Contents lists available atScienceDirect
Journal of Mathematical Analysis and Applications
www.elsevier.com/locate/jmaa
Multiple positive solutions for a Dirichlet problem involving critical Sobolev exponent
Tiexiang Li
a, Tsung-fang Wu
b,∗
aDepartment of Mathematics, Southeast University, Nanjing 211189, PR China
bDepartment of Applied Mathematics, National University of Kaohsiung, Kaohsiung 811, Taiwan
a r t i c l e i n f o a b s t r a c t
Article history:
Received 30 September 2009 Available online 16 March 2010 Submitted by J. Wei
Keywords:
Ljusternik–Schnirelmann category Multiple positive solutions Critical Sobolev exponent Nehari manifold
In this paper, we study the decomposition of the Nehari manifold via the combination of concave and convex nonlinearities. Furthermore, we use this result and the Ljusternik–
Schnirelmann category to prove that the existence of multiple positive solutions for a Dirichlet problem involving critical Sobolev exponent.
©2010 Elsevier Inc. All rights reserved.
1. Introduction
In this paper, we study the multiplicity of positive solutions for the following semilinear elliptic problem:
−
u=
fλ(
x)|
u|
q−2u+
g(
x)|
u|
p−2u inΩ,
u
=
0 in∂Ω,
(Eλ)where 1
<
q<
2<
p2∗=
N2N−2 (N3),Ω ⊂
RN is a bounded domain with smooth boundary, the parameterλ >
0 and the weight functions fλ= λ
f++
f−(f±= ±
max{
f,
0}
),g are continuous onΩ
which satisfy the following condition:(
Q)
there exist a non-empty closed setM= {
x∈ Ω |
g(
x) =
maxx∈Ωg(
x) ≡
1}
andρ >
N−
2 such that M⊂
x
∈ Ω
f(
x) >
0and
g
(
z) −
g(
x) =
O|
x−
z|
ρasx
→
zand uniformly inz∈
M.
Remark 1.1.LetMr
= {
x∈
RN|
dist(
x,
M) <
r}
forr>
0. Then by the condition(
Q)
, we may assume that there exist two positive constantsC0andr0such thatf
(
x),
g(
x) >
0 for allx∈
Mr0⊂ Ω
*
Corresponding author.E-mail addresses:txli@seu.edu.cn(T. Li),tfwu@nuk.edu.tw(T.-f. Wu).
0022-247X/$ – see front matter ©2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jmaa.2010.03.022
and
g
(
z) −
g(
x)
C0|
x−
z|
ρ for allx∈
BN(
z,
r0)
and for allz∈
M.
That the existence and multiplicity of positive solutions of Eq.
(
Eλ)
may be influenced by the concave and convex nonlinearities. These issues have been the focus of a great deal of research in recent years. When the weight functions fλ≡ λ
and g≡
1, Ambrosetti, Brezis and Cerami [1] has proved that there existsλ
0>
0 such that Eq.(
Eλ)
admits at least two positive solutions forλ ∈ (
0, λ
0)
, a positive solution forλ = λ
0 and no positive solution forλ > λ
0. Actually, Adimurthy, Pacella and Yadava [2], Ouyang and Shi [20] and Tang [22] proved that there existsλ
0>
0 such that Eq.(
Eλ)
in unit ball BN(
0;
1)
has exactly two positive solutions for allλ ∈ (
0, λ
0)
, has exactly one positive solution forλ = λ
0 and no positive solution for allλ > λ
0. When the weight functions f,
g are sign-changing, Brown and Wu [6,7], de Figueiredo, Gossez and Ubilla [14], Hsu and Lin [15] and Wu [25,26] they proved that Eq.(
Eλ)
has at least two positive solutions whenλ
is sufficiently small. Furthermore, if p=
2∗ and f,
g∈
C(Ω)
satisfy the following conditions:(
D1)
BN(
2δ
0) \
BN(δ
0) ⊂ Ω
for someδ
0>
0, where BN(
r) = {
x∈
RN| |
x| <
r}
;(
D2)
f=
f+−
f−, where f±: Ω →
Rare continuous functions and there exists a domain BN(
2δ
0) \
BN(δ
0) ⊂ Θ ⊂ Ω
of classC1such that for allx∈ Θ
, f−(
x) =
0, f+(
x) >
0 and for allx∈ Ω\Θ
, f−(
x)
0, f+(
x) =
0;(
D3)
theN-ball BN(
2δ
0) ⊂ Ω
such that for allx∈
BN(δ
0)
, 0<
g(
x) <
1 and for allx∈ Ω\
BN(
2δ
0)
, 0g(
x) <
1;(
D4)
g(
x) =
0 for allx∈
BN(
2δ
0) \
BN(δ
0)
,then Wu [27] proved that Eq.
(
Eλ)
has at least three positive solutions whenλ
is sufficiently small.The main purpose of this paper is to consider the relation between the number of solutions and the topology of the global maxima set of weight function g. Our main result improves a recent multiplicity result of [15,25,27]. The following theorem is our main result.
Theorem 1.1.Suppose that p
=
2∗. Then for eachδ <
r0there existsΛ
δ>
0such that forλ < Λ
δ, Eq.(
Eλ)
has at least catMδ(
M) +
1 positive solutions.Hereaftercatis the Ljusternik–Schnirelmann category (see e.g. [17]).
In the following sections, we proceed to proof Theorem 1.1. We use the variational methods to find positive solutions of Eq.
(
Eλ)
. Associated with Eq.(
Eλ)
, we consider the energy functional JλinH01(Ω)
,Jλ
(
u) =
12
u2H1−
1 qΩ
fλ
|
u|
qdx−
1 2∗Ω
g
|
u|
2∗dxwhere
uH1= (
Ω|∇
u|
2dx)
1/2 is the standard norm in H01(Ω)
. It is well known that the solutions of Eq.(
Eλ)
are the critical points of the energy functional JλinH01(Ω)
.This paper is organized as follows. In Section 2, we give some notations and preliminaries. In Section 3, we discuss some concentration behavior. In Section 4, we prove Theorem 1.1.
2. Notations and preliminaries
Throughout this paper, we denote by S the best Sobolev constant for the embedding ofD1,2
(
RN)
intoL2∗(
RN)
which is given byS
=
infu∈D1,2(RN)\{0}
RN
|∇
u|
2dx(
RN|
u|
2∗dx)
2/2∗>
0.
(2.1)It is well known that Sis independent of
Ω ⊂
RN in the sense that if S(Ω) =
infu∈H10(Ω)\{0}
Ω
|∇
u|
2dx(
Ω|
u|
2∗dx)
2/2∗>
0,
then S
(Ω) =
S(
RN) =
S, and the function uε(
x) = ε
(N−2)/2( ε
2+ |
x|
2)
(N−2)/2, ε >
0 andx∈ R
N,
is an extremal function for the minimum problem
(
2.
1)
. Moreover, for eachε >
0, vε(
x) = [
N(
N−
2) ε
2]
(N−2)/4( ε
2+ |
x|
2)
(N−2)/2 (2.2)which is a positive solution of critical problem:
−
u= |
u|
2∗−2u inR
N (2.3)with RN
|∇
vε|
2dx=
RN|
vε|
2∗dx=
SN2.We define the Palais–Smale (simply by (PS)) sequences, (PS)-values, and (PS)-conditions in H10
(Ω)
for Jλas follows.Definition 2.1.
(i) For
β ∈
R, a sequence{
un}
is a(
PS)
β-sequence in H10(Ω)
for Jλif Jλ(
un) = β +
o(
1)
; and Jλ(
un) =
o(
1)
; strongly in H−1(Ω)
asn→ ∞
.(ii) Jλsatisfies the
(
PS)
β-condition inH01(Ω)
if every(
PS)
β-sequence in H01(Ω)
for Jλcontains a convergent subsequence.As the energy functional Jλ is not bounded below on H10
(Ω)
, it is useful to consider the functional on the Nehari manifoldNλ
=
u
∈
H10(Ω) \{
0}
Jλ(
u),
u=
0.
Thus,u
∈
Nλif and only if u2H1−
Ω
fλ
|
u|
qdx−
Ω
g
|
u|
2∗dx=
0.
Moreover, we have the following results.
Lemma 2.2.The energy functional Jλis coercive and bounded below onNλ.
Proof. Ifu
∈
Nλ, then by the Young and Sobolev inequalities Jλ(
u) =
1N
u2H1−
1q
−
1 2∗Ω
fλ
|
u|
qdx (2.4) 1N
u2H1− λ
2∗
−
qq2∗
f+Lq∗S−q2uqH1 (2.5) 1N
u2H1−
1N
u2H1−
Dλ
2/(2−q)= −
D0λ
2/(2−q),
(2.6)whereq∗
=
2∗2−∗q andD is a positive constant depending onq,
N,
S andf+Lq∗. Thus, Jλis coercive and bounded below onNλ. 2Define
ψ
λ(
u) =
Jλ
(
u),
u=
u2H1−
Ω
fλ
|
u|
qdx−
Ω
g
|
u|
2∗dx.
Then foru
∈
Nλ,ψ
λ(
u),
u=
2u2H1−
qΩ
fλ
|
u|
qdx−
2∗Ω
g
|
u|
2∗dx= −
4N
−
2u2H1−
q−
2∗Ω
fλ
|
u|
qdx (2.7)= (
2−
q)
u2H1−
2∗−
qΩ
g
|
u|
2∗dx.
(2.8)Similarly to the method used in [26], we splitNλinto three parts:
N+λ
=
u
∈
Nλψ
λ(
u),
u>
0;
N0λ=
u
∈
Nλψ
λ(
u),
u=
0;
N−λ=
u
∈
Nλψ
λ(
u),
u<
0.
We now derive some basic properties ofN+λ,N0λandN−λ.
Lemma 2.3.Suppose that u0is a local minimizer for JλonNλand that u0
∈ /
N0λ. Then Jλ(
u0) =
0in H−1(Ω)
. Furthermore, if u0is a non-trivial nonnegative function inΩ
, then u0is a positive solution of Eq.(
Eλ)
.Proof. Similarly the argument in [8, Theorem 2.3] (or see [3]), we have Jλ
(
u0) =
0 in H−1(Ω)
. This implies u0 is a weak solution of Eq.(
Eλ)
. Now, ifu0 is a non-trivial nonnegative function inΩ
, then by [13, Lemma 2.1], we haveu0∈
L∞(Ω)
. Moreover, we can apply the Harnack inequality due to [23] in order to getu0 is positive inΩ
. 2Lemma 2.4.For each
λ >
0we have the following:(i) For any u
∈
N+λ, we have Ω fλ|
u|
qdx>
0.(ii) For any u
∈
N0λ, we have Ω fλ|
u|
qdx>
0and Ωg|
u|
2∗dx>
0.(iii) For any u
∈
N−λ, we have Ωg|
u|
2∗dx>
0.Proof. The result now follows immediately from
(
2.
7)
and(
2.
8)
. 2Let
Λ
1=
SN(2−q) 4 +q
2
f+
Lq∗
(
22∗−−qq)
(N−2)(2−q)4(
22∗∗−−2q)
. Then by an argument similar to that the proof Lemma 2.1 in [26], we have N0λ= ∅
for allλ ∈ (
0, Λ
1)
. Thus, we can writeNλ=
N+λ∪
N−λ and defineα
λ+=
infu∈N+λ Jλ
(
u)
andα
λ−=
infu∈N−λ Jλ
(
u).
Moreover, we have the following result, whose proof can be found in [27, Theorem 3.1].
Theorem 2.5.We have the following:
(i)
α
+λ<
0for allλ ∈ (
0, Λ
1)
.(ii) If
λ < Λ
2=
q2Λ
1, thenα
−λ>
c0for some c0>
0.In particular,
α
+λ=
infu∈Nλ Jλ(
u)
for allλ ∈ (
0, Λ
2)
.For each u
∈
H10(Ω) \{
0}
with Ωg|
u|
2∗dx>
0, we writetmax
=
(
2−
q)
u2H1(
2∗−
q)
Ωg|
u|
2∗dx N−42>
0.
Then we have the following lemma.
Lemma 2.6.For each u
∈
H10(Ω) \{
0}
we have the following:(i) If Ω fλ
|
u|
qdx0, then there is a unique t−=
t−(
u) >
tmaxsuch that t−u∈
N−λ and huis increasing on(
0,
t−)
and decreasing on(
t−, ∞ )
. Moreover,Jλ
t−u=
supt0 Jλ
(
tu).
(2.9)(ii) If Ω fλ
|
u|
qdx>
0, then there are unique0<
t+=
t+(
u) <
tmax<
t− such that t+u∈
N+λ, t−u∈
N−λ, hu is decreasing on(
0,
t+)
, increasing on(
t+,
t−)
and decreasing on(
t−, ∞)
. Moreover,Jλ
t+u=
inf0ttmax
Jλ
(
tu) ;
Jλ t−u=
suptt+
Jλ
(
tu).
(2.10)Proof. The proof is almost the same as the in [26, Lemma 2.4]. 2 Forc
>
0, we defineJ0c
(
u) =
12
u2H1−
c 2∗Ω
g
|
u|
2∗dx;
Nc0=
u
∈
H10(Ω)\{
0}
Jc0(
u),
u=
0.
Note thatNc0
=
N0=
N−0 forc=
1 and for eachu∈
N−λ there is a uniquetu>
0 such thattuu∈
N0. Moreover, we have the following result, whose proof can be found in [25, Lemma 5.2].Lemma 2.7.Let q∗
=
2∗2−∗q. Then for each u∈
N−λ we have the following:(i) There is a unique tc
(
u) >
0such that tc(
u)
u∈
Nc0andsup
t0 Jc0
(
tu) =
J0c tc(
u)
u=
1 Nu2H∗1
c Ωg
|
u|
2∗dx N−22.
(ii)
Jλ
(
u) (
1− λ)
N2 J0(
tuu) − λ(
2−
q)
2qf+Lq∗S−2q2−2q
and
Jλ
(
u) (
1+ λ)
N2 J0(
tuu) + λ(
2−
q)
2qf+Lq∗S−2q2−2q
.
3. Concentration behavior
First, we consider the following critical problem:
−
u= |
u|
2∗−2u inΩ,
u
∈
H10(Ω).
( E0)Associated with Eq.
(
E0)
, we consider the energy functional J∞ inH10(Ω)
, J∞(
u) =
12
u2H1−
1 2∗Ω
|
u|
2∗dx.
It is well known that inf
u∈N∞(RN)
J∞
(
u) =
infu∈N∞(Ω)J∞
(
u) =
1NSN2 for all domain
Ω ⊂ R
N,
where N∞
R
N=
u
∈
D1,2R
N\{
0}
J∞(
u),
u=
0and
N∞
(Ω) =
u
∈
H10(Ω)\{
0} (
u),
u=
0are the Nehari manifolds. Actually, infu∈N∞(Ω)J∞
(
u)
is never attained on a domainΩ
RN. Following the method of [4]and Remark 1.1, let
η ∈
C0∞(
RN)
be a radially symmetric function such that 0η
1,|∇ η |
C andη (
x) =
1, |
x|
r0/
2,
0
, |
x|
r0.
For anyz
∈
M, letwε,z
(
x) = η (
x−
z)
vε(
x−
z),
(3.1)wherevε
(
x)
as in(
2.
2)
. Then, by an argument similar to that the proof Lemma 4.2 in [16] (or see Struwe [21]), we have wε,z2H1=
SN2+
Oε
N−2uniformly inz
∈
M.
(3.2)Moreover, we have the following results.
Lemma 3.1.We have uinf∈N0
J0
(
u) =
infu∈N∞(Ω)J∞
(
u) =
1 NSN2.
Furthermore, Eq.
(
E0)
does not admit any positive solution u0such that J0(
u0) =
N1SN2.Proof. Letwε,zbe as in
(
3.
1)
and define g:
RN→
Rby g(
x) =
g(
x),
ifx∈ Ω,
0
,
ifx∈ Ω
c,
as an extension of g. Then, by Lemma 2.6, there is a uniquet0
(
wε,z) >
0 such thatt0(
wε,z)
wε,z∈
N0for allε >
0, that is t0(
wε,z)
wε,z2H1
=
Ω
g
t0(
wε,z)
wε,z2∗dxand so
t0(
wε,z)
4/(N−2)=
Ωg|
wε,z|
2∗dx wε,z2H1.
(3.3)By the definition ofvε, we get that
Ω
g
|
wε,z|
2∗dx=
BN(z,2r0)
g
(
x) η (
x−
z)
vε(
x−
z)
2∗dx=
RN
[
N(
N−
2) ε
2]
N/2g(
x+
z) η
2∗(
x) ( ε
2+ |
x|
2)
N dx.
Thus, by the condition
(
Q)
,0
1[
N(
N−
2) ε
2]
N/2RN
|
vε|
2∗dx−
Ω
g
|
wε,z|
2∗dx=
RN\BN(0,r0/2)
[
1−
g(
x+
z) η
2∗(
x)]
( ε
2+ |
x|
2)
N dx+
BN(0,r0/2)
[
1−
g(
x+
z)]
( ε
2+ |
x|
2)
N dxRN\BN(0,r0/2) 1
|
x|
2Ndx+
C0BN(0,r0/2)
|
x|
ρ( ε
2+ |
x|
2)
Ndx Nω
N ∞r0/2
r−(N+1)dr
+
C0Nω
Nε
2r0/2
0
rρ−N+1dr
= ω
N r0 2 −N+
C0Nω
Nε
2( ρ − (
N−
2))
r02
ρ−(N−2)=
C1+
C2ε
2 for allz∈
M,
(3.4)where
ω
N is the volume of the unit ball BN(
0,
1)
inRN. Then εlim→0Ω
g
|
wε,z|
2∗dx=
SN2 uniformly inz∈
M.
(3.5)From the condition
(
Q)
and the same argument of(
3.
5)
, we have εlim→0Ω
f−
|
wε,z|
2∗dx=
0 uniformly inz∈
M.
(3.6)By
(
3.
2)
,(
3.
5)
and(
3.
6)
,εlim→0t0
(
wε,z) =
1 and limε→0
wε,z2H1=
SN2 uniformly inz∈
M.
Thus, uinf∈N0
J0
(
u)
J0 t0(
wε,z)
wε,z→
1NSN2 as
ε →
0and so inf
u∈N0
J0
(
u)
infu∈N∞(Ω)J∞
(
u) =
1 NSN2.
Letu
∈
N0. Then, by Lemma 2.6(i), J0(
u) =
supt0 J0(
tu)
. Moreover, there is a uniquetu>
0 such thattuu∈
N∞(Ω)
. Thus, J0(
u)
J0(
tuu)
J∞(
tuu)
1NSN2
.
This implies infu∈N0 J0
(
u)
1NSN2. Therefore, uinf∈N0J0
(
u) =
infu∈N∞(Ω)J∞
(
u) =
1 NSN2.
Next, we will show that Eq.
(
E0)
does not admit any solution u0 such that J0(
u0) =
infu∈N0 J0(
u)
. Suppose the contrary.Then we can assume that there existsu0
∈
N0 such that J0(
u0) =
infu∈N0 J0(
u)
. Since J0(
u0) =
J0(|
u0|)
and|
u0| ∈
N0, by Lemma 2.3, we may assume thatu0 is a positive solution of Eq.(
E0)
. Moreover, by Lemma 2.6(i), J0(
u0) =
supt0 J0(
tu0)
. Thus, there is a uniquetu0>
0 such thattu0u0∈
N∞(Ω)
and so1
NSN2
=
infu∈N0
J0
(
u) =
J0(
u0)
J0(
tu0u0)
=
J∞(
tu0u0) +
t2∗ u0
2∗
Ω
(
1−
g)|
u0|
2∗dx 1NSN2
+
t2∗ u0
2∗
Ω
(
1−
g) |
u0|
2∗dx.
This implies Ω
(
1−
g) |
u0|
2∗dx=
0, which is a contradiction. This completes the proof. 2 Lemma 3.2.Suppose that{
un}
is a minimizing sequence for J0inN0. Then(i) Ω f−
|
un|
qdx=
o(
1)
; (ii) Ω(
1−
g)|
un|
2∗dx=
o(
1)
. Furthermore,{
un}
is a(
PS)
1NSN2-sequence for J∞in H01
(Ω)
.Proof. For eachn, there is a uniquetn
>
0 such thattnun∈
N∞(Ω)
, that is tn2Ω
|∇
un|
2dx=
t2n∗Ω
|
un|
2∗dx.
Then, by Lemma 2.6(i),
J0
(
un)
J0(
tnun) =
J∞(
tnun) −
tq n
q
Ω
f−
|
un|
qdx+
t2∗ n
2∗
Ω
(
1−
g) |
un|
2∗dx 1NSN2
−
tq n
q
Ω
f−
|
un|
qdx+
t2∗ n
2∗
Ω
(
1−
g) |
un|
2∗dx.
Since J0
(
un) =
1NSN2+
o(
1)
from Lemma 3.1, we have tnqq
Ω
f−
|
un|
qdx=
o(
1)
and tn2∗
2∗
Ω
(
1−
g)|
un|
2∗dx=
o(
1).
We will show that there exists c0
>
0 such thattn>
c0 for alln. Suppose the contrary. Then we may assume tn→
0 as n→ ∞
. SinceJ0
(
un) =
1NSN2
+
o(
1)
and
J0
(
un) =
1N
un2H1+
o(
1),
by Lemma 2.2,
unis uniformly bounded and sotnunH1→
0 or J∞(
tnun) →
0 and this contradicts J∞(
tnun)
N1SN2>
0.Thus,
Ω
f−
|
un|
qdx=
o(
1)
and
Ω
(
1−
g) |
un|
2∗dx=
o(
1),
this implies
Ω
|∇
un|
2dx=
Ω
|
un|
2∗dx+
o(
1)
and
J∞
(
un) =
1NSN2
+
o(
1).
Similar to the method used in [24, Lemma 7], we have
{
un}
is a(
PS)
1NSN2-sequence for J∞in H10
(Ω)
. 2 For the positive numbersd, consider the filtration of the Nehari manifoldN0 as follows:N0
(
d) =
u
∈
N0J0
(
u)
1 NSN2+
d.
Let
Φ :
H10(Ω) →
RN be a barycenter map defined byΦ(
u) =
Ωx|
u|
2∗dxΩ
|
u|
2∗dx.
(3.7)Such a map has been constructed in Bartsch and Weth [5], Coron [10], Cerami and Passaseo [11] and Clapp and Weth [12], etc. Then we have the following result.
Lemma 3.3.For each positive number
δ <
r0, there exists dδ>
0such thatΦ(
u) ∈
Mδ for all u∈
N0(
dδ).
Proof. Suppose the contrary. Then we can assume that there exist a sequence
{
un} ∈
N0 andδ
0<
r0 such that J0(
un) =
1
NSN2
+
o(
1)
andΦ(
un) / ∈
Mδ0 for alln.
(3.8)Then, by Lemma 3.2, we have
{
un}
is a(
PS)
1NSN2-sequence for J∞ inH10
(Ω)
. It follows from Lemma 2.2 that there exist a subsequence{
un}
andu0∈
H10(Ω)
such thatunu0weakly inH10
(Ω)
. SinceΩ
is a bounded domain, by the concentration- compactness principle (see Lions [18,19] or Struwe [21, Theorem 3.1]), there exist two sequences{
xn} ⊂ Ω
,{
Rn} ⊂
R+, x0∈ Ω
and a positive solutionv0∈
D1,2(R
N)
of critical problem(
2.
3)
with J∞(
v0) =
N1SN2 such thatxn→
x0 andRn→ ∞
asn→ ∞
and un(
x) −
RnN−22v0Rn
(
x−
xn)
D1,2(RN)