Open Access

Existence of solutions for semilinear elliptic equations on R N

Boundary Value Problems20132013:163

DOI: 10.1186/1687-2770-2013-163

Received: 28 March 2013

Accepted: 8 June 2013

Published: 9 July 2013

Abstract

In this paper, the existence of at least one nontrivial solution for a class of semilinear elliptic equations on R N is established by using the linking methods.

Keywords

Schrödinger equation subcritical exponent local linking

1 Introduction

In this paper we consider the question of the existence of solutions for a class of semilinear equations of the form
( P λ ) u + λ u = g ( x , u ) , x R N ,
where λ > 0 is a parameter and the nonlinearity g C ( R N × R ) is asymptotically linear, i.e.,
lim | t | g ( x , t ) t = V ( x ) , lim | x | V ( x ) = v
(1.1)
for some V ( x ) C ( R N , R ) and v R . In case this equation is considered in a bounded domain Ω R N (with, say, the Dirichlet boundary condition), there is a large amount of literature on existence and multiplicity results, with the case of resonance being of particular interest (see [13]). We recall that the problem is said to be at resonance if λ σ ( S ) , where σ ( S ) denotes the spectrum of S, the ‘asymptotic linearization’ of the problem. In other words, S : D ( S ) L 2 ( Ω ) L 2 ( Ω ) is the operator given by
S u ( x ) = u ( x ) V ( x ) u ( x ) , D ( S ) = H 0 1 ( Ω ) H 2 ( Ω ) .
(1.2)

On the other hand, a systematic study of such asymptotically linear problems set in unbounded domains or the whole space R N is more recent and presents a number of mathematical difficulties (see [4, 5]). As an example, we note that in the case of problem ( P λ ), the asymptotic linearization operator S (now defined on D ( S ) = H 2 ( R N ) ) has a much more complicated spectrum (including an essential part [ v , ) ), which in turn makes the study of this problem more challenging. In [4], motivated by the paper [5], Tehrani and Costa studied the existence of positive solutions to ( P λ ) by using the mountain pass theorem if g ( x , u ) satisfies some strong asymptotically linear conditions. Comparing with previous paper [4], in [6], Tehrani obtained the existence of a (possibly sign-changing) solution for problem ( P λ ) under essentially condition (1.1) only. In fact, he proved the following.

Theorem 1.0 [6]

Let g 0 ( x , s ) : = g ( x , s ) V ( x ) s and assume that
  1. (G)
    for every ϵ > 0 , there exists 0 b ϵ ( x ) L 2 ( R N ) such that
    | g 0 ( x , s ) | b ϵ ( x ) + ϵ | s | a.e. x R N , s R .
     

If Λ 0 or max { 0 , v } < λ < Λ and λ σ p ( S ) , then ( P λ ) has a solution in H 1 ( R N ) .

Now, one naturally asks: Are there nontrivial solutions for problem ( P λ ) if λ σ ( S ) in the above theorem? Obviously, this case is resonance. But, this problem is not easy because we face the difficulties of verifying that the energy functional satisfies the (PS) condition if we still follow the idea of [6]. Here, there is still an interesting problem: Are there nontrivial solutions for problem ( P λ ) if λ σ ( S ) and g 0 ( x , s ) (in Theorem 1.0) is more generalized superlinear? We will answer the above problems affirmatively by using Li and Willem’s local linking methods (see [7]).

Next, we recall a few basic facts in the theory of Schrödinger operators which are relevant to our discussion (see [6]).
  1. 1.

    Since lim | x | V ( x ) = v , one has σ ess ( S ) = [ v , ) .

     
  2. 2.
    The bottom of the spectrum σ ( S ) of the operator S is given by
    Λ = λ 0 = inf 0 u H 2 ( R N ) | u | 2 V ( x ) u 2 u 2 .
     
Therefore we clearly have Λ v . If Λ < v , then by using the concentration compactness principle of Lions, one shows that Λ is the principle eigenvalue of S with a positive eigenfunction Φ 0 :
S Φ 0 = λ 0 Φ 0 , Φ 0 H 2 ( R N ) , Φ 0 > 0 .
  1. 3.
    The spectrum of S in ( , v ) , namely σ ( S ) ( , v ) , is at most a countable set, which we denote by
    Λ = λ 0 < λ 1 < λ 2 < ,
     

where each λ k is an isolated eigenvalue of S of the finite multiplicity. Let E λ j denote the eigenspace of S corresponding to the eigenvalue λ j .

Now, we state our main results. In this paper, we always assume that lim | x | V ( x ) = v and v < 0 . The conditions imposed on g 0 ( x , t ) (see Theorem 1.0) are as follows:

(H1) g 0 C ( R N × R , R ) , and there are constants C 1 , C 2 0 such that
| g 0 ( x , t ) | C 1 + C 2 | t | s 1 , x R N , t R , s ( 2 , p ) ( N 3 ) ,

where p = 2 N N 2 ;

(H2) g 0 ( x , t ) = ( | t | ) , | t | 0 , uniformly on R N ;

(H3) lim | t | g 0 ( x , t ) t = + uniformly on R N ;

(H4) There is a constant θ 1 such that for all ( x , t ) R N × R and s [ 0 , 1 ] ,
θ ( g 0 ( x , t ) t 2 G 0 ( x , t ) ) ( s g 0 ( x , s t ) t 2 G 0 ( x , s t ) ) ,

where G 0 ( x , t ) = 0 t g 0 ( x , s ) d s ;

(H5) For some δ > 0 , either
G 0 ( x , t ) 0 for  | t | δ , x R N
or
G 0 ( x , t ) 0 for  | t | δ , x R N ;

(H6) lim | x | sup | t | r g 0 ( x , t ) | t | = 0 for every r > 0 .

Theorem 1.1 Assume that conditions (H1)-(H4) hold. Ifλ is an eigenvalue of S ( λ < v ) , assume also that (H5) and (H6) hold. Then the problem ( P λ ) has at least one nontrivial solution.

Remark 1.1 It follows from the condition (H3) that our nonlinearity g 0 ( x , t ) does not satisfy the classical condition of Ambrosetti and Rabinowitz:

(AR) There is μ > 2 such that 0 < μ G 0 ( x , u ) u g 0 ( x , u ) for all x R N and u 0 .

In recent years, there have been some papers devoted to replacing (AR) with more natural conditions (see [810]). But our methods are different from the references therein.

We also consider asymptotically quadratic functions. We assume that:

(H7) For every ϵ > 0 , there exists 0 b ϵ ( x ) L 2 ( R N ) such that
| g 0 ( x , s ) | b ϵ ( x ) + ϵ | s | a.e.  x R N , s R ,

and λ k < λ < λ k + 1 .

Theorem 1.2 Assume that conditions (H2), (H6), (H7) and one of the following conditions hold:

(A1) λ j < 0 < λ j + 1 , j k ;

(A2) λ j = 0 < λ j + 1 , j k for some δ > 0 ,
G 0 ( x , u ) 0 for | u | > δ , x R N ;
(A3) λ j < 0 = λ j + 1 , j k for some δ > 0 ,
G 0 ( x , u ) 0 for | u | δ , x R N .

Then problem ( P λ ) has at least one nontrivial solution.

2 Preliminaries

Let X be a Banach space with a direct sum decomposition
X = X 1 X 2 .
Consider two sequences of subspaces
X 0 1 X 1 1 X 1 , X 0 2 X 1 2 X 2
such that
X j = n N X n j , j = 1 , 2 .
For every multi-index α = ( α 1 , α 2 ) N 2 , let X α = X α 1 X α 2 . We know that
α β α 1 β 1 , α 2 β 2 .

A sequence ( α n ) N 2 is admissible if, for every α N 2 , there is m N such that n m α n α . For every I : X R , we denote by I α the function I restricted X α .

Definition 2.1 Let I be locally Lipschitz on X and c R . The functional I satisfies the ( C ) c condition if every sequence ( u α n ) such that ( α n ) is admissible and
u α n X α n , I ( u α n ) c , ( 1 + u α n ) I ( u α n ) 0

contains a subsequence which converges to a critical point of I.

Definition 2.2 Let I be locally Lipschitz on X and c R . The functional I satisfies the ( C ) condition if every sequence ( u α n ) such that ( α n ) is admissible and
u α n X α n , sup n I ( u α n ) c , ( 1 + u α n ) I ( u α n ) 0

contains a subsequence which converges to a critical point of I.

Remark 2.1 1. The ( C ) condition implies the ( C ) c condition for every c R .
  1. 2.

    When the ( C ) c sequence is bounded, then the sequence is a ( PS ) c sequence (see [11]).

     
  2. 3.
    Without loss of generality, we assume that the norm in X satisfies
    u 1 + u 2 2 = u 1 2 + u 2 2 , u j X j , j = 1 , 2 .
     
Definition 2.3 Let X be a Banach space with a direct sum decomposition
X = X 1 X 2 .
The function I C 1 ( X , R ) has a local linking at 0, with respect to ( X 1 , X 2 ) if, for some r > 0 ,
I ( u ) 0 , u X 1 , u r , I ( u ) 0 , u X 2 , u r .

Lemma 2.1 (see [7])

Suppose that I C 1 ( X , R ) satisfies the following assumptions:

(B1) I has a local linking at 0 and X 1 { 0 } ;

(B2) I satisfies ( PS ) ;

(B3) I maps bounded sets into bounded sets;

(B4) For every m N , I ( u ) , u , u X = X m 1 X 2 . Then I has at least two critical points.

Remark 2.2 Assume that I satisfies the ( C ) c condition. Then this theorem still holds.

Let X be a real Hilbert space and let I C 1 ( X , R ) . The gradient of I has the form
I ( u ) = A u + B ( u ) ,

where A is a bounded self-adjoint operator, 0 is not the essential spectrum of A, and B is a nonlinear compact mapping.

We assume that there exist an orthogonal decomposition,
X = X 1 + X 2 ,
and two sequences of finite-dimensional subspaces,
X 0 1 X 1 1 X 1 1 X 1 , X 0 2 X 1 2 X 2 ,
such that
X j = n N X n j ¯ , j = 1 , 2 , A X n j X n j , j = 1 , 2 , n N .
For every multi-index α = ( α 1 , α 2 ) N 2 , we denote by X α the space
X α 1 X α 2 ,

by p α : X X α the orthogonal projector onto X α , and by M ( L ) the Morse index of a self-adjoint operator L.

Lemma 2.2 (see [7])

I satisfies the following assumptions:
  1. (i)

    I has a local linking at 0 with respect to ( X 1 , X 2 ) ;

     
  2. (ii)
    There exists a compact self-adjoint operator B such that
    B ( u ) = B ( u ) + ( u ) , u ;
     
  3. (iii)

    A + B is invertible;

     
  4. (vi)
    For infinitely many multiple-indices α : = ( n , n ) ,
    M ( ( A + P α B ) | X α ) dim X n 2 .
     

Then I has at least two critical points.

3 The proof of main results

Proof of Theorem 1.1 (1) We shall apply Lemma 2.1 to the functional
I ( u ) = 1 2 ( | u | 2 V ( x ) | u | 2 ) + 1 2 λ u 2 Ω G 0 ( x , u )
defined on X = H 1 ( R N ) . We consider only the case λ σ ( S ) , and
G 0 ( x , u ) 0 for  | u | δ , x R N .
(3.1)

Then other case is similar and simple.

Let X 2 be a finite dimensional space spanned by the eigenfunctions corresponding to negative eigenvalues of S + λ and let X 1 be its orthogonal complement in X. Choose a Hilbertian basis e n ( n 0 ) for X and define
X n 1 = span ( e 0 , e 1 , , e n ) , n N ; X n 2 = X 2 , n N ; X 1 = n N X n 1 ¯ .
By the condition (H1) and Sobolev inequalities, it is easy to see that the functional I belongs to C 1 ( X , R ) and maps bounded sets to bounded sets.
  1. (2)

    We claim that I has a local linking at 0 with respect to ( X 1 , X 2 ) . Decompose X 1 into V + W when V = E λ , W = ( X 2 + V ) . Also, set u = v + w , u X 1 , v V , w W .

     
For the convenience of our proof, we state some facts for the norm of the whole space X. It is well known that there is an equivalent norm on X = H 1 ( R N ) such that
( | u | 2 V ( x ) | u | 2 ) = u 2 , u X 2
and
( | u | 2 V ( x ) | u | 2 ) = u 2 , u W .
By the equivalence of norm in the finite-dimensional space, there exists C > 0 such that
v C v , v V .
(3.2)
It follows from (H1) and (H2) that for any ϵ > 0 , there exists C ϵ such that
| G 0 ( x , u ) | ϵ u 2 + C ϵ | u | s .
(3.3)
Hence, we obtain
I ( u ) m u 2 + c u s + 1 ,
where m > 0 , c is a constant and hence, for r > 0 small enough,
I ( u ) 0 , u X 2 , u X r .
Let u = v + w X 1 be such that u X r 1 = δ 2 C and let
A 1 = { x R N : | w ( x ) | δ 2 } , A 2 = R N A 1 .
From (3.2), we have
| v ( x ) | v C v δ 2
for all u r 1 and x R N . On the one hand, one has | u ( x ) | | v ( x ) | + | w ( x ) | v + δ 2 δ for all x A 1 . Hence, from (H5), we obtain
A 1 G 0 ( x , u ) d x 0 .
On the other hand, we have
| u ( x ) | | v ( x ) | + | w ( x ) | δ 2 + | w ( x ) | 2 | w ( x ) |
for all x A 2 . It follows from (3.3) that
G 0 ( x , u ) ϵ u 2 + C ϵ | u | s + 1 4 ϵ w 2 + 2 s + 1 C ϵ | w | s + 1
for all x A 2 and all u X 1 with u r 1 , which implies that
G 0 ( x , u ) 4 ϵ A 2 w 2 d x + A 2 2 s + 1 C ϵ | w | s + 1 d x 4 ( C 3 ) 2 ϵ w 2 + ( 2 C 3 ) λ + 1 C ϵ w s + 1 ,
where C 3 is a constant. Hence, there exist positive constants C , C 4 and C 5 such that
I ( u ) = 1 2 w 2 A 2 G 0 ( x , u ) d x A 1 G 0 ( x , u ) d x C w 2 4 ( C 3 ) 2 ϵ w 2 ( 2 C 3 ) λ + 1 C ϵ w s + 1 A 1 G ( x , u ) d x C 4 w 2 C 5 w s + 1
for all u X 1 with u r 1 , which implies that
I ( u ) 0 , u X 1  with  u r
for 0 < r small enough.
  1. (3)
    We claim that I satisfies ( C ) c . Consider a sequence ( u α n ) such that ( u α n ) is admissible and
    u α n X α n , I ( u α n ) c , ( 1 + u α n ) I ( u α n ) 0
    (3.4)
     
and
lim n ( 1 2 g 0 ( x , u α n ) u α n G 0 ( x , u α n ) ) = c .
(3.5)
Let w α n = u α n 1 u α n . Up to a subsequence, we have
w α n w in  X , w α n w in  L loc 2 , w α n ( x ) w ( x ) a.e.  x R N .
If w = 0 , we choose a sequence { t n } [ 0 , 1 ] such that
I ( t n u α n ) = max t [ 0 , 1 ] I ( t u α n ) .
For any m > 0 , let v α n = 2 m w α n . Now, we claim that
lim n G 0 ( x , v α n ) = 0 .
Let ϵ > 0 ; for r 1 , then,
| v α n | r G 0 ( x , v α n ) d x C 6 r p 2 | v α n | r | v α n | 2 d x C 7 r p 2 | v α n | 2 2 .
Since p < 2 , we may fix r large enough such that
| | v α n | r G 0 ( x , v α n ) d x | ϵ 3
for all n. Moreover, by (H6), there exists R > 0 such that
| | v α n | r G 0 ( x , v α n ) d x | | v α n | 2 2 sup | t | r , | x | R | G 0 ( x , t ) | t 2 ϵ 3
for all n. Finally, since v α n 0 in L s ( B R ( 0 ) ) for s [ 2 , 2 ) , we can use (H1) again to derive
| | v α n | r | x | R G 0 ( x , v α n ) d x | ϵ 3

for n large enough. Combining the above three formulas, our claim holds.

So, for n large enough, 2 m u α n 1 ( 0 , 1 ) , we have
I ( t n u α n ) I ( v α n ) m ϵ m 2 ,
(3.6)

where ϵ is a small enough constant.

That is, I ( t n u α n ) . Now, I ( 0 ) = 0 , I ( u α n ) c , we know that t n [ 0 , 1 ] and
( | ( t n u α n ) | 2 V ( x ) t n 2 | u α n | 2 ) + λ t n 2 | u α n | 2 g 0 ( x , t n u α n ) t n u α n = t n d d t | t = t n I ( t u α n ) = 0 .
(3.7)
Therefore, using (H4), we have
1 2 g 0 ( x , u α n ) u α n G 0 ( x , u α n ) 1 θ ( 1 2 g 0 ( x , t n u α n ) t n u α n G 0 ( x , t n u α n ) ) + .

This contradicts (3.5).

If w 0 , then the set = { x R N : w ( x ) 0 } has a positive Lebesgue measure. For x , we have | u α n ( x ) | . Hence, by (H3), we have
g 0 ( x , u α n ( x ) ) u α n ( x ) | u α n ( x ) | 2 | w α n ( x ) | 2 .
(3.8)
From (3.4), we obtain
1 ( 1 ) ( w 0 + w = 0 ) g 0 ( x , u α n ( x ) ) u α n ( x ) | u α n ( x ) | 2 | w α n ( x ) | 2 d x .
(3.9)

By (3.8), the right-hand side of (3.9) + . This is a contradiction.

In any case, we obtain a contradiction. Therefore, { u α n } is bounded.

Next, we denote { u α n } as { u n } and prove { u n } contains a convergent subsequence.

In fact, we know that { u n } is bounded in X. Passing to a subsequence, we may assume that u n u in X. In order to establish strong convergence, it suffices to show that
u n u .

By the condition (H6) and I ( u n ) , u n u 0 , we can similarly conclude it according to the above proof of our claim.

Finally, we claim that for every m N ,
I ( u ) as  u , u X m 1 X 2 .
By (H2) and (H3), there exist large enough M and some positive constant T such that
G 0 ( x , t ) M t 2 , x R N , t T .
So, for any u X m 1 X 2 , we have
I ( t u ) = 1 2 t 2 ( | u | 2 V ( x ) | u | 2 ) + t 2 2 λ u 2 G 0 ( x , t u ) 1 2 t 2 ( | u | 2 V ( x ) | u | 2 ) + t 2 2 λ u 2 M t 2 u 2 as  t + .

Hence, our claim holds. □

Proof of Theorem 1.2 We omit the proof which depends on Lemma 2.2 and is similar to the preceding one since our result is a variant of Ding Yanheng’s Theorem 1.2 (see [12]). □

Declarations

Acknowledgements

The authors would like to thank the referees for valuable comments and suggestions in improving this paper. This work was supported by the National NSF (Grant No. 10671156) of China.

Authors’ Affiliations

(1)
Institute of Mathematics, School of Mathematics and Computer Sciences, Nanjing Normal University
(2)
School of Mathematics and Statistics, Tianshui Normal University

References

  1. Bartolo P, Benci V, Fortunato D: Abstract critical point theorems and applications to some nonlinear problems with strong resonance at infinity. Nonlinear Anal. 1983, 7: 981–1012. 10.1016/0362-546X(83)90115-3MathSciNetView ArticleGoogle Scholar
  2. Capozzi A, Lupo D, Solimini S: On the existence of a nontrivial solution to nonlinear problems at resonance. Nonlinear Anal. 1989, 13: 151–163. 10.1016/0362-546X(89)90041-2MathSciNetView ArticleGoogle Scholar
  3. Costa DG, Silva EA: On a class of resonant problems at higher eigenvalues. Differ. Integral Equ. 1995, 8: 663–671.MathSciNetGoogle Scholar
  4. Costa DG, Tehrani HT:On a class of asymptotically linear elliptic problems in R N . J. Differ. Equ. 2001, 173: 470–494. 10.1006/jdeq.2000.3944MathSciNetView ArticleGoogle Scholar
  5. Stuart CA, Zhou HS:Applying the mountain pass theorem to an asymptotically linear elliptic equation on R N . Commun. Partial Differ. Equ. 1999, 24: 1731–1758. 10.1080/03605309908821481MathSciNetView ArticleGoogle Scholar
  6. Tehrani HT:A note on asymptotically linear elliptic problems in R N . J. Math. Anal. Appl. 2002, 271: 546–554. 10.1016/S0022-247X(02)00143-9MathSciNetView ArticleGoogle Scholar
  7. Li SJ, Willem M: Applications of local linking to critical point theory. J. Math. Anal. Appl. 1995, 189: 6–32. 10.1006/jmaa.1995.1002MathSciNetView ArticleGoogle Scholar
  8. Wang ZP, Zhou HS:Positive solutions for a nonhomogeneous elliptic equation on R N without (AR) condition. J. Math. Anal. Appl. 2009, 353: 470–479. 10.1016/j.jmaa.2008.11.080MathSciNetView ArticleGoogle Scholar
  9. Liu CY, Wang ZP, Zhou HS: Asymptotically linear Schrödinger equation with potential vanishing at infinity. J. Differ. Equ. 2008, 245: 201–222. 10.1016/j.jde.2008.01.006MathSciNetView ArticleGoogle Scholar
  10. Jeanjean L, Tanaka K:A positive solution for a nonlinear Schrödinger equation on R N . Indiana Univ. Math. J. 2005, 54: 443–464. 10.1512/iumj.2005.54.2502MathSciNetView ArticleGoogle Scholar
  11. Teng KM: Existence and multiplicity results for some elliptic systems with discontinuous nonlinearities. Nonlinear Anal. 2012, 75: 2975–2987. 10.1016/j.na.2011.11.040MathSciNetView ArticleGoogle Scholar
  12. Ding YH:Some existence results of solutions for the semilinear elliptic equations on R N . J. Differ. Equ. 1995, 119: 401–425. 10.1006/jdeq.1995.1096View ArticleGoogle Scholar

Copyright

© Pei and Zhang; licensee Springer. 2013

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.