Вы находитесь на странице: 1из 14

Z. Angew. Math. Phys.

c 2014 Springer Basel



DOI 10.1007/s00033-014-0486-6

Zeitschrift f
ur angewandte
Mathematik und Physik ZAMP

The Nehari manifold for nonlocal elliptic operators involving concaveconvex


nonlinearities
Wenjing Chen and Shengbing Deng
Abstract. In this paper, we study the multiplicity of solutions to equations driven by a nonlocal integro-dierential operator
LK with homogeneous Dirichlet boundary conditions. In particular, using bering maps and Nehari manifold, we obtain
multiple solutions to the following fractional elliptic problem


()s u(x) = uq + up ,
u = 0,

u > 0 in ;
in RN \,

where is a smooth bounded set in Rn , n > 2s with s (0, 1), is a positive parameter, the exponents p and q satisfy
2n
.
0 < q < 1 < p  2s 1 with 2s = n2s
Mathematics Subject Classification. 35J20 35J60 47G20.
Keywords. Nonlocal integro-dierential operator Concaveconvex nonlinearities Nehari manifold.

1. Introduction
In this paper, we study the multiplicity of solutions to the boundary problem by the nonlocal operator

LK u(x) = uq + up , u > 0 in ;
(1.1)
u = 0,
in RN \,
where is a smooth bounded set in Rn , n > 2s with s (0, 1), is a positive parameter, the exponents
2n
(n > 2s) the fractional Sobolev exponent, and LK
p and q satisfy 0 < q < 1 < p  2s 1 with 2s = n2s
is an integro-dierential operators of nonlocal fractional type dened as follows

1
(u(x + y) + u(x y) 2u(x)) K(y)dy, x Rn .
LK u(x) =
2
Rn

Here, K : R \{0} (0, +) is a function satisfying the following properties:


K(x) L1 (Rn ) with (x) = min{|x|2 , 1};

(1.2)

there exists > 0 such that K(x)  |x|(n+2s) for any x Rn \{0};

(1.3)

K(x) = K(x) for any x Rn \{0}.

(1.4)

The authors have been partly supported by Fundamental Research Funds for the Central XDJK2015C042,
XDJK2015C043, and Doctoral Fund of Southwest University SWU114040, SWU114041. The second author has been partly
supported by Postdoctoral Fondecyt Grant 3140403.

W. Chen and S. Deng

ZAMP

A typical example for K is given by K(x) = |x|(n+2s) . In this case problem, (1.1) becomes

()s u(x) = uq + up , u > 0 in ;
in Rn \,

u = 0,

where ()s is the fractional Laplace operator, which may be dened as



u(x + y) + u(x y) 2u(x)
1
dy, x Rn .
()s u(x) =
2
|y|n+2s

(1.5)

(1.6)

Rn

Recently, a great attention has been focused on the study of fractional and nonlocal operators of
elliptic type, see, for instance, [11,1315,24] for the subcritical case and [8,12,1618] for the critical
case. In particular, Yu [24] used the Nehari manifold to establish the existence of solutions and multiple
solutions for elliptic problems involving the square root of the Laplacian. Br
andle, E. Colorado, and A. de
Pablo [3] studied the fractional Laplace equation involving concaveconvex nonlinearity for the subcritical
case. The authors in [3] used the idea of the sharmonic extension introduced by Caarelli and Silvestre
[7], which allows to express the fractional operator ()s as a Dirichlet to Neumann map.
Aim of this paper was to investigate multiple solutions to problems (1.1) and (1.5) for the subcritical
case and the critical case. In the other words, we will use the Nehari manifold method to obtain the
multiplicity of solutions to problems (1.1) and (1.5) for the following two cases:

the subcritical case: 0 < q < 1, 1 < p < 2s ;


the critical case: 0 < q < 1, p = 2s .

In problems (1.1) and (1.5), the parameter s (0, 1) is xed and the set Rn , n > 2s, is open
bounded with Lipschitz boundary. The standard Dirichlet condition u = 0 in is replaced by the
condition that the function u vanishes outside , consistently with the nonlocal character of the operator
LK . Problems (1.1) and (1.5) have variational structure, and solutions can be constructed as critical
points of an associated energy functional on some appropriate space. In order to study problems (1.1)
and (1.5), it is important to encode the boundary condition u = 0 in Rn \ in the weak formulation,
by considering also that in the norm uH s (Rn ) , the interaction between and Rn \ gives positive
contribution.
We will introduce the functional space that we will use in the following, which was introduced in [14]
by Servadei and Valdinoci. We introduce this space as follows:

X = u | u : Rn R is measurable, u| L2 (),


and (u(x) u(y)) K(x y) is in L2 (Q, dxdy)
where Q = R2n \(C C) with C = Rn \. The space X is endowed with the norm dened

1/2

uX = uL2 () + |u(x) u(y)|2 K(x y)dx dy .

(1.7)

Then, we dene
X0 = {u X : u = 0 a.e. in Rn \} .
with the norm

uX0 =

(1.8)

1/2
|u(x) u(y)|2 K(x y) dx dy

(1.9)

Concaveconvex nonlinearities

With this norm, X0 is a Hilbert space with scalar product dened as






u, v X0 =
u(x) u(y) v(x) v(y) K(x y) dx dy.

(1.10)

For this, see [14, Lemma 7]. For further details on X and X0 and also for their properties, we refer to
[10], and the references therein.
Remark 1.1. (i) According to conditions of K, by [13, Lemma 11], we know that C02 () X0 , and so,
X and X0 are nonempty. Moreover, X H s () and X0 H s (Rn ), where H s () the usual fractional
Sobolev space endowed with the norm (the so-called Gagliardo norm)

1/2

2
|g(x)

g(y)|
dx dy .
gH s () = gL2 () +
|x y|n+2s

2
s

(ii) The embedding X0  L () is continuous where 2s =

2n
n2s .

Definition 1.1. We say that u is a weak solution of problem (1.1), if u satises




q

u+ (x) + up+ (x) (x)dx
(u(x) u(y)) ((x) (y)) K(x y)dxdy =
Q

(1.11)

for all X0 , where u+ = max{u, 0}.


In the sequel, we will omit the term weak when referring to solutions that satisfy the conditions of
Denition 1.1. In fact, every weak solution of (1.1) is in L () by results of [19, Proposition 9] or [20,
Theorem 3.1].
The fact that u is a weak solution is equivalent to being a critical point of the following functional
J (u) = IK,p (u) I (u),
with
1
IK,p (u) =
2


Q

(1.12)

1
|u(x) u(y)| K(x y)dxdy
p+1
2

and
I (u) =

q+1

up+1
+ (x)dx,

(1.13)

uq+1
+ (x)dx.

(1.14)

We can see that J C 1 (X0 , R) and





J (u), X0 = (u(x) u(y)) ((x) (y)) K(x y)dxdy uq+ (x)(x)dx up+ (x)(x)dx
Q

for any X0 .
Our rst result is about the subcritical case.
Theorem 1.1. Let s (0, 1), n > 2s, and be an open bounded set of Rn with Lipschitz boundary. Let K
be a function satisfying conditions (1.2)(1.4), if 0 < q < 1, 1 < p < 2s , then there exists 1 > 0, such
that for (0, 1 ), problem (1.1) has at least two solutions.

For the critical case, since the embedding X0  L2s (Rn ) (or, in the case of the fractional Laplacian,

H (Rn )  L2s (Rn ) ) is not compact, then the energy functional does not satisfy the PalaisSmale
condition globally, but it is true for the energy functional in a suitable range related to the best fractional
s

W. Chen and S. Deng

ZAMP

critical Sobolev constant in the embedding X0  L2s (Rn ). For this, let us dene the best fractional
critical Sobolev constant SK as

|v(x) v(y)|2 K(x y)dxdy
R2n
SK = inf
for v X0 \{0}.
(1.15)

2/2s
vX0 \{0}
2
s
|v(x)|

Theorem 1.2. Let s (0, 1), n > 2s, and be an open bounded set of Rn with Lipschitz boundary. Let
K be a function satisfying conditions (1.2)(1.4), if 0 < q < 1, p = 2s , assume that
there exists u0 X0 \{0} with u0  0 a.e. in Rn , such that
sup IK,
t0

(tu0 )
2
s

<

n
s 2s
SK ,
n

(1.16)

where IK, 2s is given by (1.13). Then, there exists 2 > 0, such that for (0, 2 ), problem (1.1) has at
least two solutions.
Remark 1.2. When K(x) = |x|(n+2s) , LK is fractional Laplacian ()s , condition (1.16) can be
guaranteed by results of [16, Section 4.2], or [2, Lemma 2.11].
When s = 1, problem (1.5) reduces to a standard semilinear laplace partial dierential equation. More
precisely,

u(x) = uq + up , u > 0 in ;
(1.17)
u = 0,
on ,
n+2
where is a bounded domain in Rn (n > 2), 0 < q < 1 < p < n2
, and > 0 is a parameter.
AmbrosettiBrezisCerami [1] studied problem (1.17). They established that there exists 0 > 0 such
that problem (1.17) attains at least two solutions for (0, 0 ), has a solution for = 0 , and no
solution for > 0 . There are some results for elliptic equation involving concaveconvex nonlinearities,
see for instance [5,6,22,23,25] and the references therein.
This paper is organized as follows. In Sect. 2, we give some notations and preliminaries about Nehari
manifold and bering maps. In Sect. 3, we consider subcritical case for nonlocal elliptic operators LK
and give the proof of Theorem 1.1. In Sect. 4, we deal with the critical case and prove Theorem 1.2.

2. Some preliminary results


We will consider critical points of the function J on the Hilbert space X0 . We denote the Nehari manifold
by
N = {u X0 \{0} : J (u), u X0 = 0} ,
where , X0 is the scalar product on X0 given by (1.8). It is clear that all critical points of J must lie
on N , and N is much smaller set than X0 . So, it is easy to study functional J on N .
It is easy to see that u N if and only if


u2X0 uq+1
(x)dx

up+1
(2.1)
+
+ (x)dx = 0.

The Nehari manifold N is closely linked to the behavior of the function of the form u : t J (tu)
for t > 0 dened by


t2
tq+1
tp+1
u (t) := J (tu) = u2X0
uq+1
(x)dx

up+1
+
+ (x)dx.
2
q+1
p+1

Concaveconvex nonlinearities

Such maps are known as bering maps and were introduced by Drabek and Pohozaev in [9] and are also
discussed by Brown and Zhang [6], Brown and Wu [5].
Lemma 2.1. Let u X0 \{0}, then tu N if and only if u (t) = 0.
Proof. The result is a consequence of the fact that
u (t) = J (tu), u X0 =

1 
J (u), tu X0 .
t


Thus, the elements in N correspond to stationary points of the maps u . We note that


p
u (t) = tu2X0 tq uq+1
(x)dx

t
up+1
+
+ (x)dx,

and
u (t)

u2X0

qt

q1

(2.2)

uq+1
+ (x)dx

pt

p1

up+1
+ (x)dx.

(2.3)

By Lemma 2.1, u N if and only if u (1) = 0. Hence, for u N , by (2.2) we get




p+1

2
u (1) = uX0 q u+ (x)dx p up+1
+ (x)dx

= (1 p)

up+1
+ (x)dx

+ (1 q)

= (1 q)u2X0 (p q)

= (1 p)u2X0 + (p q)

uq+1
+ (x)dx

up+1
+ (x)dx


uq+1
+ (x)dx.

Thus, it is natural to split N into three parts corresponding to local minima, local maxima, and points
of inection, i.e.,

N+
= {u N : u (1) > 0};


N
= {u N : u (1) < 0};

N0 = {u N : u (1) = 0}.

We will prove the existence of solutions of problem (1.1) by investigating the existence of minimizers of
functional J on N . Although N is a small subset of X0 , we will see that local minimizers on Nehari
manifold N are usually critical points of J . We have the following result.
Lemma 2.2. Suppose that u0 is a local minimizer of J on N and u0 N0 . Then, u0 is a critical point
of J .
Proof. The proof is the same as that in BrownZhang [6, Theorem 2.3].

In order to understand the Nehari manifold and bering maps, let us consider the function u : R+
R dened by

up+1
+ dx.

u (t) = t1q u2X0 tpq

(2.4)

W. Chen and S. Deng

ZAMP

Fig. 1. The graphs of u (t) and u (t)

It is clear that for t > 0, tu N if and only if

u (t) =

uq+1
+ dx.

(2.5)

Moreover,
u (t) = (1 q)tq u2X0 (p q)tpq1

up+1
+ dx

and so, we can see that if tu N , then


tq u (t) = u (t).
N+

N
)

(2.6)

(2.7)

u (t)

Hence, tu
(or
if and only if
> 0 (or < 0).
0. From (2.6), u satises the following properties:
Suppose u X0 and u+
1

 p1
(1q)u2X0

(a) u has a unique critical point at t = tmax (u) =
> 0;
(pq) up+1
dx
+
(b) u is strictly increasing on (0, tmax (u)) and strictly decreasing on (tmax (u), +);
(c) lim u (t) = .
t+

Moreover, by the fact uq+1
+ dx > 0, we know that (2.5) has no solution if satises the following
condition
2(pq)

 1q 
 pq 

uXp1
1 q p1
1 q p1
q+1
0
u+ dx > u (tmax (u)) =

(2.8)

1q .
pq
pq
p1
( up+1
+ dx)

Moreover, from (2.2) and (2.5) and if satises (2.8), then u (t) > 0. It seems u (t) < 0 when is large.
Hence, tu
/ N for any t > 0. On the other hand, if satises

0 < uq+1
+ dx < u (tmax (u)),

then there are t1 and t2 with t1 < tmax (u) < t2 , such that

u (t1 ) = u (t2 ) = uq+1
and u (t1 ) > 0,
+ dx,

u (t2 ) < 0,

and (2.2) and (2.5) implies that u (t1 ) = u (t2 ) = 0. By (2.7), we have that u (t1 ) > 0, u (t2 ) < 0.
These facts imply that the bering map u has a local minimum at t1 and a local maximum at t2 such

that t1 u N+
and t2 u N . It can be seen that the graphs of u (t) and u (t) can be seen in Fig. 1.

Concaveconvex nonlinearities

3. The subcritical case: 0 < q < 1 < p < 2s 1


In this section, we consider multiple solutions of (1.1) for the subcritical case.
Lemma 3.1. There exists 1 > 0, such that for any (0, 1 ), we have N0 = .
Proof. Suppose not, that is, N0 = for any > 0. Then, for u N0 , we have J (u), u X0 = 0, and
u (1) = 0. Namely,




p+1
q+1
2
u2X0 = uq+1
ds
+
u
dx,
and
u
=
q
u
ds
+
p
up+1
X0
+
+
+
+ dx.

Then, we have
(1 q)u2X0 = (p q)

up+1
+ dx,

and

(p 1)u2X0 = (p q)

uq+1
+ dx.

(3.1)

By Holder inequality and Remark 1.1 (ii), we get that there exist two positive constants C1 , C2 such that
u2X0  C1 up+1
X0
1

and u2X0  C2 uq+1


X0 .

It yields that C11p  uX0  (C2 ) 1q . This is impossible if is suciently small. Thus, we obtain

that there exists 1 > 0 such that for any (0, 1 ), we have N0 = .
Lemma 3.2. J is coercive and bounded from below on N for (0, 1 ).
Proof. If u N , (1.12) and (2.1) imply that




1
1
1
1

uq+1
J (u) =
u2X0
+ dx.
2 p+1
q+1 p+1

By Holder inequality and Remark 1.1 (ii), we get



q+1
uq+1
+ dx  Cn,q,s,,|| uX0 ,

where Cn,q,s,,|| is a positive constant depending on n, s, q, , ||. Since 0 < q < 1 < p, then we obtain

that the functional J is coercive and bounded from below on N .

By Lemmas 3.1 and 3.2, for any (0, 1 ), we know that N = N+


N and J is coercive and
+

bounded from below on N and N . Therefore, we may dene

+ = inf J (u),
uN+

= inf J (u).
uN

We have the following result.


Proposition 3.1. If 0 < < 1 , then the functional J has a minimizer u1 in N+
and satisfies
(1) J (u1 ) = inf J (u) < 0;
uN+

(2) u1 is a solution of problem (1.1).


+
Proof. Since J is bounded from below on N+
, there exists a minimizing sequence {uk } N such that

lim J (uk ) = inf + J (u).

uN

W. Chen and S. Deng

ZAMP

Thus, by Lemma 3.2, the sequence {uk } is bounded in X0 . From [14, Lemma 7], (X0 ,  X0 ) is a Hilbert
space, and then, there exists u1 X0 such that, up to a subsequence,

(uk (x) uk (y)) ((x) (y)) K(x y)dxdy
Q

(u1 (x) u1 (y)) ((x) (y)) K(x y)dxdy

for X0

as k . Moreover, by [14, Lemma 8], up to a subsequence,


uk u1

in Lr (Rn ),

uk u1 a.e. in Rn ,

(3.2)

as k , and by [4, Theorem IV-9], there exists Lr (Rn ) such that


|uk (x)|  (x) a.e. in Rn
2n
for any 1  r < 2s = n2s
(n > 2s). Therefore, by dominated convergence theorem, we have that




q+1
p+1
(uk )q+1
dx

(u
)
dx,
and
(u
)
dx

(u1 )p+1
1
k
+
+
+
+ dx

as k .
Moreover, there exists t1 such that t1 u1 N+
and J (t1 u1 ) < 0. Hence, we have inf + J (u) < 0.
uN

Next, we show that uk u1 strongly in X0 . If not, then u1 X0 < lim inf uk X0 . Thus, for {uk } N+
,
k
we get



p

(uk )p+1
lim uk (t1 ) = lim t1 uk 2X0 tq1 (uk )q+1
+ dx t1
+ dx
k

> t1 u1 2X0

tq1

(u1 )q+1
+ dx

tp1


(u1 x)p+1
+ dx = u1 (t1 ) = 0.


That is, uk (t1 ) > 0 for k large enough. Since uk = 1 uk N+
, we can see that uk (t) < 0 for t (0, 1)

and uk (1) = 0 for all k. Then, we must have t1 > 1. On the other hand, u1 (t) is decreasing on (0, t1 ),
and so

J (t1 u1 )  J (u1 ) < lim J (uk ) = inf J (u),


k

uN+

which is a contradiction. Hence, uk u1 strongly in X0 . This implies


J (uk ) J (u1 ) = inf + J (u)
uN

as k .

Namely, u1 is a minimizer if J on N+
. Using Lemma 2.2, u1 is a solution to (1.1).
Next, we establish the existence of a local minimum for J on N
.
Proposition 3.2. If 0 < < 1 , then the functional J has a minimizer u2 in N
and satisfies
(1) J (u2 ) = inf J (u) > 0;
uN

(2) u2 is a solution of problem (1.1).

Concaveconvex nonlinearities

Proof. Since J is bounded from below on N


uk } N
, there exists a minimizing sequence {
such that
lim J (
uk ) = inf J (u).

uN

By the same argument given in the proof of Proposition 3.1, there exists u2 X0 such that, up to a
subsequence,

(
uk (x) u
k (y)) ((x) (y)) K(x y)dxdy
Q

(u2 (x) u2 (y)) ((x) (y)) K(x y)dxdy

for X0

as k , and

(
uk )q+1
+ dx

(u2 )q+1
+ dx,


and

(
uk )p+1
+ dx

(u2 )p+1
+ dx

as k . Moreover, from the analysis of bering maps u (t), we know that there exist t1 , t2 with

t1 < tmax (u) < t2 such that t1 u N+


, t2 u N , and J (t1 u)  J (tu)  J (t2 u).
uk X0 . Thus, for {
uk }
Next, we show that u
k u2 strongly in X0 . If not, then u2 X0 < lim inf 
k

N
uk )  J (t
uk ) for all t  tmax (u), and
, we have J (


t22
tq+1
tp+1
q+1
2
2
2
(u2 )+ dx
(u2 )p+1
J (t2 u2 ) = u2 X0
+ dx
2
q+1
p+1

q+1 
p+1 
2
t
t
t
q+1
p+1
uk 2X0 2
(
uk )+ dx 2
(
uk )+ dx
< lim 2 
k
2
q+1
p+1

= lim J (t2 u
k )  J (
uk ) = inf J (u),
k

uN

which is a contradiction. Hence, u


k u2 strongly in X0 . This implies
J (
uk ) J (u2 ) = inf J (u) as k .
uN

Namely, u2 is a minimizer if J on N
. Using Lemma 2.2, u2 is a solution to (1.1).

Proof of Theorem 1.1: By Propositions 3.1 and 3.2 and Lemma 2.2, we get that problem (1.1) has two

solutions u1 N+
and u2 N in X0 . Since N N = , then those two solutions are distinct. This
nishes the proof.


4. The critical case: 0 < q < 1, p = 2s 1


In this section, we consider the multiplicity of solutions to problem (1.1) for the critical case, i.e., p =
n+2s
.
2s 1 = n2s
Lemma 4.1. There exists > 0 such that N0 = for each (0, ).
Proof. The proof is analogous to the proof of Lemma 3.1.

Lemma 4.2. J is coercive and bounded from below on N for (0, ).


Proof. The proof is analogous to the proof of Lemma 3.2.

W. Chen and S. Deng

ZAMP

By Lemmas 4.1 and 4.2, for any (0, ), we know that N = N+


N and J is coercive and
+

bounded from below on N and N . Therefore, we may dene

+ = inf + J (u),

= inf J (u).

uN

uN

We have the following result.


Proposition 4.1. Let {uk } X0 be a (P S)c sequence for J with
c<

2
n
s
s 2s
SK M 2s q
n

then there exists a subsequence of {uk }, which converges strongly in X0 , where SK is defined in (1.15)
and M is a positive constant given by
(2n (n 2s)(q + 1))(1 q)
M=
4(q + 1)

(1 q)(n 2s)
4s

q+1
 2 (q+1)
s

||.

(4.1)

Proof. From Lemma 4.2, we see that {uk } is bounded in X0 . Then, up to a sequence, still denoted by
uk , there exists u X0 such that uk u weakly in X0 , that is

(uk (x) uk (y)) ((x) (y)) K(x y)dxdy
Q


(u (x) u (y)) ((x) (y)) K(x y)dxdy

for X0

as k . Moreover, by [16, Lemma 9], we have that

uk u weakly in L2s (Rn );

uk u in Lr (Rn );

uk u a.e. in Rn ,

as k , and by [4, Theorem IV-9], there exists Lr (Rn ) such that


|uk (x)|  (x) a.e. in Rn
for any 1  r < 2s =

2n
n2s

(n > 2s). Therefore, by dominated convergence theorem, we have that




q+1
(uk )+ dx (u )q+1
+ dx.

By Brezis-Lieb Lemma [21, Lemma 1.32], we get




2
|uk (x) uk (y)|2 K(x y)dxdy |uk (x) u (x) uk (y) + u (y)| K(x y)dxdy
Q

|u (x) u (y)|2 K(x y)dxdy + o(1)

and

((uk u )(x))+s dx +

(uk (x))+s dx =

(u (x))+s dx + o(1)

Concaveconvex nonlinearities

as k . Then,
J (uk ), uk X0

|uk (x) uk (y)| K(x y)dxdy

=
Q

(uk (x))q+1
+ dx

|uk (x) u (x) uk (y) + u (y)| K(x y)dxdy +

=
Q

(uk (x))q+1
+ dx

2
s

((uk u )(x))+ dx +

(uk (x))+s dx

|u (x) u (y)|2 K(x y)dxdy

2
s

(u )+ dx + o(1) + o(1)

|(uk u )(x)(uk u )(y)| K(x y)dxdy

((uk u )(x))+s dx+J (u ), u X0 + o(1).

By J (u ), u X0 = 0 and J (uk ), uk X0 0 as k , we know that


uk

u 2X0

|(uk u )(x) (uk u )(y)| K(x y)dxdy b,

(4.2)

and


((uk u )(x))+s dx b,

as k .

(4.3)

If b = 0, the proof is complete. Assuming b > 0, by the denition of SK in (1.15), we get

uk u 2X0  SK

22
2
s

((uk u )(x))+ dx

Thus, we have b  SK b 2s . That is, b  SK2s . On the other hand, we have

1
1
c = lim J (uk ) = lim uk 2X0
k
k
2
q+1
s n
 J (u ) + SK2s .
n
By the assumption that c <
0<

s 2s
n SK ,

1
(uk (x))q+1
+ dx
p+1

(uk (x))+ dx

(4.4)

we have J (u ) < 0. In particular, u = 0 and

1
1
u 2X0 <
2
2s

(u (x))+s dx +

2
s

1
q+1

(u (x))q+1
+ dx.

W. Chen and S. Deng

Then,

ZAMP



1
c = lim J (uk ) = lim J (uk ) J (uk ), uk X0
k
k
2






1
1
s
s
2
2

((uk u )(x))+s dx +
(u (x))+s dx +
(uk (x))q+1
= lim
+ dx
k
n
n
2 q+1





1
s
1
s
2

= b+
(u (x))+s dx +
(u (x))q+1
+ dx
n
n
2 q+1




n

1
1
s 2s
s
2s

 SK +
(u (x))+ dx +
(u (x))q+1
+ dx.
n
n
2 q+1

Moreover, by H
older inequality, we have


(u (x))q+1
+ dx

 ||

2
s (q+1)
2
s

q+1
2
s

2
(u (x))+s dx

Thus,
c

n
s 2s
S
n K

q+1
2




s
(q+1)
2

s
1
1
s
s n
2
2
s
s
2

+
(u (x))+ dx +
(u (x))+ dx
:= SK2s + h()
||
n
2 q+1
n

where
s
h() = 2s +
n

1
1

2 q+1


||

2
s (q+1)
2
s

(2n (n 2s)(q + 1))(1 q)


4(q + 1)

with =

21
2
s

(u (x))+ dx

We note that h() attains its minimum at 0 =


h(0 ) =

q+1

(1q)(n2s)
4s

(1 q)(n 2s)
4s

1
 2 (q+1)
s

q+1
 2 (q+1)
s

|| 2s and
2
s

2
s

|| 2s (q+1) = M 2s (q+1) .

with M given by (4.1). Therefore,


2
n
s
s 2s
SK M 2s (q+1) .
n
Then, we get a contradiction with our hypothesis. Hence, b = 0 and, we conclude that uk u strongly

in X0 . This completes the proof.

c

Proposition 4.2. There exists 2 > 0 and u0 X0 such that


sup J (tu0 ) <
t>0

2
n
s
s 2s
SK M 2s (q+1)
n

for (0, 2 ). In particular,


2
n
s
s 2s
SK M 2s (q+1)
n
where the constant M depends on q, n, s and ||, which given is by (4.1)

<

(4.5)

Concaveconvex nonlinearities
2
s

Proof. Let > 0 be such that ns SK2s M 2s (q+1) > 0 for all (0, ). By condition (1.16), we have
that there is u0 X0 \{0} such that


n
tq+1
s 2s
tq+1
q+1
0

J (tu0 )  sup IK,2s (tu0 )


(u0 )+ dx < SK
(u0 )q+1
(4.6)
+ dx.
q
+
1
n
q
+
1
t0


Let :=

tq+1
0

(u0 )q+1
+ dx
M (q+1)

2
s (q+1)
q+1

with M is a positive constant dened in (4.1). Then, for

(0, ), we nd

tq+1
0

q+1

2
s

2
s (q+1) .
(u0 )q+1
+ dx < M

Thus, we obtain that (4.5) holds. Finally, we set 2 = min{ , , }; by the analysis of bering maps
u (t) = J (tu), we get

<

2
n
s
s 2s
SK M 2s (q+1)
n

for (0, 2 ). This completes the proof.

Now, we are ready to prove Theorem 1.2.

Proof of Theorem 1.2: By Propositions 4.1 and 4.2, there exists two sequences {u+
k } and {uk } in X0
such that

+ ,
J (u+
k)

J (u+
k)0

and J (u
,
k)

J (u
k)0

as k . We observe that from the analysis of bering maps u (t), we have


+ < 0. Similar to the proof
of Propositions 3.1 and 3.2 and Theorem 1.1, problem (1.1) has two solutions u
1 N+
2 N
and u
in
+

X0 . Since N N = , then these two solutions are distinct. This nishes the proof.


References
1. Ambrosetti, A., Brezis, H., Cerami, G.: Combined eects of concave and convex nonlinearities in some elliptic problems. J.
Funct. Anal. 122, 519543 (1994)
2. Barrios B., Colorado E., Servadei R., Soria, F.: A critical fractional equation with concave-convex power nonlinearities
(2013). http://arxiv.org/abs/1306.3190
3. Br
andle, C., Colorado, E., Pablo, A.de : A concave-convex elliptic problem involving the fractional Laplacian. Proc. R.
Soc. Edinb. Sect. A Math. 143(01), 3971 (2013)
4. Brezis, H.: Analyse fonctionelle. Theorie et applications. Masson, Paris (1983)
5. Brown, K.J., Wu, T.F.: A bering map approach to a semilinear elliptic boundary value problem. Electron. J. Dier.
Equ. 69, 19 (2007)
6. Brown, K.J., Zhang, Y.: The Nehari manifold for a semilinear elliptic equation with a sign-changing weight function. J.
Dier. Equ. 193, 481499 (2003)
7. Caarelli, L., Silvestre, L.: An extension problem related to the fractional Laplacian. Commun. Partial Dier.
Equ. 32(8), 12451260 (2007)
8. Cotsiolis, A., Tavoularis, N.: Best constants for Sobolev inequalities for higher order fractional derivatives. J. Math.
Anal. Appl. 295, 225236 (2004)
9. Drbek, P., Pohozaev, S.I.: Positive solutions for the p-Laplacian: application of the bering method. Proc. R. Soc. Edinb.
Sect. A 127, 703726 (1997)
10. Di Nezza, E., Palatucci, G., Valdinoci, E.: Hitchhikers guide to the fractional Sobolev spaces. Bull. Sci. Math. 136, 521
573 (2012)
11. Fiscella, A.: Saddle point solutions for non-local elliptic operators (2012). http://arxiv.org/abs/1210.8401
12. Servadei, R.: The Yamabe equation in a non-local setting. Adv. Nonlinear Anal. 2(3), 235270 (2013)

W. Chen and S. Deng

ZAMP

13. Servadei, R., Valdinoci, E.: Lewy-Stampacchia type estimates for variational inequalities driven by (non)local operators. Rev. Mat. Iberoam. 29(3), 10911126 (2013)
14. Servadei, R., Valdinoci, E.: Mountain pass solutions for non-local elliptic operators. J. Math. Anal. Appl. 389, 887
898 (2012)
15. Servadei, R., Valdinoci, E.: Variational methods for non-local operators of elliptic type. Discret. Contin. Dyn.
Syst. 5, 21052137 (2013)
16. Servadei, R., Valdinoci, E.: The Brezis-Nirenberg result for the fractional Laplacian. Trans. Am. Math. Soc. 367, 67102
(2015)
17. Servadei, R., Valdinoci, E.: A Brezis-Nirenberg result for non-local critical equations in low dimension. Commun. Pure
Appl. Anal. 12(6), 24452464 (2013)
18. Servadei, R., Valdinoci, E.: Fractional Laplacian equations with critical Sobolev exponent (2012). http://www.math.
utexas.edu/mp arc-bin/mpa?yn=12-58
19. Servadei, R., Valdinoci, E.: Weak and viscosity solutions of the fractional Laplace equation. Publicacions Matemtiques 58(1), 133154 (2014)
20. Wei, Y., Su, X.: Multiplicity of solutions for non-local elliptic equations driven by the fractional Laplacian. Calc. Var.
Partial Dier. Equ. doi:10.1007/s00526-013-0706-5
21. Willem, M.: Minimax Theorems. Birkh
auser, Boston (1996)
22. Wu, T.F.: On semilinear elliptic equations involving concaveconvex nonlinearities and sign-changing weight function. J.
Math. Anal. Appl. 318, 253270 (2006)
23. Wu, T.F.: Multiplicity results for a semi-linear elliptic equation involving sign-changing weight function. Rocky Mt. J.
Math. 39(3), 9951011 (2009)
24. Yu, X.H.: The Nehari manifold for elliptic equation involving the square root of the Laplacian. J. Dier.
Equ. 252(2), 12831308 (2012)
25. Zhang, J.G., Liu, X.C.: The Nehari manifold for a semilinear elliptic problem with the nonlinear boundary condition. J.
Math. Anal. Appl. 400(1), 100119 (2013)
Wenjing Chen and Shengbing Deng
School of Mathematics and Statistics
Southwest University
Chongqing 400715
Peoples Republic of China
e-mail: wjchen1102@gmail.com
Shengbing Deng
Departamento de Matem
atica
Ponticia Universidad Catolica de Chile
Avda. Vicu
na Mackenna 4860
Macul, Chile
e-mail: shbdeng65@gmail.com

(Received: February 3, 2014; revised: October 8, 2014)

Вам также может понравиться