Journal of Inequalities and Applications
Volume 2010 (2010), Article ID 850602, 16 pages
doi:10.1155/2010/850602
Research Article

An Application of Category Theory to a Class of the Systems of the Superquadratic Wave Equations

1Department of Mathematics, Kunsan National University, Kunsan 573-701, South Korea
2Department of Mathematics Education, Inha University, Incheon 402-751, South Korea

Received 16 October 2009; Accepted 7 February 2010

Academic Editor: Jong Kkyungnam Kim

Copyright © 2010 Tacksun Jung and Q-Heung Choi. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

We investigate the nontrivial solutions for a class of the systems of the superquadratic nonlinear wave equations with Dirichlet boundary condition and periodic condition with a superquadratic nonlinear terms at infinity which have continuous derivatives. We approach the variational method and use the critical point theory on the manifold, in terms of the limit relative category of the sublevel subsets of the corresponding functional.

1. Introduction

We investigate the nontrivial solutions for a class of the systems of the superquadratic nonlinear wave equations with Dirichlet boundary condition and periodic condition: 𝑢 𝑡 𝑡 𝑢 𝑥 𝑥 = 𝑎 𝑣 + 𝐹 𝑢 𝜋 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) i n 2 , 𝜋 2 𝑣 × 𝑅 , 𝑡 𝑡 𝑣 𝑥 𝑥 = 𝑏 𝑢 + 𝐹 𝑣 𝜋 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) i n 2 , 𝜋 2 𝑢 ± 𝜋 × 𝑅 , 2 ± 𝜋 , 𝑡 = 𝑣 2 , 𝑡 = 0 , 𝑢 ( 𝑥 , 𝑡 + 𝜋 ) = 𝑢 ( 𝑥 , 𝑡 ) = 𝑢 ( 𝑥 , 𝑡 ) = 𝑢 ( 𝑥 , 𝑡 ) , 𝑣 ( 𝑥 , 𝑡 + 𝜋 ) = 𝑣 ( 𝑥 , 𝑡 ) = 𝑣 ( 𝑥 , 𝑡 ) = 𝑣 ( 𝑥 , 𝑡 ) , ( 1 . 1 ) where 𝐹 [ 𝜋 / 2 , 𝜋 / 2 ] × 𝑅 × 𝑅 × 𝑅 𝑅 is a superquadratic function at infinity which has continuous derivatives 𝐹 𝑟 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) , 𝐹 𝑠 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) with respect to 𝑟 , 𝑠 , for almost any ( 𝑥 , 𝑡 ) ( 𝜋 / 2 , 𝜋 / 2 ) × 𝑅 . Moreover we assume that 𝐹 satisfies the following conditions:(F1) 𝐹 ( 𝑥 , 𝑡 , 0 , 0 )  =  𝐹 𝑥 ( 𝑥 , 𝑡 , 0 , 0 )  =  𝐹 𝑡 ( 𝑥 , 𝑡 , 0 , 0 )  =  0 , 𝐹 𝑥 𝑥 ( 𝑥 , 𝑡 , 0 , 0 )  =  𝐹 𝑡 𝑡 ( 𝑥 , 𝑡 , 0 , 0 )  =  𝐹 𝑥 𝑡 ( 𝑥 , 𝑡 , 0 , 0 )   =  0 , 𝐹 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) > 0 if ( 𝑟 , 𝑠 ) ( 0 , 0 ) , i n f ( 𝑥 , 𝑡 ) ( 𝜋 / 2 , 𝜋 / 2 ) × 𝑅 , | 𝑟 | 2 + | 𝑠 | 2 = 𝑅 2 𝐹 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) > 0 ; ( F 2 ) | 𝐹 𝑟 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) | + | 𝐹 𝑠 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) | 𝑐 ( | 𝑟 | 𝜈 + | 𝑠 | 𝜈 ) for all 𝑥 , 𝑡 , 𝑟 , 𝑠 ;(F3) 𝑟 𝐹 𝑟 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) + 𝑠 𝐹 𝑠 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) 𝜇 𝐹 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) for all 𝑥 , 𝑡 , 𝑟 , 𝑠 ;(F4) | 𝐹 𝑟 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) | + | 𝐹 𝑠 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) | 𝑑 ( 𝐹 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) 𝛿 1 + 𝐹 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) 𝛿 2 ) ,

where 𝑐 > 0 , 𝑑 > 0 , 𝑅 > 0 , 𝜇 > 2 , 𝜈 > 1 and 1 / 2 < 𝛿 1 𝛿 2 1 / 𝑟 , for some 1 < 𝑟 < 2 .

As the physical model for these systems we can find crossing two beams with travelling waves, which are suspended by cable under a load. The nonlinearity 𝑢 + models the fact that cables resist expansion but do not resist compression.

Choi and Jung [13] investigate the existence and multiplicity of solutions of the single nonlinear wave equation with Dirichlet boundary condition. In [4] the authors show by critical point theory (Linking Theorem) that system (1.1) has at least one nontrivial solution ( 𝑢 , 𝑣 ) . In this paper we show by the limit relative category theory that system (1.1) has at least two nontrivial solutions ( 𝑢 , 𝑣 ) .

Let us set ( 𝑢 , 𝑣 ) = ( 𝐿 𝑢 , 𝐿 𝑣 ) , 𝐿 𝑢 = 𝑢 𝑡 𝑡 𝑢 𝑥 𝑥 . ( 1 . 2 ) Then system (1.1) can be rewritten by 1 𝑈 = 2 , 𝑈 ± 𝜋 ( 𝐴 𝑈 , 𝑈 ) + 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 2 = 0 0 , , 𝑡 𝑈 ( 𝑥 , 𝑡 + 𝜋 ) = 𝑈 ( 𝑥 , 𝑡 ) = 𝑈 ( 𝑥 , 𝑡 ) = 𝑈 ( 𝑥 , 𝑡 ) , ( 1 . 3 ) where is the gradient operator, 𝑈 = ( 𝑢 𝑣 ) , 𝐴 = ( 0 𝑎 𝑏 0 ) 𝑀 2 × 2 ( 𝑅 ) .

We note that 𝑎 𝑏 , 𝑎 𝑏 are two eigenvalues of the matrix 𝐴 = ( 0 𝑎 𝑏 0 ) .

Let 𝜆 𝑚 𝑛 be the eigenvalues of the eigenvalue problem 𝑢 𝑡 𝑡 𝑢 𝑥 𝑥 = 𝜆 𝑢 in ( 𝜋 / 2 , 𝜋 / 2 ) × 𝑅 , 𝑢 ( ± 𝜋 / 2 , 𝑡 ) = 0 , 𝑢 ( 𝑥 , 𝑡 + 𝜋 ) = 𝑢 ( 𝑥 , 𝑡 ) = 𝑢 ( 𝑥 , 𝑡 ) = 𝑢 ( 𝑥 , 𝑡 ) .

Our main result is the following.

Theorem 1.1. Assume that 𝜆 2 𝑚 𝑛 𝑎 𝑏 0 𝑚 , 𝑛 w i t h ( 𝑚 , 𝑛 ) ( 0 , 0 ) , ( 1 . 4 ) 𝑎 > 0 , 𝑏 > 0 , ( 1 . 5 ) 𝑎 𝑏 < 1 . ( 1 . 6 ) Then, for any 𝐹 with (F1), (F2), (F3) and (F4), system (1.3) has at least two nontrivial solutions ( 𝑢 , 𝑣 ) .

In Section 2, we obtain some results on the nonlinear term 𝐹 . In Section 3, we approach the variational method and recall the abstract results of the critical point theory on the manifold in terms of the limit relative category of the sublevel sets of the corresponding functional of (1.3), which plays a crucial role to prove the multiplicity result. In Section 4, we prove Theorem 1.1.

2. Some Results on the Nonlinear Term 𝐹

The eigenvalue problem for 𝑢 ( 𝑥 , 𝑡 ) 𝑢 𝑡 𝑡 𝑢 𝑥 𝑥 𝜋 = 𝜆 𝑢 i n 2 , 𝜋 2 𝑢 ± 𝜋 × 𝑅 , 2 , 𝑡 = 0 , 𝑢 ( 𝑥 , 𝑡 + 𝜋 ) = 𝑢 ( 𝑥 , 𝑡 ) = 𝑢 ( 𝑥 , 𝑡 ) = 𝑢 ( 𝑥 , 𝑡 ) ( 2 . 1 ) has infinitely many eigenvalues 𝜆 𝑚 𝑛 = ( 2 𝑛 + 1 ) 2 4 𝑚 2 ( 𝑚 , 𝑛 = 0 , 1 , 2 , ) ( 2 . 2 ) and corresponding normalized eigenfunctions 𝜙 𝑚 𝑛 ( 𝑚 , 𝑛 0 ) given by 𝜙 0 𝑛 = 2 𝜋 𝜙 c o s ( 2 𝑛 + 1 ) 𝑥 f o r 𝑛 0 , 𝑚 𝑛 = 2 𝜋 c o s 2 𝑚 𝑡 c o s ( 2 𝑛 + 1 ) 𝑥 f o r 𝑚 > 0 , 𝑛 0 . ( 2 . 3 ) Let 𝑄 be the square [ 𝜋 / 2 , 𝜋 / 2 ] × [ 𝜋 / 2 , 𝜋 / 2 ] and 𝐻 0 the Hilbert space defined by 𝐻 0 = 𝑢 𝐿 2 ( 𝑄 ) 𝑢 i s e v e n i n 𝑥 a n d 𝑡 a n d 𝑄 . 𝑢 = 0 ( 2 . 4 ) The set of functions { 𝜙 𝑚 𝑛 } is an orthonormal basis in 𝐻 0 . Let us denote an element 𝑢 , in 𝐻 0 , by 𝑢 = 𝑚 𝑛 𝜙 𝑚 𝑛 . ( 2 . 5 ) We define a Hilbert space 𝒟 as follows: 𝒟 = 𝑢 𝑚 𝑛 𝜙 𝑚 𝑛 𝑚 𝑛 𝜆 2 𝑚 𝑛 2 𝑚 𝑛 < + . ( 2 . 6 ) Then this space is a Banach space with norm 𝜆 𝑢 = 2 𝑚 𝑛 2 𝑚 𝑛 1 / 2 . ( 2 . 7 ) Let us set 𝐸 = 𝒟 × 𝒟 . We endow the Hilbert space 𝐸 with the norm ( 𝑢 , 𝑣 ) 2 𝐸 = 𝑢 2 + 𝑣 2 . ( 2 . 8 ) We are looking for the weak solutions of (1.3) in 𝒟 × 𝒟 , that is, ( 𝑢 , 𝑣 ) such that 𝑢 𝒟 , 𝑣 𝒟 , 𝐿 𝑢 = 𝑎 𝑣 + 𝐹 𝑢 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) , 𝐿 𝑣 = 𝑏 𝑢 + 𝐹 𝑣 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) . Since | 𝜆 𝑚 𝑛 | 1 for all 𝑚 , 𝑛 , we have the following lemma.

We state the lemmas. For the proofs of Lemmas 2.1, 2.2, and 2.3, we refer [4].

Lemma 2.1. (i) 𝑢 𝑢 𝐿 2 ( 𝑄 ) , where 𝑢 𝐿 2 ( 𝑄 ) denotes the 𝐿 2 norm of 𝑢 .
(ii) 𝑢 = 0 if and only if 𝑢 𝐿 2 ( 𝑄 ) = 0 .
(iii) 𝑢 𝑡 𝑡 𝑢 𝑥 𝑥 𝒟 implies 𝑢 𝒟 .

Lemma 2.2. Suppose that 𝑐 is not an eigenvalue of 𝐿 𝒟 𝐻 0 , 𝐿 𝑢 = 𝑢 𝑡 𝑡 𝑢 𝑥 𝑥 , and let 𝑓 𝐻 0 . Then one has ( 𝐿 𝑐 ) 1 𝑓 𝒟 .

By (F1) and (F3), we obtain the lower bound for 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) in the term of | 𝑢 | 𝜇 + | 𝑣 | 𝜇 .

Lemma 2.3. Assume that 𝐹 satisfies the conditions (F1) and (F3). Then there exist 𝑎 0 , 𝑏 0 𝑅 with 𝑎 0 > 0 such that 𝐹 ( 𝑥 , 𝑡 , 𝑟 , 𝑠 ) 𝑎 0 ( | 𝑟 | 𝜇 + | 𝑠 | 𝜇 ) 𝑏 0 , 𝑥 , 𝑡 , 𝑟 , 𝑠 . ( 2 . 9 )

Lemma 2.4. Assume that 𝐹 satisfies the conditions (F1), (F2), and (F3). Then (i) 𝑄 𝐹 ( 𝑥 , 𝑡 , 0 , 0 ) 𝑑 𝑥 𝑑 𝑡  =  0 ,   𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 > 0   if   ( 𝑢 , 𝑣 ) ( 0 , 0 ) ,   g r a d ( 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) ) 𝑑 𝑥 𝑑 𝑡  =  𝑜 ( ( 𝑢 , 𝑣 ) 𝐸 ) as ( 𝑢 , 𝑣 ) ( 0 , 0 ) ;(ii)there exist 𝑎 0 > 0 , 𝜇 > 2 and 𝑏 1 𝑅 such that 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 𝑎 0 ( 𝑢 , 𝑣 ) 𝜇 𝐿 𝜇 𝑏 1 ( 𝑢 , 𝑣 ) 𝐸 ; ( 2 . 1 0 ) (iii) ( 𝑢 , 𝑣 ) g r a d ( 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 𝑣 ) ) 𝑑 𝑥 𝑑 𝑡 is a compact map;(iv)if   𝑄 [ 𝑢 𝐹 𝑢 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 )  +  𝑣 𝐹 𝑣 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) ] 𝑑 𝑥 𝑑 𝑡 2 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡  =  0 ,  then   g r a d ( 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 )  =  0 ;(v)if ( 𝑢 𝑛 , 𝑣 𝑛 ) 𝐸 + and 𝑄 𝑢 𝑛 𝐹 𝑢 𝑥 , 𝑡 , 𝑢 𝑛 , 𝑣 𝑛 + 𝑣 𝑛 𝐹 𝑣 𝑥 , 𝑡 , 𝑢 𝑛 , 𝑣 𝑛 𝑑 𝑥 𝑑 𝑡 2 𝑄 𝐹 𝑥 , 𝑡 , 𝑢 𝑛 , 𝑣 𝑛 𝑑 𝑥 𝑑 𝑡 ( 𝑢 , 𝑣 ) 𝐸 0 , ( 2 . 1 1 ) then there exists ( ( 𝑢 𝑛 , 𝑣 𝑛 ) ) 𝑛 and 𝑤 𝐸 such that g r a d 𝑄 𝐹 𝑥 , 𝑦 , 𝑢 𝑛 , 𝑣 𝑛 𝑑 𝑥 𝑑 𝑡 ( 𝑢 𝑛 , 𝑣 𝑛 ) 𝐸 𝑢 𝑤 , 𝑛 , 𝑣 𝑛 ( 𝑢 𝑛 , 𝑣 𝑛 ) 𝐸 ( 0 , 0 ) . ( 2 . 1 2 )

Proof. (i) Follows from (F1) and (F2), since 1 < 𝜈 .
(ii) By Lemma 2.3, for 𝑈 = ( 𝑢 , 𝑣 ) 𝐸 , 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝑑 𝑥 𝑑 𝑡 𝑎 0 𝑈 𝜇 𝐿 𝜇 𝑑 𝑥 𝑑 𝑡 𝑏 1 , ( 2 . 1 3 ) where 𝑏 1 𝑅 . Thus (ii) holds.
(iii) Is easily obtained with standard arguments.
(iv) Is implied by (F3) and the fact that 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) > 0 for ( 𝑢 , 𝑣 ) ( 0 , 0 ) .
(v) By Lemma 2.3 and (F3), for 𝑈 = ( 𝑢 , 𝑣 ) , 𝑄 𝑢 𝐹 𝑢 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) + 𝑣 𝐹 𝑣 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 2 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 ( 𝜇 2 ) 𝑄 𝑎 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 ( 𝜇 2 ) 0 𝑈 𝜇 𝐿 𝜇 𝑏 1 . ( 2 . 1 4 )
By (F2), g r a d 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 𝐸 𝐶 𝐹 𝑈 ( 𝑥 , 𝑡 , 𝑈 ) 𝐿 𝑟 𝐶 | 𝑈 | 𝜈 𝐿 𝑟 , f o r s o m e 1 < 𝑟 < 2 ( 2 . 1 5 ) and suitable constants 𝐶 , 𝐶 . To get the conclusion it suffices to estimate | 𝑈 | 𝜈 / 𝑈 𝐸 𝐿 𝑟 in terms of 𝑈 𝜇 𝐿 𝜇 / 𝑈 𝐸 . If 𝜇 𝑟 𝜈 , then this is a consequence of Hölder inequality. If 𝜇 < 𝑟 𝜈 , by the standard interpolation arguments, it follows that | 𝑈 | 𝜈 / 𝑈 𝐸 𝐿 𝑟 𝐶 ( 𝑈 𝜇 𝐿 𝜇 / 𝑈 𝐸 ) 𝜈 / 𝜇 𝑈 𝑙 𝐸 , where 𝑙 is such that 𝑙 = 1 + 𝜈 / 𝜇 . Thus we prove (v).

Lemma 2.5. Assume that 𝐹 satisfies the conditions (F1), (F2), (F3), and (F4). Then there exist 𝜑 , 𝜓 [ 0 , + ] 𝑅 continuous and such that 𝜓 ( 𝑠 ) 𝑠 0 a s 𝑠 0 , 𝜑 ( 𝑠 ) > 0 i f 𝑠 > 0 , ( 2 . 1 6 ) (i) 𝑔 𝑟 𝑎 𝑑 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 2 𝐸 𝜓 ( 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 ) , for all ( 𝑢 , 𝑣 ) 𝐸 ,(ii) 𝑄 [ 𝑢 𝐹 𝑢 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) + 𝑣 𝐹 𝑣 ( 𝑥 , 𝑡 𝑢 , 𝑣 ) ] 𝑑 𝑥 𝑑 𝑡 2 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 𝜑 ( 𝑢 , 𝑣 ) , for all ( 𝑢 , 𝑣 ) 𝐸 .

Proof. (i) By (F4), for all 𝑈 = ( 𝑢 , 𝑣 ) 𝐸 , g r a d 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝑑 𝑥 𝑑 𝑡 𝐸 𝐹 𝑈 ( 𝑥 , 𝑡 , 𝑈 ) 𝐿 𝑟 𝐶 1 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 1 + 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 2 𝐿 𝑟 𝐶 2 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 1 𝐿 𝑟 + 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 2 𝐿 𝑟 𝐶 3 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 1 𝐿 1 1 / 𝛿 + 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 2 𝐿 2 1 / 𝛿 𝐶 4 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 1 𝐿 1 + 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝛿 2 𝐿 1 = 𝐶 5 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝑑 𝑥 𝑑 𝑡 𝛿 1 + 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝑑 𝑥 𝑑 𝑡 𝛿 2 , ( 2 . 1 7 ) where 1 < 𝑟 < 1 / 𝛿 1 , 1 / 𝛿 2 < 2 , 𝐶 1 , 𝐶 2 , 𝐶 3 , 𝐶 4 , and 𝐶 5 are constants. Since 𝛿 1 , 𝛿 2 > 1 / 2 , (i) follows.
(ii) By (F3), 𝑄 𝑢 𝐹 𝑢 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) + 𝑣 𝐹 𝑣 ( 𝑥 , 𝑡 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 2 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 ( 𝜇 2 ) 𝑄 𝑎 𝐹 ( 𝑥 , 𝑡 , 𝑈 ) 𝑑 𝑥 𝑑 ( 𝜇 2 ) 0 𝑈 𝜇 𝐿 𝜇 𝑏 1 . ( 2 . 1 8 ) Thus (ii) follows.

3. Abstract Results of Critical Point Theory

Now we are looking for the weak solutions of system (1.3). We shall approach the variational method and recall the abstract results of the critical point theory on the manifold in terms of the limit relative category of the sublevel sets of the functional of (1.3). We observe that the weak solutions of (1.3) coincide with the critical points of the corresponding functional: 𝐼 𝐸 𝑅 𝐶 1 , 1 , 𝐼 1 ( 3 . 1 ) ( 𝑈 ) = 2 𝑄 1 𝑈 𝑈 𝑑 𝑥 𝑑 𝑡 2 𝑄 ( 𝐴 𝑈 , 𝑈 ) 𝑅 2 𝑑 𝑥 𝑑 𝑡 𝑄 𝐹 ( 𝑥 , 𝑡 , 𝑢 , 𝑣 ) 𝑑 𝑥 𝑑 𝑡 . ( 3 . 2 ) Now we recall the critical point theory for strongly indefinite functional. Since the functional 𝐼 is strongly indefinite functional, it is convenient to use ( 𝑃 . 𝑆 . ) 𝑐 condition and the limit relative category which is a suitable version of ( 𝑃 . 𝑆 . ) 𝑐 condition and the relative category, respectively.

Now, we consider the critical point theory on the manifold with boundary. Let 𝐸 be a Hilbert space and let 𝑀 be the closure of an open subset of 𝐸 such that 𝑀 can be endowed with the structure of 𝐶 2 manifold with boundary. Let 𝑓 𝑊 𝑅 be a 𝐶 1 , 1 functional, where 𝑊 is an open set containing 𝑀 . For applying the usual topological methods of critical points theory we need a suitable notion of critical point for 𝑓 on 𝑀 . We recall the following notions: lower gradient of 𝑓 on 𝑀 , ( 𝑃 . 𝑆 . ) 𝑐 condition, and the limit relative category (see [4]).

Definition 3.1. If 𝑢 𝑀 , the lower gradient of 𝑓 on 𝑀 at 𝑢 is defined by g r a d 𝑀 𝑓 ( 𝑢 ) = 𝑓 ( 𝑢 ) i f 𝑢 i n t ( 𝑀 ) , 𝑓 ( 𝑢 ) + 𝑓 ( 𝑢 ) , 𝜈 ( 𝑢 ) 𝜈 ( 𝑢 ) i f 𝑢 𝜕 𝑀 , ( 3 . 3 ) where we denote by 𝜈 ( 𝑢 ) the unit normal vector to 𝜕 𝑀 at the point 𝑢 , pointing outwards. We say that 𝑢 is a lower critical for 𝑓 on 𝑀 , if g r a d 𝑀 𝑓 ( 𝑢 ) = 0 .
Since the functional 𝐼 ( 𝑢 ) (which is introduced in Section 4) is strongly indefinite, the notion of the ( 𝑃 . 𝑆 . ) 𝑐 condition and the limit relative category is a very useful tool for the proof of the main theorems.
Let 𝐸 , 𝐸 0 , 𝐸 + be the subspace of 𝐸 on which the functional 𝑈 ( 1 / 2 ) 𝑄 𝑈 𝑈 is positive definite, null, negative definite, and 𝐸 , 𝐸 0 , and 𝐸 + are mutually orthogonal. Let 𝑃 + be the projection for 𝐸 onto 𝐸 + , 𝑃 0 the one from 𝐸 onto 𝐸 0 , and 𝑃 the one from 𝐸 onto 𝐸 . Let ( 𝐸 𝑛 ) 𝑛 be a sequence of closed subspaces of 𝐸 with the conditions: 𝐸 𝑛 = 𝐸 𝑛 𝐸 0 𝐸 + 𝑛 , w h e r e 𝐸 + 𝑛 𝐸 + , 𝐸 𝑛 𝐸 𝑛 ( 3 . 4 ) ( 𝐸 + 𝑛 and 𝐸 𝑛 are subspaces of  𝐸 ) , d i m 𝐸 𝑛 < + , 𝐸 𝑛 𝐸 𝑛 + 1 , 𝑛 𝑁 𝐸 𝑛 is dense in 𝐸 .
Let 𝑃 𝐸 𝑛 be the orthogonal projections from 𝐸 onto 𝐸 𝑛 . 𝑀 𝑛 = 𝑀 𝐸 𝑛 , for any 𝑛 , and let be the closure of an open subset of 𝐸 𝑛 and have the structure of a 𝐶 2 manifold with boundary in 𝐸 𝑛 . We assume that for any 𝑛 there exists a retraction 𝑟 𝑛 𝑀 𝑀 𝑛 . For given 𝐵 𝐸 , we will write 𝐵 𝑛 = 𝐵 𝐸 𝑛 .

Definition 3.2. Let 𝑐 𝑅 . We say that 𝑓 satisfies the ( 𝑃 . 𝑆 . ) 𝑐 condition with respect to ( 𝑀 𝑛 ) 𝑛 , on the manifold with boundary 𝑀 , if for any sequence ( 𝑘 𝑛 ) 𝑛 in 𝑁 and any sequence ( 𝑢 𝑛 ) 𝑛 in 𝑀 such that 𝑘 𝑛 , 𝑢 𝑛 𝑀 𝑘 𝑛 , for all 𝑛 , 𝑓 ( 𝑢 𝑛 ) 𝑐 , g r a d 𝑀 𝑘 𝑛 𝑓 ( 𝑢 𝑛 ) 0 , there exists a subsequence of ( 𝑢 𝑛 ) 𝑛 which converges to a point 𝑢 𝑀 such that g r a d 𝑀 𝑓 ( 𝑢 ) = 0 .

Let 𝑌 be a closed subspace of 𝑀 .

Definition 3.3. Let 𝐵 be a closed subset of 𝑀 with 𝑌 𝐵 . We define the relative category c a t 𝑀 , 𝑌 ( 𝐵 ) of 𝐵 in ( 𝑀 , 𝑌 ) , as the least integer such that there exist + 1 closed subsets 𝑈 0 , 𝑈 1 , , 𝑈 with the following properties: 𝐵 𝑈 0 𝑈 1 𝑈 ; 𝑈 1 , , 𝑈 are contractible in 𝑀 ; 𝑌 𝑈 0 and there exists a continuous map 𝐹 𝑈 0 × [ 0 , 1 ] 𝑀 such that 𝐹 ( 𝑥 , 0 ) = 𝑥 𝑥 𝑈 0 , 𝐹 [ ] , ( 𝑥 , 𝑡 ) 𝑌 𝑥 𝑌 , 𝑡 0 , 1 𝐹 ( 𝑥 , 1 ) 𝑌 𝑥 𝑈 0 . ( 3 . 5 ) If such an does not exist, we say that c a t 𝑀 , 𝑌 ( 𝐵 ) = + .

Definition 3.4. Let ( 𝑋 , 𝑌 ) be a topological pair and let ( 𝑋 𝑛 ) 𝑛 be a sequence of subsets of 𝑋 . For any subset 𝐵 of 𝑋 we define the limit relative category of 𝐵 in ( 𝑋 , 𝑌 ) , with respect to ( 𝑋 𝑛 ) 𝑛 , by c a t ( 𝑋 , 𝑌 ) ( 𝐵 ) = l i m s u p 𝑛 c a t ( 𝑋 𝑛 , 𝑌 𝑛 ) 𝐵 𝑛 . ( 3 . 6 )

Now we consider a theorem which gives an estimate of the number of critical points of a functional, in terms of the limit relative category of its sublevels. The theorem is proved repeating the classical arguments, using the nonsmooth version of the classical Deformation Lemma for functions on manifolds with boundary.

Let 𝑌 be a fixed subset of 𝑀 . We set 𝑖 = 𝐵 𝑀 c a t ( 𝑀 , 𝑌 ) , 𝑐 ( 𝐵 ) 𝑖 𝑖 = i n f 𝐵 𝑖 s u p 𝑥 𝐵 𝑓 ( 𝑥 ) . ( 3 . 7 ) We have the following multiplicity theorem.

Theorem 3.5. Let 𝑖 𝑁 and assume that (1) 𝑐 𝑖 < + ,(2) s u p 𝑥 𝑌 𝑓 ( 𝑥 ) < 𝑐 𝑖 ,(3)the ( 𝑃 . 𝑆 . ) 𝑐 𝑖 condition with respect to ( 𝑀 𝑛 ) 𝑛 holds.Then there exists a lower critical point 𝑥 such that 𝑓 ( 𝑥 ) = 𝑐 𝑖 . If 𝑐 𝑖 = 𝑐 𝑖 + 1 = = 𝑐 𝑖 + 𝑘 1 = 𝑐 , ( 3 . 8 ) then c a t 𝑀 𝑥 𝑀 𝑓 ( 𝑥 ) = 𝑐 , g r a d 𝑀 𝑓 ( 𝑥 ) = 0 𝑘 . ( 3 . 9 )

Proof. Let 𝑐 = 𝑐 𝑖 ; using the ( 𝑃 . 𝑆 . ) 𝑐 condition, with respect to ( 𝑀 𝑛 ) 𝑛 , one can prove that, for any neighbourhood 𝑁 of 𝐾 𝑐 = 𝑥 𝑓 ( 𝑥 ) = 𝑐 , g r a d 𝑀 𝑓 ( 𝑥 ) = 0 , ( 3 . 1 0 ) there exist 𝑛 0 in 𝑁 and 𝛿 > 0 such that g r a d 𝑀 𝛿 for all 𝑛 𝑛 0 and all 𝑥 𝐸 𝑛 𝑁 with 𝑐 𝛿 𝑓 ( 𝑥 ) 𝑐 + 𝛿 . Moreover it is not difficult to see that, for all 𝑛 , the function 𝑓 𝑛 𝐸 𝑛 𝑅 { + } defined by 𝑓 𝑛 = 𝑓 ( 𝑥 ) , if 𝑥 𝑀 𝑛 , 𝑓 𝑛 ( 𝑥 ) = + , otherwise, is 𝜙 -convex of order two, according to the definitions of [3]. Then the conclusion follows using the same arguments of [4, 5] and the nonsmooth version of the classical Deformation Lemma.

Lemma 3.6 (Deformation Lemma). Let 𝐻 𝑅 { + } be a lower semicontinuous function and assume to be 𝜑 -convex of order 2 (see [3]). Let 𝑐 𝑅 , 𝛿 > 0 , and 𝐷 be a closed set in 𝐻 such that i n f g r a d 𝑀 ( 𝑥 ) 𝑐 𝛿 ( 𝑥 ) 𝑐 + 𝛿 , d i s t ( 𝑥 , 𝐷 ) < 𝛿 > 0 . ( 3 . 1 1 ) Then there exists 𝜖 > 0 and a continuous deformation 𝜂 𝑐 + 𝜖 𝐷 × [ 0 , 1 ] 𝑐 + 𝜖 𝐷 𝛿 ( 𝐷 𝛿 is the 𝛿 -neighborhood of 𝐷 and 𝑐 = { 𝑥 ( 𝑥 ) 0 } ) such that (i) 𝜂 ( 𝑥 , 0 ) = 𝑥 for all 𝑥 𝑐 + 𝜖 𝐷 ,(ii) 𝜂 ( 𝑥 , 𝑡 ) = 𝑥 for all 𝑥 𝑐 𝜖 𝐷 , for all 𝑡 [ 0 , 1 ] ,(iii) 𝜂 ( 𝑥 , 1 ) 𝑐 𝜖 for all 𝑥 𝑐 + 𝜖 𝐷 , for all 𝑡 [ 0 , 1 ] .

Proof. See [6, Lemmas 4 . 5 and 4.6].

Now we state the following multiplicity result (for the proof see [7, Theorem 4 . 6 ]) which will be used in the proofs of our main theorems.

Theorem 3.7. Let 𝐻 be a Hilbert space and let 𝐻 = 𝑋 1 𝑋 2 𝑋 3 , where 𝑋 1 , 𝑋 2 , 𝑋 3 are three closed subspaces of 𝐻 with 𝑋 1 , 𝑋 2 of finite dimension. For a given subspace 𝑋 of 𝐻 , let 𝑃 𝑋 be the orthogonal projection from 𝐻 onto 𝑋 . Set 𝑃 𝐶 = 𝑥 𝐻 𝑋 2 𝑥 , 1 ( 3 . 1 2 ) and let 𝑓 𝑊 𝑅 be a 𝐶 1 , 1 function defined on a neighborhood 𝑊 of 𝐶 . Let 1 < 𝜌 < 𝑅 , 𝑅 1 > 0 . We define 𝑥 Δ = 1 + 𝑥 2 𝑥 1 𝑋 1 , 𝑥 2 𝑋 2 , 𝑥 1 𝑅 1 𝑥 , 1 2 , 𝑥 𝑅 Σ = 1 + 𝑥 2 𝑥 1 𝑋 1 , 𝑥 2 𝑋 2 , 𝑥 1 𝑅 1 , 𝑥 2 𝑥 = 1 1 + 𝑥 2 𝑥 1 𝑋 1 , 𝑥 2 𝑋 2 , 𝑥 1 𝑅 1 , 𝑥 2 𝑥 = 𝑅 1 + 𝑥 2 𝑥 1 𝑋 1 , 𝑥 2 𝑋 2 , 𝑥 1 = 𝑅 1 𝑥 , 1 2 , 𝑅 𝑆 = 𝑥 𝑋 2 𝑋 3 , 𝑥 = 𝜌 𝐵 = 𝑥 𝑋 2 𝑋 3 . 𝑥 𝜌 ( 3 . 1 3 ) Assume that s u p 𝑓 ( Σ ) < i n f 𝑓 ( 𝑆 ) ( 3 . 1 4 ) and that the ( 𝑃 . 𝑆 . ) 𝑐 condition holds for 𝑓 on 𝐶 , with respect to the sequence ( 𝐶 𝑛 ) 𝑛 , for all 𝑐 [ 𝛼 , 𝛽 ] , where 𝛼 = i n f 𝑓 ( 𝑆 ) , 𝛽 = s u p 𝑓 ( Δ ) . ( 3 . 1 5 ) Moreover one assumes 𝛽 < + and 𝑓 | 𝑋 1 𝑋 3 has no critical points 𝑧 in 𝑋 1 𝑋 3 with 𝛼 𝑓 ( 𝑧 ) 𝛽 . Then there exist two lower critical points 𝑧 1 ,