topol. Methods
Nonlinear Anal. 1 (1993), no. 1, 151171.
[10] B. Pellacci: Critical points for non-differentiable functionals. Boll. Un.
Mat. Ital. B (7) 11 (1997), 733749.
[11] B. Pellacci: Critical points for some integral functionals. To appear on
Top. Meth. in Nonlinear Anal.

23
[12] S. Segura: Existence and uniqueness for L1 data of some elliptic equa-
tions with natural growth. To appear on Adv. in Diff. Eqn.

[13] G. Stampacchia: Le Probleme de Dirichlet pour les equations elliptiques


du second ordre a coefficients discontinus. Ann. Inst. Fourier Grenoble,
15 (1965), 189258.

L. Boccardo, B. Pellacci
Dipartimento di Matematica, Universita di Roma I,
Piazza A. Moro 2, 00185 Roma, Italia

24

Anda mungkin juga menyukai